Authors: Ching-Yun Ko, Pin-Yu Chen
Abstract: Modern artificial intelligence (AI) models are deployed on inference engines to optimize runtime efficiency and resource allocation, particularly for transformer-based large language models (LLMs). The vLLM project is a major open-source library to support model serving and inference. However, the current implementation of vLLM limits programmability of the internal states of deployed models. This prevents the use of popular test-time model alignment and enhancement methods. For example, it prevents the detection of adversarial prompts based on attention patterns or the adjustment of model responses based on activation steering. To bridge this critical gap, we present vLLM Hook, an opensource plug-in to enable the programming of internal states for vLLM models. Based on a configuration file specifying which internal states to capture, vLLM Hook provides seamless integration to vLLM and supports two essential features: passive programming and active programming. For passive programming, vLLM Hook probes the selected internal states for subsequent analysis, while keeping the model generation intact. For active programming, vLLM Hook enables efficient intervention of model generation by altering the selected internal states. In addition to presenting the core functions of vLLM Hook, in version 0, we demonstrate 3 use cases including prompt injection detection, enhanced retrieval-augmented retrieval (RAG), and activation steering. Finally, we welcome the community's contribution to improve vLLM Hook via https://github.com/ibm/vllm-hook.
Authors: Runyu Peng, Ruixiao Li, Mingshu Chen, Yunhua Zhou, Qipeng Guo, Xipeng Qiu
Abstract: Large Language Models (LLMs) often allocate disproportionate attention to specific tokens, a phenomenon commonly referred to as the attention sink. While such sinks are generally considered detrimental, prior studies have identified a notable exception: the model's consistent emphasis on the first token of the input sequence. This structural bias can influence a wide range of downstream applications and warrants careful consideration. Despite its prevalence, the precise mechanisms underlying the emergence and persistence of attention sinks remain poorly understood. In this work, we trace the formation of attention sinks around the first token of the input. We identify a simple mechanism, referred to as the P0 Sink Circuit, that enables the model to recognize token at position zero and induce an attention sink within two transformer blocks, without relying on any semantic information. This mechanism serves as the basis for the attention sink on position zero. Furthermore, by analyzing training traces from a 30B A3B MoE model trained from scratch, we find that this mechanism emerges early in training and becomes increasingly concentrated in the first two layers, suggesting a possible signal for tracking pre training convergence states.
Authors: Jiajun Xu, Jiageng Mao, Ang Qi, Weiduo Yuan, Alexander Romanus, Helen Xia, Vitor Campagnolo Guizilini, Yue Wang
Abstract: Vision Language Models (VLMs) are prone to errors, and identifying where these errors occur is critical for ensuring the reliability and safety of AI systems. In this paper, we propose an approach that automatically generates questions designed to deliberately induce incorrect responses from VLMs, thereby revealing their vulnerabilities. The core of this approach lies in fuzz testing and reinforcement finetuning: we transform a single input query into a large set of diverse variants through vision and language fuzzing. Based on the fuzzing outcomes, the question generator is further instructed by adversarial reinforcement fine-tuning to produce increasingly challenging queries that trigger model failures. With this approach, we can consistently drive down a target VLM's answer accuracy -- for example, the accuracy of Qwen2.5-VL-32B on our generated questions drops from 86.58\% to 65.53\% in four RL iterations. Moreover, a fuzzing policy trained against a single target VLM transfers to multiple other VLMs, producing challenging queries that degrade their performance as well.
Authors: Laha Ale, Ning Zhang, Scott A. King, Pingzhi Fan
Abstract: Deep neural networks, and more recently large-scale generative models such as large language models (LLMs) and large vision-action models (LVAs), achieve remarkable performance across diverse domains, yet their prohibitive computational cost hinders deployment in resource-constrained environments. Existing efficiency techniques offer only partial remedies: dropout improves regularization during training but leaves inference unchanged, while pruning and low-rank factorization compress models post hoc into static forms with limited adaptability. Here we introduce SWAN (Switchable Activation Networks), a framework that equips each neural unit with a deterministic, input-dependent binary gate, enabling the network to learn when a unit should be active or inactive. This dynamic control mechanism allocates computation adaptively, reducing redundancy while preserving accuracy. Unlike traditional pruning, SWAN does not simply shrink networks after training; instead, it learns structured, context-dependent activation patterns that support both efficient dynamic inference and conversion into compact dense models for deployment. By reframing efficiency as a problem of learned activation control, SWAN unifies the strengths of sparsity, pruning, and adaptive inference within a single paradigm. Beyond computational gains, this perspective suggests a more general principle of neural computation, where activation is not fixed but context-dependent, pointing toward sustainable AI, edge intelligence, and future architectures inspired by the adaptability of biological brains.
Authors: Martino Ciaperoni, Collin Leiber, Aristides Gionis, Heikki Mannila
Abstract: As datasets continue to grow in size and complexity, finding succinct yet accurate data summaries poses a key challenge. Centroid-based clustering, a widely adopted approach to address this challenge, finds informative summaries of datasets in terms of few prototypes, each representing a cluster in the data. Despite their wide adoption, the resulting data summaries often contain redundancies, limiting their effectiveness particularly in datasets characterized by a large number of underlying clusters. To overcome this limitation, we introduce the Khatri-Rao clustering paradigm that extends traditional centroid-based clustering to produce more succinct but equally accurate data summaries by postulating that centroids arise from the interaction of two or more succinct sets of protocentroids. We study two central approaches to centroid-based clustering, namely the well-established k-Means algorithm and the increasingly popular topic of deep clustering, under the lens of the Khatri-Rao paradigm. To this end, we introduce the Khatri-Rao k-Means algorithm and the Khatri-Rao deep clustering framework. Extensive experiments show that Khatri-Rao k-Means can strike a more favorable trade-off between succinctness and accuracy in data summarization than standard k-Means. Leveraging representation learning, the Khatri-Rao deep clustering framework offers even greater benefits, reducing even more the size of data summaries given by deep clustering while preserving their accuracy.
Authors: Joshua Kazdan, Noam Levi, Rylan Schaeffer, Jessica Chudnovsky, Abhay Puri, Bo He, Mehmet Donmez, Sanmi Koyejo, David Donoho
Abstract: Data duplication during pretraining can degrade generalization and lead to memorization, motivating aggressive deduplication pipelines. However, at web scale, it is unclear what constitutes a ``duplicate'': beyond surface-form matches, semantically equivalent documents (e.g. translations) may induce redundant training signals once models become sufficiently capable. Practically, this means that semantic duplicates operate increasingly like exact duplicates during training. We present evidence that duplication is scale-dependent in two ways. First, as model capability increases, cross-entropy loss gradients for semantically equivalent documents become more aligned. Smaller models, by contrast, produce gradients that reflect surface similarity (e.g., shared tokens) rather than semantic similarity. Second, we embedded all 192 million FineWeb-Edu-Dedup documents using EmbeddingGemma-300m. For moderate corpus sizes, the cosine similarity between nearest-neighbors follows an isotropic power law baseline. However, as corpus size grows to hundreds of billions of tokens, the nearest-neighbor similarities deviate sharply, indicating accelerated semantic collisions. Finally, controlled pretraining on data sampled with replacement from pools of finite unique documents shows that limited uniqueness yields mild degradation for small models, but rapidly increasing loss penalties for larger models, breaking naive scaling extrapolation. We derive explicit scaling laws that allow practitioners to estimate deviation from expected scaling due to limited semantic uniqueness of the pretraining corpus. Our results identify and resolve an unstudied source of scale-dependence, allowing for more accurate prediction at scale.
Authors: Xie Xiaohu, Liu Xiaohu, Yao Benjamin
Abstract: As large language models (LLMs) are increasingly deployed in critical decision-making systems, the lack of reliable methods to measure their uncertainty presents a fundamental trustworthiness risk. We introduce a normalized confidence score based on output anchor token probabilities: classification labels for structured tasks and self-evaluation responses (Yes/No) for open-ended generation. This enables direct detection of errors and hallucinations with minimal overhead and without external validation. We make three key contributions. First, we propose a normalized confidence score and self-evaluation framework that exposes reliable confidence estimates for error detection across seven diverse benchmark tasks and five LLMs of varying architectures and sizes. Second, our theoretical analysis reveals that supervised fine-tuning (SFT) yields well-calibrated confidence through maximum-likelihood estimation, whereas reinforcement learning methods (PPO, GRPO) and DPO induce overconfidence via reward exploitation. Third, we propose post-RL SFT with self-distillation to restore confidence reliability in RL-trained models. Empirical results demonstrated that SFT improved average confidence-correctness AUROC from 0.806 to 0.879 and reduced calibration error from 0.163 to 0.034 on Qwen3-4B, while GRPO and DPO degraded confidence reliability. We demonstrated practical value through adaptive retrieval-augmented generation (RAG) that selectively retrieves context when the model lacks confidence, using only 58\% of retrieval operations to recover 95\% of the maximum achievable accuracy gain on TriviaQA
Authors: Joohyung Lee, Kwanhyung Lee, Changhun Kim, Eunho Yang
Abstract: Electronic health records (EHR) are irregular, asynchronous multivariate time series. As time-series foundation models increasingly tokenize events rather than discretizing time, the input layout becomes a key design choice. Grids expose time$\times$variable structure but require imputation or missingness masks, risking error or sampling-policy shortcuts. Point-set tokenization avoids discretization but loses within-variable trajectories and time-local cross-variable context (Fig.1). We restore these priors in STructure-AwaRe (STAR) Set Transformer by adding parameter-efficient soft attention biases: a temporal locality penalty $-|\Delta t|/\tau$ with learnable timescales and a variable-type affinity $B_{s_i,s_j}$ from a learned feature-compatibility matrix. We benchmark 10 depth-wise fusion schedules (Fig.2). On three ICU prediction tasks, STAR-Set achieves AUC/APR of 0.7158/0.0026 (CPR), 0.9164/0.2033 (mortality), and 0.8373/0.1258 (vasopressor use), outperforming regular-grid, event-time grid, and prior set baselines. Learned $\tau$ and $B$ provide interpretable summaries of temporal context and variable interactions, offering a practical plug-in for context-informed time-series models.
Authors: Joseph Bingham, Noah Green, Saman Zonouz
Abstract: As the need for neural network-based applications to become more accurate and powerful grows, so too does their size and memory footprint. With embedded devices, whose cache and RAM are limited, this growth hinders their ability to leverage state-of-the-art neural network architectures. In this work, we propose \textbf{LegoNet}, a compression technique that \textbf{constructs blocks of weights of the entire model regardless of layer type} and clusters these induced blocks. Using blocks instead of individual values to cluster the weights, we were able to compress ResNet-50 trained for Cifar-10 and ImageNet with only 32 4x4 blocks, compressing the memory footprint by over a factor of \textbf{64x without having to remove any weights} or changing the architecture and \textbf{no loss to accuracy}, nor retraining or any data, and show how to find an arrangement of 16 4x4 blocks that gives a compression ratio of \textbf{128x with less than 3\% accuracy loss}. This was all achieved with \textbf{no need for (re)training or fine-tuning}.
Authors: Mohamed Salem
Abstract: Modern machine learning models are highly expressive but notoriously difficult to analyze statistically. In particular, while black-box predictors can achieve strong empirical performance, they rarely provide valid hypothesis tests or p-values for assessing whether individual features contain information about a target variable. This article presents a practical approach to feature-level hypothesis testing that combines the Conditional Randomization Test (CRT) with TabPFN, a probabilistic foundation model for tabular data. The resulting procedure yields finite-sample valid p-values for conditional feature relevance, even in nonlinear and correlated settings, without requiring model retraining or parametric assumptions.
Authors: Lukas Thede, Stefan Winzeck, Zeynep Akata, Jonathan Richard Schwarz
Abstract: Large language model (LLM) post-training enhances latent skills, unlocks value alignment, improves performance, and enables domain adaptation. Unfortunately, post-training is known to induce forgetting, especially in the ubiquitous use-case of leveraging third-party pre-trained models, which is typically understood as a loss of parametric or factual knowledge. We argue that this accuracy-centric view is insufficient for modern foundation models and instead define forgetting as systematic model drift that degrades behavior and user experience. In this context, we introduce \textbf{CapTrack}, a capability-centric framework for analyzing forgetting in LLMs that combines a behavioral taxonomy with an evaluation suite built on established benchmarks and targeted adaptations. Using CapTrack, we conduct a large-scale empirical study across post-training algorithms, domains, and model families, including models up to 80B parameters. We find that forgetting extends beyond parametric knowledge, with pronounced drift in robustness and default behaviors. Instruction fine-tuning induces the strongest relative drift, while preference optimization is more conservative and can partially recover lost capabilities. Differences across model families persist, and no universal mitigation emerges.
Authors: Yegor Denisov-Blanch, Joshua Kazdan, Jessica Chudnovsky, Rylan Schaeffer, Sheng Guan, Soji Adeshina, Sanmi Koyejo
Abstract: Pass@k and other methods of scaling inference compute can improve language model performance in domains with external verifiers, including mathematics and code, where incorrect candidates can be filtered reliably. This raises a natural question: can we similarly scale compute to elicit gains in truthfulness for domains without convenient verification? We show that across five benchmarks and models, surprisingly, it cannot. Even at 25x the inference cost of naive sampling, polling-style aggregation yields no consistent accuracy gains over single-sample baselines and often amplifies shared misconceptions. We find that under uncertainty, models are better at predicting what other models will say within model ensembles than at identifying what is true, revealing a separation between social prediction and truth verification. Across models and benchmarks, aggregation fails to provide a robust truth signal because language model errors are strongly correlated. The source of correlation goes beyond any individual benchmark: we show that even when conditioned on out of distribution random strings and asked to produce pseudo-random outputs, different models produce correlated outputs. Confidence-based weighting provides no benefit because self-reported confidence fails to reliably distinguish correct from incorrect answers. These results delineate a boundary for inference-time scaling: in verified domains, additional samples provide more candidates for a verifier to filter; in unverified domains, additional samples merely reinforce shared misconceptions.
Authors: Stamatis Mastromichalakis
Abstract: This paper presents OptiRoulette, a stochastic meta-optimizer that selects update rules during training instead of fixing a single optimizer. The method combines warmup optimizer locking, random sampling from an active optimizer pool, compatibility-aware learning-rate scaling during optimizer transitions, and failure-aware pool replacement. OptiRoulette is implemented as a drop-in, "torch.optim.Optimizer-compatible" component and packaged for pip installation. We report completed 10-seed results on five image-classification suites: CIFAR-100, CIFAR-100-C, SVHN, Tiny ImageNet, and Caltech-256. Against a single-optimizer AdamW baseline, OptiRoulette improves mean test accuracy from 0.6734 to 0.7656 on CIFAR-100 (+9.22 percentage points), 0.2904 to 0.3355 on CIFAR-100-C (+4.52), 0.9667 to 0.9756 on SVHN (+0.89), 0.5669 to 0.6642 on Tiny ImageNet (+9.73), and 0.5946 to 0.6920 on Caltech-256 (+9.74). Its main advantage is convergence reliability at higher targets: it reaches CIFAR-100/CIFAR-100-C 0.75, SVHN 0.96, Tiny ImageNet 0.65, and Caltech-256 0.62 validation accuracy in 10/10 runs, while the AdamW baseline reaches none of these targets within budget. On shared targets, OptiRoulette also reduces time-to-target (e.g., Caltech-256 at 0.59: 25.7 vs 77.0 epochs). Paired-seed deltas are positive on all datasets; CIFAR-100-C test ROC-AUC is the only metric not statistically significant in the current 10-seed study.
Authors: Zhengguo Li, Chaobing Zheng, Wei Wang
Abstract: Based on literature review about existing diffusion models and flow matching with a neural network to predict a predefined target from noisy data, a unified representation is first proposed for these models using two simple linear equations in this paper. Theoretical analysis of the proposed model is then presented. Our theoretical analysis shows that the correlation between the noisy data and the predicted target is sometimes weak in the existing diffusion models and flow matching. This might affect the prediction (or learning) process which plays a crucial role in all models.
Authors: Hantao Zhang, Jieke Wu, Mingda Xu, Xiao Hu, Yingxuan You, Pascal Fua
Abstract: For multivariate co-generation in scientific applications, we advocate pairwise block rather than joint modeling of all variables. This design mitigates the computational burden and data imbalance. To this end, we propose an Annealed Co-Generation (ACG) framework that replaces high-dimensional diffusion modeling with a low-dimensional diffusion model, which enables multivariate co-generation by composing pairwise variable generations. We first train an unconditional diffusion model over causal variables that are disentangled into pairs. At inference time, we recover the joint distribution by coupling these pairwise models through shared common variables, enabling coherent multivariate generation without any additional training. By employing a three-stage annealing process-Consensus, Heating, and Cooling-our method enforces consistency across shared common variables and progressively constrains each pairwise data distribution to lie on a learnable manifold, while maintaining high likelihood within each pair. We demonstrate the framework's flexibility and efficacy on two distinct scientific tasks: flow-field completion and antibody generation. All datasets and code will be made publicly available upon publication.
Authors: Sai Hao, Hao Zeng, Hongxin Wei, Bingyi Jing
Abstract: Efficiently routing queries to the optimal large language model (LLM) is crucial for optimizing the cost-performance trade-off in multi-model systems. However, most existing routers rely on single-model selection, making them susceptible to misrouting. In this work, we formulate LLM routing as the $\alpha$-VOR problem to minimize expected set size while controlling the misrouting risk, and propose a novel method -- RACER, extending base routers to output model sets that can be subsequently aggregated for improved output. In particular, RACER constructs nested model sets via augmented scoring and utilizes finite-sample concentration bounds to calibrate a threshold that allows for both variable set sizes and abstention. We theoretically prove that RACER achieves rigorous distribution-free risk control on unseen test data in a post-hoc and model-agnostic manner. Extensive experiments verify our theoretical guarantees and demonstrate that RACER consistently enhances downstream accuracy across a wide range of benchmarks.
Authors: Junde Wu, Minhao Hu, Jiayuan Zhu, Yuyuan Liu, Tianyi Zhang, Kang Li, Jingkun Chen, Jiazhen Pan, Min Xu, Yueming Jin
Abstract: We introduce \textbf{Evo}, a duality latent trajectory model that bridges autoregressive (AR) and diffusion-based language generation within a continuous evolutionary generative framework. Rather than treating AR decoding and diffusion generation as separate paradigms, Evo reconceptualizes text generation as a latent flow: each token is associated with a vector-valued embedding that evolves over a progression variable $t_i \in [0, 1]$, indicating its semantic maturity. Low $t_i$ values correspond to confident AR-like refinement, while high values invoke diffusion-style planning, allowing the model to adaptively balance AR and diffusion based on uncertainty. Theoretically, we show that both AR and diffusion models emerge as discretizations of a shared probability flow, and we derive Evo's training objective from a unified variational ELBO. The model is implemented as a time-conditioned Transformer governed by a shared vector field, trained end-to-end to jointly infer latent codes and their progression times. During decoding, Evo performs efficient, semantics-aware refinement, achieving high-quality outputs without sacrificing speed. Empirically, Evo 8B achieves state-of-the-art or highly competitive results on 15 diverse benchmarks, including reasoning (GSM8K, ARC-C), code generation (HumanEval, MBPP), and general language understanding, while maintaining fast inference speed. Our results demonstrate that Evo delivers a new paradigm for LLM design with strong generation quality, robust symbolic reasoning, and decoding efficiency.
Authors: Alana Deng, Sugitha Janarthanan, Yan Sun, Zihao Jing, Pingzhao Hu
Abstract: Multiplex Biological Networks (MBNs), which represent multiple interaction types between entities, are crucial for understanding complex biological systems. Yet, existing methods often inadequately model multiplexity, struggle to integrate structural and sequence information, and face difficulties in zero-shot prediction for unseen entities with no prior neighbourhood information. To address these limitations, we propose a novel framework for zero-shot interaction prediction in MBNs by leveraging context-aware representation learning and knowledge distillation. Our approach leverages domain-specific foundation models to generate enriched embeddings, introduces a topology-aware graph tokenizer to capture multiplexity and higher-order connectivity, and employs contrastive learning to align embeddings across modalities. A teacher-student distillation strategy further enables robust zero-shot generalization. Experimental results demonstrate that our framework outperforms state-of-the-art methods in interaction prediction for MBNs, providing a powerful tool for exploring various biological interactions and advancing personalized therapeutics.
Authors: Hejian Sang, Yuanda Xu, Zhengze Zhou, Ran He, Zhipeng Wang
Abstract: Reinforcement learning (RL) has become a key driver of progress in large language models, but scaling RL to long chain-of-thought (CoT) trajectories is increasingly constrained by backpropagation over every generated token. Even with optimized rollout engines, full-token updates can consume a large fraction of total training cost, turning token length into a hidden tax on RL. We introduce Not All Tokens Are Needed (NAT), a unified framework that makes the token budget a first-class optimization primitive. NAT updates the policy using only a selected subset of generated tokens while preserving the learning signal of full-sequence RL. The core idea is an unbiased partial-token policy-gradient estimator via Horvitz-Thompson reweighting, which ensures statistically correct gradients despite subsampling. We instantiate NAT with two simple, plug-and-play token selection schemes: Uniform Random Sampling (URS) and Random Prefix Cutting (RPC), both of which reduce forward and backward compute and memory without modifying the reward computation or rollout pipeline. Across mathematical reasoning benchmarks, NAT matches full-token GRPO performance while using as few as 50% of tokens, providing an efficient and orthogonal pathway to scaling RL beyond the limits imposed by long trajectories. In our experiments, RPC saves 18% peak GPU memory and 29% forward and backward RL training time for Qwen3-8B.
Authors: Rishabh Tiwari, Aditya Tomar, Udbhav Bamba, Monishwaran Maheswaran, Heng Yang, Michael W. Mahoney, Kurt Keutzer, Amir Gholami
Abstract: Process Reward Models (PRMs) are rapidly becoming the backbone of LLM reasoning pipelines, yet we demonstrate that state-of-the-art PRMs are systematically exploitable under adversarial optimization pressure. To address this, we introduce a three-tiered diagnostic framework that applies increasing adversarial pressure to quantify these vulnerabilities. Static perturbation analysis uncovers a fluency-logic dissociation: high invariance to surface-level style changes reward changes $<$0.1, yet inconsistent detection of logically-corrupted reasoning, with different models failing on different attack types. Adversarial optimization demonstrates that gradient-based attacks inflate rewards on invalid trajectories, with reward landscapes exhibiting wide, exploitable peaks. RL-induced reward hacking exposes the critical failure mode: policies trained on AIME problems achieve near-perfect PRM rewards ($>$0.9), while ground-truth accuracy remains low (below 4%), with 43% of reward gains attributable to stylistic shortcuts. These findings reveal that current PRMs function as fluency detectors rather than reasoning verifiers, creating systematic blind spots that undermine their use as training signals. We release PRM-BiasBench and a diagnostic toolkit to enable robustness evaluation before deployment. The code and dataset are available at https://github.com/SqueezeAILab/reward-under-attack.
Authors: Suhasnadh Reddy Veluru, Sai Teja Erukude, Viswa Chaitanya Marella
Abstract: Accurate short-term power load forecasting is important to effectively manage, optimize, and ensure the robustness of modern power systems. This paper performs an empirical evaluation of a traditional statistical model and deep learning approaches for predicting short-term energy load. Four models, namely ARIMA, LSTM, BiLSTM, and Transformer, were leveraged on the PJM Hourly Energy Consumption data. The data processing involved interpolation, normalization, and a sliding-window sequence method. Each model's forecasting performance was evaluated for the 24-hour horizon using MAE, RMSE, and MAPE. Of the models tested, the Transformer model, which relies on self-attention algorithms, produced the best results with 3.8 percent of MAPE, with performance above any model in both accuracy and robustness. These findings underscore the growing potential of attention-based architectures in accurately capturing complex temporal patterns in power consumption data.
Authors: Zexiang Liu, Xianglong He, Yangguang Li
Abstract: Large-scale flow matching models have achieved strong performance across generative tasks such as text-to-image, video, 3D, and speech synthesis. However, aligning their outputs with human preferences and task-specific objectives remains challenging. Flow-GRPO extends Group Relative Policy Optimization (GRPO) to generation models, enabling stable reinforcement learning alignment for generative systems. Since its introduction, Flow-GRPO has triggered rapid research growth, spanning methodological refinements and diverse application domains. This survey provides a comprehensive review of Flow-GRPO and its subsequent developments. We organize existing work along two primary dimensions. First, we analyze methodological advances beyond the original framework, including reward signal design, credit assignment, sampling efficiency, diversity preservation, reward hacking mitigation, and reward model construction. Second, we examine extensions of GRPO-based alignment across generative paradigms and modalities, including text-to-image, video generation, image editing, speech and audio, 3D modeling, embodied vision-language-action systems, unified multimodal models, autoregressive and masked diffusion models, and restoration tasks. By synthesizing theoretical insights and practical adaptations, this survey highlights Flow-GRPO as a general alignment framework for modern generative models and outlines key open challenges for scalable and robust reinforcement-based generation.
Authors: Ke Yu, Lu Gao
Abstract: Pavement condition data is important in providing information regarding the current state of the road network and in determining the needs of maintenance and rehabilitation treatments. However, the condition data is often incomplete due to various reasons such as sensor errors and non-periodic inspection schedules. Missing data, especially data missing systematically, presents loss of information, reduces statistical power, and introduces biased assessment. Existing methods in dealing with missing data usually discard entire data points with missing values or impute through data correlation. In this paper, we used a collective learning-based Graph Convolutional Networks, which integrates both features of adjacent sections and dependencies between observed section conditions to learn missing condition values. Unlike other variants of graph neural networks, the proposed approach is able to capture dependent relationship between the conditions of adjacent pavement sections. In the case study, pavement condition data collected from Texas Department of Transportation Austin District were used. Experiments show that the proposed model was able to produce promising results in imputing the missing data.
Authors: Yuqi Xu, Rizhen Hu, Zihan Liu, Mou Sun, Kun Yuan
Abstract: Traditional Mixture-of-Experts (MoE) training typically proceeds without any structural priors, effectively requiring the model to simultaneously train expert weights while searching for an optimal routing policy within a vast combinatorial space. This entanglement often leads to sluggish convergence and training instabilities. This paper introduces Grouter, a preemptive routing method that by distilling high-quality structures from fully-trained MoE models and serving as a fixed router for target models. By decoupling structural optimization from weight updates, Grouter significantly accelerates both the speed and quality of model convergence. To ensure the framework's versatility, we also introduce expert folding to adapt Grouter across varying model configurations and expert tuning to rebalance workloads across different data distributions. Furthermore, by leveraging the structural priors provided by preemptive routing, we can implement targeted optimizations to further enhance training throughput. Experiments demonstrate that Grouter achieves superior performance and efficiency which boosts pre-training data utilization by 4.28x and achieves up to 33.5% throughput acceleration, establishing preemptive routing as a fundamental paradigm for scalable MoE training.
Authors: Hamideh Khaleghpour, Brett McKinney
Abstract: Illicit transaction detection is often driven by transaction level attributes however, fraudulent behavior may also manifest through network structure such as central hubs, high flow intermediaries, and coordinated neighborhoods. This paper presents a time respecting, leakage safe (causal) graph feature extraction protocol for temporal transaction networks and evaluates its utility for illicit entity classification. Using the Elliptic dataset, we construct directed transaction graphs and compute interpretable structural descriptors, including degree statistics, PageRank, HITS hub or authority scores, k-core indices, and neighborhood reachability measures. To prevent look ahead bias, we additionally compute causal variants of graph features using only edges observed up to each timestep. A Random Forest classifier trained with strict temporal splits achieves strong discrimination on a held out future test period (ROC-AUC about 0.85, Average Precision about 0.54). Although transaction attributes remain the dominant predictive signal, graph derived features provide complementary interpretability and enable risk context analysis for investigation workflows. We further assess operational utility using Precision at k and evaluate probability reliability via calibration curves and Brier scores, showing that calibrated models yield better aligned probabilities for triage. Overall, the results support causal graph feature extraction as a practical and interpretable augmentation for temporal fraud detection pipelines.
Authors: V. Dolotin, A. Morozov
Abstract: Many scientific problems in the context of machine learning can be reduced to the search of polynomial answers in appropriate variables. The Hevisidization of arbitrary polynomial is actually provided by one-and-the same two-layer expression. What prevents the use of this simple idea is the fatal degeneracy of the Heaviside and sigmoid expansions, which traps the steepest-descent evolution at the bottom of canyons, close to the starting point, but far from the desired true minimum. This problem is unavoidable and can be formulated as a peculiar uncertainty principle -- the sharper the minimum, the smoother the canyons. It is a direct analogue of the usual one, which is the pertinent property of the more familiar Fourier expansion. Standard machine learning software fights with this problem empirically, for example, by testing evolutions, originated at randomly distributed starting points and then selecting the best one. Surprisingly or not, phenomena and problems, encountered in ML application to science are pure scientific and belong to physics, not to computer science. On the other hand, they sound slightly different and shed new light on the well-known phenomena -- for example, extend the uncertainty principle from Fourier and, later, wavelet analysis to a new peculiar class of nearly singular sigmoid functions.
Authors: Michal Podstawski
Abstract: Recent progress in language modeling has expanded the range of tasks that can be approached through natural language interfaces, including problems that require structured reasoning. However, it remains unclear how effectively limited-capacity language models can infer formal properties of relational structures when those structures are presented in textual form. Understanding the conditions under which structured reasoning succeeds or fails is essential for applying small models in graph-based domains. We conduct a systematic study of graph-theoretic property inference in small instruction-tuned language models, isolating the roles of input representation and reasoning strategy. Across a diverse set of local and global graph metrics, we find that structural performance is highly sensitive to how relational information is organized. Representations that preserve neighborhood structure consistently improve estimation stability and ordinal consistency, while multi-branch reasoning yields the most reliable aggregate gains across configurations. These results show that graph property inference in small language models depends critically on representational organization and inference design. Structural competence is therefore shaped not only by model scale, but by how relational information is encoded and how predictions are elicited. The findings identify practical levers for improving structured inference under constrained model capacity.
Authors: Qingsong Zou, Zhi Yan, Zhiyao Xu, Kuofeng Gao, Jingyu Xiao, Yong Jiang
Abstract: Due to the strong context-awareness capabilities demonstrated by large language models (LLMs), recent research has begun exploring their integration into smart home assistants to help users manage and adjust their living environments. While LLMs have been shown to effectively understand user needs and provide appropriate responses, most existing studies primarily focus on interpreting and executing user behaviors or instructions. However, a critical function of smart home assistants is the ability to detect when the home environment is in an anomalous state. This involves two key requirements: the LLM must accurately determine whether an anomalous condition is present, and provide either a clear explanation or actionable suggestions. To enhance the anomaly detection capabilities of next-generation LLM-based smart home assistants, we introduce SmartBench, which is the first smart home dataset designed for LLMs, containing both normal and anomalous device states as well as normal and anomalous device state transition contexts. We evaluate 13 mainstream LLMs on this benchmark. The experimental results show that most state-of-the-art models cannot achieve good anomaly detection performance. For example, Claude-Sonnet-4.5 achieves only 66.1% detection accuracy on context-independent anomaly categories, and performs even worse on context-dependent anomalies, with an accuracy of only 57.8%. More experimental results suggest that next-generation LLM-based smart home assistants are still far from being able to effectively detect and handle anomalous conditions in the smart home environment. Our dataset is publicly available at https://github.com/horizonsinzqs/SmartBench.
Authors: Sirui Li, Shuhan Xiao, Mihir Joshi, Ahmed Metwally, Daniel McDuff, Wei Wang, Yuzhe Yang
Abstract: The rise of large language models (LLMs) has shifted time series analysis from narrow analytics to general-purpose reasoning. Yet, existing benchmarks cover only a small set of health time series modalities and tasks, failing to reflect the diverse domains and extensive temporal dependencies inherent in real-world physiological modeling. To bridge these gaps, we introduce HEARTS (Health Reasoning over Time Series), a unified benchmark for evaluating hierarchical reasoning capabilities of LLMs over general health time series. HEARTS integrates 16 real-world datasets across 12 health domains and 20 signal modalities, and defines a comprehensive taxonomy of 110 tasks grouped into four core capabilities: Perception, Inference, Generation, and Deduction. Evaluating 14 state-of-the-art LLMs on more than 20K test samples reveals intriguing findings. First, LLMs substantially underperform specialized models, and their performance is only weakly related to general reasoning scores. Moreover, LLMs often rely on simple heuristics and struggle with multi-step temporal reasoning. Finally, performance declines with increasing temporal complexity, with similar failure modes within model families, indicating that scaling alone is insufficient. By making these gaps measurable, HEARTS provides a standardized testbed and living benchmark for developing next-generation LLM agents capable of reasoning over diverse health signals.
Authors: Swamynathan V P
Abstract: Test-Time Training (TTT) language models achieve theoretically infinite context windows with an O(1) memory footprint by replacing the standard exact-attention KV-cache with hidden state ``fast weights'' W_fast updated via self-supervised learning during inference. However, pure TTT architectures suffer catastrophic failures on exact-recall tasks (e.g., Needle-in-a-Haystack). Because the fast weights aggressively compress the context into an information bottleneck, highly surprising or unique tokens are rapidly overwritten and forgotten by subsequent token gradient updates. We introduce SR-TTT (Surprisal-Aware Residual Test-Time Training), which resolves this recall failure by augmenting the TTT backbone with a loss-gated sparse memory mechanism. By dynamically routing only incompressible, highly surprising tokens to a traditional exact-attention Residual Cache, SR-TTT preserves O(1) memory for low-entropy background context while utilizing exact attention exclusively for critical needles. Our complete implementation, training scripts, and pre-trained weights are open-source and available at: https://github.com/swamynathanvp/Surprisal-Aware-Residual-Test-Time-Training.
URLs: https://github.com/swamynathanvp/Surprisal-Aware-Residual-Test-Time-Training.
Authors: Paul Shepherd, Tasos Dagiuklas, Bugra Alkan, Joaquim Bastos, Jonathan Rodriguez
Abstract: This paper presents a trust aware federated learning (FL) framework for interpreting bone healing stages using spectral features derived from frequency response data. The primary objective is to address the challenge posed by either unreliable or adversarial participants in distributed medical sensing environments. The framework employs a multi-layer perceptron model trained across simulated clients using the Flower FL framework. The proposed approach integrates an Adaptive Trust Score Scaling and Filtering (ATSSSF) mechanism with exponential moving average (EMA) smoothing to assess, validate and filter client contributions.Two trust score smoothing strategies have been investigated, one with a fixed factor and another that adapts according to trust score variability. Clients with low trust are excluded from aggregation and readmitted once their reliability improves, ensuring model integrity while maintaining inclusivity. Standard classification metrics have been used to compare the performance of ATSSSF with the baseline Federated Averaging strategy. Experimental results demonstrate that adaptive trust management can improve both training stability and predictive performance by mitigating the negative effects of compromised clients while retaining robust detection capabilities. The work establishes the feasibility for adaptive trust mechanisms in federated medical sensing and identifies extension to clinical cross silo aggregation as a future research direction.
Authors: Noujoud Nadera, Hadi Majed, Stefanos Giaremis, Rola El Osta, Clint Dawson, Carola Kaiser, Hartmut Kaiser
Abstract: The coastal regions of the eastern and southern United States are impacted by severe storm events, leading to significant loss of life and properties. Accurately forecasting storm surge and wind impacts from hurricanes is essential for mitigating some of the impacts, e.g., timely preparation of evacuations and other countermeasures. Physical simulation models like the ADCIRC hydrodynamics model, which run on high-performance computing resources, are sophisticated tools that produce increasingly accurate forecasts as the resolution of the computational meshes improves. However, a major drawback of these models is the significant time required to generate results at very high resolutions, which may not meet the near real-time demands of emergency responders. The presented work introduces HURRI-GAN, a novel AI-driven approach that augments the results produced by physical simulation models using time series generative adversarial networks (TimeGAN) to compensate for systemic errors of the physical model, thus reducing the necessary mesh size and runtime without loss in forecasting accuracy. We present first results in extrapolating model bias corrections for the spatial regions beyond the positions of the water level gauge stations. The presented results show that our methodology can accurately generate bias corrections at target locations spatially beyond gauge stations locations. The model's performance, as indicated by low root mean squared error (RMSE) values, highlights its capability to generate accurate extrapolated data. Applying the corrections generated by HURRI-GAN on the ADCIRC modeled water levels resulted in improving the overall prediction on the majority of the testing gauge stations.
Authors: Liwei Hu, Guangyao Li, Wenyong Wang, Xiaoming Zhang, Yu Xiang
Abstract: Euclidean gradient descent algorithms barely capture the geometry of objective function-induced hypersurfaces and risk driving update trajectories off the hypersurfaces. Riemannian gradient descent algorithms address these issues but fail to represent complex hypersurfaces via a single classic manifold. We propose geodesic gradient descent (GGD), a generic and learning-rate-free Riemannian gradient descent algorithm. At each iteration, GGD uses an n-dimensional sphere to approximate a local neighborhood on the objective function-induced hypersurface, adapting to arbitrarily complex geometries. A tangent vector derived from the Euclidean gradient is projected onto the sphere to form a geodesic, ensuring the update trajectory stays on the hypersurface. Parameter updates are performed using the endpoint of the geodesic. The maximum step size of the gradient in GGD is equal to a quarter of the arc length on the n-dimensional sphere, thus eliminating the need for a learning rate. Experimental results show that compared with the classic Adam algorithm, GGD achieves test MSE reductions ranging from 35.79% to 48.76% for fully connected networks on the Burgers' dataset, and cross-entropy loss reductions ranging from 3.14% to 11.59% for convolutional neural networks on the MNIST dataset.
Authors: Sanjay Mishra
Abstract: Financial risk detection in Enterprise Resource Planning (ERP) systems is an important but underexplored application of machine learning. Published studies in this area tend to suffer from vague dataset descriptions, leakage-prone pipelines, and evaluation practices that inflate reported performance. This paper presents a rebuilt experimental framework for ERP financial risk detection using ensemble machine learning. The risk definition is hybrid, covering both procurement compliance anomalies and transactional fraud. A composite benchmark called ERP-RiskBench is assembled from public procurement event logs, labeled fraud data, and a new synthetic ERP dataset with rule-injected risk typologies and conditional tabular GAN augmentation. Nested cross-validation with time-aware and group-aware splitting enforces leakage prevention throughout the pipeline. The primary model is a stacking ensemble of gradient boosting methods, tested alongside linear baselines, deep tabular architectures, and an interpretable glassbox alternative. Performance is measured through Matthews Correlation Coefficient, area under the precision-recall curve, and cost-sensitive decision analysis using calibrated probabilities. Across multiple dataset configurations and a structured ablation study, the stacking ensemble achieves the best detection results. Leakage-safe protocols reduce previously inflated accuracy estimates by a notable margin. SHAP-based explanations and feature stability analysis show that procurement control features, especially three-way matching discrepancies, rank as the most informative predictors. The resulting framework provides a reproducible, operationally grounded blueprint for machine learning deployment in ERP audit and governance settings.
Authors: Minsi Ren, Wenhao Deng, Ruiqi Feng, Tailin Wu
Abstract: Stochastic differential equation (SDE)-based generative models have achieved substantial progress in conditional generation via training-free differentiable loss-guided approaches. However, existing methodologies utilizing posterior sam- pling typically confront a substantial estimation error, which results in inaccu- rate gradients for guidance and leading to inconsistent generation results. To mitigate this issue, we propose that performing an additional backward denois- ing step and Monte-Carlo sampling (ABMS) can achieve better guided diffu- sion, which is a plug-and-play adjustment strategy. To verify the effectiveness of our method, we provide theoretical analysis and propose the adoption of a dual-focus evaluation framework, which further serves to highlight the critical problem of cross-condition interference prevalent in existing approaches. We conduct experiments across various task settings and data types, mainly includ- ing conditional online handwritten trajectory generation, image inverse problems (inpainting, super resolution and gaussian deblurring) molecular inverse design and so on. Experimental results demonstrate that our approach can be effec- tively used with higher order samplers and consistently improves the quality of generation samples across all the different scenarios.
Authors: Karan Gupta, Pranav Vajreshwari, Yash Pandya, Raghav Magazine, Akshay Nambi, Ahmed Awadallah
Abstract: Agentic systems operating over large tool ecosystems must plan and execute long-horizon workflows under weak or non-verifiable supervision. While frontier models mitigate these challenges through scale and large context budgets, small language models (SLMs) remain brittle: eager tool loading saturates context, execution errors compound over time, and sparse rewards limit learning. We introduce ATLAS, a reinforcement finetuning framework that enables SLMs to operate effectively in large-scale toolspace environments by learning how to acquire context and how to execute actions. Our approach makes two key contributions. First, we treat context control and execution structure as learnable decisions, combining iterative tool loading with programmatic tool orchestration to bound context growth and stabilize long-horizon trajectories. Second, we propose rubric-based reinforcement finetuning, which decomposes task success into structured, task-aligned criteria and enables scalable training using small judge models. Across MCP benchmarks, these design choices yield large and consistent gains over generic RL baselines, allowing a 4B SLM to approach frontier-agent performance under far tighter parameter and context budgets.
Authors: Guanglin Zhou, Armin Catic, Motahare Shabestari, Matthew Young, Chaiquan Li, Katrina Poppe, Sebastiano Barbieri
Abstract: Access to electronic health records (EHRs) for digital health research is often limited by privacy regulations and institutional barriers. Synthetic EHRs have been proposed as a way to enable safe and sovereign data sharing; however, existing methods may produce records that capture overall statistical properties of real data but present inconsistencies across clinical processes and observations. We developed an integrated pipeline to make synthetic patient trajectories clinically consistent through two synergistic steps: high-fidelity generation and scalable auditing. Using the MIMIC-IV database, we trained a knowledge-grounded generative model that represents nearly 32,000 distinct clinical events, including demographics, laboratory measurements, medications, procedures, and diagnoses, while enforcing structural integrity. To support clinical consistency at scale, we incorporated an automated auditing module leveraging large language models to filter out clinical inconsistencies (e.g., contraindicated medications) that escape probabilistic generation. We generated 18,071 synthetic patient records derived from a source cohort of 180,712 real patients. While synthetic clinical event probabilities demonstrated robust agreement (mean bias effectively 0.00) and high correlation (R2=0.99) with the real counterparts, review of a random sample of synthetic records (N=20) by three clinicians identified inconsistencies in 45-60% of them. Automated auditing reduced the difference between real and synthetic data (Cohen's effect size d between 0.59 and 1.60 before auditing, and between 0.18 and 0.67 after auditing). Downstream models trained on audited data matched or even exceeded real-data performance. We found no evidence of privacy risks, with membership inference performance indistinguishable from random guessing (F1-score=0.51).
Authors: Aditya Ranganath, Hasin Us Sami, Kowshik Thopalli, Bhavya Kailkhura, Wesam Sakla
Abstract: Protein language models often take into consideration the alignment between a protein sequence and its textual description. However, they do not take structural information into consideration. Traditional methods treat sequence and structure separately, limiting the ability to exploit the alignment between the structure and protein sequence embeddings. In this paper, we introduce a sequence structure contrastive alignment framework, which learns a shared embedding space where proteins are represented consistently across modalities. By training on large-scale pairs of sequences and experimentally resolved or predicted structures, the model maximizes agreement between matched sequence structure pairs while pushing apart unrelated pairs. This alignment enables cross-modal retrieval (e.g., finding structural neighbors given a sequence), improves downstream prediction tasks such as function annotation and stability estimation, and provides interpretable links between sequence variation and structural organization. Our results demonstrate that contrastive learning can serve as a powerful bridge between protein sequences and structures, offering a unified representation for understanding and engineering proteins.
Authors: Harshala Gammulle, Lidia Morawska, Sridha Sridharan, Clinton Fookes
Abstract: Indoor air quality (IAQ) forecasting plays a critical role in safeguarding occupant health, ensuring thermal comfort, and supporting intelligent building control. However, predicting future concentrations of key pollutants such as carbon dioxide (CO2) and fine particulate matter (PM2.5) remains challenging due to the complex interplay between environmental factors and highly dynamic occupant behaviours. Traditional data driven models primarily rely on historical sensor trajectories and often fail to anticipate behaviour induced emission spikes or rapid concentration shifts. To address these limitations, we present a dual stream bi directional feedback fusion framework that jointly models indoor environmental evolution and action derived embeddings representing human activities. The proposed architecture integrates a context aware modulation mechanism that adaptively scales and shifts each stream based on a shared, evolving fusion state, enabling the model to selectively emphasise behavioural cues or long term environmental trends. Furthermore, we introduce dual timescale temporal modules that independently capture gradual CO2 accumulation patterns and short term PM2.5 fluctuations. A composite loss function combining weighted mean squared error, spike aware penalties, and uncertainty regularisation facilitates robust learning under volatile indoor conditions. Extensive validation on real-world IAQ datasets demonstrates that our approach significantly outperforms state of the art forecasting baselines while providing interpretable uncertainty estimates essential for practical deployment in smart buildings and health-aware monitoring systems.
Authors: Yunzhong Qiu, Binzhu Li, Hao Wei, Shenglin Weng, Chen Wang, Zhongyi Pei, Mingsheng Long, Jianmin Wang
Abstract: Electricity market prices exhibit extreme volatility, nonlinearity, and non-stationarity, making accurate forecasting a significant challenge. While cutting-edge time series foundation models (TSFMs) effectively capture temporal dependencies, they typically underutilize cross-variate correlations and non-periodic patterns that are essential for price forecasting. Conversely, regression models excel at capturing feature interactions but are limited to future-available inputs, ignoring crucial historical drivers that are unavailable at forecast time. To bridge this gap, we propose FutureBoosting, a novel paradigm that enhances regression-based forecasts by integrating forecasted features generated from a frozen TSFM. This approach leverages the TSFM's ability to model historical patterns and injects these insights as enriched inputs into a downstream regression model. We instantiate this paradigm into a lightweight, plug-and-play framework for electricity price forecasting. Extensive evaluations on real-world electricity market data demonstrate that our framework consistently outperforms state-of-the-art TSFMs and regression baselines, achieving reductions in Mean Absolute Error (MAE) of more than 30% at most. Through ablation studies and explainable AI (XAI) techniques, we validate the contribution of forecasted features and elucidate the model's decision-making process. FutureBoosting establishes a robust, interpretable, and effective solution for practical market participation, offering a general framework for enhancing regression models with temporal context.
Authors: Jingyuan Feng, Andrew Gambardella, Gouki Minegishi, Takeshi Kojima, Yusuke Iwasawa, Yutaka Matsuo
Abstract: Current safety alignment methods encode safe behavior implicitly within model parameters, creating a fundamental opacity: we cannot easily inspect why a model refuses a request, nor intervene when its safety judgments fail. We propose Safe Transformer, a modular approach that augments pre-trained language models by inserting a discrete information bottleneck containing an explicit safety bit between transformer layers. The safety bit serves as both an interpretable signal of the model's safety classification and a controllable switch: through contrastive training, the model learns disentangled representations where the safety bit governs the behavioral mode - producing helpful responses when $s=1$ and refusals when $s=0$ - while additional unsupervised bits $u$ encode semantic content for generation. Additional unsupervised bits in the information bottleneck allow semantic information to flow through, preserving the model's generation capabilities. This design achieves both interpretability (the safety decision is directly readable) and controllability (the safety bit can be manually overridden), requiring only lightweight fine-tuning without pre-training from scratch. In red-team benchmarks, Safe Transformer achieves near-zero Attack Success Rate, substantially outperforming base models and safety fine-tuning baselines.
Authors: Ramchand Kumaresan
Abstract: Over two billion Apple devices ship with a Neural Processing Unit (NPU) - the Apple Neural Engine (ANE) - yet this accelerator remains largely unused for large language model workloads. CoreML, Apple's public ML framework, imposes opaque abstractions that prevent direct ANE programming and do not support on-device training. We present Orion, to our knowledge the first open end-to-end system that combines direct ANE execution, a compiler pipeline, and stable multi-step training with checkpoint resume in a single native runtime, bypassing CoreML entirely via Apple's private _ANEClient and _ANECompiler APIs. Building on prior characterization work by maderix, we extend public knowledge of ANE constraints to a catalog of 20 restrictions on MIL IR programs, memory layout, compilation limits, and numerical behavior, including 14 previously undocumented constraints discovered during Orion development. Orion includes a compiler that lowers a graph IR through five optimization passes to ANE-native MIL and a runtime that manages IOSurface-backed zero-copy tensor I/O, program caching, and delta compilation for weight updates. Because the ANE bakes weights at compile time, naive training normally requires full recompilation per step (~4.2 s). We show that compiled programs can instead be updated by unloading, patching weight files, and reloading, bypassing ANECCompile() and reducing recompilation from 4,200 ms to 494 ms per step (8.5x), yielding a 3.8x training speedup. On an M4 Max, Orion achieves 170+ tokens/s for GPT-2 124M inference and demonstrates stable training of a 110M-parameter transformer on TinyStories for 1,000 steps in 22 minutes with zero NaN occurrences. We also present LoRA adapter-as-input, enabling hot-swap of adapters via IOSurface inputs without recompilation.
Authors: Jiefu Zhang, Yang Xu, Vaneet Aggarwal
Abstract: Navigating safely through dense crowds requires collision avoidance that generalizes beyond the densities seen during training. Learning-based crowd navigation can break under out-of-distribution crowd sizes due to density-sensitive observation normalization and social-cost scaling, while analytical solvers often remain safe but freeze in tight interactions. We propose a reinforcement learning approach for dense, variable-density navigation that attains zero-shot density generalization using a density-invariant observation encoding with density-randomized training and physics-informed proxemic reward shaping with density-adaptive scaling. The encoding represents the distance-sorted $K$ nearest pedestrians plus bounded crowd summaries, keeping input statistics stable as crowd size grows. Trained with $N\!\in\![11,16]$ pedestrians in a $3\mathrm{m}\times3\mathrm{m}$ arena and evaluated up to $N\!=\!21$ pedestrians ($1.3\times$ denser), our policy reaches the goal in $>99\%$ of episodes and achieves $86\%$ collision-free success in random crowds, with markedly less freezing than analytical methods and a $>\!60$-point collision-free margin over learning-based benchmark methods. Codes are available at \href{https://github.com/jznmsl/PSS-Social}{https://github.com/jznmsl/PSS-Social}.
URLs: https://github.com/jznmsl/PSS-Social, https://github.com/jznmsl/PSS-Social
Authors: Dongheon Lee, Seokju Yun, Jaegyun Im, Youngmin Ro
Abstract: Recent Super-Resolution~(SR) methods mainly adopt Transformers for their strong long-range modeling capability and exceptional representational capacity. However, most SR Transformers rely heavily on relative positional bias~(RPB), which prevents them from leveraging hardware-efficient attention kernels such as FlashAttention. This limitation imposes a prohibitive computational burden during both training and inference, severely restricting attempts to scale SR Transformers by enlarging the training patch size or the self-attention window. Consequently, unlike other domains that actively exploit the inherent scalability of Transformers, SR Transformers remain heavily focused on effectively utilizing limited receptive fields. In this paper, we propose Rank-factorized Implicit Neural Bias~(RIB), an alternative to RPB that enables FlashAttention in SR Transformers. Specifically, RIB approximates positional bias using low-rank implicit neural representations and concatenates them with pixel content tokens in a channel-wise manner, turning the element-wise bias addition in attention score computation into a dot-product operation. Further, we introduce a convolutional local attention and a cyclic window strategy to fully leverage the advantages of long-range interactions enabled by RIB and FlashAttention. We enlarge the window size up to \textbf{96$\times$96} while jointly scaling the training patch size and the dataset size, maximizing the benefits of Transformers in the SR task. As a result, our network achieves \textbf{35.63\,dB PSNR} on Urban100$\times$2, while reducing training and inference time by \textbf{2.1$\times$} and \textbf{2.9$\times$}, respectively, compared to the RPB-based SR Transformer~(PFT).
Authors: Zhiying Jiang, Raihan Seraj, Marcos Villagra, Bidhan Roy
Abstract: Training frontier-scale diffusion models often requires substantial computational resources concentrated in tightly coupled clusters, limiting participation to well-resourced institutions. While Decentralized Diffusion Models (DDM) enable training multiple experts in isolation, existing approaches require 1176 GPU-days and homogeneous training objectives across all experts. We present an efficient framework that reduces resource requirements while supporting heterogeneous training objectives. Our approach combines three contributions: (1) a heterogeneous decentralized training paradigm that allows experts to use different objectives (DDPM and Flow Matching), unified at inference time via a deterministic schedule-aware conversion into a common velocity space without retraining; (2) pretrained checkpoint conversion from ImageNet-DDPM to Flow Matching objectives, accelerating convergence and enabling initialization without objective-specific pretraining; and (3) PixArt-alpha's efficient AdaLN-Single architecture, reducing parameters while maintaining quality. Experiments on LAION-Aesthetics show that, relative to the training scale reported for prior DDM work, our approach reduces compute from 1176 to 72 GPU-days (16x) and data from 158M to 11M (14x). Under aligned inference settings, our heterogeneous 2DDPM:6FM configuration achieves better FID (11.88 vs. 12.45) and higher intra-prompt diversity (LPIPS 0.631 vs. 0.617) than the homogeneous 8FM baseline. By eliminating synchronization requirements and enabling mixed DDPM/FM objectives, our framework lowers infrastructure requirements for decentralized generative model training.
Authors: Xiaoxuan Liang, Saeid Naderiparizi, Yunpeng Liu, Berend Zwartsenberg, Frank Wood
Abstract: Constrained generative modeling is fundamental to applications such as robotic control and autonomous driving, where models must respect physical laws and safety-critical constraints. In real-world settings, these constraints rarely take the form of simple linear inequalities, but instead complex feasible regions that resemble road maps or other structured spatial domains. We propose a constrained generation framework that generates samples directly within such feasible regions while preserving realism. Our method fine-tunes a pretrained generative model to enforce constraints while maintaining generative fidelity. Experimentally, our method exhibits characteristics distinct from existing fine-tuning and training-free constrained baselines, revealing a new compromise between constraint satisfaction and sampling quality.
Authors: Jianyuan Zhong, Kaibo Wang, Ding Ding, Zijin Feng, Haoli Bai, Yang Xiang, Jiacheng Sun, Qiang Xu
Abstract: Group Relative Policy Optimization (GRPO) is highly effective for post-training autoregressive (AR) language models, yet its direct application to diffusion large language models (dLLMs) often triggers reward collapse. We identify two sources of incompatibility. First, GRPO relies on importance ratios defined by sequence probabilities, which are intractable in dLLMs and must be estimated (e.g., via ELBO-based or mean-field likelihood proxies), yielding inherently noisy ratios. Second, standard GRPO's formulation is not designed for estimated ratios: its conditional clipping can be anomalously bypassed by model-agnostic estimation noise, producing gradient spikes, while its fixed group-size normalization amplifies gradient-magnitude fluctuations under high-variance ratio estimates. We show these effects form a self-reinforcing instability loop that drives policy drift and further increases ratio variance. To break this loop, we propose StableDRL, a reformulation of GRPO tailored for dLLMs that uses (i) unconditional clipping to suppress outlier-induced spikes and (ii) self-normalization to constrain updates within the convex hull of per-sample gradients. We further extend StableDRL to block-wise diffusion models via a staircase attention mechanism.
Authors: Minjae Kang, Jaehyung Kim
Abstract: Large Language Models (LLMs), despite advances in instruction tuning, often fail to follow complex user instructions. Activation steering techniques aim to mitigate this by manipulating model internals, but have a potential risk of oversteering, where excessive emphasis on the instruction degrades task accuracy and overall text quality. To address this, we introduce DIRECTER (Dynamic rejection steering), a novel steering method that dynamically modulates steering strength by scaling the KV cache without extra dataset. DIRECTER couples steering with a plausibility-guided decoding loop, which adaptively adjusts steering strength at each step by comparing the steered output distribution to the original. If the steered output is deemed implausible, steering strength is progressively weakened. This strength modulation is guided by a lightweight, one-time attention sensitivity analysis that ranks layers by their influence on model representations. Extensive evaluations show that DIRECTER significantly enhances instruction-following capabilities across diverse benchmarks, improving accuracy by up to 6.5% over baselines without the common trade-offs in generation quality or task fidelity. The proposed dynamic, plausibility-guided control during activation steering further demonstrates its potential as a general mechanism for mitigating oversteering that is compatible with existing baselines.
Authors: Xiaoyang Hou, Junqi Liu, Chence Shi, Xin Liu, Zhi Yang, Jian Tang
Abstract: Protein sequence design must balance designability, defined as the ability to recover a target backbone, with multiple, often competing, developability properties such as solubility, thermostability, and expression. Existing approaches address these properties through post hoc mutation, inference-time biasing, or retraining on property-specific subsets, yet they are target dependent and demand substantial domain expertise or careful hyperparameter tuning. In this paper, we introduce ProtAlign, a multi-objective preference alignment framework that fine-tunes pretrained inverse folding models to satisfy diverse developability objectives while preserving structural fidelity. ProtAlign employs a semi-online Direct Preference Optimization strategy with a flexible preference margin to mitigate conflicts among competing objectives and constructs preference pairs using in silico property predictors. Applied to the widely used ProteinMPNN backbone, the resulting model MoMPNN enhances developability without compromising designability across tasks including sequence design for CATH 4.3 crystal structures, de novo generated backbones, and real-world binder design scenarios, making it an appealing framework for practical protein sequence design.
Authors: Xin T. Tong, Yanyan Wang, Liang Yan
Abstract: The ensemble Kalman filter (EnKF) is widely used for data assimilation in high-dimensional systems, but its performance often deteriorates for strongly nonlinear dynamics due to the structural mismatch between the Kalman update and the underlying system behavior. In this work, we propose a latent autoencoder ensemble Kalman filter (LAE-EnKF) that addresses this limitation by reformulating the assimilation problem in a learned latent space with linear and stable dynamics. The proposed method learns a nonlinear encoder--decoder together with a stable linear latent evolution operator and a consistent latent observation mapping, yielding a closed linear state-space model in the latent coordinates. This construction restores compatibility with the Kalman filtering framework and allows both forecast and analysis steps to be carried out entirely in the latent space. Compared with existing autoencoder-based and latent assimilation approaches that rely on unconstrained nonlinear latent dynamics, the proposed formulation emphasizes structural consistency, stability, and interpretability. We provide a theoretical analysis of learning linear dynamics on low-dimensional manifolds and establish generalization error bounds for the proposed latent model. Numerical experiments on representative nonlinear and chaotic systems demonstrate that the LAE-EnKF yields more accurate and stable assimilation than the standard EnKF and related latent-space methods, while maintaining comparable computational cost and data-driven.
Authors: Saadet M\"uzehher Eren
Abstract: We propose a quantum implicit neural representation (QINR)-based autoencoder (AE) and variational autoencoder (VAE) for image reconstruction and generation tasks. Our purpose is to demonstrate that the QINR in VAEs and AEs can transform information from the latent space into highly rich, periodic, and high-frequency features. Additionally, we aim to show that the QINR-VAE can be more stable than various quantum generative adversarial network (QGAN) models in image generation because it can address the low diversity problem. Our quantum-classical hybrid models consist of a classical convolutional neural network (CNN) encoder and a quantum-based QINR decoder. We train the QINR-AE/VAE with binary cross-entropy with logits (BCEWithLogits) as the reconstruction loss. For the QINR-VAE, we additionally employ Kullback-Leibler divergence for latent regularization with beta/capacity scheduling to prevent posterior collapse. We introduce learnable angle-scaling in data reuploading to address optimization challenges. We test our models on the MNIST, E-MNIST, and Fashion-MNIST datasets to reconstruct and generate images. Our results demonstrate that the QINR structure in VAE can produce a wider variety of images with a small amount of data than various generative models that have been studied. We observe that the generated and reconstructed images from the QINR-VAE/AE are clear with sharp boundaries and details. Overall, we find that the addition of QINR-based quantum layers into the AE/VAE frameworks enhances the performance of reconstruction and generation with a constrained set of parameters.
Authors: Yiqun Zhang, Zhanpei Huang, Mingjie Zhao, Chuyao Zhang, Yang Lu, Yuzhu Ji, Fangqing Gu, An Zeng
Abstract: Unlabeled streaming data are usually collected to describe dynamic systems, where concept drift detection is a vital prerequisite to understanding the evolution of systems. However, the drifting concepts are usually imbalanced in most real cases, which brings great challenges to drift detection. That is, the dominant statistics of large clusters can easily mask the drifting of small cluster distributions (also called small concepts), which is known as the `masking effect'. Considering that most existing approaches only detect the overall existence of drift under the assumption of balanced concepts, two critical problems arise: 1) where the small concept is, and 2) how to detect its drift. To address the challenging concept drift detection for imbalanced data, we propose Imbalanced Cluster Descriptor-based Drift Detection (ICD3) approach that is unbiased to the imbalanced concepts. This approach first detects imbalanced concepts by employing a newly designed multi-distribution-granular search, which ensures that the distribution of both small and large concepts is effectively captured. Subsequently, it trains a One-Cluster Classifier (OCC) for each identified concept to carefully monitor their potential drifts in the upcoming data chunks. Since the detection is independently performed for each concept, the dominance of large clusters is thus circumvented. ICD3 demonstrates highly interpretability by specifically locating the drifted concepts, and is robust to the changing of the imbalance ratio of concepts. Comprehensive experiments with multi-aspect ablation studies conducted on various benchmark datasets demonstrate the superiority of ICD3 against the state-of-the-art counterparts.
Authors: Pablo Guill\'en, Enrique Frias-Martinez
Abstract: Alzheimer disease (AD) diagnosis and prognosis increasingly rely on machine learning (ML) models. Although these models provide good results, clinical adoption is limited by the need for technical expertise and the lack of trustworthy and consistent model explanations. SHAP (SHapley Additive exPlanations) is com-monly used to interpret AD models, but existing studies tend to focus on explanations for isolated tasks, providing little evidence about their robustness across disease stages, model architectures, or prediction objectives. This paper proposes a multi-level explainability framework that measures the coherence, stabil-ity and consistency of explanations by integrating: (1) within-model coherence metrics between feature importance and SHAP, (2) SHAP stability across AD boundaries, and (3) SHAP cross-task consistency be-tween diagnosis and prognosis. Using AutoML to optimize classifiers on the NACC dataset, we trained four diagnostic and four prognostic models covering the standard AD progression stages. Stability was then evaluated using correlation metrics, top-k feature overlap, SHAP sign consistency, and domain-level contribution ratios. Results show that cognitive and functional markers dominate SHAP explanations in both diagnosis and prognosis. SHAP-SHAP consistency between diagnostic and prognostic models was high across all classifiers, with 100% sign stability and minimal shifts in explanatory magnitude. Domain-level contributions also remained stable, with only minimal increases in genetic features for prognosis. These results demonstrate that SHAP explanations can be quantitatively vali-dated for robustness and transferability, providing clinicians with more reliable interpretations of ML pre-dictions.
Authors: Hadi Salloum, Maximilian Mifsud Bonici, Sinan Ibrahim, Pavel Osinenko, Alexei Kornaev
Abstract: Physics-Informed Neural Networks (PINNs) enforce governing equations by penalizing PDE residuals at interior collocation points, but standard collocation strategies - uniform sampling and residual-based adaptive refinement - can oversample smooth regions, produce highly correlated point sets, and incur unnecessary training cost. We reinterpret collocation selection as a coreset construction problem: from a large candidate pool, select a fixed-size subset that is simultaneously informative (high expected impact on reducing PDE error) and diverse (low redundancy under a space-time similarity notion). We formulate this as a QUBO/BQM objective with linear terms encoding residual-based importance and quadratic terms discouraging redundant selections. To avoid the scalability issues of dense k-hot QUBOs, we propose a sparse graph-based BQM built on a kNN similarity graph and an efficient repair procedure that enforces an exact collocation budget. We further introduce hybrid coverage anchors to guarantee global PDE enforcement. We evaluate the method on the 1D time-dependent viscous Burgers equation with shock formation and report both accuracy and end-to-end time-to-accuracy, including a timing breakdown of selection overhead. Results demonstrate that sparse and hybrid formulations reduce selection overhead relative to dense QUBOs while matching or improving accuracy at fixed collocation budgets.
Authors: Serio Agriesti (Department of Technology, Management and Economics, Technical University of Denmark, Lyngby, Denmark), Guido Cantelmo (Department of Technology, Management and Economics, Technical University of Denmark, Lyngby, Denmark), Francisco Camara Pereira (Department of Technology, Management and Economics, Technical University of Denmark, Lyngby, Denmark)
Abstract: Building machine-learning models for estimating traffic flows from OD matrices requires an appropriate design of the training process and a training dataset spanning over multiple regimes and dynamics. As machine-learning models rely heavily on historical data, their predictions are typically accurate only when future traffic patterns resemble those observed during training. However, their performance often degrades when there is a significant statistical discrepancy between historical and future conditions. This issue is particularly relevant in traffic forecasting when predictions are required for modified versions of the network, where the underlying graph structure changes due to events such as maintenance, public demonstrations, flooding, or other extreme disruptions. Ironically, these are precisely the situations in which reliable traffic predictions are most needed. In the presented work, we combine a machine-learning model (graph convolutional neural network) with a meta-learning architecture to train the former to quickly adapt to new graph structures and demand patterns, so that it may easily be applied to scenarios in which changes in the road network (the graph) and the demand (the node features) happen simultaneously. Our results show that the use of meta-learning allows the graph neural network to quickly adapt to unseen graphs (network closures) and OD matrixes while easing the burden of designing a training dataset that covers all relevant patterns for the practitioners. The proposed architecture achieves a R^2 of around 0.85 over unseen closures and OD matrixes.
Authors: Julien Amblard, Niklas Groll, Matthew Tait, Mark Law, G\"urkan Sin, Alessandra Russo
Abstract: Over the past decade, Artificial Intelligence has significantly advanced, mostly driven by large-scale neural approaches. However, in the chemical process industry, where safety is critical, these methods are often unsuitable due to their brittleness, and lack of explainability and interpretability. Furthermore, open-source real-world datasets containing historical failures are scarce in this domain. In this paper, we investigate an approach for predicting failures in chemical processes using symbolic machine learning and conduct a feasibility study in the context of an ethylene oxidation process. Our method builds on a state-of-the-art symbolic machine learning system capable of learning predictive models in the form of probabilistic rules from context-dependent noisy examples. This system is a general-purpose symbolic learner, which makes our approach independent of any specific chemical process. To address the lack of real-world failure data, we conduct our feasibility study leveraging data generated from a chemical process simulator. Experimental results show that symbolic machine learning can outperform baseline methods such as random forest and multilayer perceptron, while preserving interpretability through the generation of compact, rule-based predictive models. Finally, we explain how such learned rule-based models could be integrated into agents to assist chemical plant operators in decision-making during potential failures.
Authors: Jericho Cain
Abstract: Neural network representations are often analyzed as vectors in a fixed Euclidean space. However, their coordinates are not uniquely defined. If a hidden representation is transformed by an invertible linear map, the network function can be preserved by applying the inverse transformation to downstream weights. Representations are therefore defined only up to invertible linear transformations. We study neural representation spaces from this geometric viewpoint and treat them as vector spaces with a gauge freedom under the general linear group. Within this framework, commonly used similarity measures such as cosine similarity become metric-dependent quantities whose values can change under coordinate transformations that leave the model function unchanged. This provides a common interpretation for several observations in the literature, including cosine-similarity instability, anisotropy in embedding spaces, and the appeal of representation comparison methods such as SVCCA and CKA. Experiments on multilayer perceptrons and convolutional networks confirm that inserting invertible transformations into trained models can substantially distort cosine similarity and nearest-neighbor structure while leaving predictions unchanged. These results indicate that analysis of neural representations should focus either on quantities that are invariant under this gauge freedom or on explicitly chosen canonical coordinates.
Authors: Bulent Soykan
Abstract: The Job Shop Scheduling Problem (JSSP) is commonly formulated as a disjunctive graph in which nodes represent operations and edges encode technological precedence constraints as well as machine-sharing conflicts. Most existing reinforcement learning approaches model this graph as homogeneous, merging job-precedence and machine-contention edges into a single relation type. Such a simplification overlooks the intrinsic heterogeneity of the problem structure and may lead to the loss of critical relational information. To address this limitation, we propose the Heterogeneous Graph Transformer (HGT)-Scheduler, a reinforcement learning framework that models the JSSP as a heterogeneous graph. The proposed architecture leverages a Heterogeneous Graph Transformer to capture type-specific relational patterns through edge-type-dependent attention mechanisms applied to precedence and contention relations. The scheduling policy is trained using Proximal Policy Optimization. The effectiveness of the proposed method is evaluated on the Fisher--Thompson benchmark instances. On the FT06 instance, the HGT-Scheduler achieves an optimality gap of 8.4\%, statistically outperforming both an identical architecture that ignores edge types ($p = 0.011$) and a standard Graph Isomorphism Network baseline. On the larger FT10 instance, the approach demonstrates favorable scalability. However, under a 50,000-step training limit, the performance of heterogeneous and homogeneous graph models is comparable, suggesting that edge-type awareness requires longer training horizons for larger problem instances. Ablation analyses further indicate that a three-layer attention architecture provides the best performance. Overall, the results confirm that explicitly modeling distinct edge semantics improves the learning of effective scheduling policies.
Authors: Sayeem Bin Zaman, Fahim Hafiz, Riasat Azim
Abstract: Spatial transcriptomics (ST) enables mapping gene expression with spatial context but is severely affected by high sparsity and technical noise, which conceals true biological signals and hinders downstream analyses. To address these challenges, SpatialMagic was proposed, which is a hybrid imputation model combining MAGIC-based graph diffusion with transformer-based spatial self-attention. The long-range dependencies in the gene expression are captured by graph diffusion, and local neighborhood structure is captured by spatial attention models, which allow for recovering the missing expression values, retaining spatial consistency. Across multiple platforms, SpatialMagic consistently outperforms existing baselines, including MAGIC and attention-based models, achieving peak Adjusted Rand Index (ARI) scores in clustering accuracy of 0.3301 on high-resolution Stereo-Seq data, 0.3074 on Slide-Seq, and 0.4216 on the Sci-Space dataset. Beyond quantitative improvements, SpatialMagic substantially enhances downstream biological analyses by improving the detection of both up- and down-regulated genes while maintaining regulatory consistency across datasets. The pathway enrichment analysis of the recovered genes indicates that they are involved in consistent processes across key metabolic, transport, and neural signaling pathways, suggesting that the framework improves data quality while preserving biological interpretability. Overall, SpatialMagic's hybrid diffusion attention strategy and refinement module outperform state-of-the-art baselines on quantitative metrics and provide a better understanding of the imputed data by preserving tissue architecture and uncovering biologically relevant genes. The source code and datasets are provided in the following link: https://github.com/sayeemzzaman/SpatialMAGIC
Authors: Gregor Baer
Abstract: Evaluating time series attribution methods is difficult because real-world datasets rarely provide ground truth for which time points drive a prediction. A common workaround is to generate synthetic data where class-discriminating features are placed at known locations, but each study currently reimplements this from scratch. We introduce xaitimesynth, a Python package that provides reusable infrastructure for this evaluation approach. The package generates synthetic time series following an additive model where each sample is a sum of background signal and a localized, class-discriminating feature, with the feature window automatically tracked as a ground truth mask. A fluent data generation API and YAML configuration format allow flexible and reproducible dataset definitions for both univariate and multivariate time series. The package also provides standard localization metrics, including AUC-PR, AUC-ROC, Relevance Mass Accuracy, and Relevance Rank Accuracy. xaitimesynth is open source and available at https://github.com/gregorbaer/xaitimesynth.
Authors: Marawan Yakout, Tannistha Maiti, Monira Majhabeen, Tarry Singh
Abstract: Data scarcity is a primary obstacle in developing robust Machine Learning (ML) models for detecting rapidly intensifying tropical cyclones. Traditional data augmentation techniques (rotation, flipping, brightness adjustment) fail to preserve the physical consistency and high-intensity gradients characteristic of rare Category 4-equivalent events, which constitute only 0.14\% of our dataset (202 of 140,514 samples). We propose a physics-informed diffusion model based on the Context-UNet architecture to generate synthetic, multi-spectral satellite imagery of extreme weather events. Our model is conditioned on critical atmospheric parameters such as average wind speed, type of Ocean and stage of development (early, mature, late etc) -- the known drivers of rapid intensification. Using a controlled pre-generated noise sampling strategy and mixed-precision training, we generated $16\times16$ wind-field samples that are cropped from multi-spectral satellite imagery which preserve realistic spatial autocorrelation and physical consistency. Results demonstrate that our model successfully learns discriminative features across ten distinct context classes, effectively mitigating the data bottleneck. Specifically, we address the extreme class imbalance in our dataset, where Class 4 (Ocean 2, early stage with average wind speed 50kn hurricane) contains only 202 samples compared to 79,768 samples in Class 0. This generative framework provides a scalable solution for augmenting training datasets for operational weather detection algorithms. The average Results yield an average Log-Spectral Distance (LSD) of 4.5dB, demonstrating a scalable framework for enhancing operational weather detection algorithms.
Authors: Mai Pham, Vikrant Vaze, Peter Chin
Abstract: Deep reinforcement learning agents frequently suffer from premature convergence, where early entropy collapse causes the policy to discard exploratory behaviors before discovering globally optimal strategies. We introduce Optimistic Policy Regularization (OPR), a lightweight mechanism designed to preserve and reinforce historically successful trajectories during policy optimization. OPR maintains a dynamic buffer of high-performing episodes and biases learning toward these behaviors through directional log-ratio reward shaping and an auxiliary behavioral cloning objective. When instantiated on Proximal Policy Optimization (PPO), OPR substantially improves sample efficiency on the Arcade Learning Environment. Across 49 Atari games evaluated at the 10-million step benchmark, OPR achieves the highest score in 22 environments despite baseline methods being reported at the standard 50-million step horizon. Beyond arcade benchmarks, OPR also generalizes to the CAGE Challenge 2 cyber-defense environment, surpassing the competition-winning Cardiff agent while using the same PPO architecture. These results demonstrate that anchoring policy updates to empirically successful trajectories can improve both sample efficiency and final performance.
Authors: Irene Wang, Vishnu Varma Venkata, Arvind Krishnamurthy, Divya Mahajan
Abstract: The growing scale of deep learning demands distributed training frameworks that jointly reason about parallelism, memory, and network topology. Prior works often rely on heuristic or topology-agnostic search, handling communication and memory separately. Without per-device memory awareness, these methods typically ensure feasibility post hoc by sharding parameters and activations across many devices, increasing synchronization, inflating communication, and underutilizing compute-limiting scalability and efficiency on real datacenter networks. We present NEST, a network-, compute-, and memory-aware device placement framework that unifies model parallelism, topology modeling, and memory feasibility via structured dynamic programming. NEST's DP operates on operator graphs with tensor and expert parallel configurations, explicit allreduce latencies across hierarchical or arbitrary networks, and memory/compute profiles. By factoring parallelism across tensor, pipeline, data, and expert dimensions, NEST defines a principled search space for hybrid strategies while jointly optimizing co-location, network latency, and memory feasibility. Evaluations across diverse hardware and networks show NEST achieves up to 2.43 times higher throughput, better memory efficiency, and improved scalability over state-of-the-art baselines, providing a foundation for co-designing parallelization strategies and datacenter interconnects for next-generation AI infrastructure. The source code of NEST is available at: https://github.com/scai-tech/Nest
Authors: Wenjing Chen, Chengyuan Qian, Shuo Xing, Yi Zhou, Victoria Crawford
Abstract: In this paper, we study cooperative multi-agent reinforcement learning (MARL) where the joint reward exhibits submodularity, which is a natural property capturing diminishing marginal returns when adding agents to a team. Unlike standard MARL with additive rewards, submodular rewards model realistic scenarios where agent contributions overlap (e.g., multi-drone surveillance, collaborative exploration). We provide the first formal framework for this setting and develop algorithms with provable guarantees on sample efficiency and regret bound. For known dynamics, our greedy policy optimization achieves a $1/2$-approximation with polynomial complexity in the number of agents $K$, overcoming the exponential curse of dimensionality inherent in joint policy optimization. For unknown dynamics, we propose a UCB-based learning algorithm achieving a $1/2$-regret of $O(H^2KS\sqrt{AT})$ over $T$ episodes.
Authors: Dhruman Gupta (Ashoka University), Yashas Shende (Ashoka University), Aritra Das (Ashoka University), Chanda Grover Kamra (Ashoka University), Debayan Gupta (Ashoka University)
Abstract: Subsurface ore detection is of paramount importance given the gradual depletion of shallow mineral resources in recent years. It is crucial to explore approaches that go beyond the limitations of traditional geological exploration methods. One such promising new method is joint magnetic and gravitational inversion. Given magnetic and gravitational data on a surface, jointly reconstructing the underlying densities that generate them remains an ill-posed inverse problem. Although joint inversion of multiple properties mitigates the non-uniqueness problem in magnetic and gravitational data, deterministic algorithms converge to a single regularized solution and thus do not capture the distribution of possible solutions. Similarly, most machine learning based techniques predict a single solution without modelling the entire distribution. In this paper, we introduce a novel framework that reframes 3D gravity and magnetic joint inversion as a rectified flow on the Noddyverse dataset, the largest physics-based dataset for inversion. We introduce a Ginzburg-Landau (GL) regularizer, a generalized version of the Ising model that aids in ore identification, enabling physics-aware training. We also propose a guidance methodology based on GL theory that can be used as a plug-and-play module with existing unconditional denoisers. Lastly, we also train and release a VAE for the 3D densities, which facilitates downstream work in the field.
Authors: Yanjun Chen, Yirong Sun, Hanlin Wang, Xinming Zhang, Xiaoyu Shen, Wenjie Li, Wei Zhang
Abstract: Cooperative multi-agent reinforcement learning (MARL) systems powered by large language models (LLMs) are frequently optimized via sparse terminal-only feedback. This shared signal entangles upstream decisions, obstructing accurate decision-level credit assignment. To address this trajectory-level diffusion, we introduce Contextual Counterfactual Credit Assignment (\textbf{\texttt{C3}}). Instead of distributing rewards across an entire episode, \textbf{\texttt{C3}} isolates the causal impact of individual messages by freezing the exact transcript-derived context, evaluating context-matched alternatives via fixed-continuation replay, and applying a leave-one-out (LOO) baseline. This localized intervention extracts unbiased, low-variance marginal advantages for standard policy-gradient optimization. Evaluated across five mathematical and coding benchmarks under matched budgets, \textbf{\texttt{C3}} improves terminal performance over established baselines. Mechanistic diagnostics further show that these gains are accompanied by higher credit fidelity, lower contextual variance, and stronger inter-agent causal dependence. Our code is available at https://github.com/EIT-EAST-Lab/C3.
Authors: Mingi Kang, Zai Yang, Jeova Farias Sales Rocha Neto
Abstract: Activation functions are fundamental to deep neural networks, governing gradient flow, optimization stability, and representational capacity. Within historic deep architectures, while ReLU has been the dominant choice for the activation function, modern transformer-based models increasingly are adopting smoother alternatives such as GELU and other self-gated alternatives. Despite their empirical success, the mathematical relationships among these functions and the principles underlying their effectiveness remains only partially understood. We introduce IGLU, a parametric activation function derived as a scale mixture of GELU gates under a half-normal mixing distribution. This derivation yields a closed-form expression whose gating component is exactly the Cauchy CDF, providing a principled one-parameter family that continuously interpolates between identity-like and ReLU-like behavior via a single sharpness parameter $\sigma$. Unlike GELU's Gaussian gate, IGLU's heavy-tailed Cauchy gate decays polynomially in the negative tail, guaranteeing non-zero gradients for all finite inputs and offering greater robustness to vanishing gradients. We further introduce IGLU-Approx, a computationally efficient rational approximation of IGLU expressed entirely in terms of ReLU operations that eliminates transcendental function evaluation. Through evaluations on CIFAR-10, CIFAR-100, and WikiText-103 across ResNet-20, ViT-Tiny, and GPT-2 Small, IGLU achieves competitive or superior performance on both vision and language datasets against ReLU and GELU baselines, with IGLU-Approx recovering this performance at substantially reduced computational cost. In particular, we show that employing a heavy-tailed gate leads to considerable performance gains in heavily imbalanced classification datasets.
Authors: Abdulrahman Alswaidan, Jeffrey D. Varner
Abstract: Attention heads retrieve: given a query, they return a softmax-weighted average of stored values. We show that this computation is one step of gradient descent on a classical energy function, and that Langevin sampling from the corresponding distribution yields \emph{stochastic attention}: a training-free sampler controlled by a single temperature. Lowering the temperature gives exact retrieval; raising it gives open-ended generation. Because the energy gradient equals the attention map, no score network, training loop, or learned model is required. We validate on four domains (64 to 4,096 dimensions). At generation temperature, stochastic attention is 2.6 times more novel and 2.0 times more diverse than the best learned baseline (a variational autoencoder trained on the same patterns), while matching a Metropolis-corrected gold standard. A simple signal-to-noise rule selects the operating temperature for any dimension. The approach requires no architectural changes and extends naturally to retrieval-augmented generation and in-context learning.
Authors: Gyujun Jeong (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Sungwon Cho (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Minji Shon (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Namhoon Kim (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Woohyun Hwang (Semiconductor Research and Development, Samsung Electronics Co., Ltd, South Korea), Kwangyou Seo (Semiconductor Research and Development, Samsung Electronics Co., Ltd, South Korea), Suhwan Lim (Semiconductor Research and Development, Samsung Electronics Co., Ltd, South Korea), Wanki Kim (Semiconductor Research and Development, Samsung Electronics Co., Ltd, South Korea), Daewon Ha (Semiconductor Research and Development, Samsung Electronics Co., Ltd, South Korea), Prasanna Venkatesan (NVIDIA, Santa Clara, CA, USA), Kihang Youn (NVIDIA, Santa Clara, CA, USA), Ram Cherukuri (NVIDIA, Santa Clara, CA, USA), Yiyi Wang (NVIDIA, Santa Clara, CA, USA), Suman Datta (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Asif Khan (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA), Shimeng Yu (School of Electrical and Computer Engineering, Georgia Institute of Technology, GA, USA)
Abstract: Ferroelectric field-effect transistors (FeFET)-based vertical NAND (Fe-VNAND) has emerged as a promising candidate to overcome z-scaling limitations with lower programming voltages. However, the data retention of 3D Fe-VNAND is hindered by the complex interaction between charge detrapping and ferroelectric depolarization. Developing optimized device designs requires exploring an extensive parameter space, but the high computational cost of conventional Technology Computer-Aided Design (TCAD) tools makes such wide-scale optimization impractical. To overcome these simulation barriers, we present a Physics-Informed Neural Operator (PINO)-based AI surrogate model designed for high-efficiency prediction of threshold voltage (Vth) shifts and retention behavior. By embedding fundamental physical principles into the learning architecture, our PINO framework achieves a speedup exceeding 10000x compared to TCAD while maintaining physical accuracy. This study demonstrates the model's effectiveness on a single FeFET configuration, serving as a pathway toward modeling the retention loss mechanisms.
Authors: Jeffrey Dale, James Keller, Aquila Galusha
Abstract: Streaming clustering is a domain that has become extremely relevant in the age of big data, such as in network traffic analysis or in processing continuously-running sensor data. Furthermore, possibilistic models offer unique benefits over approaches from the literature, especially with the introduction of a "fuzzifier" parameter that controls how quickly typicality degrades as one gets further from cluster centers. We propose a single-pass possibilistic clustering (SPC) algorithm that is effective and easy to apply to new datasets. Key contributions of SPC include the ability to model non-spherical clusters, closed-form footprint updates over arbitrarily sized damped windows, and the employment of covariance union from the multiple hypothesis tracking literature to merge two cluster mean and covariance estimates. SPC is validated against five other streaming clustering algorithm on the basis of cluster purity and normalized mutual information.
Authors: Yan-Ying Chen, Dule Shu, Matthew Hong, Andrew Taber, Jonathan Li, Matthew Klenk
Abstract: Large Language Models (LLMs) have demonstrated impressive capabilities in a wide range of code generation tasks. However, generating code for certain domains remains challenging. One such domain is Computer-Aided Design (CAD) program, where the goal is to produce scripted parametric models that define object geometry for precise design and manufacturing applications. A key challenge in LLM-based CAD program generation is the limited geometric complexity of generated shapes compared to those found in real-world industrial designs. This shortfall is in part due to the lack of diversity in the available CAD program training data. To address this, we propose a novel data augmentation paradigm that prompts an LLM to generate CAD programs conditioned on a reference surface program and a modeling procedure - an idea inspired by practices in industrial design. By varying the reference surface using a collection of organic shapes, our method enriches the geometric distribution of generated CAD models. In particular, it introduces edges and faces defined by spline-based curvature, which are typically missing or underrepresented in existing open-source CAD program datasets. Experiments show that our method produces CAD samples with significantly greater geometric diversity and a higher resemblance to industry-grade CAD designs in terms of the proportion of organic shape primitives. This enhancement makes our CAD data augmentation approach a useful tool for training LLMs and other deep learning models in CAD generation.
Authors: Jim Achterberg, Marcel Haas, Bram van Dijk, Marco Spruit
Abstract: Tree ensembles such as XGBoost are often preferred for discriminative tasks in mixed-type tabular data, due to their inductive biases, minimal hyperparameter tuning, and training efficiency. We argue that these qualities, when leveraged correctly, can make for better generative models as well. As such, we present XGenBoost, a pair of generative models based on XGBoost: i) a Denoising Diffusion Implicit Model (DDIM) with XGBoost as score-estimator suited for smaller datasets, and ii) a hierarchical autoregressive model whose conditionals are learned via XGBoost classifiers, suited for large-scale tabular synthesis. The architectures follow from the natural constraints imposed by tree-based learners, e.g., in the diffusion model, combining Gaussian and multinomial diffusion to leverage native categorical splits and avoid one-hot encoding while accurately modelling mixed data types. In the autoregressive model, we use a fixed-order factorization, a hierarchical classifier to impose ordinal inductive biases when modelling numerical features, and de-quantization based on empirical quantile functions to model the non-continuous nature of most real-world tabular datasets. Through two benchmarks, one containing smaller and the other larger datasets, we show that our proposed architectures outperform previous neural- and tree-based generative models for mixed-type tabular synthesis at lower training cost.
Authors: Nandan Kumar Jha, Brandon Reagen
Abstract: We introduce NerVE, a unified eigenspectral framework for understanding how feed-forward networks (FFNs) in large language models (LLMs) organize and regulate information flow in high-dimensional latent space. Despite FFNs dominating the parameter budget, their high-dimensional dynamics remain poorly understood. NerVE addresses this gap through lightweight, memory-efficient tracking of eigenspectrum dynamics via four complementary metrics: Spectral Entropy (dispersion), Participation Ratio (effective dimensionality), Eigenvalue Early Enrichment (top-heaviness), and Jensen-Shannon divergence (distributional shifts). Our key insight is that FFN nonlinearities reinject variance across eigenmodes, fundamentally governing latent dimension utilization, and that optimizer geometry strongly modulates the extent of this variance reinjection. We validate NerVE across model scales, and diverse architectural and optimizer configurations, each uniquely shaping FFN dynamics: normalization schemes controlling variance flow; FFN weight geometries constraining latent space; positional encoding and activation functions regulating information flow; and optimizer choices redistributing effective capacity across depth. Across these settings, NerVE consistently recovers stable spectral signatures that correlate with model's generalization ability and respond predictably to design choices, generalizing beyond transformer to MLP-Mixer architectures, providing actionable insights for architectural and optimizer choices beyond trial-and-error.
Authors: Zhixu Du, Krishna Teja Chitty-Venkata, Murali Emani, Venkatram Vishwanath, Hai Helen Li, Yiran Chen
Abstract: Mixture-of-experts (MoE) is a common approach for increasing parameter capacity, but applying MoE to state space model (SSM) token mixers can multiply the cost of the recurrent state update. We study how to introduce expert specialization into selective SSMs while preserving computational efficiency. We show that MoE--SSM can refer to two designs: (1) MoE over separated SSMs, which maintains multiple state trajectories and thus scales compute with the number of experts; and (2) MoE-parameterized SSM, which mixes experts in parameter space, maintains a single state trajectory, and evaluates the recurrence once. Our method, Switch Mamba (Swimba), follows the second design by routing over expert-produced SSM streams. Theoretically, we establish well-definedness and stability for MoE-parameterized SSMs and characterize the relationship between the two designs. Empirically, we evaluate Swimba on standard benchmark tasks and measure real-time throughput and latency. Under matched FLOPs, Swimba achieves slightly better average performance than the baseline, with a small slowdown in real-time latency and throughput. Overall, these results suggest that parameter-space MoE can increase SSM capacity while keeping the dominant recurrence cost fixed.
Authors: Milad Shirani, Pete H. Gueldner, Murat Khidoyatov, Jeremy L. Warren, Federica Ninno
Abstract: In this work, we study the mechanical behavior of solids with microstructure using the framework of Cosserat elasticity with a single unit director. This formulation captures the coupling between deformation and orientational fields that arises in many structured materials. To compute equilibrium configurations of such media, we develop two complementary computational approaches: a finite element formulation based on variational principles and a neural network-based solver that directly minimizes the total potential energy. The neural architecture is constructed to respect the fundamental kinematic structure of the theory. In particular, it enforces frame invariance of the energy, satisfies the unit-length constraint on the director field, and represents deformation and director fields through separate networks to preserve their kinematic independence in the variational setting. Beyond satisfying balance laws, however, physically admissible solutions must also correspond to stable energy minimizers. To assess this requirement, we derive the quasiconvexity condition, rank-one convexity condition, and the Legendre-Hadamard inequalities for the Cosserat model and formulate them in a manner suitable for evaluating neural network predictions. These necessary stability conditions provide a physics-based validation framework: network outputs that violate these necessary conditions cannot correspond to stable energy minimizers and can therefore be rejected. In this way, we integrate classical variational stability theory with modern machine-learning solvers, establishing a computational workflow in which equilibrium solutions are not only learned but also assessed for energetic consistency.
Authors: Ege C. Kaya, Mahsa Ghasemi, Abolfazl Hashemi
Abstract: Many distributional quantities in reinforcement learning are intrinsically joint across actions, including distributions of gaps and probabilities of superiority. However, the classical Markov decision process (MDP) formalism specifies only marginal laws and leaves the joint law of counterfactual one-step outcomes across multiple possible actions at a state unspecified. We study coupled-dynamics environments with a multi-action generative interface which can sample counterfactual one-step outcomes for multiple actions under shared exogenous randomness. We propose joint MDPs (JMDPs) as a formalism for such environments by augmenting an MDP with a multi-action sample transition model which specifies a coupling of one-step counterfactual outcomes, while preserving standard MDP interaction as marginal observations. We adopt and formalize a one-step coupling regime where dependence across actions is confined to immediate counterfactual outcomes at the queried state. In this regime, we derive Bellman operators for $n$th-order return moments, providing dynamic programming and incremental algorithms with convergence guarantees.
Authors: Yuhang Song, Naima Abrar Shami, Romaric Duvignau, Vasiliki Kalavri
Abstract: As graphs scale to billions of nodes and edges, graph Machine Learning workloads are constrained by the cost of multi-hop traversals over exponentially growing neighborhoods. While various system-level and algorithmic optimizations have been proposed to accelerate Graph Neural Network (GNN) pipelines, data management and movement remain the primary bottlenecks at scale. In this paper, we explore whether graph sparsification, a well-established technique that reduces edges to create sparser neighborhoods, can serve as a lightweight pre-processing step to address these bottlenecks while preserving accuracy on node classification tasks. We develop an extensible experimental framework that enables systematic evaluation of how different sparsification methods affect the performance and accuracy of GNN models. We conduct the first comprehensive study of GNN training and inference on sparsified graphs, revealing several key findings. First, sparsification often preserves or even improves predictive performance. As an example, random sparsification raises the accuracy of the GAT model by 6.8% on the PubMed graph. Second, benefits increase with scale, substantially accelerating both training and inference. Our results show that the K-Neighbor sparsifier improves model serving performance on the Products graph by 11.7x with only a 0.7% accuracy drop. Importantly, we find that the computational overhead of sparsification is quickly amortized, making it practical for very large graphs.
Authors: Xin Zhang, Xingyu Li, Rongguang Wang, Ruizhong Miao, Zheng Wang, Dan Roth, Chenyang Li
Abstract: Accurate chart comprehension represents a critical challenge in advancing multimodal learning systems, as extensive information is compressed into structured visual representations. However, existing vision-language models (VLMs) frequently struggle to generalize on unseen charts because it requires abstract, symbolic, and quantitative reasoning over structured visual representations. In this work, we introduce Chart-RL, an effective reinforcement learning (RL) method that employs mathematically verifiable rewards to enhance chart question answering in VLMs. Our experiments demonstrate that Chart-RL consistently outperforms supervised fine-tuning (SFT) across different chart understanding benchmarks, achieving relative improvements of 16.7% on MutlChartQA, and 11.5% on ChartInsights. We conduct robustness analysis, where Chart-RL achieves enhanced performance in 18 of 25 perturbed chart categories, demonstrating strong consistency and reasoning capability across visual variations. Furthermore, we demonstrate that task difficulty and inherent complexity are more critical than data quantity in RL training. For instance, Chart-RL trained on merely 10 complex chart-query examples significantly outperforms models trained on over 6,000 simple examples. Additionally, training on challenging reasoning tasks not only improves in-domain generalization relative to simpler tasks, but also facilitate strong transfer to out-of-domain visual mathematical problems.
Authors: Ruipeng Zhang, Hongzhan Yu, Ya-Chien Chang, Chenghao Li, Henrik I. Christensen, Sicun Gao
Abstract: Quadruped locomotion provides a natural setting for understanding when model-free learning can outperform model-based control design, by exploiting data patterns to bypass the difficulty of optimizing over discrete contacts and the combinatorial explosion of mode changes. We give a principled analysis of why imitation learning with quadrupeds can be inherently effective in a small data regime, based on the structure of its limit cycles, Poincar\'e return maps, and local numerical properties of neural networks. The understanding motivates a new imitation learning method that regulates the alignment between variations in a latent space and those over the output actions. Hardware experiments confirm that a few seconds of demonstration is sufficient to train various locomotion policies from scratch entirely offline with reasonable robustness.
Authors: Jiwoo Yoon, Kyumin Choi, Jaewoong Choi
Abstract: Conditional Optimal Transport (COT) problem aims to find a transport map between conditional source and target distributions while minimizing the transport cost. Recently, these transport maps have been utilized in conditional generative modeling tasks to establish efficient mappings between the distributions. However, classical COT inherits a fundamental limitation of optimal transport, i.e., sensitivity to outliers, which arises from the hard distribution matching constraints. This limitation becomes more pronounced in a conditional setting, where each conditional distribution is estimated from a limited subset of data. To address this, we introduce the Conditional Unbalanced Optimal Transport (CUOT) framework, which relaxes conditional distribution-matching constraints through Csisz\'ar divergence penalties while strictly preserving the conditioning marginals. We establish a rigorous formulation of the CUOT problem and derive its dual and semi-dual formulations. Based on the semi-dual form, we propose Conditional Unbalanced Optimal Transport Maps (CUOTM), an outlier-robust conditional generative model built upon a triangular $c$-transform parameterization. We theoretically justify the validity of this parameterization by proving that the optimal triangular map satisfies the $c$-transform relationships. Our experiments on 2D synthetic and image-scale datasets demonstrate that CUOTM achieves superior outlier robustness and competitive distribution-matching performance compared to existing COT-based baselines, while maintaining high sampling efficiency.
Authors: Addison Kalanther, Sanika Bharvirkar, Shankar Sastry, Chinmay Maheshwari
Abstract: Multi-agent reinforcement learning (MARL) is increasingly used to design learning-enabled agents that interact in shared environments. However, training MARL algorithms in general-sum games remains challenging: learning dynamics can become unstable, and convergence guarantees typically hold only in restricted settings such as two-player zero-sum or fully cooperative games. Moreover, when agents have heterogeneous and potentially conflicting preferences, it is unclear what system-level objective should guide learning. In this paper, we propose a new MARL pipeline called Near-Potential Policy Optimization (NePPO) for computing approximate Nash equilibria in mixed cooperative--competitive environments. The core idea is to learn a player-independent potential function such that the Nash equilibrium of a cooperative game with this potential as the common utility approximates a Nash equilibrium of the original game. To this end, we introduce a novel MARL objective such that minimizing this objective yields the best possible potential function candidate and consequently an approximate Nash equilibrium of the original game. We develop an algorithmic pipeline that minimizes this objective using zeroth-order gradient descent and returns an approximate Nash equilibrium policy. We empirically show the superior performance of this approach compared to popular baselines such as MAPPO, IPPO and MADDPG.
Authors: Tong Yang, Moonkyung Ryu, Chih-Wei Hsu, Guy Tennenholtz, Yuejie Chi, Craig Boutilier, Bo Dai
Abstract: Controllable diffusion generation often relies on various heuristics that are seemingly disconnected without a unified understanding. We bridge this gap with Diffusion Controller (DiffCon), a unified control-theoretic view that casts reverse diffusion sampling as state-only stochastic control within (generalized) linearly-solvable Markov Decision Processes (LS-MDPs). Under this framework, control acts by reweighting the pretrained reverse-time transition kernels, balancing terminal objectives against an $f$-divergence cost. From the resulting optimality conditions, we derive practical reinforcement learning methods for diffusion fine-tuning: (i) f-divergence-regularized policy-gradient updates, including a PPO-style rule, and (ii) a regularizer-determined reward-weighted regression objective with a minimizer-preservation guarantee under the Kullback-Leibler (KL) divergence. The LS-MDP framework further implies a principled model form: the optimal score decomposes into a fixed pretrained baseline plus a lightweight control correction, motivating a side-network parameterization conditioned on exposed intermediate denoising outputs, enabling effective gray-box adaptation with a frozen backbone. Experiments on Stable Diffusion v1.4 across supervised and reward-driven finetuning show consistent gains in preference-alignment win rates and improved quality-efficiency trade-offs versus gray-box baselines and even the parameter-efficient white-box adapter LoRA.
Authors: Yuki Shibukawa, Koichi Tanaka, Yuta Saito, Shinji Ito
Abstract: A matching platform is a system that matches different types of participants, such as companies and job-seekers. In such a platform, merely maximizing the number of matches can result in matches being concentrated on highly popular participants, which may increase dissatisfaction among other participants, such as companies, and ultimately lead to their churn, reducing the platform's profit opportunities. To address this issue, we propose a novel online learning problem, Combinatorial Allocation Bandits (CAB), which incorporates the notion of *arm satisfaction*. In CAB, at each round $t=1,\dots,T$, the learner observes $K$ feature vectors corresponding to $K$ arms for each of $N$ users, assigns each user to an arm, and then observes feedback following a generalized linear model (GLM). Unlike prior work, the learner's objective is not to maximize the number of positive feedback, but rather to maximize the arm satisfaction. For CAB, we provide an upper confidence bound algorithm that achieves an approximate regret upper bound, which matches the existing lower bound for the special case. Furthermore, we propose a TS algorithm and provide an approximate regret upper bound. Finally, we conduct experiments on synthetic data to demonstrate the effectiveness of the proposed algorithms compared to other methods.
Authors: Xiangjie Xiao, Cong Zhang, Wen Song, Zhiguang Cao
Abstract: Neural approaches to the Flexible Job Shop Scheduling Problem (FJSP), particularly those based on deep reinforcement learning (DRL), have gained growing attention in recent years. However, existing methods rely on complex feature-engineered state representations (i.e., often requiring more than 20 handcrafted features) and graph-biased neural architectures. To reduce modeling complexity and advance a more generalizable framework for FJSP, we introduce \textsc{ReSched}, a minimalist DRL framework that rethinks both the scheduling formulation and model design. First, by revisiting the Markov Decision Process (MDP) formulation of FJSP, we condense the state space to just four essential features, eliminating historical dependencies through a subproblem-based perspective. Second, we employ Transformer blocks with dot-product attention, augmented by three lightweight but effective architectural modifications tailored to scheduling tasks. Extensive experiments show that \textsc{ReSched} outperforms classical dispatching rules and state-of-the-art DRL methods on FJSP. Moreover, \textsc{ReSched} also generalizes well to the Job Shop Scheduling Problem (JSSP) and the Flexible Flow Shop Scheduling Problem (FFSP), achieving competitive performance against neural baselines specifically designed for these variants.
Authors: Jiayi Wang, John Gounley, Heidi Hanson
Abstract: In cross-silo federated learning (FL), sensitive text datasets remain confined to local organizations due to privacy regulations, making repeated training for each downstream task both communication-intensive and privacy-demanding. A promising alternative is to generate differentially private (DP) synthetic datasets that approximate the global distribution and can be reused across tasks. However, pretrained large language models (LLMs) often fail under domain shift, and federated finetuning is hindered by computational heterogeneity: only resource-rich clients can update the model, while weaker clients are excluded, amplifying data skew and the adverse effects of DP noise. We propose a flexible participation framework that adapts to client capacities. Strong clients perform DP federated finetuning, while weak clients contribute through a lightweight DP voting mechanism that refines synthetic text. To ensure the synthetic data mirrors the global dataset, we apply control codes (e.g., labels, topics, metadata) that represent each client's data proportions and constrain voting to semantically coherent subsets. This two-phase approach requires only a single round of communication for weak clients and integrates contributions from all participants. Experiments show that our framework improves distribution alignment and downstream robustness under DP and heterogeneity.
Authors: Zhiji Yang, Mei Huang, Xinyu Li, Xianli Pan, Qi Wang, Jianhua Zhao
Abstract: Anomaly detection is a crucial machine-learning task with wide-ranging applications. Deep Support Vector Data Description (Deep SVDD) is a prominent deep one-class method, but it is vulnerable to hypersphere collapse, often relies on heuristic choices for hypersphere parameters, and provides limited interpretability. To address these issues, we propose Interpretable Maximum Margin Deep Anomaly Detection (IMD-AD), which leverages a small set of labeled anomalies and a maximum margin objective to stabilize training and improve discrimination. It is inherently resilient to hypersphere collapse. Furthermore, we prove an equivalence between hypersphere parameters and the network's final-layer weights, which allows the center and radius to be learned end-to-end as part of the model and yields intrinsic interpretability and visualizable outputs. We further develop an efficient training algorithm that jointly optimizes representation, margin, and final-layer parameters. Extensive experiments and ablation studies on image and tabular benchmarks demonstrate that IMD-AD empirically improves detection performance over several state-of-the-art baselines while providing interpretable decision diagnostics.
Authors: Woogyeol Jin, Taywon Min, Yongjin Yang, Swanand Ravindra Kadhe, Yi Zhou, Dennis Wei, Nathalie Baracaldo, Kimin Lee
Abstract: On-policy distillation is a promising approach for transferring knowledge between language models, where a student learns from dense token-level signals along its own trajectories. This framework typically uses reverse KL divergence, encouraging the student to match the teacher's high-confidence predictions. However, we show that the mode-seeking property of reverse KL reduces generation diversity and yields unstable learning signals when the teacher distribution has high entropy. To address this, we introduce Entropy-Aware On-Policy Distillation. Our key idea is augmenting the standard reverse KL objective with forward KL when teacher entropy is high, capturing the full range of plausible outputs while retaining precise imitation elsewhere. It balances mode-seeking precision with mode-covering robustness without sacrificing on-policy training efficiency. Experiments show that our method maintains generation diversity (sustained token-level entropy) and improves student-teacher alignment (lower forward KL on high-entropy tokens). Across six math reasoning benchmarks, this yields Pass@8 accuracy gains of +1.37 for Qwen3-0.6B-Base, +2.39 for Qwen3-1.7B-Base, and +5.05 for Qwen3-4B-Base compared to baseline on-policy distillation methods. These results demonstrate that accounting for teacher uncertainty is essential for maintaining diversity and achieving effective knowledge transfer.
Authors: Michael Hauri, Friedemann Zenke
Abstract: Model-based reinforcement learning (MBRL) agents operating in high-dimensional observation spaces, such as Dreamer, rely on learning abstract representations for effective planning and control. Existing approaches typically employ reconstruction-based objectives in the observation space, which can render representations sensitive to task-irrelevant details. Recent alternatives trade reconstruction for auxiliary action prediction heads or view augmentation strategies, but perform worse in the Crafter environment than reconstruction-based methods. We close this gap between Dreamer and reconstruction-free models by introducing a JEPA-style predictor defined on continuous, deterministic representations. Our method matches Dreamer's performance on Crafter, demonstrating effective world model learning on this benchmark without reconstruction objectives.
Authors: Muhammad Khalifa, Zohaib Khan, Omer Tafveez, Hao Peng, Lu Wang
Abstract: Reward hacking is a form of misalignment in which models overoptimize proxy rewards without genuinely solving the underlying task. Precisely measuring reward hacking occurrence remains challenging because true task rewards are often expensive or impossible to compute. We introduce Countdown-Code, a minimal environment where models can both solve a mathematical reasoning task and manipulate the test harness. This dual-access design creates a clean separation between proxy rewards (test pass/fail) and true rewards (mathematical correctness), enabling accurate measurement of reward-hacking rates. Using this environment, we study reward hacking in open-weight LLMs and find that such behaviors can be unintentionally learned during supervised fine-tuning (SFT) when even a small fraction of reward-hacking trajectories leak into training data. As little as 1\% contamination in distillation SFT data is sufficient for models to internalize reward hacking which resurfaces during subsequent reinforcement learning (RL). We further show that RL amplifies misalignment and drives its generalization beyond the original domain. We open-source our environment and code to facilitate future research on reward hacking in LLMs. Our results reveal a previously underexplored pathway through which reward hacking can emerge and persist in LLMs, underscoring the need for more rigorous validation of synthetic SFT data. Code is available at https://github.com/zohaib-khan5040/Countdown-Code.
Authors: Tao Shi, Liangming Chen, Long Jin, Mengchu Zhou
Abstract: In the training of neural networks, adaptive moment estimation (Adam) typically converges fast but exhibits suboptimal generalization performance. A widely accepted explanation for its defect in generalization is that it often tends to converge to sharp minima. To enhance its ability to find flat minima, we propose its new variant named inverse Adam (InvAdam). The key improvement of InvAdam lies in its parameter update mechanism, which is opposite to that of Adam. Specifically, it computes element-wise multiplication of the first-order and second-order moments, while Adam computes the element-wise division of these two moments. This modification aims to increase the step size of the parameter update when the elements in the second-order moments are large and vice versa, which helps the parameter escape sharp minima and stay at flat ones. However, InvAdam's update mechanism may face challenges in convergence. To address this challenge, we propose dual Adam (DualAdam), which integrates the update mechanisms of both Adam and InvAdam, ensuring convergence while enhancing generalization performance. Additionally, we introduce the diffusion theory to mathematically demonstrate InvAdam's ability to escape sharp minima. Extensive experiments are conducted on image classification tasks and large language model (LLM) fine-tuning. The results validate that DualAdam outperforms Adam and its state-of-the-art variants in terms of generalization performance. The code is publicly available at https://github.com/LongJin-lab/DualAdam.
Authors: Subhojyoti Mukherjee, Stefano Petrangeli, Branislav Kveton, Trung Bui, Franck Dernoncourt, Arko Mukherjee
Abstract: Direct prompt-based editing often fails on complex transformations because vague and subjective prompts often require nuanced understanding of what should be changed in the image. Our core intuition is that leveraging compositional image editing tools rather than direct prompting profits from structured agent-level planning with explicit reasoning, leading to better results. This structured planning framework enables efficient offline RL post-training on quality-scored trajectories to improve performance. We present a tool-based agentic RL post-training framework that addresses this through structured planning with chain-of-thought reasoning. Our key contributions include: (1) A tool-based agentic planning methodology that combines a compositional library of orthogonal primitive transformations, structured context representation, and explicit per-step reasoning to decompose complex styling into interpretable tool sequences. (2) A synthetic data generation pipeline producing three large-scale datasets (each $\sim$10K trajectories) with reasoning chains, plans, and quality scores, as no existing datasets provide such supervision. Our datasets and code are publicly available at the HuggingFace repository. (3) Offline RL training methods for learning planners with reasoning as our core algorithmic contributions, which consistently improve over the Edit-Only baseline in visual quality and instruction following. (4) Comprehensive evaluation across 4B and 8B parameter Qwen3-VL models showing that our methods outperform other baselines in the majority of compositional tasks, validated by human evaluations.
Authors: Hemanth Saratchandran, Simon Lucey
Abstract: We present a theoretical analysis of the Jacobian of an attention block within a transformer, showing that it is governed by the query, key, and value projections that define the attention mechanism. Leveraging this insight, we introduce a method that systematically alters the spectral properties of each attention layer to reduce the Jacobian's condition number, thereby improving the overall conditioning of the attention layers within a transformer network. We empirically show that this improved Jacobian conditioning translates to enhanced performance in practice. Our approach is simple, broadly applicable, and can be easily integrated as a drop-in replacement for a wide range of existing attention mechanisms. We validate its effectiveness across diverse transformer architectures and tasks, demonstrating consistent improvements in performance.
Authors: Yuxuan Han, Meng-Hao Guo, Zhengning Liu, Wenguang Chen, Shi-Min Hu
Abstract: Optimizing GPU kernels manually is a challenging and time-consuming task. With the rapid development of LLMs, automated GPU kernel optimization is gradually becoming a tangible reality. However, current LLM-driven automated optimization methods narrowly focus on machine learning applications, such as PyTorch operator optimization, while overlooking broader domains like sparse matrix operations in scientific computing. Extending to these broader applications brings new challenges for the benchmark and algorithm. Therefore, developing a general-purpose automated kernel optimization method becomes our primary focus. In this paper, we address the absence of systematic evaluation for multi-scenario settings by introducing MSKernelBench, which spans multiple scenarios, including fundamental algebraic operations, common LLM kernels, sparse matrix operators, and scientific computing routines, each supporting both FP32 and BF16 precision. Building on this benchmark, we introduce CUDAMaster, a multi-agent, hardware-aware system for kernel optimization that leverages profiling information and automatically constructs the full compilation and execution toolchain. Experimental results demonstrate that CUDAMaster achieves significant speedups across most operators, outperforming Astra by about 35%. In several cases, its performance matches or surpasses that of highly optimized, closed-source libraries such as cuBLAS. A demo showcasing the original and optimized code for each operator is available at https://hanyx2021.github.io/MSKernelBenchDemo/.
Authors: Haonan Xu, Yang Yang
Abstract: Out-of-distribution (OOD) detection is a well-known challenge due to deep models often producing overconfident. In this paper, we reveal a key insight that trained classifiers tend to rely on sparse parameter contribution patterns, meaning that only a few dominant parameters drive predictions. This brittleness can be exploited by OOD inputs that anomalously trigger these parameters, resulting in overconfident predictions. To address this issue, we propose a simple yet effective method called Shaping Parameter Contribution Patterns (SPCP), which enhances OOD detection robustness by encouraging the classifier to learn boundary-oriented dense contribution patterns. Specifically, SPCP operates during training by rectifying excessively high parameter contributions based on a dynamically estimated threshold. This mechanism promotes the classifier to rely on a broader set of parameters for decision-making, thereby reducing the risk of overconfident predictions caused by anomalously triggered parameters, while preserving in-distribution (ID) performance. Extensive experiments under various OOD detection setups verify the effectiveness of SPCP.
Authors: Zhaoyang Ren, Qilin Li
Abstract: High-fidelity nonlinear finite-element (FE) simulations of reinforced-concrete (RC) structures are still costly, especially in parametric settings where loading positions vary. We develop a dual-graph spatiotemporal GNN surrogate to approximate the time histories of RC beams under four-point bending. To generate training data, we run a parametric Abaqus campaign that independently shifts the two loading blocks on a mesh-aligned grid and exports full-field responses at fixed normalized loading levels. The model rolls out autoregressively and jointly predicts nodal displacements, element wise von Mises stress, element-wise equivalent plastic strain (PEEQ), and the global vertical reaction force in a single multi-task setup. A key motivation is the peak loss introduced when element quantities are forced through node-based representations. We therefore couple node- and element-level dynamics using two recurrent graph branches: a node-level graph convolutional gated recurrent unit (GConvGRU) for kinematics and an element-level GConvGRU for history-dependent internal variables, with global force predicted through pooling on the element branch. In controlled ablations, removing the Element to Node to Element pathway improves peak-sensitive prediction in localized high-gradient stress/PEEQ regions without degrading global load displacement trends. After training, the surrogate produces full trajectories at a fraction of the cost of nonlinear FE, enabling faster parametric evaluation and design exploration.
Authors: Jilong Liu, Yonghui Yang, Pengyang Shao, Haokai Ma, Wei Qin, Richang Hong
Abstract: Direct Preference Optimization (DPO) aligns large language models by optimizing pairwise preferences and has shown remarkable effectiveness as a simple and scalable alternative to RLHF. However, in practice, preference data are often noisy. Existing robust variants of DPO mainly rely on uniform objective modifications or global reweighting. While partially effective, these methods treat noisy samples as a homogeneous source of uncertainty and fail to distinguish between different noise types, leading to sub-optimal alignment robustness. In this work, we show that robust preference alignment benefits from addressing different noise types with targeted interventions rather than uniform regularization. We propose winsorized Direct Preference Optimization~(wDPO), a robust LLM alignment approach with hierarchical winsorization. Specifically, wDPO adopts a reward-free hierarchical intervention strategy that leverages only signals already available during DPO training. It first uses the implicit margin from DPO log-ratio to identify heterogeneous noise patterns without relying on external reward models. For hard noise, wDPO performs a data-level intervention by sparsely correcting strongly inconsistent preference pairs. For ambiguous comparisons, it applies a gradient-level intervention through soft winsorization, capping extreme losses in the high-loss tail to prevent weakly informative samples from dominating gradient updates. Extensive experiments on PKU-SafeRLHF and multiple external safety benchmarks demonstrate that wDPO consistently improves preference alignment quality and robustness over vanilla DPO and strong DPO-family baselines, with particularly pronounced gains under controlled label-flip noise.
Authors: Yair Ashlagi, Roi Livni, Shay Moran, Tom Waknine
Abstract: Margin-based learning, exemplified by linear and kernel methods, is one of the few classical settings where generalization guarantees are independent of the number of parameters. This makes it a central case study in modern highly over-parameterized learning. We ask what minimal mathematical structure underlies this phenomenon. We begin with a simple margin-based problem in arbitrary metric spaces: concepts are defined by a center point and classify points according to whether their distance lies below $r$ or above $R$. We show that whenever $R>3r$, this class is learnable in \emph{any} metric space. Thus, sufficiently large margins make learnability depend only on the triangle inequality, without any linear or analytic structure. Our first main result extends this phenomenon to concepts defined by bounded linear combinations of distance functions, and reveals a sharp threshold: there exists a universal constant $\gamma>0$ such that above this margin the class is learnable in every metric space, while below it there exist metric spaces where it is not learnable at all. We then ask whether margin-based learnability can always be explained via an embedding into a linear space -- that is, reduced to linear classification in some Banach space through a kernel-type construction. We answer this negatively by developing a structural taxonomy of Banach spaces: if a Banach space is learnable for some margin parameter $\gamma\geq 0$, then it is learnable for all such $\gamma$, and in infinite-dimensional spaces the sample complexity must scale polynomially in $1/\gamma$. Specifically, it must grow as $(1/\gamma)^p$ for some $p\ge 2$, and every such rate is attainable.
Authors: Chuxue Cao, Honglin Lin, Zhanping Zhong, Xin Gao, Mengzhang Cai, Conghui He, Sirui Han, Lijun Wu
Abstract: Large Language Models (LLMs) have demonstrated strong general capabilities, yet their deployment in finance remains challenging due to dense domain-specific terminology, stringent numerical reasoning requirements, and low tolerance for factual errors. We conduct a controlled empirical study showing that in specialized vertical domains, performance is largely determined by the quality and difficulty/verifiability profile of post-training data. We introduce \textbf{ODA-Fin-SFT-318k}, constructed via multi-stage distillation and verification to produce high-quality Chain-of-Thought supervision, and \textbf{ODA-Fin-RL-12k}, curated for hard-but-verifiable tasks that balance reward precision and task diversity. Using standard SFT and RL pipelines, we show that high-quality CoT distillation establishes a robust foundation during SFT, while difficulty- and verifiability-aware sampling improves RL generalization. Evaluated on nine benchmarks spanning general financial tasks, sentiment analysis, and numerical reasoning, our ODA-Fin-RL-8B consistently surpasses open-source state-of-the-art (SOTA) financial LLMs of comparable size. We release our ODA-Fin-SFT-318k and ODA-Fin-RL-12k datasets, along with trained models to advance data-centric financial AI research.
Authors: Kavyansh Tyagi, Vishwas Rathi, Puneet Goyal
Abstract: Accurate and efficient 3D medical image segmentation is essential for clinical AI, where models must remain reliable under stringent memory, latency, and data availability constraints. Transformer-based methods achieve strong accuracy but suffer from excessive parameters, high FLOPs, and limited generalization. We propose LightMedSeg, a modular UNet-style segmentation architecture that integrates anatomical priors with adaptive context modeling. Anchor-conditioned FiLM modulation enables anatomy-aware feature calibration, while a local structural prior module and texture-aware routing dynamically allocate representational capacity to boundary-rich regions. Computational redundancy is minimized through ghost and depthwise convolutions, and multi-scale features are adaptively fused via a learned skip router with anchor-relative spatial position bias. Despite requiring only 0.48M parameters and 14.64~GFLOPs, LightMedSeg achieves segmentation accuracy within a few Dice points of heavy transformer baselines. Therefore, LightMedSeg is a deployable and data-efficient solution for 3D medical image segmentation. Code will be released publicly upon acceptance.
Authors: Andrea Giuseppe Di Francesco, Andrea Rubbi, Pietro Li\`o
Abstract: Predicting how cells respond to genetic perturbations is fundamental to understanding gene function, disease mechanisms, and therapeutic development. While recent deep learning approaches have shown promise in modeling single-cell perturbation responses, they struggle to generalize across cell types and perturbation contexts due to limited contextual information during generation. We introduce PT-RAG (Perturbation-aware Two-stage Retrieval-Augmented Generation), a novel framework that extends Retrieval-Augmented Generation beyond traditional language-model applications to cellular biology. Unlike standard RAG systems designed for text retrieval with pre-trained LLMs, perturbation retrieval lacks established similarity metrics and requires learning what constitutes relevant context, making differentiable retrieval essential. PT-RAG addresses this through a two-stage pipeline: first, retrieving candidate perturbations $K$ using GenePT embeddings, then adaptively refining the selection through Gumbel-Softmax discrete sampling conditioned on both the cell state and the input perturbation. This cell-type-aware differentiable retrieval enables end-to-end optimization of the retrieval objective jointly with generation. On the Replogle-Nadig single-gene perturbation dataset, we demonstrate that PT-RAG outperforms both STATE and vanilla RAG under identical experimental conditions, with the strongest gains in distributional similarity metrics ($W_1$, $W_2$). Notably, vanilla RAG's dramatic failure is itself a key finding: it demonstrates that differentiable, cell-type-aware retrieval is essential in this domain, and that naive retrieval can actively harm performance. Our results establish retrieval-augmented generation as a promising paradigm for modelling cellular responses to gene perturbation. The code to reproduce our experiments is available at https://github.com/difra100/PT-RAG_ICLR.
Authors: Zixuan Yu, Zhenheng Tang, Tongliang Liu, Chengqi Zhang, Xiaowen Chu, Bo Han
Abstract: Despite the integration of search tools, Deep Search Agents often suffer from a misalignment between reasoning-driven queries and the underlying web indexing structures. Existing frameworks treat the search engine as a static utility, leading to queries that are either too coarse or too granular to retrieve precise evidence. We propose WeDas, a Web Content Distribution Aware framework that incorporates search-space structural characteristics into the agent's observation space. Central to our method is the Query-Result Alignment Score, a metric quantifying the compatibility between agent intent and retrieval outcomes. To overcome the intractability of indexing the dynamic web, we introduce a few-shot probing mechanism that iteratively estimates this score via limited query accesses, allowing the agent to dynamically recalibrate sub-goals based on the local content landscape. As a plug-and-play module, WeDas consistently improves sub-goal completion and accuracy across four benchmarks, effectively bridging the gap between high-level reasoning and low-level retrieval.
Authors: Chia-Fu Lin, Yi-Ju Tseng
Abstract: Second primary cancer (SPC), a new cancer in patients different from previously diagnosed, is a growing concern due to improved cancer survival rates. Early prediction of SPC is essential to enable timely clinical interventions. This study focuses on lung cancer survivors treated in Taiwanese hospitals, where the limited size and geographic scope of local datasets restrict the effectiveness and generalizability of traditional machine learning approaches. To address this, we incorporate external data from the publicly available US-based Surveillance, Epidemiology, and End Results (SEER) program, significantly increasing data diversity and scale. However, the integration of multi-source datasets presents challenges such as feature inconsistency and privacy constraints. Rather than naively merging data, we proposed a loss fusion horizontal federated learning (LF2L) framework that can enable effective cross-institutional collaboration while preserving institutional privacy by avoiding data sharing. Using both common and unique features and balancing their contributions through a shared loss mechanism, our method demonstrates substantial improvements in the prediction performance of SPC. Experiment results show statistically significant improvements in AUROC and AUPRC when compared to localized, horizontal federated, and centralized learning baselines. This highlights the importance of not only acquiring external data but also leveraging it effectively to enhance model performance in real-world clinical model development.
Authors: Madhurima Panja, Grace Younes, Tanujit Chakraborty
Abstract: Chaotic time series are notoriously difficult to forecast. Small uncertainties in initial conditions amplify rapidly, while strong nonlinearities and regime dependent variability constrain predictability. Although modern deep learning often delivers strong short horizon accuracy, its black box nature limits scientific insight and practical trust in settings where understanding the underlying dynamics matters. To address this gap, we propose two complementary symbolic forecasters that learn explicit, interpretable algebraic equations from chaotic time series data. Symbolic Neural Forecaster (SyNF) adapts a neural network based equation learning architecture to the forecasting setting, enabling fully differentiable discovery of compact and interpretable algebraic relations. The Symbolic Tree Forecaster (SyTF) builds on evolutionary symbolic regression to search directly over equation structures under a principled accuracy complexity trade off. We evaluate both approaches in a rolling window nowcasting setting with one step ahead forecasting using several accuracy metrics and compare against a broad suite of baselines spanning classical statistical models, tree ensembles, and modern deep learning architectures. Numerical experiments cover a benchmark of 132 low dimensional chaotic attractors and two real world chaotic time series, namely weekly dengue incidence in San Juan and the Nino 3.4 sea surface temperature index. Across datasets, symbolic forecasters achieve competitive one step ahead accuracy while providing transparent equations that reveal salient aspects of the underlying dynamics.
Authors: Ninda Nurseha Amalina, Heungjo An
Abstract: Patient no-shows disrupt outpatient clinic operations, reduce productivity, and may delay necessary care. Clinics often adopt overbooking or double-booking to mitigate these effects. However, poorly calibrated policies can increase congestion and waiting times. Most existing methods rely on fixed heuristics and fail to adapt to real-time scheduling conditions or patient-specific no-show risk. To address these limitations, we propose an adaptive outpatient double-booking framework that integrates individualized no-show prediction with multi-objective reinforcement learning. The scheduling problem is formulated as a Markov decision process, and patient-level no-show probabilities estimated by a Multi-Head Attention Soft Random Forest model are incorporated in the reinforcement learning state. We develop a Multi-Policy Proximal Policy Optimization method equipped with a Multi-Policy Co-Evolution Mechanism. Under this mechanism, we propose a novel {\tau} rule based on Kullback-Leibler divergence that enables selective knowledge transfer among behaviorally similar policies, improving convergence and expanding the diversity of trade-offs. In addition, SHapley Additive exPlanations is used to interpret both the predicted no-show risk and the agent's scheduling decisions. The proposed framework determines when to single-book, double-book, or reject appointment requests, providing a dynamic and data-driven alternative to conventional outpatient scheduling policies.
Authors: Pavan Karjol, Kumar Shubham, Prathosh AP
Abstract: Continuous symmetries are fundamental to many scientific and learning problems, yet they are often unknown a priori. Existing symmetry discovery approaches typically search directly in the space of transformation generators or rely on learned augmentation schemes. We propose a fundamentally different perspective based on spectral structure. We introduce a framework for discovering continuous one-parameter subgroups using the Generalized Fourier Transform (GFT). Our central observation is that invariance to a subgroup induces structured sparsity in the spectral decomposition of a function across irreducible representations. Instead of optimizing over generators, we detect symmetries by identifying this induced sparsity pattern in the spectral domain. We develop symmetry detection procedures on maximal tori, where the GFT reduces to multi-dimensional Fourier analysis through their irreducible representations. Across structured tasks, including the double pendulum and top quark tagging, we demonstrate that spectral sparsity reliably reveals one-parameter symmetries. These results position spectral analysis as a principled and interpretable alternative to generator-based symmetry discovery.
Authors: Nilesh Jain, Rohit Yadav, Sagar Kotian, Claude AI
Abstract: We present AutoResearch-RL, a framework in which a reinforcement learning agent conducts open-ended neural architecture and hyperparameter research without human supervision, running perpetually until a termination oracle signals convergence or resource exhaustion. At each step the agent proposes a code modification to a target training script, executes it under a fixed wall clock time budget, observes a scalar reward derived from validation bits-per-byte (val-bpb), and updates its policy via Proximal Policy Optimisation (PPO). The key design insight is the separation of three concerns: (i) a frozen environment (data pipeline, evaluation protocol, and constants) that guarantees fair cross-experiment comparison; (ii) a mutable target file (train.py) that represents the agent's editable state; and (iii) a meta-learner (the RL agent itself) that accumulates a growing trajectory of experiment outcomes and uses them to inform subsequent proposals. We formalise this as a Markov Decision Process, derive convergence guarantees under mild assumptions, and demonstrate empirically on a single GPU nanochat pretraining benchmark that AutoResearch-RL discovers configurations that match or exceed hand-tuned baselines after approximately 300 overnight iterations, with no human in the loop.
Authors: Yiming Sun, Qi Cheng, Licheng Liu, Runlong Yu, Yiqun Xie, Xiaowei Jia
Abstract: This paper proposes a new method for crop yield prediction, which is essential for developing management strategies, informing insurance assessments, and ensuring long-term food security. Although existing data-driven approaches have shown promise in this domain, their performance often degrades when applied across large geographic regions and long time periods. This limitation arises from two key challenges: (1) difficulty in jointly capturing short-term and long-term temporal patterns, and (2) inability to effectively accommodate spatial data variability in agricultural systems. Ignoring these issues often leads to unreliable predictions for specific regions or years, which ultimately affects policy decisions and resource allocation. In this paper, we propose a new predictive framework to address these challenges. First, we introduce a new backbone model architecture that captures both short-term daily-scale crop growth dynamics and long-term dependencies across years. To further improve generalization across diverse spatial regions, we augment this model with a retrieval-based adaptation strategy. Recognizing the substantial yield variation across years, we design a novel retrieval-and-refinement pipeline that adjusts retrieved samples by removing cross-year bias not explained by input features. Our experiments on real-world county-level corn yield data over 630 counties in the US demonstrate that our method consistently outperforms different types of baselines. The results also verify the effectiveness of the retrieval-based augmentation method in improving model robustness under spatial heterogeneity.
Authors: Angad Singh Ahuja
Abstract: Robustness under latent distribution shift remains challenging in partially observable reinforcement learning. We formalize a focused setting where an adversary selects a hidden initial latent distribution before the episode, termed an adversarial latent-initial-state POMDP. Theoretically, we prove a latent minimax principle, characterize worst-case defender distributions, and derive approximate best-response inequalities with finite-sample concentration bounds that make the optimization and sampling terms explicit. Empirically, using a Battleship benchmark, we demonstrate that targeted exposure to shifted latent distributions reduces average robustness gaps between Spread and Uniform distributions from 10.3 to 3.1 shots at equal budget. Furthermore, iterative best-response training exhibits budget-sensitive behavior that is qualitatively consistent with the theorem-guided diagnostics once one accounts for discounted PPO surrogates and finite-sample noise. Ultimately, we show that for latent-initial-state problems, the framework yields a clean evaluation game and useful theorem-motivated diagnostics while also making clear where implementation-level surrogates and optimization limits enter.
Authors: Lujing Zhang, Daniel Hsu, Sivaraman Balakrishnan
Abstract: Multi-group learning is a learning task that focuses on controlling predictors' conditional losses over specified subgroups. We propose ShakyPrepend, a method that leverages tools inspired by differential privacy to obtain improved theoretical guarantees over existing approaches. Through numerical experiments, we demonstrate that ShakyPrepend adapts to both group structure and spatial heterogeneity. We provide practical guidance for deploying multi-group learning algorithms in real-world settings.
Authors: Truong Xuan Khanh, Truong Quynh Hoa
Abstract: Neural networks often rely on spurious shortcuts for many epochs before discovering structured representations. However, the mechanism governing when this transition occurs and whether its timing can be predicted remains unclear. Prior work shows that gradient descent converges to low norm solutions and that neural networks exhibit simplicity bias, but neither explains the timescale of the transition from shortcut features to structured representations. We introduce the Norm-Hierarchy Transition (NHT) framework, which explains delayed representation learning as the slow traversal of a hierarchy of parameter norms during regularized optimization. When multiple interpolating solutions exist with different norms, weight decay gradually moves the model from high norm shortcut solutions toward lower norm structured representations. We derive a tight bound showing that the transition delay grows logarithmically with the ratio between shortcut and structured norms. Experiments on modular arithmetic, CIFAR-10 with spurious features, CelebA, and Waterbirds support the predictions of the framework. The results suggest that grokking, shortcut learning, and delayed feature discovery arise from a common mechanism based on norm hierarchy traversal during training.
Authors: Antonio De Santis, Schrasing Tong, Marco Brambilla, Lalana Kagal
Abstract: Concept Bottleneck Models (CBMs) aim for ante-hoc interpretability by learning a bottleneck layer that predicts interpretable concepts before the decision. State-of-the-art approaches typically select which concepts to learn via human specification, open knowledge graphs, prompting an LLM, or using general CLIP concepts. However, concepts defined a-priori may not have sufficient predictive power for the task or even be learnable from the available data. As a result, these CBMs often significantly trail their black-box counterpart when controlling for information leakage. To address this, we introduce a novel CBM pipeline named Mechanistic CBM (M-CBM), which builds the bottleneck directly from a black-box model's own learned concepts. These concepts are extracted via Sparse Autoencoders (SAEs) and subsequently named and annotated on a selected subset of images using a Multimodal LLM. For fair comparison and leakage control, we also introduce the Number of Contributing Concepts (NCC), a decision-level sparsity metric that extends the recently proposed NEC metric. Across diverse datasets, we show that M-CBMs consistently surpass prior CBMs at matched sparsity, while improving concept predictions and providing concise explanations. Our code is available at https://github.com/Antonio-Dee/M-CBM.
Authors: Yuanyun Zhang, Shi Li
Abstract: Clinical machine learning models are increasingly trained using large scale, multimodal foundation paradigms, yet deployment environments often differ systematically from the data generating settings used during training. Such shifts arise from heterogeneous measurement policies, documentation practices, and institutional workflows, leading to representation entanglement between physiologic signal and practice specific artifacts. In this work, we propose a practice invariant representation learning framework for multimodal clinical prediction. We model clinical observations as arising from latent physiologic factors and environment dependent processes, and introduce an objective that jointly optimizes predictive performance while suppressing environment predictive information in the learned embedding. Concretely, we combine supervised risk minimization with adversarial environment regularization and invariant risk penalties across hospitals. Across multiple longitudinal EHR prediction tasks and cross institution evaluations, our method improves out of distribution AUROC by up to 2 to 3 points relative to masked pretraining and standard supervised baselines, while maintaining in distribution performance and improving calibration. These results demonstrate that explicitly accounting for systematic distribution shift during representation learning yields more robust and transferable clinical models, highlighting the importance of structural invariance alongside architectural scale in healthcare AI.
Authors: Sean Gunn, Jorio Cocola, Oliver De Candido, Vaggos Chatziafratis, Paul Hand
Abstract: Generative models have emerged as powerful priors for solving inverse problems. These models typically represent a class of natural signals using a single fixed complexity or dimensionality. This can be limiting: depending on the problem, a fixed complexity may result in high representation error if too small, or overfitting to noise if too large. We develop tunable-complexity priors for diffusion models, normalizing flows, and variational autoencoders, leveraging nested dropout. Across tasks including compressed sensing, inpainting, denoising, and phase retrieval, we show empirically that tunable priors consistently achieve lower reconstruction errors than fixed-complexity baselines. In the linear denoising setting, we provide a theoretical analysis that explicitly characterizes how the optimal tuning parameter depends on noise and model structure. This work demonstrates the potential of tunable-complexity generative priors and motivates both the development of supporting theory and their application across a wide range of inverse problems.
Authors: Yucheng Xing, Xin Wang
Abstract: Long-horizon wildfire risk forecasting requires generating probabilistic spatial fields under sparse event supervision while maintaining computational efficiency across multiple prediction horizons. Extending diffusion models to multi-step forecasting typically repeats the denoising process independently for each horizon, leading to redundant computation. We introduce N-Tree Diffusion (NT-Diffusion), a hierarchical diffusion model designed for long-horizon wildfire risk forecasting. Fire occurrences are represented as continuous Fire Risk Maps (FRMs), which provide a smoothed spatial risk field suitable for probabilistic modeling. Instead of running separate diffusion trajectories for each predicted timestamp, NT-Diffusion shares early denoising stages and branches at later levels, allowing horizon-specific refinement while reducing redundant sampling. We evaluate the proposed framework on a newly collected real-world wildfire dataset constructed for long-horizon probabilistic prediction. Results indicate that NT-Diffusion achieves consistent accuracy improvements and reduced inference cost compared to baseline forecasting approaches.
Authors: Mohammed Alnemari, Rizwan Qureshi, Nader Begrazadah
Abstract: Neural scaling laws describe how model performance improves as a power law with size, but existing work focuses on models above 100M parameters. The sub-20M regime -- where TinyML and edge AI operate -- remains unexamined. We train 90 models (22K--19.8M parameters) across two architectures (plain ConvNet, MobileNetV2) on CIFAR-100, varying width while holding depth and training fixed. Both follow approximate power laws in error rate: $\alpha = 0.156 \pm 0.002$ (ScaleCNN) and $\alpha = 0.106 \pm 0.001$ (MobileNetV2) across five seeds. Since prior work fit cross-entropy loss rather than error rate, direct exponent comparison is approximate; with that caveat, these are 1.4--2x steeper than $\alpha \approx 0.076$ for large language models. The power law does not hold uniformly: local exponents decay with scale, and MobileNetV2 saturates at 19.8M parameters ($\alpha_{\mathrm{local}} = 0.006$). Error structure also changes. Jaccard overlap between error sets of the smallest and largest ScaleCNN is only 0.35 (25 seed pairs, $\pm 0.004$) -- compression changes which inputs are misclassified, not merely how many. Small models concentrate capacity on easy classes (Gini: 0.26 at 22K vs. 0.09 at 4.7M) while abandoning the hardest (bottom-5 accuracy: 10% vs. 53%). Counter to expectation, the smallest models are best calibrated (ECE = 0.013 vs. peak 0.110 at mid-size). Aggregate accuracy is therefore misleading for edge deployment; validation must happen at the target model size.
Authors: Hieu Le, Oguz Bedir, Mostafa Ibrahim, Jian Tao, Sabit Ekin
Abstract: Reconfigurable Intelligent Surfaces promise to transform wireless environments, yet practical deployment is hindered by the prohibitive overhead of Channel State Information (CSI) estimation and the dimensionality explosion inherent in centralized optimization. This paper proposes a Hierarchical Multi-Agent Reinforcement Learning (HMARL) framework for the control of mechanically reconfigurable reflective surfaces in millimeter-wave (mmWave) systems. We introduce a "CSI-free" paradigm that substitutes pilot-based channel estimation with readily available user localization data. To manage the massive combinatorial action space, the proposed architecture utilizes Multi-Agent Proximal Policy Optimization (MAPPO) under a Centralized Training with Decentralized Execution (CTDE) paradigm. The proposed architecture decomposes the control problem into two abstraction levels: a high-level controller for user-to-reflector allocation and decentralized low-level controllers for low-level focal point optimization. Comprehensive ray-tracing evaluations demonstrate that the framework achieves 2.81-7.94 dB RSSI improvements over centralized baselines, with the performance advantage widening as system complexity increases. Scalability analysis reveals that the system maintains sustained efficiency, exhibiting minimal per-user performance degradation and stable total power utilization even when user density doubles. Furthermore, robustness validation confirms the framework's viability across varying reflector aperture sizes (45-99 tiles) and demonstrates graceful performance degradation under localization errors up to 0.5 m. By eliminating CSI overhead while maintaining high-fidelity beam-focusing, this work establishes HMARL as a practical solution for intelligent mmWave environments.
Authors: Siddhartha Laghuvarapu, Ying Jin, Jimeng Sun
Abstract: The success of deep generative models in scientific discovery requires not only the ability to generate novel candidates but also reliable guarantees that these candidates indeed satisfy desired properties. Recent conformal-prediction methods offer a path to such guarantees, but its application to generative modeling in drug discovery is limited by budget constraints, lack of oracle access, and distribution shift. To this end, we introduce ConfHit, a distribution-free framework that provides validity guarantees under these conditions. ConfHit formalizes two central questions: (i) Certification: whether a generated batch can be guaranteed to contain at least one hit with a user-specified confidence level, and (ii) Design: whether the generation can be refined to a compact set without weakening this guarantee. ConfHit leverages weighted exchangeability between historical and generated samples to eliminate the need for an experimental oracle, constructs multiple-sample density-ratio weighted conformal p-value to quantify statistical confidence in hits, and proposes a nested testing procedure to certify and refine candidate sets of multiple generated samples while maintaining statistical guarantees. Across representative generative molecule design tasks and a broad range of methods, ConfHit consistently delivers valid coverage guarantees at multiple confidence levels while maintaining compact certified sets, establishing a principled and reliable framework for generative modeling.
Authors: Scott Aaronson, Lin Lin Lee, Jiawei Li
Abstract: Explaining out-of-distribution generalization has been a central problem in epistemology since Goodman's "grue" puzzle in 1946. Today it's a central problem in machine learning, including AI alignment. Here we propose a principled account of OOD generalization with three main ingredients. First, the world is always presented to experience not as an amorphous mass, but via distinguished features (for example, visual and auditory channels). Second, Occam's Razor favors hypotheses that are "sparse," meaning that they depend on as few features as possible. Third, sparse hypotheses will generalize from a training to a test distribution, provided the two distributions sufficiently overlap on their restrictions to the features that are either actually relevant or hypothesized to be. The two distributions could diverge arbitrarily on other features. We prove a simple theorem that formalizes the above intuitions, generalizing the classic sample complexity bound of Blumer et al. to an OOD context. We then generalize sparse classifiers to subspace juntas, where the ground truth classifier depends solely on a low-dimensional linear subspace of the features.
Authors: Xuxing Chen, Yun He, Jiayi Xu, Minhui Huang, Xiaoyi Liu, Boyang Liu, Fei Tian, Xiaohan Wei, Rong Jin, Sem Park, Bo Long, Xue Feng
Abstract: In machine learning, the goal of multi-task learning (MTL) is to optimize multiple objectives together. Recent works, for example, Multiple Gradient Descent Algorithm (MGDA) and its variants, show promising results with dynamically adjusted weights for different tasks to mitigate conflicts that may potentially degrade the performance on certain tasks. Despite the empirical success of MGDA-type methods, one major limitation of such methods is their computational inefficiency, as they require access to all task gradients. In this paper we introduce MARIGOLD, a unified algorithmic framework for efficiently solving MTL problems. Our method reveals that multi-task gradient balancing methods have a hierarchical structure, in which the model training and the gradient balancing are coupled during the whole optimization process and can be viewed as a bi-level optimization problem. Moreover, we showcase that the bi-level problem can be solved efficiently by leveraging zeroth-order method. Extensive experiments on both public datasets and industrial-scale datasets demonstrate the efficiency and superiority of our method.
Authors: Rian Atri
Abstract: Legal teams increasingly use machine learning to triage large volumes of contractual evidence, but many models are opaque, non-deterministic, and difficult to align with frameworks such as HIPAA or NERC-CIP. We study a simple, reproducible alternative based on deterministic dual encoders and transparent fuzzy triage bands. We train a RoBERTa-base dual encoder with a 512-dimensional projection and cosine similarity on the ACORD benchmark for graded clause retrieval, then fine-tune it on a CUAD-derived binary compliance dataset. Across five random seeds (40-44) on a single NVIDIA A100 GPU, the model achieves ACORD-style retrieval performance of NDCG@5 0.38-0.42, NDCG@10 0.45-0.50, and 4-star Precision@5 about 0.37 on the test split. On CUAD-derived binary labels, it achieves AUC 0.98-0.99 and F1 0.22-0.30 depending on positive-class weighting, outperforming majority and random baselines in a highly imbalanced setting with a positive rate of about 0.6%. We then map scalar compliance scores into three regions: auto-noncompliant, auto-compliant, and human-review. Thresholds are tuned on validation data to maximize automatic decision coverage subject to an empirical error-rate constraint of at most 2% over auto-decided examples. The result is a seed-stable system summarized by a small number of scalar parameters. We argue that deterministic encoders, calibrated fuzzy bands, and explicit error constraints provide a practical middle ground between hand-crafted rules and opaque large language models, supporting explainable evidence triage, reproducible audit trails, and concrete mappings to legal review concepts.
Authors: Ruixin Guo, Xinyu Li, Hao Zhou, Yang Zhou, Ruoming Jin
Abstract: Linear autoencoders (LAEs) have gained increasing popularity in recommender systems due to their simplicity and strong empirical performance. Most LAE models, including the Emphasized Denoising Linear Autoencoder (EDLAE) introduced by (Steck, 2020), use quadratic loss during training. However, the original EDLAE only provides closed-form solutions for the hyperparameter choice $b = 0$, which limits its capacity. In this work, we generalize EDLAE objective into a Decoupled Expected Quadratic Loss (DEQL). We show that DEQL simplifies the process of deriving EDLAE solutions and reveals solutions in a broader hyperparameter range $b > 0$, which were not derived in Steck's original paper. Additionally, we propose an efficient algorithm based on Miller's matrix inverse theorem to ensure the computational tractability for the $b > 0$ case. Empirical results on benchmark datasets show that the $b > 0$ solutions provided by DEQL outperform the $b = 0$ EDLAE baseline, demonstrating that DEQL expands the solution space and enables the discovery of models with better testing performance.
Authors: Ran Cheng
Abstract: Catastrophic forgetting remains a central challenge in continual learning (CL), yet lacks a unified information-theoretic explanation for why some architectures forget catastrophically while others do not. We introduce \emph{Context Channel Capacity} ($C_\mathrm{ctx}$), the mutual information between a CL architecture's context signal and its generated parameters, and prove that zero forgetting requires $C_\mathrm{ctx} \geq H(T)$, where $H(T)$ is the task identity entropy. We establish an \emph{Impossibility Triangle} -- zero forgetting, online learning, and finite parameters cannot be simultaneously satisfied by sequential state-based learners -- and show that conditional regeneration architectures (HyperNetworks) bypass this triangle by redefining parameters as function values rather than states. We validate this framework across 8 CL methods on Split-MNIST (1,130+ experiments over 86 days, 4 seeds each), showing that $C_\mathrm{ctx}$ perfectly predicts forgetting behavior: methods with $C_\mathrm{ctx} = 0$ (NaiveSGD, EWC, SI, LwF, CFlow) exhibit catastrophic forgetting (6--97\%), while methods with $C_\mathrm{ctx} \approx 1$ (HyperNetwork) achieve zero forgetting (98.8\% ACC). We further propose \emph{Wrong-Context Probing} (P5), a practical diagnostic protocol for measuring $C_\mathrm{ctx}$, and extend the framework to CIFAR-10 via a novel \emph{Gradient Context Encoder} that closes the oracle gap from 23.3pp to 0.7pp. A systematic taxonomy of 15+ closed research directions -- including the Hebbian null result (frozen random features outperform learned features), CFlow's $\theta_0$-memorizer phenomenon, and the $S_N$ symmetry barrier to column specialization -- provides the community with precisely diagnosed negative results. Our central design principle: \emph{architecture over algorithm} -- the context pathway must be structurally unbypassable.
Authors: Shuzhang Zhong, Baotong Lu, Qi Chen, Chuanjie Liu, Fan Yang, Meng Li
Abstract: Large language model-based deep research agents have been increasingly popular for addressing long-horizon information-seeking tasks, but they often incur high end-to-end latency due to extensive reasoning and frequent tool use. Speculation frameworks aim to reduce latency by overlapping action execution with reasoning; however, existing approaches typically rely on uniform speculation strategies and strict action matching, which limits inference speedups and robustness. In this work, we revisit the speculate-verify paradigm for deep research agents through the lens of action heterogeneity. We show that \textit{Search} and \textit{Visit} actions exhibit fundamentally different reasoning and model capacity requirements: entropy-based analysis reveals that Search decisions have higher uncertainty and benefit significantly from explicit reasoning, whereas Visit decisions have lower entropy and depend primarily on model capacity. Motivated by this dual-process characteristic, we propose DualSpec, a heterogeneous speculation framework equipped with a lightweight, confidence-based semantic verifier. Experiments across multiple models and benchmarks demonstrate that DualSpec achieves up to 3.28$\times$ end-to-end speedup while maintaining accuracy comparable to fully reasoning agents.
Authors: Charles Luo
Abstract: Transformer architectures excel at sequential modeling yet remain fundamentally limited by correlational learning - they capture spurious associations induced by latent confounders rather than invariant causal mechanisms. We identify this as an epistemological challenge: standard Transformers conflate static background factors (intrinsic identity, style, context) with dynamic causal flows (state evolution, mechanism), leading to catastrophic out-of-distribution failure. We propose OrthoFormer, a causally grounded architecture that embeds instrumental variable estimation directly into Transformer blocks via neural control functions. Our framework rests on four theoretical pillars: Structural Directionality (time-arrow enforcement), Representation Orthogonality (latent-noise separation), Causal Sparsity (Markov Blanket approximation), and End-to-End Consistency (gradient- detached stage separation). We prove that OrthoFormer achieves bias strictly less than OLS for any valid instrument lag, with residual bias decaying geometrically as O(\r{ho}k ). We characterize the bias-variance-exogeneity trilemma inherent in self-instrumenting and identify the neural forbidden regression - where removing gradient detachment improves prediction loss while destroying causal validity. Experiments confirm all theoretical predictions. OrthoFormer represents a paradigm shift from correlational to causal sequence modeling, with implications for robustness, interpretability, and reliable decision-making under distribution shift.
Authors: Suorong Yang, Fangjian Su, Hai Gan, Ziqi Ye, Jie Li, Baile Xu, Furao Shen, Soujanya Poria
Abstract: Dynamic Data selection aims to accelerate training by prioritizing informative samples during online training. However, existing methods typically rely on task-specific handcrafted metrics or static/snapshot-based criteria to estimate sample importance, limiting scalability across learning paradigms and making it difficult to capture the evolving utility of data throughout training. To address this challenge, we propose Data Agent, an end-to-end dynamic data selection framework that formulates data selection as a training-aware sequential decision-making problem. The agent learns a sample-wise selection policy that co-evolves with model optimization, guided by a composite reward that integrates loss-based difficulty and confidence-based uncertainty signals. The reward signals capture complementary objectives of optimization impact and information gain, together with a tuning-free adaptive weighting mechanism that balances these signals over training. Extensive experiments across a wide range of datasets and architectures demonstrate that Data Agent consistently accelerates training while preserving or improving performance, e.g., reducing costs by over 50\% on ImageNet-1k and MMLU with lossless performance. Moreover, its dataset-agnostic formulation and modular reward make it plug-and-play across tasks and scenarios, e.g., robustness to noisy datasets, highlighting its potential in real-world scenarios.
Authors: Yi Tian, Kaiqing Zhang, Russ Tedrake, Suvrit Sra
Abstract: We study the problem of state representation learning for control from partial and potentially high-dimensional observations. We approach this problem via cost-driven state representation learning, in which we learn a dynamical model in a latent state space by predicting cumulative costs. In particular, we establish finite-sample guarantees on finding a near-optimal representation function and a near-optimal controller using the learned latent model for infinite-horizon time-invariant Linear Quadratic Gaussian (LQG) control. We study two approaches to cost-driven representation learning, which differ in whether the transition function of the latent state is learned explicitly or implicitly. The first approach has also been investigated in Part I of this work, for finite-horizon time-varying LQG control. The second approach closely resembles MuZero, a recent breakthrough in empirical reinforcement learning, in that it learns latent dynamics implicitly by predicting cumulative costs. A key technical contribution of this Part II is to prove persistency of excitation for a new stochastic process that arises from the analysis of quadratic regression in our approach, and may be of independent interest.
Authors: Yael S. Elmatad
Abstract: Gradient boosting still dominates Transformers on tabular benchmarks. Our tokenizer uses a deliberately simplistic discretized vocabulary so we can highlight how even basic tokenization unlocks the power of attention on tabular features, yet it already outperforms tuned gradient boosting when combined with Gaussian smoothing. Our solution discretizes environmental context while smoothing labels with adaptive Gaussians, yielding calibrated PDFs. On 600K entities (5M training examples) we outperform tuned XGBoost by 10.8% (35.94s vs 40.31s median MAE) and achieve KS=0.0045 with the adaptive-sigma checkpoint selected to minimize KS rather than median MAE. Ablations confirm architecture matters: losing sequential ordering costs about 2.0%, dropping the time-delta tokens costs about 1.8%, and a stratified calibration analysis reveals where miscalibration persists.
Authors: Mingxin Zhang, Xiaofeng Dai, Yu Yao, Ziqi Yin
Abstract: In this study, we present a conditional diffusion-transformer framework for generating ensembles of three-dimensional Escherichia coli genome conformations guided by Hi-C contact maps. Instead of producing a single deterministic structure, we formulate genome reconstruction as a conditional generative modeling problem that samples heterogeneous conformations whose ensemble-averaged contacts are consistent with the input Hi-C data. A synthetic dataset is constructed using coarse-grained molecular dynamics simulations to generate chromatin ensembles and corresponding Hi-C maps under circular topology. Our models operate in a latent diffusion setting with a variational autoencoder that preserves per-bin alignment and supports replication-aware representations. Hi-C information is injected through a transformer-based encoder and cross-attention, enforcing a physically interpretable one-way constraint from Hi-C to structure. The model is trained using a flow-matching objective for stable optimization. On held-out ensembles, generated structures reproduce the input Hi-C distance-decay and structural correlation metrics while maintaining substantial conformational diversity, demonstrating the effectiveness of diffusion-based generative modeling for ensemble-level 3D genome reconstruction.
Authors: Clayton Kerce, Alexis Fox
Abstract: While transformers achieve strong performance, their internal decision-making processes remain opaque. We investigate whether architectural constraints can enforce interpretability by design through architectural stream independence: maintaining a token stream (carrying symbolic structure) and contextual semantics in separated streams that remain independently observable throughout processing, with integration delayed until output. We validate this principle through the Late Fusion Architecture (LFA), which demonstrates interpretable symbolic heads through all the final layers, while standard transformers show dissolution by the third of six layers; we quantify this effect by introducing the Token-Position Dependence Score (PDS), with $PDS_{max}$ = 0.276 and 0.058, respectively. Crucially, intervention experiments demonstrate functional modularity: suppressing LFA's recency heads causes minimal semantic damage (Cohen's d = -0.158) versus catastrophic entanglement in baselines (d = -0.672). LFA demonstrates that architectural constraints improve underlying learning mechanisms, averaging 42% stability versus 19% and 11% for baseline comparisons, with extremes from 50% on LFA's best pairs (6 of 12 heads position-invariant) down to 0% complete collapse in over-constrained cases. By preventing premature entanglement, architectural independence steers models toward semantic understanding over positional heuristics, establishing interpretability as an architectural design criterion enforceable through structural constraints rather than post-hoc analysis.
Authors: Eman Khalid, Moimma Ali Khan, Zarmeena Ali, Abdullah Illyas, Muhammad Usman, Saoud Ahmed
Abstract: High-dimensional time series forecasting suffers from severe overfitting when the number of predictors exceeds available observations, making standard local projection methods unstable and unreliable. We propose an enhanced Random Subspace Local Projection (RSLP) framework designed to deliver robust impulse response estimation in the presence of hundreds of correlated predictors. The method introduces weighted subspace aggregation, category-aware subspace sampling, adaptive subspace size selection, and a bootstrap inference procedure tailored to dependent data. These enhancements substantially improve estimator stability at longer forecast horizons while providing more reliable finite-sample inference. Experiments on synthetic data, macroeconomic indicators, and the FRED-MD dataset demonstrate a 33 percent reduction in estimator variability at horizons h >= 3 through adaptive subspace size selection. The bootstrap inference procedure produces conservative confidence intervals that are 14 percent narrower at policy-relevant horizons in very high-dimensional settings (FRED-MD with 126 predictors) while maintaining proper coverage. The framework provides practitioners with a principled approach for incorporating rich information sets into impulse response analysis without the instability of traditional high-dimensional methods.
Authors: Jianlu Shen, Fu Feng, Jiaze Xu, Yucheng Xie, Jiaqi Lv, Xin Geng
Abstract: Transferring pre-trained knowledge from a source model to a target model of a different architectural size is a key challenge for flexible and efficient model scaling. However, current parameter-space methods treat Small-to-Large (S2L) and Large-to-Small (L2S) scaling as separate, incompatible problems, focusing on parameter synthesis and selection, respectively. This fragmented perspective has resulted in specialized tools, hindering a unified, bidirectional framework. In this paper, we propose BoT (Bidirectional knowledge Transfer), the first size-agnostic framework to unify S2L and L2S scaling. Our core insight is to treat model weights as continuous signals, where models of different sizes represent distinct discretizations of the transferable knowledge. This multi-resolution perspective directly casts S2L and L2S scaling as the signal processing operations of upsampling and downsampling, naturally leading to the adoption of the Discrete Wavelet Transform (DWT) and its Inverse (IDWT). BoT leverages the recursive nature of wavelets, using the decomposition level as a dynamic scaling factor to bridge disparate model sizes in a parameter-free and computationally efficient manner. Extensive experiments on DeiT, BERT, and GPT demonstrate significant pre-training FLOPs savings (up to 67.1% for S2L, 52.8% for L2S) and state-of-the-art performance on benchmarks like GLUE and SQuAD.
Authors: Matea Marinova, Shashi Raj Pandey, Junya Shiraishi, Martin Voigt Vejling, Valentin Rakovic, Petar Popovski
Abstract: In this work, we present OCLADS, a novel communication framework with continual learning (CL) for Internet of Things (IoT) anomaly detection (AD) when operating in non-stationary environments. As the statistical properties of the observed data change with time, the on-device inference model becomes obsolete, which necessitates strategic model updating. OCLADS keeps track of data distribution shifts to timely update the on-device IoT AD model. To do so, OCLADS introduces two mechanisms during the interaction between the resource-constrained IoT device and an edge server (ES): i) an intelligent sample selection mechanism at the device for data transmission, and ii) a distribution-shift detection mechanism at the ES for model updating. Experimental results with TinyML demonstrate that our proposed framework achieves high inference accuracy while realizing a significantly smaller number of model updates compared to the baseline schemes.
Authors: Chieh-Hsin Lai, Bac Nguyen, Naoki Murata, Yuhta Takida, Toshimitsu Uesaka, Yuki Mitsufuji, Stefano Ermon, Molei Tao
Abstract: Drifting models train one-step generators by optimizing a mean-shift discrepancy induced by a kernel between the data and model distributions, with Laplace kernels used by default in practice. At each point, this discrepancy compares the kernel-weighted displacement toward nearby data samples with the corresponding displacement toward nearby model samples, yielding a transport direction for generated samples. In this paper, we make its relationship to the score-matching principle behind diffusion models precise by showing that drifting admits a score-based formulation on kernel-smoothed distributions. For Gaussian kernels, the population mean-shift field coincides with the score difference between the Gaussian-smoothed data and model distributions. This identity follows from Tweedie's formula, which links the score of a Gaussian-smoothed density to the corresponding conditional mean, and implies that Gaussian-kernel drifting is exactly a score-matching-style objective on smoothed distributions. It also clarifies the connection to Distribution Matching Distillation (DMD): both methods use score-mismatch transport directions, but drifting realizes the score signal nonparametrically from kernel neighborhoods, whereas DMD uses a pretrained diffusion teacher. Beyond Gaussians, we derive an exact decomposition for general radial kernels, and for the Laplace kernel we prove rigorous error bounds showing that drifting remains an accurate proxy for score matching in low-temperature and high-dimensional regimes.
Authors: Heungjo An
Abstract: Advancing autonomous green technologies in solar photovoltaic (PV) systems is key to improving sustainability and efficiency in renewable energy production. This study presents a reinforcement learning (RL)-based framework to autonomously optimize the cleaning schedules of PV panels in arid regions, where soiling from dust and other airborne particles significantly reduces energy output. By employing advanced RL algorithms, Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC), the framework dynamically adjusts cleaning intervals based on uncertain environmental conditions. The proposed approach was applied to a case study in Abu Dhabi, UAE, demonstrating that PPO outperformed SAC and traditional simulation optimization (Sim-Opt) methods, achieving up to 13% cost savings by dynamically responding to weather uncertainties. The results highlight the superiority of flexible, autonomous scheduling over fixed-interval methods, particularly in adapting to stochastic environmental dynamics. This aligns with the goals of autonomous green energy production by reducing operational costs and improving the efficiency of solar power generation systems. This work underscores the potential of RL-driven autonomous decision-making to optimize maintenance operations in renewable energy systems. In future research, it is important to enhance the generalization ability of the proposed RL model, while also considering additional factors and constraints to apply it to different regions.
Authors: Jianlu Shen, Fu Feng, Yucheng Xie, Jiaqi Lv, Xin Geng
Abstract: Transferring knowledge by fine-tuning large-scale pre-trained networks has become a standard paradigm for downstream tasks, yet the knowledge of a pre-trained model is tightly coupled with monolithic architecture, which restricts flexible reuse across models of varying scales. In response to this challenge, recent approaches typically resort to either parameter selection, which fails to capture the interdependent structure of this knowledge, or parameter prediction using generative models that depend on impractical access to large network collections. In this paper, we empirically demonstrate that a model's foundational, task-agnostic knowledge, its "learngene", is encoded within the low-frequency components of its weights, and can be efficiently inherited by downstream models. Based on this insight, we propose FRONT (FRequency dOmain kNowledge Transfer), a novel framework that uses the Discrete Cosine Transform (DCT) to isolate the low-frequency "learngene". This learngene can be seamlessly adapted to initialize models of arbitrary size via simple truncation or padding, a process that is entirely training-free. For enhanced performance, we propose an optional low-cost refinement process that introduces a spectral regularizer to further improve the learngene's transferability. Extensive experiments demonstrate that FRONT achieves the state-of-the-art performance, accelerates convergence by up to 15 times in vision tasks, and reduces training FLOPs by an average of 40.5% in language tasks.
Authors: Hongjie Jiang, Yifei Tang, Shuqiang Wang
Abstract: Brain activity is intrinsically a neural dynamic process constrained by anatomical space. This leads to significant variations in spatial distribution patterns and correlation patterns of neural activity across variable and heterogeneous scenarios. However, dominant brain functional network construction methods, which relies on pre-defined brain atlases and linear assumptions, fails to precisely capture varying neural activity patterns in heterogeneous scenarios. This limits the consistency and generalizability of the brain functional networks constructed by dominant methods. Here, a neural dynamics-informed pre-trained framework is proposed for personalized brain functional network construction. The proposed framework extracts personalized representations of neural activity patterns in heterogeneous scenarios. Personalized brain functional networks are obtained by utilizing these representations to guide brain parcellation and neural activity correlation estimation. Systematic evaluations were employed on 18 datasets across tasks, such as virtual neural modulation and abnormal neural circuit identification. Experimental results demonstrate that the proposed framework attains superior performance in heterogeneous scenarios. Overall, the proposed framework challenges the dominant brain functional network construction method.
Authors: Tony Zahtila, Ettore Saetta, Murray Cutforth, Davy Brouzet, Diego Rossinelli, Gianluca Iaccarino
Abstract: Accurate and predictive scale-resolving simulations of laser-ignited rocket engines are highly time-consuming because the problem includes turbulent fuel-oxidizer mixing dynamics, laser-induced energy deposition, and high-speed flame growth. This is conflated with the large design space primarily corresponding to the laser operating conditions and target location. To enable rapid exploration and uncertainty quantification, we propose a data-driven surrogate modeling approach that combines convolutional autoencoders (cAEs) with neural ordinary differential equations (neural ODEs). The present target application of an ML-based surrogate model to leading-edge multi-physics turbulence simulation is part of a paradigm shift in the deployment of surrogate models towards increasing real-world complexity. Sequentially, the cAE spatially compresses high-dimensional flow fields into a low-dimensional latent space, wherein the system's temporal dynamics are learned via neural ODEs. Once trained, the model generates fast spatiotemporal predictions from initial conditions and specified operating inputs. By learning a surrogate to replace the entirety of the time-evolving simulation, the cost of predicting an ignition trial is reduced by several orders of magnitude, allowing efficient exploration of the input parameter space. Further, as the current framework yields a spatiotemporal field prediction, appraisal of the model output's physical grounding is more tractable. This approach marks a significant step toward real-time digital twins for laser-ignited rocket combustors and represents surrogate modeling in a complex system context.
Authors: Ramin Akbari, Milad Afshari, Vishnu Naresh Boddeti
Abstract: Concept erasure aims to remove unwanted attributes, such as social or demographic factors, from learned representations, while preserving their task-relevant utility. While the goal of concept erasure is protection against all adversaries, existing methods remain vulnerable to nonlinear ones. This vulnerability arises from their failure to fully capture the complex, nonlinear statistical dependencies between learned representations and unwanted attributes. Moreover, although the existence of a trade-off between utility and erasure is expected, its progression during the erasure process, i.e., the cost of erasure, remains unstudied. In this work, we introduce Obliviator, a post-hoc erasure method designed to fully capture nonlinear statistical dependencies. We formulate erasure from a functional perspective, leading to an optimization problem involving a composition of kernels that lacks a closed-form solution. Instead of solving this problem in a single shot, we adopt an iterative approach that gradually morphs the feature space to achieve a more utility-preserving erasure. Unlike prior methods, Obliviator guards unwanted attribute against nonlinear adversaries. Our gradual approach quantifies the cost of nonlinear guardedness and reveals the dynamics between attribute protection and utility-preservation over the course of erasure. The utility-erasure trade-off curves obtained by Obliviator outperform the baselines and demonstrate its strong generalizability: its erasure becomes more utility-preserving when applied to the better-disentangled representations learned by more capable models.
Authors: Naqcho Ali Mehdi, Amir Ali
Abstract: Automated electrocardiogram (ECG) classification is essential for early detection of cardiovascular diseases. While recent approaches have increasingly relied on deep neural networks with complex architectures, we demonstrate that careful data preprocessing, class balancing, and a simplified convolutional neural network combined with a variational autoencoder (CNN-VAE) architecture can achieve competitive performance with significantly reduced model complexity. Using the publicly available PTB XL dataset, we achieve 87.01% binary accuracy and 0.7454 weighted F1-score across five diagnostic classes (CD, HYP, MI, NORM, STTC) with only 197,093 trainable parameters. Our work emphasises the importance of data-centric machine learning practices over architectural complexity, demonstrating that systematic preprocessing and balanced training strategies are critical for medical signal classification. We identify challenges in minority class detection (particularly hypertrophy) and provide insights for future improvements in handling imbalanced ECG datasets. Index Terms: ECG classification, convolutional neural networks, class balancing, data preprocessing, variational autoencoders, PTB-XL dataset
Authors: Zhenwei Wang, Tiehua Zhang, Ning Xue, Ender Ozcan, Ling Wang, Ruibin Bai
Abstract: Over the past decade, neural network solvers powered by generative artificial intelligence have garnered significant attention in the domain of vehicle routing problems (VRPs), owing to their exceptional computational efficiency and superior reasoning capabilities. In particular, autoregressive solvers integrated with reinforcement learning have emerged as a prominent trend. However, much of the existing work emphasizes large-scale generalization of neural approaches while neglecting the limited robustness of attention-based methods across heterogeneous distributions of problem parameters. Their improvements over heuristic search remain largely restricted to hand-curated, fixed-distribution benchmarks. Furthermore, these architectures tend to degrade significantly when node representations are highly similar or when tasks involve long decision horizons. To address the aforementioned limitations, we propose a novel fusion neural network framework that employs a discrete noise graph diffusion model to learn the underlying constraints of vehicle routing problems and generate a constraint assignment matrix. This matrix is subsequently integrated adaptively into the feature representation learning and decision process of the autoregressive solver, serving as a graph structure mask that facilitates the formation of solutions characterized by both global vision and local feature integration. To the best of our knowledge, this work represents the first comprehensive experimental investigation of neural network model solvers across a 378-combinatorial space spanning four distinct dimensions within the CVRPlib public dataset. Extensive experimental evaluations demonstrate that our proposed fusion model effectively captures and leverages problem constraints, achieving state-of-the-art performance across multiple benchmark datasets.
Authors: Haiteng Wang, Yikang Li, Yunfei Zhu, Jingheng Yan, Lei Ren, Laurence T. Yang
Abstract: Accurate analysis of industrial time-series big data is critical for the Prognostics and Health Management (PHM) of industrial equipment. While recent advancements in Large Language Models (LLMs) have shown promise in time-series analysis, existing methods typically focus on single-modality adaptations, failing to exploit the complementary nature of temporal signals, frequency-domain visual representations, and textual knowledge information. In this paper, we propose TS-MLLM, a unified multi-modal large language model framework designed to jointly model temporal signals, frequency-domain images, and textual domain knowledge. Specifically, we first develop an Industrial time-series Patch Modeling branch to capture long-range temporal dynamics. To integrate cross-modal priors, we introduce a Spectrum-aware Vision-Language Model Adaptation (SVLMA) mechanism that enables the model to internalize frequency-domain patterns and semantic context. Furthermore, a Temporal-centric Multi-modal Attention Fusion (TMAF) mechanism is designed to actively retrieve relevant visual and textual cues using temporal features as queries, ensuring deep cross-modal alignment. Extensive experiments on multiple industrial benchmarks demonstrate that TS-MLLM significantly outperforms state-of-the-art methods, particularly in few-shot and complex scenarios. The results validate our framework's superior robustness, efficiency, and generalization capabilities for industrial time-series prediction.
Authors: Hans Farrell Soegeng, Sarthak Ketanbhai Modi, Thomas Peyrin
Abstract: Interpretable machine learning is essential in high-stakes domains where decision-making requires accountability, transparency, and trust. While rule-based models offer global and exact interpretability, learning rule sets that simultaneously achieve high predictive performance and low, human-understandable complexity remains challenging. To address this, we introduce TT-Sparse, a flexible neural building block that leverages differentiable truth tables as nodes to learn sparse, effective connections. A key contribution of our approach is a new soft TopK operator with straight-through estimation for learning discrete, cardinality-constrained feature selection in an end-to-end differentiable manner. Crucially, the forward pass remains sparse, enabling efficient computation and exact symbolic rule extraction. As a result, each node (and the entire model) can be transformed exactly into compact, globally interpretable DNF/CNF Boolean formulas via Quine-McCluskey minimization. Extensive empirical results across 28 datasets spanning binary, multiclass, and regression tasks show that the learned sparse rules exhibit superior predictive performance with lower complexity compared to existing state-of-the-art methods.
Authors: Jiajun He, Zongyu Guo, Zhaoyang Jia, Xiaoyi Zhang, Jiahao Li, Xiao Li, Bin Li, Jos\'e Miguel Hern\'andez-Lobato, Yan Lu
Abstract: Modern visual generative models acquire rich visual knowledge through large-scale training, yet existing visual representations (such as pixels, latents, or tokens) remain external to the model and cannot directly exploit this knowledge for compact storage or reuse. In this work, we introduce a new visual representation framework that encodes a signal as a function, which is parametrized by low-rank adaptations attached to a frozen visual generative model. Such implicit representations of visual signals, \textit{e.g.}, an 81-frame video, can further be hashed into a single compact vector, achieving strong perceptual video compression at extremely low bitrates. Beyond basic compression, the functional nature of this representation enables inference-time scaling and control, allowing additional refinement on the compression performance. More broadly, as the implicit representations directly act as a function of the generation process, this suggests a unified framework bridging visual compression and generation.
Authors: Chang Su, Zhongkai Hao, Zhizhou Zhang, Zeyu Xia, Youjia Wu, Hang Su, Jun Zhu
Abstract: Large language models (LLMs) with reasoning abilities have demonstrated growing promise for tackling complex scientific problems. Yet such tasks are inherently domain-specific, unbounded and open-ended, demanding exploration across vast and flexible solution spaces. Existing approaches, whether purely learning-based or reliant on carefully designed workflows, often suffer from limited exploration efficiency and poor generalization. To overcome these challenges, we present HELIX -- a Hierarchical Evolutionary reinforcement Learning framework with In-context eXperiences. HELIX introduces two key novelties: (i) a diverse yet high-quality pool of candidate solutions that broadens exploration through in-context learning, and (ii) reinforcement learning for iterative policy refinement that progressively elevates solution quality. This synergy enables the discovery of more advanced solutions. On the circle packing task, HELIX achieves state-of-the-art result with a sum of radii of 2.63598308 using only a 14B model. Across standard machine learning benchmarks, HELIX further surpasses GPT-4o with a carefully engineered pipeline, delivering an average F1 improvement of 5.95 points on the Adult and Bank Marketing datasets.
Authors: Mohammad Nooraiepour, Jakub Wiktor Both, Teeratorn Kadeethum, Saeid Sadeghnejad
Abstract: Partial differential equations (PDEs) govern physical phenomena across the full range of scientific scales, yet their computational solution remains one of the defining challenges of modern science. This critical review examines two mature but epistemologically distinct paradigms for PDE solution, classical numerical methods and machine learning approaches, through a unified evaluative framework organized around six fundamental computational challenges. Classical methods are assessed for their structure-preserving properties, rigorous convergence theory, and scalable solver design; their persistent limitations in high-dimensional and geometrically complex settings are characterized precisely. Machine learning approaches are introduced under a taxonomy organized by the degree to which physical knowledge is incorporated and subjected to the same critical evaluation applied to classical methods. Classical methods are deductive -- errors are bounded by quantities derivable from PDE structure and discretization parameters -- while machine learning methods are inductive -- accuracy depends on statistical proximity to the training distribution. This epistemological distinction is the primary criterion governing responsible method selection. We identify three genuine complementarities between the paradigms and develop principles for hybrid design, including a framework for the structure inheritance problem that addresses when classical guarantees propagate through hybrid couplings, and an error budget decomposition that separates discretization, neural approximation, and coupling contributions. We further assess emerging frontiers, including foundation models, differentiable programming, quantum algorithms, and exascale co-design, evaluating each against the structural constraints that determine whether current barriers are fundamental or contingent on engineering progress.
Authors: Yuanhao Pu, Defu Lian, Enhong Chen
Abstract: The Consistency property between surrogate losses and evaluation metrics has been extensively studied to ensure that minimizing a loss leads to metric optimality. However, the direct relationship between different evaluation metrics remains significantly underexplored. This theoretical gap results in the "Metric Mismatch" frequently observed in industrial applications, where gains in offline validation metrics fail to translate into online performance. To bridge this disconnection, this paper proposes a unified theoretical framework designed to quantify the relationships between metrics. We categorize metrics into different classes to facilitate a comparative analysis across different mathematical forms and interrogates these relationships through Bayes-Optimal Set and Regret Transfer. Through this framework, we provide a new perspective on identifying the structural asymmetry in regret transfer, enabling the design of evaluation systems that are theoretically guaranteed to align offline improvements with online objectives.
Authors: Anirudh Satheesh, Pankaj Kumar Barman, Washim Uddin Mondal, Vaneet Aggarwal
Abstract: We study infinite-horizon Constrained Markov Decision Processes (CMDPs) with general policy parameterizations and multi-layer neural network critics. Existing theoretical analyses for constrained reinforcement learning largely rely on tabular policies or linear critics, which limits their applicability to high-dimensional and continuous control problems. We propose a primal-dual natural actor-critic algorithm that integrates neural critic estimation with natural policy gradient updates and leverages Neural Tangent Kernel (NTK) theory to control function-approximation error under Markovian sampling, without requiring access to mixing-time oracles. We establish global convergence and cumulative constraint violation rates of $\tilde{\mathcal{O}}(T^-1/4)$ up to approximation errors induced by the policy and critic classes. Our results provide the first such guarantees for CMDPs with general policies and multi-layer neural critics, substantially extending the theoretical foundations of actor-critic methods beyond the linear-critic regime.
Authors: Pablo M. Bern\'a
Abstract: Greedy algorithms are central to sparse approximation and stage-wise learning methods such as matching pursuit and boosting. It is known that the Power-Relaxed Greedy Algorithm with step sizes $m^{-\alpha}$ may fail to converge when $\alpha>1$ in general Hilbert spaces. In this work, we revisit this phenomenon from a sparse learning perspective. We study realizable regression problems with controlled feature coherence and derive explicit lower bounds on the residual norm, showing that over-decaying step-size schedules induce structural stagnation even in low-dimensional sparse settings. Numerical experiments confirm the theoretical predictions and illustrate the role of feature coherence. Our results provide insight into step-size design in greedy sparse learning.
Authors: Darius Catrina, Christian Bepler, Samuel Sledzieski, Rohit Singh
Abstract: Unlike the predictable scaling laws in natural language processing and computer vision, protein language models (PLMs) scale poorly: for many tasks, models within the same family plateau or even decrease in performance, with mid-sized models often outperforming the largest in the family. We introduce Reverse Distillation, a principled framework that decomposes large PLM representations into orthogonal subspaces guided by smaller models of the same family. The resulting embeddings have a nested, Matryoshka-style structure: the first k dimensions of a larger model's embedding are exactly the representation from the smaller model. This ensures that larger reverse-distilled models consistently outperform smaller ones. A motivating intuition is that smaller models, constrained by capacity, preferentially encode broadly-shared protein features. Reverse distillation isolates these shared features and orthogonally extracts additional contributions from larger models, preventing interference between the two. On ProteinGym benchmarks, reverse-distilled ESM-2 variants outperform their respective baselines at the same embedding dimensionality, with the reverse-distilled 15 billion parameter model achieving the strongest performance. Our framework is generalizable to any model family where scaling challenges persist. Code and trained models are available at https://github.com/rohitsinghlab/plm_reverse_distillation.
URLs: https://github.com/rohitsinghlab/plm_reverse_distillation.
Authors: Jinshan Liu, Ken Li, Jiazhe Wei, Bin Shi, Bo Dong
Abstract: Federated Graph Learning (FedGL) is vulnerable to malicious attacks, yet developing a truly effective and stealthy attack method remains a significant challenge. Existing attack methods suffer from low attack success rates, high computational costs, and are easily identified and smoothed by defense algorithms. To address these challenges, we propose \textbf{FedShift}, a novel two-stage "Hide and Find" distributed adversarial attack. In the first stage, before FedGL begins, we inject a learnable and hidden "shifter" into part of the training data, which subtly pushes poisoned graph representations toward a target class's decision boundary without crossing it, ensuring attack stealthiness during training. In the second stage, after FedGL is complete, we leverage the global model information and use the hidden shifter as an optimization starting point to efficiently find the adversarial perturbations. During the final attack, we aggregate these perturbations from multiple malicious clients to form the final effective adversarial sample and trigger the attack. Extensive experiments on six large-scale datasets demonstrate that our method achieves the highest attack effectiveness compared to existing advanced attack methods. In particular, our attack can effectively evade 3 mainstream robust federated learning defense algorithms and converges with a time cost reduction of over 90\%, highlighting its exceptional stealthiness, robustness, and efficiency.
Authors: Xingrui Gu, Haixi Zhang
Abstract: Financial time-series forecasting is a high-stakes problem where regime shifts and shocks make point-accurate yet overconfident models dangerous. We propose Uncertainty-Gated Generative Modeling (UGGM), which treats uncertainty as an internal control signal that gates (i) representation via gated reparameterization, (ii) propagation via similarity and confidence routing, and (iii) generation via uncertainty-controlled predictive distributions, together with uncertainty-driven regularization and calibration to curb miscalibration. Instantiated on Weak Innovation AutoEncoder (WIAE-GPF), our UG-WIAE-GPF significantly improves risk-sensitive forecasting, delivering a 63.5\% MSE reduction on NYISO (0.3508 $\rightarrow$ 0.1281), with improved robustness under shock intervals (mSE: 0.2739 $\rightarrow$ 0.1748).
Authors: Christopher Brix, Julia Walczak, Nils Lommen, Thomas Noll
Abstract: Solving quantifier-free non-linear real arithmetic (NRA) problems is a computationally hard task. To tackle this problem, prior work proposed a promising approach based on gradient descent. In this work, we extend their ideas and combine LLMs and GPU acceleration to obtain an efficient technique. We have implemented our findings in the novel SMT solver GANRA (GPU Accelerated solving of Nonlinear Real Arithmetic problems). We evaluate GANRA on two different NRA benchmarks and demonstrate significant improvements over the previous state of the art. In particular, on the Sturm-MBO benchmark, we can prove satisfiability for more than five times as many instances in less than 1/20th of the previous state-of-the-art runtime.
Authors: Zongqian Li, Shaohan Huang, Zewen Chi, Yixuan Su, Lexin Zhou, Li Dong, Nigel Collier, Furu Wei
Abstract: Modern code generation models exhibit longer outputs, accelerated capability growth, and changed training dynamics, rendering traditional training methodologies, algorithms, and datasets ineffective for improving their performance. To address these training bottlenecks, we propose MicroCoder-GRPO, an improved Group Relative Policy Optimization approach with three innovations: conditional truncation masking to improve long output potential while maintaining training stability, diversity-determined temperature selection to maintain and encourage output diversity, and removal of KL loss with high clipping ratios to facilitate solution diversity. MicroCoder-GRPO achieves up to 17.6% relative improvement over strong baselines on LiveCodeBench v6, with more pronounced gains under extended context evaluation. Additionally, we release MicroCoder-Dataset, a more challenging training corpus that achieves 3x larger performance gains than mainstream datasets on LiveCodeBench v6 within 300 training steps, and MicroCoder-Evaluator, a robust framework with approximately 25% improved evaluation accuracy and around 40% faster execution. Through comprehensive analysis across more than thirty controlled experiments, we reveal 34 training insights across seven main aspects, demonstrating that properly trained models can achieve competitive performance with larger counterparts.
Authors: Jinzhou Tan, Gabriel Adineera, Jinoh Kim
Abstract: We present ProgAgent, a continual reinforcement learning (CRL) agent that unifies progress-aware reward learning with a high-throughput, JAX-native system architecture. Lifelong robotic learning grapples with catastrophic forgetting and the high cost of reward specification. ProgAgent tackles these by deriving dense, shaped rewards from unlabeled expert videos through a perceptual model that estimates task progress across initial, current, and goal observations. We theoretically interpret this as a learned state-potential function, delivering robust guidance in line with expert behaviors. To maintain stability amid online exploration - where novel, out-of-distribution states arise - we incorporate an adversarial push-back refinement that regularizes the reward model, curbing overconfident predictions on non-expert trajectories and countering distribution shift. By embedding this reward mechanism into a JIT-compiled loop, ProgAgent supports massively parallel rollouts and fully differentiable updates, rendering a sophisticated unified objective feasible: it merges PPO with coreset replay and synaptic intelligence for an enhanced stability-plasticity balance. Evaluations on ContinualBench and Meta-World benchmarks highlight ProgAgent's advantages: it markedly reduces forgetting, boosts learning speed, and outperforms key baselines in visual reward learning (e.g., Rank2Reward, TCN) and continual learning (e.g., Coreset, SI) - surpassing even an idealized perfect memory agent. Real-robot trials further validate its ability to acquire complex manipulation skills from noisy, few-shot human demonstrations.
Authors: Caihao Sun, Mingqi Yuan, Shiyuan Wang, Jiayu Chen
Abstract: Loss of plasticity refers to the progressive inability of a model to adapt to new tasks and poses a fundamental challenge for continual learning. While this phenomenon has been extensively studied in homogeneous neural architectures, such as multilayer perceptrons, its mechanisms in structurally heterogeneous, attention-based models such as Vision Transformers (ViTs) remain underexplored. In this work, we present a systematic investigation of loss of plasticity in ViTs, including a fine-grained diagnosis using local metrics that capture parameter diversity and utilization. Our analysis reveals that stacked attention modules exhibit increasing instability that exacerbates plasticity loss, while feed-forward network modules suffer even more pronounced degradation. Furthermore, we evaluate several approaches for mitigating plasticity loss. The results indicate that methods based on parameter re-initialization fail to recover plasticity in ViTs, whereas approaches that explicitly regulate the update process are more effective. Motivated by this insight, we propose ARROW, a geometry-aware optimizer that preserves plasticity by adaptively reshaping gradient directions using an online curvature estimate for the attention module. Extensive experiments show that ARROW effectively improves plasticity and maintains better performance on newly encountered tasks.
Authors: Zaid Abdullah, Merouane Debbah, Symeon Chatzinotas, Bjorn Ottersten
Abstract: Deep-learning (DL)-based precoding in multi-user multiple-input single-output (MU-MISO) systems involves training DL models to map features derived from channel coefficients to labels derived from precoding weights. Traditionally, complex-valued channel and precoder coefficients are parameterized using either their real and imaginary components or their amplitude and phase. However, precoding performance depends on magnitudes of inner products between channel and precoding vectors, which are invariant to global phase rotations. Conventional representations fail to exploit this symmetry, leading to inefficient learning and degraded generalization. To address this, we propose a DL framework based on complex projective space (CPS) parameterizations of both the wireless channel and the weighted minimum mean squared error (WMMSE) precoder vectors. By removing the global phase redundancies inherent in conventional representations, the proposed framework enables the DL model to learn geometry-aligned and physically distinct channel-precoder mappings. Two CPS parameterizations based on real-valued embeddings and complex hyperspherical coordinates are investigated and benchmarked against two baseline methods. Simulation results demonstrate substantial improvements in sum-rate performance and generalization, with negligible increase in model complexity.
Authors: Th\'eo Vincent, Kevin Gerhardt, Yogesh Tripathi, Habib Maraqten, Adam White, Martha White, Jan Peters, Carlo D'Eramo
Abstract: Temporal-difference (TD) learning is highly effective at controlling and evaluating an agent's long-term outcomes. Most approaches in this paradigm implement a semi-gradient update to boost the learning speed, which consists of ignoring the gradient of the bootstrapped estimate. While popular, this type of update is prone to divergence, as Baird's counterexample illustrates. Gradient TD methods were introduced to overcome this issue, but have not been widely used, potentially due to issues with learning speed compared to semi-gradient methods. Recently, iterated TD learning was developed to increase the learning speed of TD methods. For that, it learns a sequence of action-value functions in parallel, where each function is optimized to represent the application of the Bellman operator over the previous function in the sequence. While promising, this algorithm can be unstable due to its semi-gradient nature, as each function tracks a moving target. In this work, we modify iterated TD learning by computing the gradients over those moving targets, aiming to build a powerful gradient TD method that competes with semi-gradient methods. Our evaluation reveals that this algorithm, called Gradient Iterated Temporal-Difference learning, has a competitive learning speed against semi-gradient methods across various benchmarks, including Atari games, a result that no prior work on gradient TD methods has demonstrated.
Authors: Abduragim Shtanchaev, Albina Ilina, Yazid Janati, Arip Asadulaev, Martin Tak\'ac, Eric Moulines
Abstract: Pretrained diffusion models serve as effective priors for Bayesian inverse problems. These priors enable zero-shot generation by sampling from the conditional distribution, which avoids the need for task-specific retraining. However, a major limitation of existing methods is their reliance on surrogate likelihoods that require vector-Jacobian products at each denoising step, creating a substantial computational burden. To address this, we introduce a lightweight likelihood surrogate that eliminates the need to calculate gradients through the denoiser network. This enables us to handle diverse inverse problems without backpropagation overhead. Experiments confirm that using our method, the inference cost drops dramatically. At the same time, our approach delivers the highest results in multiple tasks. Broadly speaking, we propose the fastest and Pareto optimal method for Bayesian inverse problems.
Authors: Jean Erik Delanois, Aditya Ahuja, Giri P. Krishnan, Maxim Bazhenov
Abstract: Artificial neural networks are often overconfident, undermining trust because their predicted probabilities do not match actual accuracy. Inspired by biological sleep and the role of spontaneous replay in memory and learning, we introduce Sleep Replay Consolidation (SRC), a novel calibration approach. SRC is a post-training, sleep-like phase that selectively replays internal representations to update network weights and improve calibration without supervised retraining. Across multiple experiments, SRC is competitive with and complementary to standard approaches such as temperature scaling. Combining SRC with temperature scaling achieves the best Brier score and entropy trade-offs for AlexNet and VGG19. These results show that SRC provides a fundamentally novel approach to improving neural network calibration. SRC-based calibration offers a practical path toward more trustworthy confidence estimates and narrows the gap between human-like uncertainty handling and modern deep networks.
Authors: Noah Golowich, Fan Chen, Dhruv Rohatgi, Raghav Singhal, Carles Domingo-Enrich, Dylan J. Foster, Akshay Krishnamurthy
Abstract: Inference-time methods that aggregate and prune multiple samples have emerged as a powerful paradigm for steering large language models, yet we lack any principled understanding of their accuracy-cost tradeoffs. In this paper, we introduce a route to rigorously study such approaches using the lens of *particle filtering* algorithms such as Sequential Monte Carlo (SMC). Given a base language model and a *process reward model* estimating expected terminal rewards, we ask: *how accurately can we sample from a target distribution given some number of process reward evaluations?* Theoretically, we identify (1) simple criteria enabling non-asymptotic guarantees for SMC; (2) algorithmic improvements to SMC; and (3) a fundamental limit faced by all particle filtering methods. Empirically, we demonstrate that our theoretical criteria effectively govern the *sampling error* of SMC, though not necessarily its final *accuracy*, suggesting that theoretical perspectives beyond sampling may be necessary.
Authors: Colin Aitken, Rajat Masiwal, Adam Marchakitus, Katherine Kowal, Mayank Gupta, Tyler Yang, Amir Jina, Pedram Hassanzadeh, William R. Boos, Michael Kremer
Abstract: Hundreds of millions of farmers make high-stakes decisions under uncertainty about future weather. Forecasts can inform these decisions, but available choices and their risks and benefits vary between farmers. We introduce a decision-theory framework for designing useful forecasts in settings where the forecaster cannot prescribe optimal actions because farmers' circumstances are heterogeneous. We apply this framework to the case of seasonal onset of monsoon rains, a key date for planting decisions and agricultural investments in many tropical countries. We develop a system for tailoring forecasts to the requirements of this framework by blending systematically benchmarked artificial intelligence (AI) weather prediction models with a new "evolving farmer expectations" statistical model. This statistical model applies Bayesian inference to historical observations to predict time-varying probabilities of first-occurrence events throughout a season. The blended system yields more skillful Indian monsoon forecasts at longer lead times than its components or any multi-model average. In 2025, this system was deployed operationally in a government-led program that delivered subseasonal monsoon onset forecasts to 38 million Indian farmers, skillfully predicting that year's early-summer anomalous dry period. This decision-theory framework and blending system offer a pathway for developing climate adaptation tools for large vulnerable populations around the world.
Authors: Lizhi Ma, Yi-Xiang Hu, Yihui Ren, Feng Wu, Xiang-Yang Li
Abstract: Databricks job orchestration systems (e.g., LeJOT) reduce cloud costs by selecting low-priced compute configurations while meeting latency and dependency constraints. Accurate execution-time prediction under heterogeneous instance types and non-stationary runtime conditions is therefore critical. Existing pipelines rely on static, manually engineered features that under-capture runtime effects (e.g., partition pruning, data skew, and shuffle amplification), and predictive signals are scattered across logs, metadata, and job scripts-lengthening update cycles and increasing engineering overhead. We present LeJOT-AutoML, an agent-driven AutoML framework that embeds large language model agents throughout the ML lifecycle. LeJOT-AutoML combines retrieval-augmented generation over a domain knowledge base with a Model Context Protocol toolchain (log parsers, metadata queries, and a read-only SQL sandbox) to analyze job artifacts, synthesize and validate feature-extraction code via safety gates, and train/select predictors. This design materializes runtime-derived features that are difficult to obtain through static analysis alone. On enterprise Databricks workloads, LeJOT-AutoML generates over 200 features and reduces the feature-engineering and evaluation loop from weeks to 20-30 minutes, while maintaining competitive prediction accuracy. Integrated into the LeJOT pipeline, it enables automated continuous model updates and achieves 19.01% cost savings in our deployment setting through improved orchestration.
Authors: Sajib Debnath, Md. Uzzal Mia
Abstract: The reliable operation of modern power grids requires probabilistic load forecasts with well-calibrated uncertainty estimates. However, existing deep learning models produce overconfident point predictions that fail catastrophically under extreme weather distributional shifts. This study proposes a Bayesian Transformer (BT) framework that integrates three complementary uncertainty mechanisms into a PatchTST backbone: Monte Carlo Dropout for epistemic parameter uncertainty, variational feed-forward layers with log-uniform weight priors, and stochastic attention with learnable Gaussian noise perturbations on pre-softmax logits, representing, to the best of our knowledge, the first application of Bayesian attention to probabilistic load forecasting. A seven-level multi-quantile pinball-loss prediction head and post-training isotonic regression calibration produce sharp, near-nominally covered prediction intervals. Evaluation of five grid datasets (PJM, ERCOT, ENTSO-E Germany, France, and Great Britain) augmented with NOAA covariates across 24, 48, and 168-hour horizons demonstrates state-of-the-art performance. On the primary benchmark (PJM, H=24h), BT achieves a CRPS of 0.0289, improving 7.4% over Deep Ensembles and 29.9% over the deterministic LSTM, with 90.4% PICP at the 90% nominal level and the narrowest prediction intervals (4,960 MW) among all probabilistic baselines. During heat-wave and cold snap events, BT maintained 89.6% and 90.1% PICP respectively, versus 64.7% and 67.2% for the deterministic LSTM, confirming that Bayesian epistemic uncertainty naturally widens intervals for out-of-distribution inputs. Calibration remained stable across all horizons (89.8-90.4% PICP), while ablation confirmed that each component contributed a distinct value. The calibrated outputs directly support risk-based reserve sizing, stochastic unit commitment, and demand response activation.
Authors: Zihao Zheng, Hangyu Cao, Sicheng Tian, Jiayu Chen, Maoliang Li, Xinhao Sun, Hailong Zou, Zhaobo Zhang, Xuanzhe Liu, Donggang Cao, Hong Mei, Xiang Chen
Abstract: Vision-Language-Action (VLA) models are dominant in embodied intelligence but are constrained by inference overheads. While model quantization alleviates these bottlenecks for edge deployment, static quantization approaches remain suboptimal for VLAs due to two critical challenges: (1) Temporal-dynamic sensitivity, where fixed precision wastes resources by ignoring stage-varying error tolerances; and (2) Real-time allocation, where identifying real-time sensitivity to guide bit allocation remains unsolved. To address these challenges, we propose DyQ-VLA, a dynamic quantization framework for VLAs. Specifically, a sensitivity-aware switching strategy leverages real-time kinematic proxies to trigger the bit-width switch, while a kinematic-guided module dynamically allocates the optimal bit-width. Experiments show that DyQ-VLA requires only 30.9% of the original memory footprint while maintaining 99.5% of its original performance, achieving 1.49x simulation and up to 1.43x real-world speedups.
Authors: Mohammed Omer Shakeel Ahmed
Abstract: Large healthcare institutions typically operate multiple business intelligence (BI) teams segmented by domain, including clinical performance, fundraising, operations, and compliance. Due to HIPAA, FERPA, and IRB restrictions, these teams face challenges in sharing patient-level data needed for analytics. To mitigate this, A metric aggregation table is proposed, which is a precomputed, privacy-compliant summary. These abstractions enable decision-making without direct access to sensitive data. However, even aggregated metrics can inadvertently lead to privacy risks if constructed without rigorous safeguards. A modular AI framework is proposed that evaluates SQL-based metric definitions for potential overexposure using both semantic and syntactic features. Specifically, the system parses SQL queries into abstract syntax trees (ASTs), extracts sensitive patterns (e.g., fine-grained GROUP BY on ZIP code or gender), and encodes the logic using pretrained CodeBERT embeddings. These are fused with structural features and passed to an XGBoost classifier trained to assign risk scores. Queries that surpass the risk threshold (e.g., > 0.85) are flagged and returned with human-readable explanations. This enables proactive governance, preventing statistical disclosure before deployment. This implementation demonstrates strong potential for cross-departmental metric sharing in healthcare while maintaining compliance and auditability. The system also promotes role-based access control (RBAC), supports zero-trust data architectures, and aligns with national data modernization goals by ensuring that metric pipelines are explainable, privacy-preserving, and AI-auditable by design. Unlike prior works that rely on runtime data access to flag privacy violations, the proposed framework performs static, explainable detection at the query-level, enabling pre-execution protection and audit readiness
Authors: Yusong Wang, Chuang Yang, Jiawei Wang, Xiaohang Xu, Jiayi Xu, Dongyuan Li, Chuan Xiao, Renhe Jiang
Abstract: Human mobility generation aims to synthesize plausible trajectory data, which is widely used in urban system research. While Large Language Model-based methods excel at generating routine trajectories, they struggle to capture deviated mobility during large-scale societal events. This limitation stems from two critical gaps: (1) the absence of event-annotated mobility datasets for design and evaluation, and (2) the inability of current frameworks to reconcile competitions between users' habitual patterns and event-imposed constraints when making trajectory decisions. This work addresses these gaps with a twofold contribution. First, we construct the first event-annotated mobility dataset covering three major events: Typhoon Hagibis, COVID-19, and the Tokyo 2021 Olympics. Second, we propose ELLMob, a self-aligned LLM framework that first extracts competing rationales between habitual patterns and event constraints, based on Fuzzy-Trace Theory, and then iteratively aligns them to generate trajectories that are both habitually grounded and event-responsive. Extensive experiments show that ELLMob wins state-of-the-art baselines across all events, demonstrating its effectiveness. Our codes and datasets are available at https://github.com/deepkashiwa20/ELLMob.
Authors: Boris Kriuk, Fedor Kriuk
Abstract: Reliable real-time estimation of atmospheric turbulence intensity remains an open challenge for aircraft operating across diverse altitude bands, particularly over oceanic, polar, and data-sparse regions that lack operational nowcasting infrastructure. Classical spectral models encode climatological averages rather than the instantaneous atmospheric state, and generic ML regressors offer adaptivity but provide no guarantee that predictions respect fundamental scaling laws. This paper introduces the Physically-Structured Turbulence Network (PSTNet), a lightweight architecture that embeds physics directly into its structure. PSTNet couples four components: (i) a zero-parameter backbone derived from Monin-Obukhov theory, (ii) a regime-gated mixture of specialist sub-networks supervised by Richardson-number-derived soft targets, (iii) Feature-wise Linear Modulation layers conditioning hidden representations on local air-density ratio, and (iv) a Kolmogorov output layer enforcing inertial-subrange scaling as an architectural constraint. The entire model contains only 552 learnable parameters, requiring fewer than 2.5 kB of storage and executing in under 12s on a Cortex-M7 microcontroller. We validate PSTNet on 340 paired six-degree-of-freedom guidance simulations spanning three vehicle classes (Mach 2.8, 4.5, and 8.0) and six operational categories with real-time satellite weather ingestion. PSTNet achieves a mean miss-distance improvement of +2.8% with a 78% win rate and a statistically significant effect size. Our results demonstrate that encoding domain physics as architectural priors yields a more efficient and interpretable path to turbulence estimation accuracy than scaling model capacity, establishing PSTNet as a viable drop-in replacement for legacy look-up tables in resource-constrained, safety-critical on-board guidance systems.
Authors: Qianyu Yang, Yang Liu, Jiaqi Li, Jun Bai, Hao Chen, Kaiyuan Chen, Tiliang Duan, Jiayun Dong, Xiaobo Hu, Zixia Jia, Yang Liu, Tao Peng, Yixin Ren, Ran Tian, Zaiyuan Wang, Yanglihong Xiao, Gang Yao, Lingyue Yin, Ge Zhang, Chun Zhang, Jianpeng Jiao, Zilong Zheng, Yuan Gong
Abstract: As language models (LMs) evolve from chat assistants to long-horizon agents capable of multi-step reasoning and tool use, existing benchmarks remain largely confined to structured or exam-style tasks that fall short of real-world professional demands. To this end, we introduce \$OneMillion-Bench \$OneMillion-Bench, a benchmark of 400 expert-curated tasks spanning Law, Finance, Industry, Healthcare, and Natural Science, built to evaluate agents across economically consequential scenarios. Unlike prior work, the benchmark requires retrieving authoritative sources, resolving conflicting evidence, applying domain-specific rules, and making constraint decisions, where correctness depends as much on the reasoning process as the final answer. We adopt a rubric-based evaluation protocol scoring factual accuracy, logical coherence, practical feasibility, and professional compliance, focused on expert-level problems to ensure meaningful differentiation across agents. Together, \$OneMillion-Bench provides a unified testbed for assessing agentic reliability, professional depth, and practical readiness in domain-intensive scenarios.
Authors: Bhavesh Kumar, Dylan Feng, Leonard Tang
Abstract: Multimodal judges struggle to ground decisions in visual evidence. We present MJ1, a multimodal judge trained with reinforcement learning that enforces visual grounding through a structured grounded verification chain (observations $\rightarrow$ claims $\rightarrow$ verification $\rightarrow$ evaluation $\rightarrow$ scoring) and a counterfactual consistency reward that penalizes position bias. Even without training, our mechanism improves base-model accuracy on MMRB2 by +3.8 points on Image Editing and +1.7 on Multimodal Reasoning. After training, MJ1, with only 3B active parameters, achieves 77.0% accuracy on MMRB2 and surpasses orders-of-magnitude larger models like Gemini-3-Pro. These results show that grounded verification and consistency-based training substantially improve multimodal judgment without increasing model scale.
Authors: Theo X. Olausson, Jo\~ao Monteiro, Michal Klein, Marco Cuturi
Abstract: Maximum inner product search (MIPS) is a crucial subroutine in machine learning, requiring the identification of key vectors that align best with a given query. We propose amortized MIPS: a learning-based approach that trains neural networks to directly predict MIPS solutions, amortizing the computational cost of matching queries (drawn from a fixed distribution) to a fixed set of keys. Our key insight is that the MIPS value function, the maximal inner product between a query and keys, is also known as the support function of the set of keys. Support functions are convex, 1-homogeneous and their gradient w.r.t. the query is exactly the optimal key in the database. We approximate the support function using two complementary approaches: (1) we train an input-convex neural network (SupportNet) to model the support function directly; the optimal key can be recovered via (autodiff) gradient computation, and (2) we regress directly the optimal key from the query using a vector valued network (KeyNet), bypassing gradient computation entirely at inference time. To learn a SupportNet, we combine score regression with gradient matching losses, and propose homogenization wrappers that enforce the positive 1-homogeneity of a neural network, theoretically linking function values to gradients. To train a KeyNet, we introduce a score consistency loss derived from the Euler theorem for homogeneous functions. Our experiments show that learned SupportNet or KeyNet achieve high match rates and open up new directions to compress databases with a specific query distribution in mind.
Authors: Peishen Yan, Yang Hua, Hao Wang, Jiaru Zhang, Xiaoyu Wu, Tao Song, Haibing Guan
Abstract: Federated fine-tuning of large language models (LLMs) with low-rank adaptation (LoRA) offers a communication-efficient and privacy-preserving solution for task-specific adaptation. Naive aggregation of LoRA modules introduces noise due to mathematical incorrectness when averaging the downsampling and upsampling matrices independently. However, existing noise-free aggregation strategies inevitably compromise the structural expressiveness of LoRA, limiting its ability to retain client-specific adaptations by either improperly reconstructing the low-rank structure or excluding partially trainable components. We identify this problem as loss of training momentum, where LoRA updates fail to accumulate effectively across rounds, resulting in slower convergence and suboptimal performance. To address this, we propose FedMomentum, a novel framework that enables structured and momentum-preserving LoRA aggregation via singular value decomposition (SVD). Specifically, after aggregating low-rank updates in a mathematically correct manner, FedMomentum applies SVD to extract the dominant components that capture the main update directions. These components are used to reconstruct the LoRA modules with the same rank, while residual components can be retained and later merged into the backbone to preserve semantic information and ensure robustness. Extensive experiments across multiple tasks demonstrate that FedMomentum consistently outperforms prior state-of-the-art methods in convergence speed and final accuracy.
Authors: Jingwei Li, Xinran Gu, Jingzhao Zhang
Abstract: A data mixture refers to how different data sources are combined to train large language models, and selecting an effective mixture is crucial for optimal downstream performance. Existing methods either conduct costly searches directly on the target model or rely on mixture scaling laws that fail to extrapolate well to large model sizes. We address these limitations by introducing a compute-efficient pipeline for data mixture scaling. First, we propose CAMEL, a capacity-aware mixture law that models validation loss with the nonlinear interplay between model size and mixture. We also introduce a loss-to-benchmark prediction law that estimates benchmark accuracy from validation loss, enabling end-to-end performance prediction for the target model. Next, we study how to allocate a fixed compute budget across model scales to fit the law and reduce prediction error. Finally, we apply our method to Mixture-of-Experts models with up to 7B-A150M parameters to fit the law, and verify the optimal mixture derived from the law by extrapolating to a 55B-A1.2B target model. Compared to prior methods, we reduces mixture optimization costs by 50\% and improves downstream benchmark performance by up to 3\%.
Authors: Zhengyu Li, Xiangfei Qiu, Yuhan Zhu, Xingjian Wu, Jilin Hu, Chenjuan Guo, Bin Yang
Abstract: Exogenous variables offer valuable supplementary information for predicting future endogenous variables. Forecasting with exogenous variables needs to consider both past-to-future dependencies (i.e., temporal correlations) and the influence of exogenous variables on endogenous variables (i.e., channel correlations). This is pivotal when future exogenous variables are available, because they may directly affect the future endogenous variables. Many methods have been proposed for time series forecasting with exogenous variables, focusing on modeling temporal and channel correlations. However, most of them use a two-step strategy, modeling temporal and channel correlations separately, which limits their ability to capture joint correlations across time and channels. Furthermore, in real-world scenarios, time series are frequently affected by various forms of noises, underscoring the critical importance of robustness in such correlations modeling. To address these limitations, we propose GCGNet, a Graph-Consistent Generative Network for time series forecasting with exogenous variables. Specifically, GCGNet first employs a Variational Generator to produce coarse predictions. A Graph Structure Aligner then further guides it by evaluating the consistency between the generated and true correlations, where the correlations are represented as graphs, and are robust to noises. Finally, a Graph Refiner is proposed to refine the predictions to prevent degeneration and improve accuracy. Extensive experiments on 12 real-world datasets demonstrate that GCGNet outperforms state-of-the-art baselines.
Authors: Jiayu Huang, Xiaohu Wu, Tiantian He, Qicheng Lao
Abstract: Large Language Models (LLMs) are pivotal in natural language processing. The impracticality of full fine-tuning has prompted Parameter-Efficient Fine-Tuning (PEFT) methods like Low-Rank Adaptation (LoRA), optimizing low-rank matrices A and B. In distributed scenarios where privacy constraints necessitate Federated Learning (FL), however, the integration of LoRA is often unstable. Specifically, we identify that aggregating updates from multiple clients introduces statistical variance that scales with the client count, causing gradient collapse when using high-rank adapters. Existing scaling factor candidates, such as the one used by Rank-Stabilized LoRA, ignore the interaction caused by the aggregation process. To bridge this gap, this paper introduces Stabilized Federated LoRA (SFed-LoRA), a framework that theoretically characterizes the interaction between adapter rank and federated aggregation. We derive an optimal scaling factor designed to effectively mitigate the aggregation error accumulating across N clients. By correcting the scaling mismatch inherent in previous approaches, SFed-LoRA restores the efficacy of high-rank adaptation without altering the original model architecture or increasing inference latency. Extensive experiments in diverse tasks, model architectures, and heterogeneous data distributions are conducted to validate our results. We demonstrate that SFed-LoRA prevents high-rank collapse, and achieves significantly improved stability and faster convergence compared with state-of-the-art baselines for high-rank adaptation.
Authors: Kevin Dradjat, Massinissa Hamidi, Blaise Hanczar
Abstract: Accurate phenotype prediction from RNA sequencing (RNA-seq) data is essential for diagnosis, biomarker discovery, and personalized medicine. Deep learning models have demonstrated strong potential to outperform classical machine learning approaches, but their performance relies on large, well-annotated datasets. In transcriptomics, such datasets are frequently limited, leading to over-fitting and poor generalization. Knowledge transfer from larger, more general datasets can alleviate this issue. However, transferring information across RNA-seq datasets remains challenging due to heterogeneous preprocessing pipelines and differences in target phenotypes. In this study, we propose a deep learning-based domain adaptation framework that enables effective knowledge transfer from a large general dataset to a smaller one for cancer type classification. The method learns a domain-invariant latent space by jointly optimizing classification and domain alignment objectives. To ensure stable training and robustness in data-scarce scenarios, the framework is trained with an adversarial approach with appropriate regularization. Both supervised and unsupervised approach variants are explored, leveraging labeled or unlabeled target samples. The framework is evaluated on three large-scale transcriptomic datasets (TCGA, ARCHS4, GTEx) to assess its ability to transfer knowledge across cohorts. Experimental results demonstrate consistent improvements in cancer and tissue type classification accuracy compared to non-adaptive baselines, particularly in low-data scenarios. Overall, this work highlights domain adaptation as a powerful strategy for data-efficient knowledge transfer in transcriptomics, enabling robust phenotype prediction under constrained data conditions.
Authors: Weiyu Huang, Pengle Zhang, Xiaolu Zhang, Jun Zhou, Jun Zhu, Jianfei Chen
Abstract: Structured pruning reduces LLM inference cost by removing low-importance architectural components. This can be viewed as learning a multiplicative gate for each component under an l0 sparsity constraint. Due to the discreteness of the l0 norm, prior work typically adopts stochastic hard-concrete relaxations to enable differentiable optimization; however, this stochasticity can introduce a train--test mismatch when sampled masks are discretized for deployment and restricts masks to a bounded, near-binary range. To address this, we propose Deterministic Differentiable Pruning (DDP), a mask-only optimization method that eliminates stochasticity by directly optimizing a deterministic soft surrogate of the discrete l0 objective. Compared with prior approaches, DDP offers greater expressiveness, reduced train--test mismatch, and faster convergence. We apply our method to several dense and MoE models, including Qwen3-32B and Qwen3-30B-A3B, achieving a performance loss as small as 1% on downstream tasks while outperforming previous methods at 20% sparsity. We further demonstrate end-to-end inference speedups in realistic deployment settings with vLLM.
Authors: Irene Iele, Floriano Caprio, Paolo Soda, Matteo Tortora
Abstract: Forecasting physiological signals can support proactive monitoring and timely clinical intervention by anticipating critical changes in patient status. In this work, we address multivariate multi-horizon forecasting of physiological time series by jointly predicting heart rate, oxygen saturation, pulse rate, and respiratory rate at forecasting horizons of 15, 30, and 60 seconds. We propose a hybrid quantum-classical architecture that integrates a Variational Quantum Circuit (VQC) within a recurrent neural backbone. A GRU encoder summarizes the historical observation window into a latent representation, which is then projected into quantum angles used to parameterize the VQC. The quantum layer acts as a learnable non-linear feature mixer, modeling cross-variable interactions before the final prediction stage. We evaluate the proposed approach on the BIDMC PPG and Respiration dataset under a Leave-One-Patient-Out protocol. The results show competitive accuracy compared with classical and deep learning baselines, together with greater robustness to noise and missing inputs. These findings suggest that hybrid quantum layers can provide useful inductive biases for physiological time series forecasting in small-cohort clinical settings.
Authors: Paulius Rauba, Claudio Fanconi, Mihaela van der Schaar
Abstract: Tiny Recursive Models (TRMs) have recently demonstrated remarkable performance on ARC-AGI, showing that very small models can compete against large foundation models through a two-step refinement mechanism that updates an internal reasoning state $z$ and the predicted output $y$. Naturally, such refinement is of interest for any predictor; it is therefore natural to wonder whether the TRM mechanism could be effectively re-adopted in autoregressive models. However, TRMs cannot be simply compared to standard models because they lack causal predictive structures and contain persistent latent states that make it difficult to isolate specific performance gains. In this paper, we propose the Autoregressive TRM and evaluate it on small autoregressive tasks. To understand its efficacy, we propose a suite of models that gradually transform a standard Transformer to a Tiny Autoregressive Recursive Model in a controlled setting that fixes the block design, token stream, and next-token objective. Across compute-matched experiments on character-level algorithmic tasks, we surprisingly find that there are some two-level refinement baselines that show strong performance. Contrary to expectations, we find no reliable performance gains from the full Autoregressive TRM architecture. These results offer potential promise for two-step refinement mechanisms more broadly but caution against investing in the autoregressive TRM-specific model as a fruitful research direction.
Authors: Chang Han, Yijie Hu, Jingling Liu
Abstract: Autoregressive decoding remains a primary bottleneck in large language model (LLM) serving, motivating speculative decoding methods that reduce expensive teacher-model invocations by verifying multiple candidate tokens per step. Tree-structured speculation further increases parallelism, but is often brittle when ported across heterogeneous backends and accelerator stacks, where attention masking, KV-cache layouts, and indexing semantics are not interchangeable. We present EAGLE-Pangu, a reproducible system that ports EAGLE-3-style tree speculative decoding to a Pangu teacher backend on Ascend NPUs. EAGLE-Pangu contributes (i) an explicit branch/commit cache manager built on the Cache API, (ii) accelerator-safe tree tensorization that removes undefined negative indices by construction and validates structural invariants, and (iii) a fused-kernel-compatible teacher verification path with a debuggable eager fallback. On 240 turns from MT-Bench and HumanEval-style prompts, EAGLE-Pangu improves end-to-end decoding throughput by 1.27x on average, up to 2.46x at p99, over teacher-only greedy decoding in the fused-kernel performance path. We also provide a fused-kernel-free reference path with structured traces and invariant checks to support reproducible debugging and ablation across execution modes and tree budgets.
Authors: Guangnian Wan, Xinyin Ma, Gongfan Fang, Xinchao Wang
Abstract: Understanding and addressing potential safety alignment risks in large language models (LLMs) is critical for ensuring their safe and trustworthy deployment. In this paper, we highlight an insidious safety threat: a compromised LLM can maintain a facade of proper safety alignment while covertly generating harmful content. To achieve this, we finetune the model to understand and apply a steganographic technique. At inference time, we input a prompt that contains a steganographically embedded malicious target question along with a plaintext cover question. The model, in turn, produces a target response similarly embedded within a benign-looking cover response. In this process, human observers only see the model being prompted with a cover question and generating a corresponding cover response, while the malicious content is hidden from view. We demonstrate this invisible safety threat on GPT-4.1 despite the OpenAI finetuning API's safeguards. The finetuned model produces steganographic malicious outputs in response to hidden malicious prompts, while the user interface displays only a fully benign cover interaction. We also replicate the attack on three open-source models, Llama-3.3-70B-Instruct, Phi-4, and Mistral-Small-24B-Base-2501, confirming the generality of our method. We quantitatively evaluate our method on the AdvBench dataset, using Llama-Guard-3-8B for content safety classification. Across all four models, all stegotexts containing malicious content are incorrectly classified as safe.
Authors: Zhongjian Qiao, Jiafei Lyu, Boxiang Lyu, Yao Shu, Siyang Gao, Shuang Qiu
Abstract: Model-based offline reinforcement learning (RL) aims to enhance offline RL with a dynamics model that facilitates policy exploration. However, \textit{model exploitation} could occur due to inevitable model errors, degrading algorithm performance. Adversarial model learning offers a theoretical framework to mitigate model exploitation by solving a maximin formulation. Within such a paradigm, RAMBO~\citep{rigter2022rambo} has emerged as a representative and most popular method that provides a practical implementation with model gradient. However, we empirically reveal that severe Q-value underestimation and gradient explosion can occur in RAMBO with only slight hyperparameter tuning, suggesting that it tends to be overly conservative and suffers from unstable model updates. To address these issues, we propose \textbf{RO}bust value-aware \textbf{M}odel learning with \textbf{I}mplicitly differentiable adaptive weighting (ROMI). Instead of updating the dynamics model with model gradient, ROMI introduces a novel robust value-aware model learning approach. This approach requires the dynamics model to predict future states with values close to the minimum Q-value within a scale-adjustable state uncertainty set, enabling controllable conservatism and stable model updates. To further improve out-of-distribution (OOD) generalization during multi-step rollouts, we propose implicitly differentiable adaptive weighting, a bi-level optimization scheme that adaptively achieves dynamics- and value-aware model learning. Empirical results on D4RL and NeoRL datasets show that ROMI significantly outperforms RAMBO and achieves competitive or superior performance compared to other state-of-the-art methods on datasets where RAMBO typically underperforms. Code is available at https://github.com/zq2r/ROMI.git.
Authors: Aurelio Raffa Ugolini, Jessica Leoni, Valentina Breschi, Damiano Paniccia, Francesco Aldo Tucci, Luigi Capone, Mara Tanelli
Abstract: We present a novel Explainable methodology for Condition Monitoring, relying on healthy data only. Since faults are rare events, we propose to focus on learning the probability distribution of healthy observations only, and detect Anomalies at runtime. This objective is achieved via the definition of probabilistic measures of deviation from nominality, which allow to detect and anticipate faults. The Bayesian perspective underpinning our approach allows us to perform Uncertainty Quantification to inform decisions. At the same time, we provide descriptive tools to enhance the interpretability of the results, supporting the deployment of the proposed strategy also in safety-critical applications. The methodology is validated experimentally on two use cases: a publicly available benchmark for Predictive Maintenance, and a real-world Helicopter Transmission dataset collected over multiple years. In both applications, the method achieves competitive detection performance with respect to state-of-the-art anomaly detection methods.
Authors: Yunhui Liu, Qizhuo Xie, Yinfeng Chen, Xudong Jin, Tao Zheng, Bin Chong, Tieke He
Abstract: Graph anomaly detection (GAD) aims to identify nodes that deviate from normal patterns in structure or features. While recent GNN-based approaches have advanced this task, they struggle with two major challenges: 1) homophily disparity, where nodes exhibit varying homophily at both class and node levels; and 2) limited scalability, as many methods rely on costly whole-graph operations. To address them, we propose SAGAD, a Scalable and Adaptive framework for GAD. SAGAD precomputes multi-hop embeddings and applies reparameterized Chebyshev filters to extract low- and high-frequency information, enabling efficient training and capturing both homophilic and heterophilic patterns. To mitigate node-level homophily disparity, we introduce an Anomaly Context-Aware Adaptive Fusion, which adaptively fuses low- and high-pass embeddings using fusion coefficients conditioned on Rayleigh Quotient-guided anomalous subgraph structures for each node. To alleviate class-level disparity, we design a Frequency Preference Guidance Loss, which encourages anomalies to preserve more high-frequency information than normal nodes. SAGAD supports mini-batch training, achieves linear time and space complexity, and drastically reduces memory usage on large-scale graphs. Theoretically, SAGAD ensures asymptotic linear separability between normal and abnormal nodes under mild conditions. Extensive experiments on 10 benchmarks confirm SAGAD's superior accuracy and scalability over state-of-the-art methods.
Authors: Mingxi Zou, Jiaxiang Chen, Junfan Li, Langzhang Liang, Qifan Wang, Xu Yinghui, Zenglin Xu
Abstract: Preference-based alignment methods (e.g., RLHF, DPO) typically optimize a single scalar objective, implicitly averaging over heterogeneous human preferences. In practice, systematic annotator and user-group disagreement makes mean-reward maximization brittle and susceptible to proxy over-optimization. We propose **Disagreement-Aware Alignment via Risk-Constrained Decoding (DARC)**, a retraining-free inference-time method that frames response selection as distributionally robust, risk-sensitive decision making. Given multiple preference samples or scalable disagreement proxies, DARC reranks candidates by maximizing a *KL-robust (entropic)* satisfaction objective, and provides simple deployment controls that cap or penalize the corresponding entropic risk premium relative to the mean, enabling explicit risk budgets without retraining. We provide theoretical characterization linking this decoding rule to principled pessimism and KL-based distributionally robust optimization. Experiments on alignment benchmarks show that DARC reduces disagreement and tail risk while maintaining competitive average quality under noisy, heterogeneous feedback.
Authors: Lukas K\"onig, Manuel Kuhn, David Kappel, Anand Subramoney
Abstract: Existing frameworks for gradient-based training of spiking neural networks face a trade-off: discrete-time methods using surrogate gradients support arbitrary neuron models but introduce gradient bias and constrain spike-time resolution, while continuous-time methods that compute exact gradients require analytical expressions for spike times and state evolution, restricting them to simple neuron types such as Leaky Integrate and Fire (LIF). We introduce the Eventax framework, which resolves this trade-off by combining differentiable numerical ODE solvers with event-based spike handling. Built in JAX, our frame-work uses Diffrax ODE-solvers to compute gradients that are exact with respect to the forward simulation for any neuron model defined by ODEs . It also provides a simple API where users can specify just the neuron dynamics, spike conditions, and reset rules. Eventax prioritises modelling flexibility, supporting a wide range of neuron models, loss functions, and network architectures, which can be easily extended. We demonstrate Eventax on multiple benchmarks, including Yin-Yang and MNIST, using diverse neuron models such as Leaky Integrate-and-fire (LIF), Quadratic Integrate-and-fire (QIF), Exponential integrate-and-fire (EIF), Izhikevich and Event-based Gated Recurrent Unit (EGRU) with both time-to-first-spike and state-based loss functions, demonstrating its utility for prototyping and testing event-based architectures trained with exact gradients. We also demonstrate the application of this framework for more complex neuron types by implementing a multi-compartment neuron that uses a model of dendritic spikes in human layer 2/3 cortical Pyramidal neurons for computation. Code available at https://github.com/efficient-scalable-machine-learning/eventax.
URLs: https://github.com/efficient-scalable-machine-learning/eventax.
Authors: Jiayang Gao, Tianyi Zheng, Jiayang Zou, Fengxiang Yang, Shice Liu, Luyao Fan, Zheyu Zhang, Hao Zhang, Jinwei Chen, Peng-Tao Jiang, Bo Li, Jia Wang
Abstract: Classifier-Free Guidance (CFG) is a cornerstone of modern conditional diffusion models, yet its reliance on the fixed or heuristic dynamic guidance weight is predominantly empirical and overlooks the inherent dynamics of the diffusion process. In this paper, we provide a rigorous theoretical analysis of the Classifier-Free Guidance. Specifically, we establish strict upper bounds on the score discrepancy between conditional and unconditional distributions at different timesteps based on the diffusion process. This finding explains the limitations of fixed-weight strategies and establishes a principled foundation for time-dependent guidance. Motivated by this insight, we introduce \textbf{Control Classifier-Free Guidance (C$^2$FG)}, a novel, training-free, and plug-in method that aligns the guidance strength with the diffusion dynamics via an exponential decay control function. Extensive experiments demonstrate that C$^2$FG is effective and broadly applicable across diverse generative tasks, while also exhibiting orthogonality to existing strategies.
Authors: Thanapol Phungtua-eng, Yoshitaka Yamamoto
Abstract: Long-term time series forecasting (LTSF) is widely recognized as a central challenge in data mining and machine learning. LTSF has increasingly evolved into a benchmark-driven ''GAME,'' where models are ranked, compared, and declared state-of-the-art based primarily on marginal reductions in aggregated pointwise error metrics such as MSE and MAE. Across a small set of canonical datasets and fixed forecasting horizons, progress is communicated through leaderboard-style tables in which lower numerical scores define success. In this GAME, what is measured becomes what is optimized, and incremental error reduction becomes the dominant currency of advancement. We argue that this metric-centric regime is not merely incomplete, but structurally misaligned with the broader objectives of forecasting. In real-world settings, forecasting often prioritizes preserving temporal structure, trend stability, seasonal coherence, robustness to regime shifts, and supporting downstream decision processes. Optimizing aggregate pointwise error does not necessarily imply modeling these structural properties. As a result, leaderboard improvement may increasingly reflect specialization in benchmark configurations rather than a deeper understanding of temporal dynamics. This paper revisits LTSF evaluation as a foundational question in data science: what does it mean to measure forecasting progress? We propose a multi-dimensional evaluation perspective that integrates statistical fidelity, structural coherence, and decision-level relevance. By challenging the current metric monoculture, we aim to redirect attention from winning benchmark tables toward advancing meaningful, context-aware forecasting.
Authors: Yunhui Liu, Yongchao Liu, Yinfeng Chen, Chuntao Hong, Tao Zheng, Tieke He
Abstract: Hierarchical knowledge structures are ubiquitous across real-world domains and play a vital role in organizing information from coarse to fine semantic levels. While such structures have been widely used in taxonomy systems, biomedical ontologies, and retrieval-augmented generation, their potential remains underexplored in the context of Text-Rich Networks (TRNs), where each node contains rich textual content and edges encode semantic relationships. Existing methods for learning on TRNs often focus on flat semantic modeling, overlooking the inherent hierarchical semantics embedded in textual documents. To this end, we propose TIER (Hierarchical \textbf{T}axonomy-\textbf{I}nformed R\textbf{E}presentation Learning on Text-\textbf{R}ich Networks), which first constructs an implicit hierarchical taxonomy and then integrates it into the learned node representations. Specifically, TIER employs similarity-guided contrastive learning to build a clustering-friendly embedding space, upon which it performs hierarchical K-Means followed by LLM-powered clustering refinement to enable semantically coherent taxonomy construction. Leveraging the resulting taxonomy, TIER introduces a cophenetic correlation coefficient-based regularization loss to align the learned embeddings with the hierarchical structure. By learning representations that respect both fine-grained and coarse-grained semantics, TIER enables more interpretable and structured modeling of real-world TRNs. We demonstrate that our approach significantly outperforms existing methods on multiple datasets across diverse domains, highlighting the importance of hierarchical knowledge learning for TRNs.
Authors: Sidharth Sinha, Anson Bastos, Xuchao Zhang, Akshay Nambi, Chetan Bansal, Saravan Rajmohan
Abstract: Large language models (LLMs) excel in open domains but struggle in specialized settings with limited data and evolving knowledge. Existing domain adaptation practices rely heavily on manual trial-and-error processes, incur significant hyperparameter complexity, and are highly sensitive to data and user preferences, all under the high cost of LLM training. Moreover, the interactions and transferability of hyperparameter choices across models/domains remain poorly understood, making adaptation gains uncertain even with substantial effort. To solve these challenges, we present AutoAdapt, a novel end-to-end automated framework for efficient and reliable LLM domain adaptation. AutoAdapt leverages curated knowledge bases from literature and open-source resources to reduce expert intervention. To narrow the search space, we design a novel multi-agent debating system in which proposal and critic agents iteratively interact to align user intent and incorporate data signals and best practices into the planning process. To optimize hyperparameters under tight budgets, we propose AutoRefine, a novel LLM-based surrogate that replaces costly black-box search. Across 10 tasks, AutoAdapt achieves a 25% average relative accuracy improvement over state-of-the-art Automated Machine Learning baselines with minimal overhead.
Authors: Yeonsik Park, Hyeonseong Kim, Seungkyu Choi
Abstract: Post-training quantization (PTQ) has emerged as a prevailing technique for deploying large language models (LLMs) efficiently in terms of both memory and computation, across edge devices and server platforms. Existing PTQ methods primarily aim to reduce precision in weights and activations by mitigating quantization errors caused by channel-wise outlier activations (e.g., pre-quantization scaling, online transformations, or low-rank error reconstruction). Among these approaches, error reconstruction with low-rank adaptation (LoRA) has proven particularly effective, as it introduces a lightweight auxiliary computation path without requiring heavy optimization or additional online layers. However, prior studies reveal severe accuracy degradation under W4A4 settings, and conventional low-rank adaptations rely on two sequential factors, necessitating intermediate quantization during inference and thereby limiting low-precision efficiency. In this work, we propose SERQ, a saliency-aware error reconstruction method for low-bit LLM inference that employs a single low-rank compensation matrix. SERQ preserves efficient 4-bit matrix multiplication in linear layers by jointly mitigating quantization errors arising from both activation and weight saliency through three stages: (1) static activation flattening, (2) saliency-aware error reconstruction, and (3) offline weight permutation. The method incurs additional computation only for low-rank error reconstruction via a single decomposition, while all other operations are performed offline, thereby keeping latency overhead minimal. Empirically, SERQ outperforms prior error reconstruction methods under both W4A8 and W4A4 settings, and achieves higher accuracy than state-of-the-art rotation-based W4A4 approaches, while substantially reducing calibration complexity.
Authors: Tingting Chen, Feng Chu, Jiantong Zhang
Abstract: Service region design determines the geographic coverage of service networks, shaping long-term operational performance. Capital and operational constraints preclude simultaneous large-scale deployment, requiring expansion to proceed sequentially. The resulting challenge is to determine when and where to invest under demand uncertainty, balancing intertemporal trade-offs between early and delayed investment and accounting for network effects whereby each deployment reshapes future demand through inter-regional connectivity. This study addresses a sequential service region design (SSRD) problem incorporating two practical yet underexplored factors: a $k$-region constraint that limits the number of regions investable per period and a stochastic spillover effect linking investment decisions to demand evolution. The resulting problem requires sequencing regional portfolios under uncertainty, leading to a combinatorial explosion in feasible investment sequences. To address this challenge, we propose a solution framework that integrates real options analysis (ROA) with a Transformer-based Proximal Policy Optimization (TPPO) algorithm. ROA evaluates the intertemporal option value of investment sequences, while TPPO learns sequential policies that directly generate high option-value sequences without exhaustive enumeration. Numerical experiments on realistic multi-region settings demonstrate that TPPO converges faster than benchmark DRL methods and consistently identifies sequences with superior option value. Case studies and sensitivity analyses further confirm robustness and provide insights on investment concurrency, regional prioritization, and the increasing benefits of adaptive expansion via our approach under stronger spillovers and dynamic market conditions.
Authors: Jonas Landsgesell, Pascal Knoll
Abstract: Prior-Data Fitted Networks (PFNs), such as TabPFN and TabICL, have revolutionized tabular deep learning by leveraging in-context learning for tabular data. These models are meant as foundation models for classification and regression settings and promise to greatly simplify deployment in practical settings because their performance is unprecedented (in terms of mean squared error or $R^2$, when measured on common benchmarks like TabArena or TALENT). However, we see an important weakness of current benchmarks for the regression setting: the current benchmarks focus on evaluating win rates and performance using metrics like (root) mean squared error or $R^2$. Therefore, these leaderboards (implicitly and explicitly) push researchers to optimize for machine learning pipelines which elicit a good mean value estimate. The main problem is that this approach only evaluates a point estimate (namely the mean estimator which is the Bayes estimator associated with the mean squared error loss). In this article we discuss the application of proper scoring rules for evaluating the goodness of probabilistic forecasts in distributional regression. We also propose to enhance common machine learning benchmarks with metrics for probabilistic regression. To improve the status quo and make the machine learning community aware of scoring rules for probabilistic regression, we advocate to use the continuous ranked probability score (CRPS) in benchmarks for probabilistic regression. However, we also illustrate that the choice of the scoring rule changes the inductive bias of the trained model. We, therefore, advocate for finetuning or promptable tabular foundation models.
Authors: Patrick Wilhelm, Odej Kao
Abstract: In asynchronous federated learning (FL), client devices send updates to a central server at varying times based on their computational speed, often using stale versions of the global model. This staleness can degrade the convergence and accuracy of the global model. Previous work, such as AsyncFedED, proposed an adaptive aggregation method using Euclidean distance to measure staleness. In this paper, we extend this approach by exploring alternative distance metrics to more accurately capture the effect of gradient staleness. We integrate these metrics into the aggregation process and evaluate their impact on convergence speed, model performance, and training stability under heterogeneous clients and non-IID data settings. Our results demonstrate that certain metrics lead to more robust and efficient asynchronous FL training, offering a stronger foundation for practical deployment.
Authors: Dai Shi, Luke Thompson, Andi Han, Peiyan Hu, Junbin Gao, Jos\'e Miguel Hern\'andez-Lobato
Abstract: In this paper, we explore how our recently developed Wiener Chaos Expansion (WCE)-based neural operator (NO) can be applied to singular stochastic partial differential equations, e.g., the dynamic $\boldsymbol{\Phi}^4_2$ model simulated in the recent works. Unlike the previous WCE-NO which solves SPDEs by simply inserting Wick-Hermite features into the backbone NO model, we leverage feature-wise linear modulation (FiLM) to appropriately capture the dependency between the solution of singular SPDE and its smooth remainder. The resulting WCE-FiLM-NO shows excellent performance on $\boldsymbol{\Phi}^4_2$, as measured by relative $L_2$ loss, out-of-distribution $L_2$ loss, and autocorrelation score; all without the help of renormalisation factor. In addition, we also show the potential of simulating $\boldsymbol{\Phi}^4_3$ data, which is more aligned with real scientific practice in statistical quantum field theory. To the best of our knowledge, this is among the first works to develop an efficient data-driven surrogate for the dynamical $\boldsymbol{\Phi}^4_3$ model.
Authors: Chang Li, Tshihao Tsu, Yaren Zhang, Chao Xue, Xiaodong He
Abstract: Large language models are increasingly trained as heterogeneous systems spanning multiple domains, expert partitions, and agentic pipelines, yet prevalent proximal objectives operate at a single scale and lack a principled mechanism for coupling token-level, trajectory-level, and higher-level hierarchical stability control. To bridge this gap, we derive the Aggregational Policy Censoring Objective (APC-Obj), the first exact unconstrained reformulation of sample-based TV-TRPO, establishing that clipping-based surrogate design and trust-region optimization are dual formulations of the same problem. Building on this foundation, we develop Fiber Bundle Gating (FBG), an algebraic framework that organizes sampled RL data as a fiber bundle and decomposes ratio gating into a base-level gate on trajectory aggregates and a fiber-level gate on per-token residuals, with provable first-order agreement with the true RL objective near on-policy. From APC-Obj and FBG we derive Fibration Policy Optimization (or simply, FiberPO), a concrete objective whose Jacobian is block-diagonal over trajectories, reduces to identity at on-policy, and provides better update direction thus improving token efficiency. The compositional nature of the framework extends beyond the trajectory-token case: fibrations compose algebraically into a Fibration Gating Hierarchy (FGH) that scales the same gating mechanism to arbitrary hierarchical depth without new primitives, as demonstrated by FiberPO-Domain, a four-level instantiation with independent trust-region budgets at the domain, prompt group, trajectory, and token levels. Together, these results connect the trust-region theory, a compositional algebraic structure, and practical multi-scale stability control into a unified framework for LLM policy optimization.
Authors: Magnus Ross, Nel Swanepoel, Akish Luintel, Emma McGuire, Ingemar J. Cox, Steve Harris, Vasileios Lampos
Abstract: Timely transition from intravenous (IV) to oral antibiotic therapy shortens hospital stays, reduces catheter-related infections, and lowers healthcare costs, yet one in five patients in England remain on IV antibiotics despite meeting switching criteria. Clinical decision support systems can improve switching rates, but approaches that learn from historical decisions reproduce the delays and inconsistencies of routine practice. We propose using neural processes to model vital sign trajectories probabilistically, predicting switch-readiness by comparing forecasts against clinical guidelines rather than learning from past actions, and ranking patients to prioritise clinical review. The design yields interpretable outputs, adapts to updated guidelines without retraining, and preserves clinical judgement. Validated on MIMIC-IV (US intensive care, 6,333 encounters) and UCLH (a large urban academic UK hospital group, 10,584 encounters), the system selects 2.2-3.2$\times$ more relevant patients than random. Our results demonstrate that forecasting patient physiology offers a principled foundation for decision support in antibiotic stewardship.
Authors: Prakash Kumbhakar, Shrey Srivastava, Haroon R Lone
Abstract: Federated Learning (FL) suffers significant performance degradation in real-world deployments characterized by moderate to extreme statistical heterogeneity (non-IID client data). While global aggregation strategies promote broad generalization, they often fail to capture the diversity of local data distributions, leading to suboptimal personalization. We address this problem with FedPrism, a framework that uses two main strategies. First, it uses a Prism Decomposition method that builds each client's model from three parts: a global foundation, a shared group part for similar clients, and a private part for unique local data. This allows the system to group similar users together automatically and adapt if their data changes. Second, we include a Dual-Stream design that runs a general model alongside a local specialist. The system routes predictions between the general model and the local specialist based on the specialist's confidence. Through systematic experiments on non-IID data partitions, we demonstrate that FedPrism exceeds static aggregation and hard-clustering baselines, achieving significant accuracy gains under high heterogeneity. These results establish FedPrism as a robust and flexible solution for federated learning in heterogeneous environments, effectively balancing generalizable knowledge with adaptive personalization.
Authors: Antonia Hager, Sven Nebendahl, Alexej Klushyn, Jasper Krauser, Torleiv H. Bryne, Tor Arne Johansen
Abstract: Airborne Magnetic Anomaly Navigation (MagNav) provides a jamming-resistant and robust alternative to satellite navigation but requires the real-time compensation of the aircraft platform's large and dynamic magnetic interference. State-of-the-art solutions often rely on extensive offline calibration flights or pre-training, creating a logistical barrier to operational deployment. We present a fully adaptive MagNav architecture featuring a "cold-start" capability that identifies and compensates for the aircraft's magnetic signature entirely in-flight. The proposed method utilizes an extended Kalman filter with an augmented state vector that simultaneously estimates the aircraft's kinematic states as well as the coefficients of the physics-based Tolles-Lawson calibration model and the parameters of a Neural Network to model aircraft interferences. The Kalman filter update is mathematically equivalent to an online Natural Gradient descent, integrating superior convergence and data efficiency of state-of-the-art second-order optimization directly into the navigation filter. To enhance operational robustness, the neural network is constrained to a residual learning role, modeling only the nonlinearities uncorrected by the explainable physics-based calibration baseline. Validated on the MagNav Challenge dataset, our framework effectively bounds inertial drift using a magnetometer-only feature set. The results demonstrate navigation accuracy comparable to state-of-the-art models trained offline, without requiring prior calibration flights or dedicated maneuvers.
Authors: Yuxiang Zhang, Enyan Dai
Abstract: Graph Neural Networks (GNNs) have demonstrated remarkable success across diverse tasks. However, their generalization capability is often hindered by spurious correlations between node features and labels in the graph. Our analysis reveals that GNNs tend to exploit imperceptible statistical correlations in training data, even when such correlations are unreliable for prediction. To address this challenge, we propose the Spurious Correlation Learning Graph Neural Network (SCL-GNN), a novel framework designed to enhance generalization on both Independent and Identically Distributed (IID) and Out-of-Distribution (OOD) graphs. SCL-GNN incorporates a principled spurious correlation learning mechanism, leveraging the Hilbert-Schmidt Independence Criterion (HSIC) to quantify correlations between node representations and class scores. This enables the model to identify and mitigate irrelevant but influential spurious correlations effectively. Additionally, we introduce an efficient bi-level optimization strategy to jointly optimize modules and GNN parameters, preventing overfitting. Extensive experiments on real-world and synthetic datasets demonstrate that SCL-GNN consistently outperforms state-of-the-art baselines under various distribution shifts, highlighting its robustness and generalization capabilities.
Authors: Zahra Jafari, Azadeh Zamanifar, Amirfarhad Farhadi
Abstract: Accurate and interpretable mortality risk prediction in intensive care units (ICUs) remains a critical challenge due to the irregular temporal structure of electronic health records (EHRs), the complexity of longitudinal disease trajectories, and the lack of clinically grounded explanations in many data-driven models. To address these challenges, we propose \textit{TA-RNN-Medical-Hybrid}, a time-aware and knowledge-enriched deep learning framework that jointly models longitudinal clinical sequences and irregular temporal dynamics through explicit continuous-time encoding, along with standardized medical concept representations. The proposed framework extends time-aware recurrent modeling by integrating explicit continuous-time embeddings that operate independently of visit indexing, SNOMED-based disease representations, and a hierarchical dual-level attention mechanism that captures both visit-level temporal importance and feature/concept-level clinical relevance. This design enables accurate mortality risk estimation while providing transparent and clinically meaningful explanations aligned with established medical knowledge. We evaluate the proposed approach on the MIMIC-III critical care dataset and compare it against strong time-aware and sequential baselines. Experimental results demonstrate that TA-RNN-Medical-Hybrid consistently improves predictive performance in terms of AUC, accuracy, and recall-oriented F$_2$-score. Moreover, qualitative analysis shows that the model effectively decomposes mortality risk across time and clinical concepts, yielding interpretable insights into disease severity, chronicity, and temporal progression. Overall, the proposed framework bridges the gap between predictive accuracy and clinical interpretability, offering a scalable and transparent solution for high-stakes ICU decision support systems.
Authors: Yilin Wen, Yi Guo, Bo Zhao, Wei Qi, Zechun Hu, Colin Jones, Jian Sun
Abstract: Real-world optimization problems are often constrained by complex physical laws that limit computational scalability. These constraints are inherently tied to complex regions, and thus learning models that incorporate physical and geometric knowledge, i.e., physics-informed machine learning (PIML), offer a promising pathway for efficient solution. Here, we introduce PolyFormer, which opens a new direction for PIML in prescriptive optimization tasks, where physical and geometric knowledge is not merely used to regularize learning models, but to simplify the problems themselves. PolyFormer captures geometric structures behind constraints and transforms them into efficient polytopic reformulations, thereby decoupling problem complexity from solution difficulty and enabling off-the-shelf optimization solvers to efficiently produce feasible solutions with acceptable optimality loss. Through evaluations across three important problems (large-scale resource aggregation, network-constrained optimization, and optimization under uncertainty), PolyFormer achieves computational speedups up to 6,400-fold and memory reductions up to 99.87%, while maintaining solution quality competitive with or superior to state-of-the-art methods. These results demonstrate that PolyFormer provides an efficient and reliable solution for scalable constrained optimization, expanding the scope of PIML to prescriptive tasks in scientific discovery and engineering applications.
Authors: Chaewon Moon, Dongkuk Si, Chulhee Yun
Abstract: We study the implicit bias of Sharpness-Aware Minimization (SAM) when training $L$-layer linear diagonal networks on linearly separable binary classification. For linear models ($L=1$), both $\ell_\infty$- and $\ell_2$-SAM recover the $\ell_2$ max-margin classifier, matching gradient descent (GD). However, for depth $L = 2$, the behavior changes drastically -- even on a single-example dataset. For $\ell_\infty$-SAM, the limit direction depends critically on initialization and can converge to $\mathbf{0}$ or to any standard basis vector, in stark contrast to GD, whose limit aligns with the basis vector of the dominant data coordinate. For $\ell_2$-SAM, we show that although its limit direction matches the $\ell_1$ max-margin solution as in the case of GD, its finite-time dynamics exhibit a phenomenon we call "sequential feature amplification", in which the predictor initially relies on minor coordinates and gradually shifts to larger ones as training proceeds or initialization increases. Our theoretical analysis attributes this phenomenon to $\ell_2$-SAM's gradient normalization factor applied in its perturbation, which amplifies minor coordinates early and allows major ones to dominate later, giving a concrete example where infinite-time implicit-bias analyses are insufficient. Synthetic and real-data experiments corroborate our findings.
Authors: Shubham Aggarwal, Lokendra Kumar
Abstract: The dense output projection in multi-head attention scales quadratically with model dimension, contributing significantly to parameter count, memory footprint, and inference cost. We propose replacing this projection with a fixed, parameter-free Walsh Hadamard Transform followed by a lightweight learnable affine rescaling, eliminating approximately 25 percent of attention parameters per block while preserving global cross head interaction through an orthogonal, norm-preserving transformation. Across different model sizes, we demonstrate that this structured substitution maintains comparable or slightly superior downstream task performance on standard benchmarks, while achieving up to 7 percent aggregate parameter reduction, 8.9 percent peak memory savings, and 6.6 percent throughput improvement at scale, with efficiency gains growing monotonically with model size, batch size, and sequence length. Interestingly, we observe that structured Hadamard-based models exhibit a steeper validation loss curve relative to training FLOPs compared to their dense counterparts, suggesting more favorable compute utilization during training.
Authors: Marcin Kostrzewa, Krzysztof Galus, Maciej Zi\k{e}ba
Abstract: We present a new method for generating plausible counterfactual explanations for time series classification problems. The approach performs gradient-based optimization directly in the input space. To enforce plausibility, we integrate soft-DTW (dynamic time warping) alignment with $k$-nearest neighbors from the target class, which effectively encourages the generated counterfactuals to adopt a realistic temporal structure. The overall optimization objective is a multi-faceted loss function that balances key counterfactual properties. It incorporates losses for validity, sparsity, and proximity, alongside the novel soft-DTW-based plausibility component. We conduct an evaluation of our method against several strong reference approaches, measuring the key properties of the generated counterfactuals across multiple dimensions. The results demonstrate that our method achieves competitive performance in validity while significantly outperforming existing approaches in distributional alignment with the target class, indicating superior temporal realism. Furthermore, a qualitative analysis highlights the critical limitations of existing methods in preserving realistic temporal structure. This work shows that the proposed method consistently generates counterfactual explanations for time series classifiers that are not only valid but also highly plausible and consistent with temporal patterns.
Authors: Gustavo A. Dorrego
Abstract: Standard Gradient Descent and its modern variants assume local, Markovian weight updates, making them highly susceptible to noise and overfitting. This limitation becomes critically severe in extremely imbalanced datasets such as financial fraud detection where dominant class gradients systematically overwrite the subtle signals of the minority class. In this paper, we introduce a novel optimization algorithm grounded in Fractional Calculus. By isolating the core memory engine of the generalized fractional derivative, the Weighted Fractional Weyl Integral, we replace the instantaneous gradient with a dynamically weighted historical sequence. This fractional memory operator acts as a natural regularizer. Empirical evaluations demonstrate that our method prevents overfitting in medical diagnostics and achieves an approximately 40 percent improvement in PR-AUC over classical optimizers in financial fraud detection, establishing a robust bridge between pure fractional topology and applied Machine Learning.
Authors: Dongsu Lee, Daehee Lee, Amy Zhang
Abstract: Despite remarkable achievements in single-agent offline reinforcement learning (RL), multi-agent RL (MARL) has struggled to adopt this paradigm, largely persisting with on-policy training and self-play from scratch. One reason for this gap comes from the instability of non-linear value decomposition, leading prior works to avoid complex mixing networks in favor of linear value decomposition (e.g., VDN) with value regularization used in single-agent setups. In this work, we analyze the source of instability in non-linear value decomposition within the offline MARL setting. Our observations confirm that they induce value-scale amplification and unstable optimization. To alleviate this, we propose a simple technique, scale-invariant value normalization (SVN), that stabilizes actor-critic training without altering the Bellman fixed point. Empirically, we examine the interaction among key components of offline MARL (e.g., value decomposition, value learning, and policy extraction) and derive a practical recipe that unlocks its full potential.
Authors: Daniil Karzanov, Marcin Detyniecki
Abstract: Deep neural networks for image classification often exhibit overconfidence on out-of-distribution (OOD) samples. To address this, we introduce Geometrically Constrained Outlier Synthesis (GCOS), a training-time regularization framework aimed at improving OOD robustness during inference. GCOS addresses a limitation of prior synthesis methods by generating virtual outliers in the hidden feature space that respect the learned manifold structure of in-distribution (ID) data. The synthesis proceeds in two stages: (i) a dominant-variance subspace extracted from the training features identifies geometrically informed, off-manifold directions; (ii) a conformally-inspired shell, defined by the empirical quantiles of a nonconformity score from a calibration set, adaptively controls the synthesis magnitude to produce boundary samples. The shell ensures that generated outliers are neither trivially detectable nor indistinguishable from in-distribution data, facilitating smoother learning of robust features. This is combined with a contrastive regularization objective that promotes separability of ID and OOD samples in a chosen score space, such as Mahalanobis or energy-based. Experiments demonstrate that GCOS outperforms state-of-the-art methods using standard energy-based inference on near-OOD benchmarks, defined as tasks where outliers share the same semantic domain as in-distribution data. As an exploratory extension, the framework naturally transitions to conformal OOD inference, which translates uncertainty scores into statistically valid p-values and enables thresholds with formal error guarantees, providing a pathway toward more predictable and reliable OOD detection.
Authors: Th\'eo Zangato, Aomar Osmani, Pegah Alizadeh
Abstract: Meta-Reinforcement Learning addresses the critical limitations of conventional Reinforcement Learning in multi-task and non-stationary environments by enabling fast policy adaptation and improved generalization. We introduce a novel Meta-RL framework that integrates a bi-level optimization scheme with a hybrid actor-critic architecture specially designed to enhance sample efficiency and inter-task adaptability. To improve knowledge transfer, we meta-learn a shared state feature extractor jointly optimized across actor and critic networks, providing efficient representation learning and limiting overfitting to individual tasks or dominant profiles. Additionally, we propose a parameter-sharing mechanism between the outer- and inner-loop actor networks, to reduce redundant learning and accelerate adaptation during task revisitation. The approach is validated on a real-world Building Energy Management Systems dataset covering nearly a decade of temporal and structural variability, for which we propose a task preparation method to promote generalization. Experiments demonstrate effective task adaptation and better performance compared to conventional RL and Meta-RL methods.
Authors: Jes\'us S\'anchez Ochoa, Enrique Tom\'as Mart\'inez Beltr\'an, Alberto Huertas Celdr\'an
Abstract: In recent years, Artificial Intelligence has become a powerful partner for complex tasks such as data analysis, prediction, and problem-solving, yet its lack of transparency raises concerns about its reliability. In sensitive domains such as healthcare or cybersecurity, ensuring transparency, trustworthiness, and robustness is essential, since the consequences of wrong decisions or successful attacks can be severe. Prior neuron-level interpretability approaches are primarily descriptive, task-dependent, or require retraining, which limits their use as systematic, reusable tools for evaluating internal robustness across architectures and domains. To overcome these limitations, this work proposes SYNAPSE, a systematic, training-free framework for understanding and stress-testing the internal behavior of Transformer models across domains. It extracts per-layer [CLS] representations, trains a lightweight linear probe to obtain global and per-class neuron rankings, and applies forward-hook interventions during inference. This design enables controlled experiments on internal representations without altering the original model, thereby allowing weaknesses, stability patterns, and label-specific sensitivities to be measured and compared directly across tasks and architectures. Across all experiments, SYNAPSE reveals a consistent, domain-independent organization of internal representations, in which task-relevant information is encoded in broad, overlapping neuron subsets. This redundancy provides a strong degree of functional stability, while class-wise asymmetries expose heterogeneous specialization patterns and enable label-aware analysis. In contrast, small structured manipulations in weight or logit space are sufficient to redirect predictions, highlighting complementary vulnerability profiles and illustrating how SYNAPSE can guide the development of more robust Transformer models.
Authors: Adrian Garcia-Casta\~neda, Jon Irureta, Jon Imaz, Aizea Lojo
Abstract: Class Incremental Learning (CIL) poses a fundamental challenge: maintaining a balance between the plasticity required to learn new tasks and the stability needed to prevent catastrophic forgetting. While expansion-based methods effectively mitigate forgetting by adding task-specific parameters, they suffer from uncontrolled architectural growth and memory overhead. In this paper, we propose a novel dynamic scaling framework that adaptively manages model capacity through a cyclic "GRow, Assess, ComprEss" (GRACE) strategy. Crucially, we supplement backbone expansion with a novel saturation assessment phase that evaluates the utilization of the model's capacity. This assessment allows the framework to make informed decisions to either expand the architecture or compress the backbones into a streamlined representation, preventing parameter explosion. Experimental results demonstrate that our approach achieves state-of-the-art performance across multiple CIL benchmarks, while reducing memory footprint by up to a 73% compared to purely expansionist models.
Authors: Dongfang Li, Zixuan Liu, Gang Lin, Baotian Hu, Min Zhang
Abstract: The quadratic complexity of the attention mechanism and the substantial memory footprint of the Key-Value (KV) cache present severe computational and memory challenges for Large Language Models (LLMs) processing long contexts. Existing retrieval-based methods often compromise semantic integrity through fixed-size chunking and suffer from inefficient linear scanning. In this paper, we propose LycheeCluster, a novel method for efficient KV cache management. LycheeCluster preserves local semantic coherence via boundary-aware chunking and constructs a recursive hierarchical index rooted in the triangle inequality. This design transforms cache retrieval from a linear scan into a theoretically bounded, logarithmic-time pruning process, while a lazy update strategy supports efficient streaming generation. Experiments demonstrate that LycheeCluster achieves up to a 3.6x end-to-end inference speedup with negligible degradation in model performance, outperforming state-of-the-art KV cache management methods (e.g., Quest, ClusterKV). We will release our code and kernels after publication.
Authors: L. Juli\'an Lechuga L\'opez, Tim G. J. Rudner, Farah E. Shamout
Abstract: Safe predictions are a crucial requirement for integrating predictive models into clinical decision support systems. One approach for ensuring trustworthiness is to enable models' ability to express their uncertainty about individual predictions. However, current machine learning models frequently lack reliable uncertainty estimation, hindering real-world deployment. This is further observed in multimodal settings, where the goal is to enable effective information fusion. In this work, we propose $\texttt{MedCertAIn}$, a predictive uncertainty framework that leverages multimodal clinical data for in-hospital risk prediction to improve model performance and reliability. We design data-driven priors over neural network parameters using a hybrid strategy that considers cross-modal similarity in self-supervised latent representations and modality-specific data corruptions. We train and evaluate the models with such priors using clinical time-series and chest X-ray images from the publicly-available datasets MIMIC-IV and MIMIC-CXR. Our results show that $\texttt{MedCertAIn}$ significantly improves predictive performance and uncertainty quantification compared to state-of-the-art deterministic baselines and alternative Bayesian methods. These findings highlight the promise of data-driven priors in advancing robust, uncertainty-aware AI tools for high-stakes clinical applications.
Authors: Fabio Valerio Massoli, Andrey Kuzmin, Arash Behboodi
Abstract: Chain-of-Thought (CoT) prompting improves LLM accuracy on complex tasks but often increases token usage and inference cost. Existing "Budget Forcing" methods reducing cost via fine-tuning with heuristic length penalties, suppress both essential reasoning and redundant filler. We recast efficient reasoning as a lossy compression problem under the Information Bottleneck (IB) principle, and identify a key theoretical gap when applying naive IB to transformers: attention violates the Markov property between prompt, reasoning trace, and response. To resolve this issue, we model CoT generation under the Conditional Information Bottleneck (CIB) principle, where the reasoning trace Z acts as a computational bridge that contains only the information about the response Y that is not directly accessible from the prompt X. This yields a general Reinforcement Learning objective: maximize task reward while compressing completions under a prior over reasoning traces, subsuming common heuristics (e.g., length penalties) as special cases (e.g., uniform priors). In contrast to naive token-counting-based approaches, we introduce a semantic prior that measures token cost by surprisal under a language model prior. Empirically, our CIB objective prunes cognitive bloat while preserving fluency and logic, improving accuracy at moderate compression and enabling aggressive compression with minimal accuracy drop.
Authors: Weizheng Zhang, Xunjie Xie, Hao Pan, Xiaowei Duan, Bingteng Sun, Qiang Du, Lin Lu
Abstract: While Physics-Informed Neural Networks (PINNs) offer a mesh-free approach to solving PDEs, standard point-wise residual minimization suffers from convergence pathologies in topologically complex domains like Triply Periodic Minimal Surfaces (TPMS). The locality bias of point-wise constraints fails to propagate global information through tortuous channels, causing unstable gradients and conservation violations. To address this, we propose the Multi-scale Weak-form PINN (MUSA-PINN), which reformulates PDE constraints as integral conservation laws over hierarchical spherical control volumes. We enforce continuity and momentum conservation via flux-balance residuals on control surfaces. Our method utilizes a three-scale subdomain strategy-comprising large volumes for long-range coupling, skeleton-aware meso-scale volumes aligned with transport pathways, and small volumes for local refinement-alongside a two-stage training schedule prioritizing continuity. Experiments on steady incompressible flow in TPMS geometries show MUSA-PINN outperforms state-of-the-art baselines, reducing relative errors by up to 93% and preserving mass conservation.
Authors: Eric Parish, Anthony Gruber, Patrick Blonigan, Irina Tezaur
Abstract: We propose neural network operator inference (NN-OpInf): a structure-preserving, composable, and minimally restrictive operator inference framework for the non-intrusive reduced-order modeling of dynamical systems. The approach learns latent dynamics from snapshot data, enforcing local operator structure such as skew-symmetry, (semi-)positive definiteness, and gradient preservation, while also reflecting complex dynamics by supporting additive compositions of heterogeneous operators. We present practical training strategies and analyze computational costs relative to linear and quadratic polynomial OpInf (P-OpInf). Numerical experiments across several nonlinear and parametric problems demonstrate improved accuracy, stability, and robustness over P-OpInf and prior NN-ROM formulations, particularly when the dynamics are not well represented by polynomial models. These results suggest that NN-OpInf can serve as an effective drop-in replacement for P-OpInf when the dynamics to be modeled contain non-polynomial nonlinearities, offering potential gains in accuracy and out-of-distribution performance at the expense of higher training computational costs and a more difficult, non-convex learning problem.
Authors: Paul Hofman, Timo L\"ohr, Maximilian Muschalik, Yusuf Sale, Eyke H\"ullermeier
Abstract: A reliable representation of uncertainty is essential for the application of modern machine learning methods in safety-critical settings. In this regard, the use of credal sets (i.e., convex sets of probability distributions) has recently been proposed as a suitable approach to representing epistemic uncertainty. However, as with other approaches to epistemic uncertainty, training credal predictors is computationally complex and usually involves (re-)training an ensemble of models. The resulting computational complexity prevents their adoption for complex models such as foundation models and multi-modal systems. To address this problem, we propose an efficient method for credal prediction that is grounded in the notion of relative likelihood and inspired by techniques for the calibration of probabilistic classifiers. For each class label, our method predicts a range of plausible probabilities in the form of an interval. To produce the lower and upper bounds of these intervals, we propose a technique that we refer to as decalibration. Extensive experiments show that our method yields credal sets with strong performance across diverse tasks, including coverage-efficiency evaluation, out-of-distribution detection, and in-context learning. Notably, we demonstrate credal prediction on models such as TabPFN and CLIP -- architectures for which the construction of credal sets was previously infeasible.
Authors: Michelle Espranita Liman, \"Ozg\"un Turgut, Alexander M\"uller, Eimo Martens, Daniel Rueckert, Philip M\"uller
Abstract: Electrocardiography (ECG) is a low-cost, widely used modality for diagnosing electrical abnormalities like atrial fibrillation by capturing the heart's electrical activity. However, it cannot directly measure cardiac morphological phenotypes, such as left ventricular ejection fraction (LVEF), which typically require echocardiography (Echo). Predicting these phenotypes from ECG would enable early, accessible health screening. Existing self-supervised methods suffer from a representational mismatch by aligning ECGs to single-view Echos, which only capture local, spatially restricted anatomical snapshots. To address this, we propose Echo2ECG, a multimodal self-supervised learning framework that enriches ECG representations with the heart's morphological structure captured in multi-view Echos. We evaluate Echo2ECG as an ECG feature extractor on two clinically relevant tasks that fundamentally require morphological information: (1) classification of structural cardiac phenotypes across three datasets, and (2) retrieval of Echo studies with similar morphological characteristics using ECG queries. Our extracted ECG representations consistently outperform those of state-of-the-art unimodal and multimodal baselines across both tasks, despite being 18x smaller than the largest baseline. These results demonstrate that Echo2ECG is a robust, powerful ECG feature extractor. Our code is accessible at https://github.com/michelleespranita/Echo2ECG.
Authors: Prajit T Rajendran, Fabio Arnez, Huascar Espinoza, Agnes Delaborde, Chokri Mraidha
Abstract: In high stakes environments, agents relying purely on imitation learning or reinforcement learning often struggle to avoid safety-critical errors during exploration. Existing reinforcement learning approaches for environments such as chess require hundreds of thousands of episodes and substantial computational resources to converge. Imitation learning, on the other hand, is more sample efficient but is brittle under distributional shift and lacks mechanisms for proactive risk avoidance. In this work, we propose Oracle-Guided Soft Shielding (OGSS), a simple yet effective framework for safer decision-making, enabling safe exploration by learning a probabilistic safety model from oracle feedback in an imitation learning setting. Focusing on the domain of chess, we train a model to predict strong moves based on past games, and separately learn a blunder prediction model from Stockfish evaluations to estimate the tactical risk of each move. During inference, the agent first generates a set of candidate moves and then uses the blunder model to determine high-risk options, and uses a utility function combining the predicted move likelihood from the policy model and the blunder probability to select actions that strike a balance between performance and safety. This enables the agent to explore and play competitively while significantly reducing the chance of tactical mistakes. Across hundreds of games against a strong chess engine, we compare our approach with other methods in the literature, such as action pruning, SafeDAgger, and uncertainty-based sampling. Our results demonstrate that OGSS variants maintain a lower blunder rate even as the agent's exploration ratio is increased by several folds, highlighting its ability to support broader exploration without compromising tactical soundness.
Authors: Swetha Ganesh, Vaneet Aggarwal
Abstract: While standard reinforcement learning optimizes a single reward signal, many applications require optimizing a nonlinear utility $f(J_1^\pi,\dots,J_M^\pi)$ over multiple objectives, where each $J_m^\pi$ denotes the expected discounted return of a distinct reward function. A common approach is concave scalarization, which captures important trade-offs such as fairness and risk sensitivity. However, nonlinear scalarization introduces a fundamental challenge for policy gradient methods: the gradient depends on $\partial f(J^\pi)$, while in practice only empirical return estimates $\hat J$ are available. Because $f$ is nonlinear, the plug-in estimator is biased ($\mathbb{E}[\partial f(\hat J)] \neq \partial f(\mathbb{E}[\hat J])$), leading to persistent gradient bias that degrades sample complexity. In this work we identify and overcome this bias barrier in concave-scalarized multi-objective reinforcement learning. We show that existing policy-gradient methods suffer an intrinsic $\widetilde{\mathcal{O}}(\epsilon^{-4})$ sample complexity due to this bias. To address this issue, we develop a Natural Policy Gradient (NPG) algorithm equipped with a multi-level Monte Carlo (MLMC) estimator that controls the bias of the scalarization gradient while maintaining low sampling cost. We prove that this approach achieves the optimal $\widetilde{\mathcal{O}}(\epsilon^{-2})$ sample complexity for computing an $\epsilon$-optimal policy. Furthermore, we show that when the scalarization function is second-order smooth, the first-order bias cancels automatically, allowing vanilla NPG to achieve the same $\widetilde{\mathcal{O}}(\epsilon^{-2})$ rate without MLMC. Our results provide the first optimal sample complexity guarantees for concave multi-objective reinforcement learning under policy-gradient methods.
Authors: Songyang Chen, Youfang Lin, Yu Liu, Shuai Zheng, Lei Zou
Abstract: Unsupervised graph alignment aims to find the node correspondence across different graphs without any anchor node pairs. Despite the recent efforts utilizing deep learning-based techniques, such as the embedding and optimal transport (OT)-based approaches, we observe their limitations in terms of model accuracy-efficiency tradeoff. By focusing on the exploitation of local and global graph information, we formalize them as the ``local representation, global alignment'' paradigm, and present a new ``global representation and alignment'' paradigm to resolve the mismatch between the two phases in the alignment process. We then propose \underline{Gl}obal representation and \underline{o}ptimal transport-\underline{b}ased \underline{Align}ment (\texttt{GlobAlign}), and its variant, \texttt{GlobAlign-E}, for better \underline{E}fficiency. Our methods are equipped with the global attention mechanism and a hierarchical cross-graph transport cost, able to capture long-range and implicit node dependencies beyond the local graph structure. Furthermore, \texttt{GlobAlign-E} successfully closes the time complexity gap between representative embedding and OT-based methods, reducing OT's cubic complexity to quadratic terms. Through extensive experiments, our methods demonstrate superior performance, with up to a 20\% accuracy improvement over the best competitor. Meanwhile, \texttt{GlobAlign-E} achieves the best efficiency, with an order of magnitude speedup against existing OT-based methods.
Authors: Tommaso Giorgi, Pierriccardo Olivieri, Keyue Jiang, Laura Toni, Matteo Papini
Abstract: Learning compact state representations in Markov Decision Processes (MDPs) has proven crucial for addressing the curse of dimensionality in large-scale reinforcement learning (RL) problems. Existing principled approaches leverage structural priors on the MDP by constructing state representations as linear combinations of the state-graph Laplacian eigenvectors. When the transition graph is unknown or the state space is prohibitively large, the graph spectral features can be estimated directly via sample trajectories. In this work, we prove an upper bound on the approximation error of linear value function approximation under the learned spectral features. We show how this error scales with the algebraic connectivity of the state-graph, grounding the approximation quality in the topological structure of the MDP. We further bound the error introduced by the eigenvector estimation itself, leading to an end-to-end error decomposition across the representation learning pipeline. Additionally, our expression of the Laplacian operator for the RL setting, although equivalent to existing ones, prevents some common misunderstandings, of which we show some examples from the literature. Our results hold for general (non-uniform) policies without any assumptions on the symmetry of the induced transition kernel. We validate our theoretical findings with numerical simulations on gridworld environments.
Authors: Ismail Lamaakal, Chaymae Yahyati, Khalid El Makkaoui, Ibrahim Ouahbi, Yassine Maleh
Abstract: Deployed machine learning systems face distribution drift, yet most monitoring pipelines stop at alarms and leave the response underspecified under labeling, compute, and latency constraints. We introduce Drift2Act, a drift-to-action controller that treats monitoring as constrained decision-making with explicit safety. Drift2Act combines a sensing layer that maps unlabeled monitoring signals to a belief over drift types with an active risk certificate that queries a small set of delayed labels from a recent window to produce an anytime-valid upper bound $U_t(\delta)$ on current risk. The certificate gates operation: if $U_t(\delta) \le \tau$, the controller selects low-cost actions (e.g., recalibration or test-time adaptation); if $U_t(\delta) > \tau$, it activates abstain/handoff and escalates to rollback or retraining under cooldowns. In a realistic streaming protocol with label delay and explicit intervention costs, Drift2Act achieves near-zero safety violations and fast recovery at moderate cost on WILDS Camelyon17, DomainNet, and a controlled synthetic drift stream, outperforming alarm-only monitoring, adapt-always adaptation, schedule-based retraining, selective prediction alone, and an ablation without certification. Overall, online risk certification enables reliable drift response and reframes monitoring as decision-making with safety.
Authors: Andr\'es Ortiz, Nicol\'as J. Gallego-Molina, Carmen Jim\'enez-Mesa, Juan M. G\'orriz, Javier Ram\'irez
Abstract: Multi-Layer Perceptrons (MLPs) rely on pre-defined, fixed activation functions, imposing a static inductive bias that forces the network to approximate complex topologies solely through increased depth and width. Kolmogorov-Arnold Networks (KANs) address this limitation through edge-centric learnable functions, yet their formulation suffers from quadratic parameter scaling and architectural rigidity that hinders the effective integration of standard regularization techniques. This paper introduces the DualFlexKAN (DFKAN), a flexible architecture featuring a dual-stage mechanism that independently controls pre-linear input transformations and post-linear output activations. This decoupling enables hybrid networks that optimize the trade-off between expressiveness and computational cost. Unlike standard formulations, DFKAN supports diverse basis function families, including orthogonal polynomials, B-splines, and radial basis functions, integrated with configurable regularization strategies that stabilize training dynamics. Comprehensive evaluations across regression benchmarks, physics-informed tasks, and function approximation demonstrate that DFKAN outperforms both MLPs and conventional KANs in accuracy, convergence speed, and gradient fidelity. The proposed hybrid configurations achieve superior performance with one to two orders of magnitude fewer parameters than standard KANs, effectively mitigating the parameter explosion problem while preserving KAN-style expressiveness. DFKAN provides a principled, scalable framework for incorporating adaptive non-linearities, proving particularly advantageous for data-efficient learning and interpretable function discovery in scientific applications.
Authors: Riccardo De Monte, Matteo Cederle, Gian Antonio Susto
Abstract: State-of-the-art deep reinforcement learning (RL) methods have achieved remarkable performance in continuous control tasks, yet their computational complexity is often incompatible with the constraints of resource-limited hardware, due to their reliance on replay buffers, batch updates, and target networks. The emerging paradigm of streaming deep RL addresses this limitation through purely online updates, achieving strong empirical performance on standard benchmarks. In this work, we propose two novel streaming deep RL algorithms, Streaming Soft Actor-Critic (S2AC) and Streaming Deterministic Actor-Critic (SDAC), explicitly designed to be compatible with state-of-the-art batch RL methods, making them particularly suitable for on-device finetuning applications such as Sim2Real transfer. Both algorithms achieve performance comparable to state-of-the-art streaming baselines on standard benchmarks without requiring tedious hyperparameter tuning. Finally, we further investigate the practical challenges of transitioning from batch to streaming learning during finetuning and propose concrete strategies to tackle them.
Authors: Federico Giannini, Sandro D'Andrea, Emanuele Della Valle
Abstract: Concept drift, temporal dependence, and catastrophic forgetting represent major challenges when learning from data streams. While Streaming Machine Learning and Continual Learning (CL) address these issues separately, recent efforts in Streaming Continual Learning (SCL) aim to unify them. In this work, we introduce MAGIC Net, a novel SCL approach that integrates CL-inspired architectural strategies with recurrent neural networks to tame temporal dependence. MAGIC Net continuously learns, looks back at past knowledge by applying learnable masks over frozen weights, and expands its architecture when necessary. It performs all operations online, ensuring inference availability at all times. Experiments on synthetic and real-world streams show that it improves adaptation to new concepts, limits memory usage, and mitigates forgetting.
Authors: Valentin Heyraud, Zachary Weller-Davies, Jules Tilly
Abstract: We derive integral formulas that simplify the Vector Spherical Tensor Product recently introduced by Xie et al., which generalizes the Gaunt tensor product to antisymmetric couplings. In particular, we obtain explicit closed-form expressions for the antisymmetric analogues of the Gaunt coefficients. This enables us to simulate the Clebsch-Gordan tensor product using a single Vector Spherical Tensor Product, yielding a $9\times$ reduction in the required tensor product evaluations. Our results enable efficient and practical implementations of the Vector Spherical Tensor Product, paving the way for applications of this generalization of Gaunt tensor products in $\mathrm{SO}(3)$-equivariant neural networks. Moreover, we discuss how the Gaunt and the Vector Spherical Tensor Products allow to control the expressivity-runtime tradeoff associated with the usual Clebsch-Gordan Tensor Products. Finally, we investigate low rank decompositions of the normalizations of the considered tensor products in view of their use in equivariant neural networks.
Authors: Dyah Adila, Hanna Mazzawi, Benoit Dherin, Xavier Gonzalvo
Abstract: Adapting pre-trained models to specialized tasks often leads to catastrophic forgetting, where new knowledge overwrites foundational capabilities. Existing methods either compromise performance on the new task or struggle to balance training stability with efficient reuse of pre-trained knowledge. We introduce a novel function-preserving expansion method that resolves this dilemma. Our technique expands model capacity by replicating pre-trained parameters within transformer submodules and applying a scaling correction that guarantees the expanded model is mathematically identical to the original at initialization, enabling stable training while exploiting existing knowledge. Empirically, our method eliminates the trade-off between plasticity and stability, matching the performance of full fine-tuning on downstream tasks without any degradation of the model's original capabilities. Furthermore, we demonstrate the modularity of our approach, showing that by selectively expanding a small subset of layers we can achieve the same performance as full fine-tuning at a fraction of the computational cost.
Authors: Fenix W. Huang, Henning S. Mortveit, Christian M. Reidys
Abstract: In this article the authors develop an intrinsic measure for quantifying heterogeneity in training data for supervised learning. This measure is the variance of a random variable which factors through the influences of pairs of training points. The variance is shown to capture data heterogeneity and can thus be used to assess if a sample is a mixture of distributions. The authors prove that the data itself contains key information that supports a partitioning into blocks. Several proof of concept studies are provided that quantify the connection between variance and heterogeneity for EMNIST image data and synthetic data. The authors establish that variance is maximal for equal mixes of distributions, and detail how variance-based data purification followed by conventional training over blocks can lead to significant increases in test accuracy.
Authors: Andrzej Cichocki, Piergiulio Tempesta
Abstract: We introduce a comprehensive theoretical and algorithmic framework that bridges formal group theory and group entropies with modern machine learning, paving the way for an infinite, flexible family of Mirror Descent (MD) optimization algorithms. Our approach exploits the rich structure of group entropies, which are generalized entropic functionals governed by group composition laws, encompassing and significantly extending all trace-form entropies such as the Shannon, Tsallis, and Kaniadakis families. By leveraging group-theoretical mirror maps (or link functions) in MD, expressed via multi-parametric generalized logarithms and their inverses (group exponentials), we achieve highly flexible and adaptable MD updates that can be tailored to diverse data geometries and statistical distributions. To this end, we introduce the notion of \textit{mirror duality}, which allows us to seamlessly switch or interchange group-theoretical link functions with their inverses, subject to specific learning rate constraints. By tuning or learning the hyperparameters of the group logarithms enables us to adapt the model to the statistical properties of the training distribution, while simultaneously ensuring desirable convergence characteristics via fine-tuning. This generality not only provides greater flexibility and improved convergence properties, but also opens new perspectives for applications in machine learning and deep learning by expanding the design of regularizers and natural gradient algorithms. We extensively evaluate the validity, robustness, and performance of the proposed updates on large-scale, simplex-constrained quadratic programming problems.
Authors: Tiago Rodrigues de Almeida, Eduardo Gutierrez Maestro, Oscar Martinez Mozos
Abstract: In this paper, we present a context-free unsupervised approach based on a self-conditioned GAN to learn different modes from 2D trajectories. Our intuition is that each mode indicates a different behavioral moving pattern in the discriminator's feature space. We apply this approach to the problem of trajectory forecasting. We present three different training settings based on self-conditioned GAN, which produce better forecasters. We test our method in two data sets: human motion and road agents. Experimental results show that our approach outperforms previous context-free methods in the least representative supervised labels while performing well in the remaining labels. In addition, our approach outperforms globally in human motion, while performing well in road agents.
Authors: Bingxiang He, Yuxin Zuo, Zeyuan Liu, Shangziqi Zhao, Zixuan Fu, Junlin Yang, Cheng Qian, Kaiyan Zhang, Yuchen Fan, Ganqu Cui, Xiusi Chen, Youbang Sun, Xingtai Lv, Xuekai Zhu, Li Sheng, Ran Li, Huan-ang Gao, Yuchen Zhang, Bowen Zhou, Zhiyuan Liu, Ning Ding
Abstract: Unsupervised reinforcement learning with verifiable rewards (URLVR) offers a pathway to scale LLM training beyond the supervision bottleneck by deriving rewards without ground truth labels. Recent works leverage model intrinsic signals, showing promising early gains, yet their potential and limitations remain unclear. In this work, we revisit URLVR and provide a comprehensive analysis spanning taxonomy, theory and extensive experiments. We first classify URLVR methods into intrinsic versus external based on reward sources, then establish a unified theoretical framework revealing that all intrinsic methods converge toward sharpening the model's initial distribution This sharpening mechanism succeeds when initial confidence aligns with correctness but fails catastrophically when misaligned. Through systematic experiments, we show intrinsic rewards consistently follow a rise-then-fall pattern across methods, with collapse timing determined by model prior rather than engineering choices. Despite these scaling limits, we find intrinsic rewards remain valuable in test-time training on small datasets, and propose Model Collapse Step to measure model prior, serving as a practical indicator for RL trainability. Finally, we explore external reward methods that ground verification in computational asymmetries, showing preliminary evidence they may escape the confidence-correctness ceiling. Our findings chart boundaries for intrinsic URLVR while motivating paths toward scalable alternatives.
Authors: Yang Cai, Vineet Gupta, Zun Li, Aranyak Mehta
Abstract: The celebrated Myerson--Satterthwaite theorem shows that in bilateral trade, no mechanism can be simultaneously fully efficient, Bayesian incentive compatible (BIC), and budget balanced (BB). This naturally raises the question of how closely the gains from trade (GFT) achievable by a BIC and BB mechanism can approximate the first-best (fully efficient) benchmark. The optimal BIC and BB mechanism is typically complex and highly distribution-dependent, making it difficult to characterize directly. Consequently, much of the literature analyzes simpler mechanisms such as the Random-Offerer (RO) mechanism and establishes constant-factor guarantees relative to the first-best GFT. An important open question concerns the worst-case performance of the RO mechanism relative to first-best (FB) efficiency. While it was originally hypothesized that the approximation ratio $\frac{\text{GFT}_{\text{FB}}}{\text{GFT}_{\text{RO}}}$ is bounded by $2$, recent work provided counterexamples to this conjecture: Cai et al. proved that the ratio can be strictly larger than $2$, and Babaioff et al. exhibited an explicit example with ratio approximately $2.02$. In this work, we employ AlphaEvolve, an AI-guided evolutionary search framework, to explore the space of value distributions. We identify a new worst-case instance that yields an improved lower bound of $\frac{\text{GFT}_{\text{FB}}}{\text{GFT}_{\text{RO}}} \ge \textbf{2.0749}$. This establishes a new lower bound on the worst-case performance of the Random-Offerer mechanism, demonstrating a wider efficiency gap than previously known.
Authors: Yiannis Papageorgiou, Yannis Thomas, Ramin Khalili, Iordanis Koutsopoulos
Abstract: Can we find a network architecture for ML model training so as to optimize training loss (and thus, accuracy) in Split Federated Learning (SFL)? And can this architecture also reduce training delay and communication overhead? While accuracy is not influenced by how we split the model in ordinary, state-of-the-art SFL, in this work we answer the questions above in the affirmative. Recent Hierarchical SFL (HSFL) architectures adopt a three-tier training structure consisting of clients, (local) aggregators, and a central server. In this architecture, the model is partitioned at two partitioning layers into three sub-models, which are executed across the three tiers. Despite their merits, HSFL architectures overlook the impact of the partitioning layers and client-to-aggregator assignments on accuracy, delay, and overhead. This work explicitly captures the impact of the partitioning layers and client-to-aggregator assignments on accuracy, delay and overhead by formulating a joint optimization problem. We prove that the problem is NP-hard and propose the first accuracy-aware heuristic algorithm that explicitly accounts for model accuracy, while remaining delay-efficient. Simulation results on public datasets show that our approach can improve accuracy by 3%, while reducing delay by 20% and overhead by 50%, compared to state-of-the-art SFL and HSFL schemes.
Authors: Azul Garza, Ren\'ee Rosillo, Rodrigo Mendoza-Smith, David Salinas, Andrew Robert Williams, Arjun Ashok, Mononito Goswami, Jos\'e Mart\'in Ju\'arez
Abstract: Recent advances in time-series forecasting increasingly rely on pre-trained foundation-style models. While these models often claim broad generalization, existing evaluation protocols provide limited evidence. Indeed, most current benchmarks use static train-test splits that can easily lead to contamination as foundation models can inadvertently train on test data or perform model selection using test scores, which can inflate performance. We introduce Impermanent, a live benchmark that evaluates forecasting models under open-world temporal change by scoring forecasts sequentially over time on continuously updated data streams, enabling the study of temporal robustness, distributional shift, and performance stability rather than one-off accuracy on a frozen test set. Impermanent is instantiated on GitHub open-source activity, providing a naturally live and highly non-stationary dataset shaped by releases, shifting contributor behavior, platform/tooling changes, and external events. We focus on the top 400 repositories by star count and construct time series from issues opened, pull requests opened, push events, and new stargazers, evaluated over a rolling window with daily updates, alongside standardized protocols and leaderboards for reproducible, ongoing comparison. By shifting evaluation from static accuracy to sustained performance, Impermanent takes a concrete step toward assessing when and whether foundation-level generalization in time-series forecasting can be meaningfully claimed. Code and a live dashboard are available at https://github.com/TimeCopilot/impermanent and https://impermanent.timecopilot.dev.
URLs: https://github.com/TimeCopilot/impermanent, https://impermanent.timecopilot.dev.
Authors: Fei Wang, Jiangnan Yang, Junjie Chen, Yuxin Liu, Kun Li, Yanyan Wei, Dan Guo, Meng Wang
Abstract: Web-based platforms are becoming a primary channel for psychological support, yet most LLM-driven chatbots remain opaque, single-stage, and weakly grounded in established therapeutic practice, limiting their usefulness for web applications that promote digital well-being. To address this gap, we present \textbf{XInsight}, a counseling-inspired multi-agent framework that models psychological support as a stage-consistent workflow aligned with the classical \textit{Exploration-Insight-Action} paradigm. Building on structured client representations, XInsight orchestrates specialized agents under a unified \textit{Reason-Intervene-Reflect} cycle: an Exploration agent organizes background and concerns into a structured Case Conceptualization Form, a Routing agent performs Adaptive Therapeutic Routing (ATR) across SFBT, CBT, and MBCT, a unified Therapeutic agent executes school-consistent submodules, and a Consolidation agent guides review, skill integration, and relapse-prevention planning. A Recording agent continuously transforms open-ended web dialogues into standardized psychological artifacts, including case formulations, therapeutic records, and relapse-prevention plans, enhancing interpretability, continuity, and accountability. To support rigorous and transparent assessment, we introduce \textbf{XInsight-Bench} with a Scale-Guided LLM Evaluation (SGLE) protocol that combines therapy-specific clinical scales with general counseling criteria. Experiments show improved paradigm alignment, multi-therapy integration, interaction depth, and interpretability over existing multi-agent counseling systems, indicating that XInsight provides a practical blueprint for integrating counseling-inspired support agents into web applications for digital well-being.
Authors: Hailing Cheng, Yafang Yang, Hemeng Tao, Fengyu Zhang
Abstract: Model calibration and debiasing are fundamental to the reliability and fairness of large scale recommendation systems. We introduce the Isotonic Layer, a novel, differentiable framework that integrates piecewise linear fitting directly into neural architectures. By partitioning the feature space into discrete segments and optimizing non negative slopes via a constrained dot product mechanism, we enforce a global monotonic inductive bias. This ensures model outputs remain logically consistent with critical features such as latent relevance, recency, or quality scores. We further generalize this architecture by parameterizing segment wise slopes as learnable embeddings. This enables the model to adaptively capture context specific distortions, such as position based CTR bias through specialized isotonic profiles. Our approach utilizes a dual task formulation that decouples the recommendation objective into latent relevance estimation and bias aware calibration. A major contribution of this work is the ability to perform highly granular, customized calibration for arbitrary combinations of context features, a level of control difficult to achieve with traditional non parametric methods. We also extend this to Multi Task Learning environments with dedicated embeddings for distinct objectives. Extensive empirical evaluations on real world datasets and production AB tests demonstrate that the Isotonic Layer effectively mitigates systematic bias and enhances calibration fidelity, significantly outperforming production baselines in both predictive accuracy and ranking consistency.
Authors: Jonas Rohweder, Subhabrata Dutta, Iryna Gurevych
Abstract: Contemporary studies have uncovered many puzzling phenomena in the neural information processing of Transformer-based language models. Building a robust, unified understanding of these phenomena requires disassembling a model within the scope of its training. While the intractable scale of pretraining corpora limits a bottom-up investigation in this direction, simplistic assumptions of the data generation process limit the expressivity and fail to explain complex patterns. In this work, we use probabilistic context-free grammars (PCFGs) to generate synthetic corpora that are faithful and computationally efficient proxies for web-scale text corpora. We investigate the emergence of three mechanistic phenomena: induction heads, function vectors, and the Hydra effect, under our designed data generation process, as well as in the checkpoints of real-world language models. Our findings suggest that hierarchical structures in the data generation process serve as the X-factor in explaining the emergence of these phenomena. We provide the theoretical underpinnings of the role played by hierarchy in the training dynamics of language models. In a nutshell, our work is the first of its kind to provide a unified explanation behind the emergence of seemingly unrelated mechanistic phenomena in LLMs, augmented with efficient synthetic tooling for future interpretability research.
Authors: Nikita Sorokin, Ivan Sedykh, Valentin Malykh
Abstract: Retrieval-augmented code generation often conditions the decoder on large retrieved code snippets. This ties online inference cost to repository size and introduces noise from long contexts. We present Hierarchical Embedding Fusion (HEF), a two-stage approach to repository representation for code completion. First, an offline cache compresses repository chunks into a reusable hierarchy of dense vectors using a small fuser model. Second, an online interface maps a small number of retrieved vectors into learned pseudo-tokens that are consumed by the code generator. This replaces thousands of retrieved tokens with a fixed pseudo-token budget while preserving access to repository-level information. On RepoBench and RepoEval, HEF with a 1.8B-parameter pipeline achieves exact-match accuracy comparable to snippet-based retrieval baselines, while operating at sub-second median latency on a single A100 GPU. Compared to graph-based and iterative retrieval systems in our experimental setup, HEF reduces median end-to-end latency by 13 to 26 times. We also introduce a utility-weighted likelihood signal for filtering training contexts and report ablation studies on pseudo-token budget, embedding models, and robustness to harmful retrieval. Overall, these results indicate that hierarchical dense caching is an effective mechanism for low-latency, repository-aware code completion.
Authors: Siyuan Wang, Lei Lei, Pranav Maheshwari, Sam Bellefeuille, Kan Zheng, Dusit Niyato
Abstract: Multi-agent deep reinforcement learning (DRL) has emerged as a promising approach for radio resource allocation (RRA) in cellular vehicle-to-everything (C-V2X) networks. However, the multifaceted challenges inherent to multi-agent reinforcement learning (MARL) - including non-stationarity, coordination difficulty, large action spaces, partial observability, and limited robustness and generalization - are often intertwined, making it difficult to understand their individual impact on performance in vehicular environments. Moreover, existing studies typically rely on different baseline MARL algorithms, and a systematic comparison of their capabilities in addressing specific challenges in C-V2X RRA remains lacking. In this paper, we bridge this gap by formulating C-V2X RRA as a sequence of multi-agent interference games with progressively increasing complexity, each designed to isolate a key MARL challenge. Based on these formulations, we construct a suite of learning tasks that enable controlled evaluation of performance degradation attributable to each challenge. We further develop large-scale, diverse training and testing datasets using SUMO-generated highway traces to capture a wide range of vehicular topologies and corresponding interference patterns. Through extensive benchmarking of representative MARL algorithms, we identify policy robustness and generalization across diverse vehicular topologies as the dominant challenge in C-V2X RRA. We further show that, on the most challenging task, the best-performing actor-critic method outperforms the value-based approach by 42%. By emphasizing the need for zero-shot policy transfer to both seen and unseen topologies at runtime, and by open-sourcing the code, datasets, and interference-game benchmark suite, this work provides a systematic and reproducible foundation for evaluating and advancing MARL algorithms in vehicular networks.
Authors: Sourav Panda, Shreyash Kale, Tanmay Ambadkar, Abhinav Verma, Jonathan Dodge
Abstract: The research community lacks a middle ground between StarCraft IIs full game and its mini-games. The full-games sprawling state-action space renders reward signals sparse and noisy, but in mini-games simple agents saturate performance. This complexity gap hinders steady curriculum design and prevents researchers from experimenting with modern Reinforcement Learning algorithms in RTS environments under realistic compute budgets. To fill this gap, we present the Two-Bridge Map Suite, the first entry in an open-source benchmark series we purposely engineered as an intermediate benchmark to sit between these extremes. By disabling economy mechanics such as resource collection, base building, and fog-of-war, the environment isolates two core tactical skills: long-range navigation and micro-combat. Preliminary experiments show that agents learn coherent maneuvering and engagement behaviors without imposing full-game computational costs. Two-Bridge is released as a lightweight, Gym-compatible wrapper on top of PySC2, with maps, wrappers, and reference scripts fully open-sourced to encourage broad adoption as a standard benchmark.
Authors: Sanjay Srinivasan
Abstract: Fecal-contaminated water causes diseases and even death. Current microbial water safety tests require pathogen incubation, taking 24-72 hours and costing \$20-\$50 per test. This paper presents a solution (DeepScope) exceeding UNICEF's ideal Target Product Profile requirements for presence/absence testing, with an estimated per-test cost of \$0.44. By eliminating the need for pathogen incubation, DeepScope reduces testing time by over 98\%. In DeepScope, a dataset of microscope images of bacteria and water samples was assembled. An innovative augmentation technique, generating up to 21 trillion images from a single microscope image, was developed. Four convolutional neural network models were developed using transfer learning and regularization techniques, then evaluated on a field-test dataset comprising 100,000 microscope images of unseen, real-world water samples collected from fourteen different water sources across Sammamish, WA. Precision-recall analysis showed the DeepScope model achieves 93\% accuracy, with precision of 90\% and recall exceeding 94\%. The DeepScope model was deployed on a web server, and mobile applications for Android and iOS were developed, enabling Internet-based or smartphone-based water safety testing, with results obtained in seconds.
Authors: Fali Wang, Chenglin Weng, Xianren Zhang, Siyuan Hong, Hui Liu, Suhang Wang
Abstract: The growing demand for automated graph algorithm reasoning has attracted increasing attention in the large language model (LLM) community. Recent LLM-based graph reasoning methods typically decouple task descriptions from graph data, generate executable code augmented by retrieval from technical documentation, and refine the code through debugging. However, we identify two key limitations in existing approaches: (i) they treat technical documentation as flat text collections and ignore its hierarchical structure, leading to noisy retrieval that degrades code generation quality; and (ii) their debugging mechanisms focus primarily on runtime errors, yet ignore more critical logical errors. To address them, we propose {\method}, an \textit{agentic hierarchical retrieval-augmented coding framework} that exploits the document hierarchy through top-down traversal and early pruning, together with a \textit{self-debugging coding agent} that iteratively refines code using automatically generated small-scale test cases. To enable comprehensive evaluation of complex graph reasoning, we introduce a new dataset, {\dataset}, covering small-scale, large-scale, and composite graph reasoning tasks. Extensive experiments demonstrate that our method achieves higher task accuracy and lower inference cost compared to baselines\footnote{The code is available at \href{https://github.com/FairyFali/GraphSkill}{\textcolor{blue}{https://github.com/FairyFali/GraphSkill}}.}.
URLs: https://github.com/FairyFali/GraphSkill, https://github.com/FairyFali/GraphSkill
Authors: Madjid Sadallah
Abstract: Location-based recommender systems have achieved considerable sophistication, yet none provides a formal, lattice-theoretic representation of prerequisite dependencies among points of interest -- the semantic reality that meaningfully experiencing certain locations presupposes contextual knowledge gained from others -- nor the structural guarantees that such a representation entails. We introduce Exploration Space Theory (EST), a formal framework that transposes Knowledge Space Theory into location-based recommendation. We prove that the valid user exploration states -- the order ideals of a surmise partial order on points of interest -- form a finite distributive lattice and a well-graded learning space; Birkhoff's representation theorem, combined with the structural isomorphism between lattices of order ideals and concept lattices, connects the exploration space canonically to Formal Concept Analysis. These structural results yield four direct consequences: linear-time fringe computation, a validity certificate guaranteeing that every fringe-guided recommendation is a structurally sound next step, sub-path optimality for dynamic-programming path generation, and provably existing structural explanations for every recommendation. Building on these foundations, we specify the Exploration Space Recommender System (ESRS) -- a memoized dynamic program over the exploration lattice, a Bayesian state estimator with beam approximation and EM parameter learning, an online feedback loop enforcing the downward-closure invariant, an incremental surmise-relation inference pipeline, and three cold-start strategies, the structural one being the only approach in the literature to provide a formal validity guarantee conditional on the correctness of the inferred surmise relation. All results are established through proof and illustrated on a fully traced five-POI numerical example.
Authors: Soroush Mokhtari, Muhammad Tayyab Asif, Sergiy Zubatiy
Abstract: Online grocery shopping presents unique challenges for sequential recommendations due to repetitive purchase patterns and complex item relationships within the baskets. Unlike traditional e-commerce, grocery recommendations must capture both complementary item associations and temporal dependencies across shopping sessions. To address these challenges in Amazon's online grocery business, we propose T-REX, a novel transformer architecture that generates personalized category-level suggestions by learning both short-term basket dependencies and long-term user preferences. Our approach introduces three key innovations: (1) an efficient sampling strategy utilizing dynamic sequence splitting for sparse shopping patterns, (2) an adaptive positional encoding scheme for temporal patterns, and (3) a category-level modeling approach that reduces dimensionality while maintaining recommendation quality. Although masked language modeling techniques like BERT4Rec excel at capturing item relations, they prove less suitable for next basket generation due to information leakage issues. In contrast, T-REX's causal masking approach better aligns with the sequential nature of basket generation, enabling more accurate next-basket predictions. Experiments on large-scale grocery offline data and online A/B tests show significant improvement over existing systems.
Authors: Heng Zhang
Abstract: Robust perception in brains is often attributed to high-dimensional population activity together with local plasticity mechanisms that reinforce recurring structure. In contrast, most modern image recognition systems are trained by error backpropagation and end-to-end gradient optimization, which are not naturally aligned with local computation and local plasticity. We introduce RECAP (Reservoir Computing with Hebbian Co-Activation Prototypes), a bio-inspired learning strategy for robust image classification that couples untrained reservoir dynamics with a self-organizing Hebbian prototype readout. RECAP discretizes time-averaged reservoir responses into activation levels, constructs a co-activation mask over reservoir unit pairs, and incrementally updates class-wise prototype matrices via a Hebbian-like potentiation-decay rule. Inference is performed by overlap-based prototype matching. The method avoids error backpropagation and is naturally compatible with online prototype updates. We illustrate the resulting robustness behavior on MNIST-C, where RECAP remains robust under diverse corruptions without exposure to corrupted training samples.
Authors: Ci Zhang, Zhaojun Ding, Chence Yang, Jun Liu, Xiaoming Zhai, Shaoyi Huang, Beiwen Li, Xiaolong Ma, Jin Lu, Geng Yuan
Abstract: Pruning-based unlearning has recently emerged as a fast, training-free, and data-independent approach to remove undesired concepts from diffusion models. It promises high efficiency and robustness, offering an attractive alternative to traditional fine-tuning or editing-based unlearning. However, in this paper we uncover a hidden danger behind this promising paradigm. We find that the locations of pruned weights, typically set to zero during unlearning, can act as side-channel signals that leak critical information about the erased concepts. To verify this vulnerability, we design a novel attack framework capable of reviving erased concepts from pruned diffusion models in a fully data-free and training-free manner. Our experiments confirm that pruning-based unlearning is not inherently secure, as erased concepts can be effectively revived without any additional data or retraining. Extensive experiments on diffusion-based unlearning based on concept related weights lead to the conclusion: once the critical concept-related weights in diffusion models are identified, our method can effectively recover the original concept regardless of how the weights are manipulated. Finally, we explore potential defense strategies and advocate safer pruning mechanisms that conceal pruning locations while preserving unlearning effectiveness, providing practical insights for designing more secure pruning-based unlearning frameworks.
Authors: Yanjun Ji, Zhao-Yun Chen, Marco Roth, David A. Kreplin, Christian Schiffer, Martin King, Oliver Anton, M. Sahnawaz Alam, Markus Krutzik, Dennis Willsch, Ludwig Mathey, Frank K. Wilhelm, Guo-Ping Guo
Abstract: Quantum deep learning (QDL) explores the use of both quantum and quantum-inspired resources to determine when deep learning's core capabilities, such as expressivity, generalization, and scalability, can be enhanced based on specific resource constraints. Distinct from broader quantum machine learning, QDL emphasizes compositional depth at the pipeline level and the integration of quantum or quantum-inspired components within end-to-end workflows. This review provides an operational definition of QDL and introduces a taxonomy comprising four primary paradigms: hybrid quantum-classical models, quantum deep neural networks, quantum algorithms for deep learning primitives, and quantum-inspired classical algorithms. Theoretical principles are connected to advanced architectures, software toolchains, and experimental demonstrations across superconducting, trapped-ion, photonic, semiconductor spin, and neutral-atom systems, as well as quantum annealers. Claims of quantum advantage are critically assessed by distinguishing provable complexity-theoretic separations from empirical observations. The analysis characterizes trade-offs between model expressivity, trainability, and classical simulability, while systematically detailing the bottlenecks imposed by optimization landscapes, input-output access models, and hardware constraints. Applications are surveyed in domains encompassing image classification, natural language processing, scientific discovery, quantum data processing, and quantum optimal control, underscoring fair benchmarking against optimized classical counterparts and a comprehensive assessment of resource requirements. This review serves as a tutorial entry point for graduate students while guiding readers to specialized literature. It concludes with a verification-aware roadmap to transition QDL from near-term demonstrations to scalable and fault-tolerant implementations.
Authors: Hassan Wasswa, Hussein Abbass, Timothy Lynar
Abstract: The increasing incidence of IoT-based botnet attacks has driven interest in advanced learning models for detection. Recent efforts have focused on leveraging attention mechanisms to model long-range feature dependencies and Graph Neural Networks (GNNs) to capture relationships between data instances. Since GNNs require graph-structured input, tabular NetFlow data must be transformed accordingly. This study evaluates how the choice of the method for constructing the graph-structured dataset impacts the classification performance of a GNN model. Five methods--k-Nearest Neighbors, Mutual Nearest Neighbors, Shared Nearest Neighbor, Gabriel Graph, and epsilon-radius Graph--were evaluated in this research. To reduce the computational burden associated with high-dimensional data, a Variational Autoencoder (VAE) is employed to project the original features into a lower-dimensional latent space prior to graph generation. Subsequently, a Graph Attention Network (GAT) is trained on each graph to classify traffic in the N-BaIoT dataset into three categories: Normal, Mirai, and Gafgyt. The results indicate that using Gabriel graph achieves the highest detection performance with an accuracy of 97.56% while SNN recorded the lowest performance with an accuracy as low as 78.56%.
Authors: Kejing Lu, Zhenpeng Pan, Jianbin Qin, Yoshiharu Ishikawa, Chuan Xiao
Abstract: Approximate Nearest Neighbor Search (ANNS) is fundamental to modern AI applications. Most existing solutions optimize query efficiency but fail to align with the practical requirements of modern workloads. In this paper, we outline six critical demands of modern AI applications: high query efficiency, fast indexing, low memory footprint, scalability to high dimensionality, robustness across varying retrieval sizes, and support for online insertions. To satisfy all these demands, we introduce Projection-Augmented Graph (PAG), a new ANNS framework that integrates projection techniques into a graph index. PAG reduces unnecessary exact distance computations through asymmetric comparisons between exact and approximate distances as guided by projection-based statistical tests. Three key components are designed and unified to the graph index to optimize indexing and searching. Experiments on six modern datasets demonstrate that PAG consistently achieves superior query per second (QPS)-recall performance -- up to 5x faster than HNSW -- while offering fast indexing speed and moderate memory footprint. PAG remains robust as dimensionality and retrieval size increase and naturally supports online insertions.
Authors: Bikram De, Habib Irani, Vangelis Metsis
Abstract: Data augmentation is a crucial technique for training robust deep learning models for human motion, where annotated datasets are often scarce. However, generic augmentation methods often ignore the underlying geometric and kinematic constraints of the human body, risking the generation of unrealistic motion patterns that can degrade model performance. Furthermore, the conventional approach of training a single generalist model on a dataset expanded with a mixture of all available transformations does not fully exploit the unique learning signals provided by each distinct augmentation type. We challenge this convention by introducing a novel training paradigm, EnsAug, that strategically uses augmentation to foster model diversity within an ensemble. Our method involves training an ensemble of specialists, where each model learns from the original dataset augmented by only a single, distinct geometric transformation. Experiments on sign language and human activity recognition benchmarks demonstrate that our diversified ensemble methodology significantly outperforms the standard practice of training one model on a combined augmented dataset and achieves state-of-the-art accuracy on two sign language and one human activity recognition dataset while offering greater modularity and efficiency. Our primary contribution is the empirical validation of this training strategy, establishing an effective baseline for leveraging data augmentation in skeletal motion analysis.
Authors: Toan Nguyen, Yang Liu, Celso De Melo, Flora D. Salim
Abstract: Continual VideoQA with multimodal LLMs is hindered by interference between tasks and the prohibitive cost of storing task-specific prompts. We introduce HyperTokens, a transformer-based token generator that produces fine-tuning tokens on demand, giving explicit control over prompt updates while keeping memory fixed. To suppress forgetting, we propose meta-inspired regularisers that look ahead to avoid task-specific sharp directions and anchor the evolving generator to prior tasks. We further connect our objective to sharpness-aware optimisation, providing insight into why it encourages flatter cross-task minima and improves retention. Beyond regularisation, HyperTokens exploits lightweight auxiliary multimodal supervision through shared generation weights; guided by a causal perspective, we design feasible objectives and surrogate mutual-information losses to regularise anti-causal cross-modal directions. Across two standard continual VideoQA benchmarks, HyperTokens achieves higher average accuracy with substantially lower forgetting. Finally, we introduce a challenging cross-modal ImageQA->VideoQA protocol and show that HyperTokens enables robust continual transfer in this setting.
Authors: Shivam Pande, Nicolas Nadisic, Francisco Mederos-Henry, Aleksandra Pizurica
Abstract: Spectroscopic imaging (SI) has become central to heritage science because it enables non-invasive, spatially resolved characterisation of materials in artefacts. In particular, attenuated total reflection Fourier transform infrared microscopy (ATR-$\mu$FTIR) is widely used to analyse painting cross-sections, where a spectrum is recorded at each pixel to form a hyperspectral image (HSI). Interpreting these data is difficult: spectra are often mixtures of several species in heterogeneous, multi-layered and degraded samples, and current practice still relies heavily on manual comparison with reference libraries. This workflow is slow, subjective and hard to scale. We propose an unsupervised CNN autoencoder for blind unmixing of ATR-$\mu$FTIR HSIs, estimating endmember spectra and their abundance maps while exploiting local spatial structure through patch-based modelling. To reduce sensitivity to atmospheric and acquisition artefacts across $>1500$ bands, we introduce a weighted spectral angle distance (WSAD) loss with automatic band-reliability weights derived from robust measures of spatial flatness, neighbour agreement and spectral roughness. Compared with standard SAD training, WSAD improves interpretability in contamination-prone spectral regions. We demonstrate the method on an ATR-$\mu$FTIR cross-section from the Ghent Altarpiece attributed to the Van Eyck brothers.
Authors: Vishak K Bhat, Eric A. F. Reinhardt, Sergei Gleyzer
Abstract: Due to a high rate of overall data generation relative to data generation of interest, the CMS experiment at the Large Hadron Collider uses a combination of hardware- and software-based triggers to select data for capture. Accurate momentum calculation is crucial for improving the efficiency of the CMS trigger systems, enabling better classification of low- and high- momentum particles and reducing false triggers. This paper explores the use of Graph Neural Networks (GNNs) for the momentum estimation task. We present two graph construction methods and apply a GNN model to leverage the inherent graph structure of the data. In this paper firstly, we show that the GNN outperforms traditional models like TabNet in terms of Mean Absolute Error (MAE), demonstrating its effectiveness in capturing complex dependencies within the data. Secondly we show that the dimension of the node feature is crucial for the efficiency of GNN.
Authors: Diana Susan Joseph, Pranav M Pawar, Raja Muthalagu, Mithun Mukharjee
Abstract: Performing a timely and accurate identification of crop diseases is vital to maintain agricultural productivity and food security. The current work presents a hybrid few-shot learning model that integrates Explainable Artificial Intelligence (XAI) and Few-Shot Learning (FSL) to address the challenge of identifying and classifying the stages of disease of the diseases of maize, rice, and wheat leaves under limited annotated data conditions. The proposed model integrates Siamese and Prototypical Networks within an episodic training paradigm to effectively learn discriminative disease features from a few examples. To ensure model transparency and trustworthiness, Gradient-weighted Class Activation Mapping (Grad-CAM) is employed for visualizing key decision regions in the leaf images, offering interpretable insights into the classification process. Experimental evaluations on custom few-shot datasets developed in the study prove that the model consistently achieves high accuracy, precision, recall, and F1-scores, frequently exceeding 92% across various disease stages. Comparative analyses against baseline FSL models further confirm the superior performance and explainability of the proposed approach. The framework offers a promising solution for real-world, data-constrained agricultural disease monitoring applications.
Authors: Jiajin Tang, Gaoyang, Wenjie Wang, Sibei Yang, Xing Chen
Abstract: With the rapid advancement of data science, charts have evolved from simple numerical presentation tools to essential instruments for insight discovery and decision-making support. However, current chart data intelligence exhibits significant limitations in deep research capabilities, with existing methods predominantly addressing shallow tasks such as visual recognition or factual question-answering, rather than the complex reasoning and high-level data analysis that deep research requires. This limitation stems from two primary technical bottlenecks: at the training level, existing post-training techniques exhibit deficiencies in handling multi-dimensional reward signal interference and heterogeneous data gradient conflicts, preventing models from achieving balanced development across multiple capability dimensions; at the evaluation level, current methods remain limited to factual retrieval and basic computation, failing to assess end-to-end analytic reasoning and other deep research capabilities. To address the training challenge, we propose PRPO, which performs parallel optimization across reward dimensions and capability partitioning across data types, effectively disentangling conflicts between heterogeneous data and multi-dimensional reward signals while ensuring optimization stability. For the evaluation challenge, we construct MCDR-Bench based on the ``error uniqueness principle," transforming subjective generation assessment into objective error identification through controllable error injection, enabling quantifiable evaluation of deep research capabilities. Experimental validation confirms that the proposed PRPO and MCDR-Bench jointly establish a unified framework that systematically advances chart deep research through enhanced collaborative training and objective evaluation.
Authors: Francis Osswald (IPHC), Mohammed Chahbaoui (UNISTRA), Xinyi Liang (SU)
Abstract: Image reconstruction in the presence of severe degradation remains a challenging inverse problem, particularly in beam diagnostics for high-energy physics accelerators. As modern facilities demand precise detection of beam halo structures to control losses, traditional analysis tools have reached their performance limits. This work reviews existing image-processing techniques for data cleaning, contour extraction, and emittance reconstruction, and introduces a novel approach based on convolutional filtering and neural networks with optimized early-stopping strategies in order to control overfitting. Despite the absence of training datasets, the proposed unsupervised framework achieves robust denoising and high-fidelity reconstruction of beam emittance images under low signal-to-noise conditions. The method extends measurable amplitudes beyond seven standard deviations, enabling unprecedented halo resolution.
Authors: Joohyung Lee, Changhun Kim, Hyunsu Kim, Kwanhyung Lee, Juho Lee
Abstract: Self-supervised learning (SSL) typically learns representations invariant to semantic-preserving augmentations. While effective for recognition, enforcing strong invariance can suppress transformation-dependent structure that is useful for robustness to geometric perturbations and spatially sensitive transfer. A growing body of work, therefore, augments invariance-based SSL with equivariance objectives, but these objectives are often imposed on the same final representation. We empirically observe a trade-off in this coupled setting: pushing equivariance regularization toward deeper layers improves equivariance scores but degrades ImageNet-1k linear evaluation, motivating a layer-decoupled design. Motivated by this trade-off, we propose Soft Equivariance Regularization (SER), a plug-in regularizer that decouples where invariance and equivariance are enforced: we keep the base SSL objective unchanged on the final embedding, while softly encouraging equivariance on an intermediate spatial token map via analytically specified group actions $\rho_g$ applied directly in feature space. SER learns/predicts no per-sample transformation codes/labels, requires no auxiliary transformation-prediction head, and adds only 1.008x training FLOPs. On ImageNet-1k ViT-S/16 pretraining, SER improves MoCo-v3 by +0.84 Top-1 in linear evaluation under a strictly matched 2-view setting and consistently improves DINO and Barlow Twins; under matched view counts, SER achieves the best ImageNet-1k linear-eval Top-1 among the compared invariance+equivariance add-ons. SER further improves ImageNet-C/P by +1.11/+1.22 Top-1 and frozen-backbone COCO detection by +1.7 mAP. Finally, applying the same layer-decoupling recipe to existing invariance+equivariance baselinesimproves their accuracy, suggesting layer decoupling as a general design principle for combining invariance and equivariance.
Authors: Gongjie Zhang, Wenhao Li, Quanhao Qian, Jiuniu Wang, Deli Zhao, Shijian Lu, Ran Xu
Abstract: Multimodal Large Language Models (MLLMs) that directly process RGB inputs for tasks like 3D localization and navigation have shown remarkable potential. However, we argue that these RGB-only approaches are fundamentally flawed in their ability to generalize across cameras. By ignoring camera parameters, they entangle an object's physical properties with the camera's perspective, creating an irresolvable ambiguity. We show this leads MLLMs to overfit to the training camera distribution, rather than learning true and generalizable 3D geometric principles. To address this, we propose Camera-Aware MLLM framework for spatial MLLMs. It learns generalizable spatial reasoning by: (i) injecting camera intrinsics via a dense embedding that conditions each visual token; (ii) introducing a camera-aware data augmentation strategy that synthetically varies camera parameters, forcing the model to disentangle camera properties from scene content; and (iii) distilling geometric priors from a 3D vision foundation model. Extensive experiments demonstrate that camera-aware MLLMs substantially outperform their naive counterparts, particularly in cross-camera generalization tests on spatially-grounded tasks, indicating that camera-awareness is not only beneficial but also a prerequisite for robust and generalizable spatial intelligence in MLLMs.
Authors: Jinsu Hong, Chetraj Pandey, Berkay Aydin
Abstract: Current solar flare predictions often lack precise quantification of their reliability, resulting in frequent false alarms, particularly when dealing with datasets skewed towards extreme events. To improve the trustworthiness of space weather forecasting, it is crucial to establish confidence intervals for model predictions. Conformal prediction, a machine learning framework, presents a promising avenue for this purpose by constructing prediction intervals that ensure valid coverage in finite samples without making assumptions about the underlying data distribution. In this study, we explore the application of conformal prediction to regression tasks in space weather forecasting. Specifically, we implement full-disk solar flare prediction using images created from magnetic field maps and adapt four pre-trained deep learning models to incorporate three distinct methods for constructing confidence intervals: conformal prediction, quantile regression, and conformalized quantile regression. Our experiments demonstrate that conformalized quantile regression achieves higher coverage rates and more favorable average interval lengths compared to alternative methods, underscoring its effectiveness in enhancing the reliability of solar weather forecasting models.
Authors: Uday Bondhugula, Akshay Baviskar, Navdeep Katel, Vimal Patel, Anoop JS, Arnab Dutta
Abstract: We present the design and implementation of PolyBlocks, a modular and reusable MLIR-based compiler infrastructure for AI programming frameworks and AI chips. PolyBlocks is based on pass pipelines that compose transformations on loop nests and SSA, primarily relying on lightweight affine access analysis; the transformations are stitched together in specialized ways to realize high-performance code automatically by the use of analytical cost models and heuristics. The optimizations in these passes include multi-level tiling, fusion, on-chip scratchpad usage, mapping matmuls and convolutions to matrix units, fusing the attention layer, and several other transformations for parallelism and locality. They have been developed in a way that makes it easy to build PolyBlocks-based compilers to target new chips, reusing much of the infrastructure. PolyBlocks' design and architecture enable fully automatic code generation from high-level frameworks to low-level target-specific intrinsics. Experimental results from evaluating PolyBlocks-powered just-in-time compilation for PyTorch and JAX targeting NVIDIA GPUs show that it is able to match or outperform Torch Inductor and XLA in several cases, although the latter rely on a combination of vendor libraries and code generation. For individual operators like matmuls and convolutions, PolyBlocks-generated code is competitive with the best vendor-tuned libraries or hand-written kernels.
Authors: Srikumar Nayak
Abstract: Credit risk scoring must support high-stakes lending decisions where data distributions change over time, probability estimates must be reliable, and group-level fairness is required. While modern machine learning models improve default prediction accuracy, they often produce poorly calibrated scores under distribution shift and may create unfair outcomes when trained without explicit constraints. This paper proposes Calibrated Credit Intelligence (CCI), a deployment-oriented framework that combines (i) a Bayesian neural risk scorer to capture epistemic uncertainty and reduce overconfident errors, (ii) a fairnessconstrained gradient boosting model to control group disparities while preserving strong tabular performance, and (iii) a shiftaware fusion strategy followed by post-hoc probability calibration to stabilize decision thresholds in later time periods. We evaluate CCI on the Home Credit Credit Risk Model Stability benchmark using a time-consistent split to reflect real-world drift. Compared with strong baselines (LightGBM, XGBoost, CatBoost, TabNet, and a standalone Bayesian neural model), CCI achieves the best overall trade-off between discrimination, calibration, stability, and fairness. In particular, CCI reaches an AUC-ROC of 0.912 and an AUC-PR of 0.438, improves operational performance with Recall@1%FPR = 0.509, and reduces calibration error (Brier score 0.087, ECE 0.015). Under temporal shift, CCI shows a smaller AUC-PR drop from early to late periods (0.017), and it lowers group disparities (demographic parity gap 0.046, equal opportunity gap 0.037) compared to unconstrained boosting. These results indicate that CCI produces risk scores that are accurate, reliable, and more equitable under realistic deployment conditions.
Authors: Jinhong Wang, Matei C. Ignuta-Ciuncanu, Ricardo F. Martinez-Botas, Teng Cao
Abstract: Solving flow through porous media is a crucial step in the topology optimisation of cold plates, a key component in modern thermal management. Traditional computational fluid dynamics (CFD) methods, while accurate, are often prohibitively expensive for large and complex geometries. In contrast, data-driven surrogate models provide a computationally efficient alternative, enabling rapid and reliable predictions. In this study, we develop a machine-learning framework for predicting steady-state flow through porous media governed by the Navier-Stokes-Brinkman equations. We implement and compare three model architectures-convolutional autoencoder (AE), U-Net, and Fourier Neural Operator (FNO)-evaluating their predictive performance. To enhance physics consistency, we incorporate physics-informed loss functions. Our results demonstrate that FNO outperforms AE and U-Net, achieving a mean squared error (MSE) as low as 0.0017 while providing speedups of up to 1000 times compared to CFD. Additionally, the mesh-invariant property of FNO emphasizes its suitability for topology optimisation tasks, where varying mesh resolutions are required. This study highlights the potential of machine learning to accelerate fluid flow predictions in porous media, offering a scalable alternative to traditional numerical methods.
Authors: Hsiang Hsu, Eric Lei, Chun-Fu Chen
Abstract: Inference-time alignment effectively steers large language models (LLMs) by generating multiple candidates from a reference model and selecting among them with an imperfect reward model. However, current strategies face a fundamental dilemma: ``optimistic'' approaches like Best-of-$N$ suffer from reward hacking, while ``pessimistic'' regularized methods often stifle the exploration needed to discover high-quality responses. In this work, we formalize this trade-off through the lens of regret minimization, demonstrating that the optimal strategy depends critically on the tail behavior of the reward distribution. We show theoretically that light-tailed regimes favor optimism to unearth high-quality outliers, whereas heavy-tailed regimes require pessimism to guard against reward mis-calibration in the extremes. Guided by this insight, we introduce Best-of-Tails (BoT), an adaptive inference-time alignment framework that uses Tsallis divergence as a tunable regularizer to provide a finer granularity of interpolation between these extremes. BoT uses the Hill estimator to characterize reward-tail heaviness on a per-prompt basis and dynamically adjusts its selection rule to balance exploration gains against alignment error. Across math, multiple-choice reasoning, and human-preference evaluations, BoT improves alignment performance across a range of reference and reward model configurations relative to fixed-strategy baselines.
Authors: Luben M. C. Cabezas, Sabina J. Sloman, Bruno M. Resende, Fanyi Wu, Michele Caprio, Rafael Izbicki
Abstract: Conformal prediction delivers prediction intervals with distribution-free coverage, but its intervals can look overconfident in regions where the model is extrapolating, because standard conformal scores do not explicitly represent epistemic uncertainty. Credal methods, by contrast, make epistemic effects visible by working with sets of plausible predictive distributions, but they are typically model-based and lack calibration guarantees. We introduce CREDO, a simple "credal-then-conformalize" recipe that combines both strengths. CREDO first builds an interpretable credal envelope that widens when local evidence is weak, then applies split conformal calibration on top of this envelope to guarantee marginal coverage without further assumptions. This separation of roles yields prediction intervals that are interpretable: their width can be decomposed into aleatoric noise, epistemic inflation, and a distribution-free calibration slack. We provide a fast implementation based on trimming extreme posterior predictive endpoints, prove validity, and show on benchmark regressions that CREDO maintains target coverage while improving sparsity adaptivity at competitive efficiency.
Authors: Hangyi Zhao
Abstract: We study contextual bilateral trade under full feedback when trader valuations have bounded density but infinite variance. We first extend the self-bounding property of Bachoc et al. (ICML 2025) from bounded to real-valued valuations, showing that the expected regret of any price $\pi$ satisfies $\mathbb{E}[g(m,V,W) - g(\pi,V,W)] \le L|m-\pi|^2$ under bounded density alone. Combining this with truncated-mean estimation, we prove that an epoch-based algorithm achieves regret $\widetilde{O}(T^{1-2\beta(p-1)/(\beta p + d(p-1))})$ when the noise has finite $p$-th moment for $p \in (1,2)$ and the market value function is $\beta$-H\"older, and we establish a matching $\Omega(\cdot)$ lower bound via Assouad's method with a smoothed moment-matching construction. Our results characterize the exact minimax rate for this problem, interpolating between the classical nonparametric rate at $p=2$ and the trivial linear rate as $p \to 1^+$.
Authors: Mirza Samad Ahmed Baig, Syeda Anshrah Gillani
Abstract: Discovering compact governing equations from experimental observations is one of the defining objectives of quantitative science, yet practical discovery pipelines routinely fail when measurements are noisy, relevant state variables are unobserved, or multiple symbolic structures explain the data equally well within statistical uncertainty. Here we introduce SymLang (Symmetry-constrained Language-guided equation discovery), a unified framework that brings together three previously separate ideas: (i) typed symmetry-constrained grammars that encode dimensional analysis, group-theoretic invariance, and parity constraints as hard production rules, eliminating on average 71.3% of candidate expression trees before any fitting; (ii) language-model-guided program synthesis in which a fine-tuned 7B-parameter proposer, conditioned on interpretable data descriptors, efficiently navigates the constrained search space; and (iii) MDL-regularized Bayesian model selection coupled with block-bootstrap stability analysis that quantifies structural uncertainty rather than committing to a single best equation. Across 133 dynamical systems spanning classical mechanics, electrodynamics, thermodynamics, population dynamics, and nonlinear oscillators, SymLang achieves an exact structural recovery rate of 83.7% under 10% observational noise - a 22.4 percentage-point improvement over the next-best baseline - while reducing out-of-distribution extrapolation error by 61% and near-eliminating conservation-law violations (3.1 x 10-3 vs. 187.3 x 10-3 physical drift for the closest competitor). In all tested regimes the framework correctly identifies structural degeneracy, reporting it explicitly rather than returning a confidently wrong single equation. The framework is fully open-source and reproducible, providing a principled pathway from raw data to interpretable, physically auditable symbolic laws.
Authors: Yi Yang, Xiangyu Chang, Pei-yu Chen
Abstract: As machine learning (ML) systems increasingly shape access to credit, jobs, and other opportunities, the fairness of algorithmic decisions has become a central concern. Yet it remains unclear when enforcing fairness constraints in these systems genuinely improves outcomes for affected groups or instead leads to "leveling down," making one or both groups worse off. We address this question in a unified, population-level (Bayes) framework for binary classification under prevalent group fairness notions. Our Bayes approach is distribution-free and algorithm-agnostic, isolating the intrinsic effect of fairness requirements from finite-sample noise and from training and intervention specifics. We analyze two deployment regimes for ML classifiers under common legal and governance constraints: attribute-aware decision-making (sensitive attributes available at decision time) and attribute-blind decision-making (sensitive attributes excluded from prediction). We show that, in the attribute-aware regime, fair ML necessarily (weakly) improves outcomes for the disadvantaged group and (weakly) worsens outcomes for the advantaged group. In contrast, in the attribute-blind regime, the impact of fairness is distribution-dependent: fairness can benefit or harm either group and may shift both groups' outcomes in the same direction, leading to either leveling up or leveling down. We characterize the conditions under which these patterns arise and highlight the role of "masked" candidates in driving them. Overall, our results provide structural guidance on when pursuing algorithmic fairness is likely to improve group outcomes and when it risks systemic leveling down, informing fair ML design and deployment choices.
Authors: Moin Amin-Naseri, Hannah Kim, Estevam Hruschka
Abstract: The extraction of structured information from raw text is a fundamental component of many NLP applications, including document retrieval, ranking, and relevance estimation. High-quality extractions often require domain-specific accuracy, up-to-date understanding of specialized taxonomies, and the ability to incorporate emerging jargon and rare outliers. In many domains--such as medical, legal, and HR--the extraction model must also adapt to shifting terminology and benefit from explicit reasoning over structured knowledge. We propose DySECT, a Dynamic Self-Evolving Extraction and Curation Toolkit, which continually improves as it is used. The system incrementally populates a versatile, self-expanding knowledge base (KB) with triples extracted by the LLM. The KB further enriches itself through the integration of probabilistic knowledge and graph-based reasoning, gradually accumulating domain concepts and relationships. The enriched KB then feeds back into the LLM extractor via prompt tuning, sampling of relevant few-shot examples, or fine-tuning using KB-derived synthetic data. As a result, the system forms a symbiotic closed-loop cycle in which extraction continuously improves knowledge, and knowledge continuously improves extraction.
Authors: Bojan Deraji\'c, Sebastian Bernhard, Wolfgang H\"onig
Abstract: Safe navigation of autonomous robots remains one of the core challenges in the field, especially in dynamic and uncertain environments. One of the prevalent approaches is safety filtering based on control barrier functions (CBFs), which are easy to deploy but difficult to design. Motivated by the shortcomings of existing learning- and model-based methods, we propose a simple yet effective neural CBF design method for safe robot navigation in dynamic environments. We employ the idea of a composite CBF, where multiple neural CBFs are combined into a single CBF. The individual CBFs are trained via the Hamilton-Jacobi reachability framework to approximate the optimal safe set for single moving obstacles. Additionally, we use the residual neural architecture, which guarantees that the estimated safe set does not intersect with the corresponding failure set. The method is extensively evaluated in simulation experiments for a ground robot and a quadrotor, comparing it against several baseline methods. The results show improved success rates of up to 18\% compared to the best baseline, without increasing the conservativeness of the motion. Also, the method is demonstrated in hardware experiments for both types of robots.
Authors: Sofiane Ouaari, Jules Kreuer, Nico Pfeifer
Abstract: DNA foundation models have become transformative tools in bioinformatics and healthcare applications. Trained on vast genomic datasets, these models can be used to generate sequence embeddings, dense vector representations that capture complex genomic information. These embeddings are increasingly being shared via Embeddings-as-a-Service (EaaS) frameworks to facilitate downstream tasks, while supposedly protecting the privacy of the underlying raw sequences. However, as this practice becomes more prevalent, the security of these representations is being called into question. This study evaluates the resilience of DNA foundation models to model inversion attacks, whereby adversaries attempt to reconstruct sensitive training data from model outputs. In our study, the model's output for reconstructing the DNA sequence is a zero-shot embedding, which is then fed to a decoder. We evaluated the privacy of three DNA foundation models: DNABERT-2, Evo 2, and Nucleotide Transformer v2 (NTv2). Our results show that per-token embeddings allow near-perfect sequence reconstruction across all models. For mean-pooled embeddings, reconstruction quality degrades as sequence length increases, though it remains substantially above random baselines. Evo 2 and NTv2 prove to be most vulnerable, especially for shorter sequences with reconstruction similarities > 90%, while DNABERT-2's BPE tokenization provides the greatest resilience. We found that the correlation between embedding similarity and sequence similarity was a key predictor of reconstruction success. Our findings emphasize the urgent need for privacy-aware design in genomic foundation models prior to their widespread deployment in EaaS settings. Training code, model weights and evaluation pipeline are released on: https://github.com/not-a-feature/DNA-Embedding-Inversion.
URLs: https://github.com/not-a-feature/DNA-Embedding-Inversion.
Authors: Alireza Mousavi-Hosseini, Murat A. Erdogdu
Abstract: We study post-training linear autoregressive models with outcome and process rewards. Given a context $\boldsymbol{x}$, the model must predict the response $\boldsymbol{y} \in Y^N$, a sequence of length $N$ that satisfies a $\gamma$ margin condition, an extension of the standard separability to sequences. We prove that on test samples where the base model achieves a non-trivial likelihood $\alpha$, a variant of policy gradient (PG) can achieve likelihood $1 - \varepsilon$ with an essentially minimax optimal number of reward queries $\tilde{O}((\alpha^{-1} + \varepsilon^{-1})/\gamma^2)$. However, a barrier arises for going beyond the support of the base model. We prove that the overall expected error after post-training with outcome rewards is governed by a property of the base model called the Likelihood Quantile (LQ), and that variants of PG, while minimax optimal, may require a number of reward queries exponential in $N$ to go beyond this support, regardless of the pre-training algorithm. To overcome this barrier, we study post-training with a process reward model, and demonstrate how PG variants in this setting avoid the curse of dimensionality in $N$ via dependence on a token-level LQ. Along the way, we prove that under the margin condition, SGD with adaptive learning rate (LR) achieves a near optimal test error for statistical learning, and PG with adaptive LR achieves a near optimal number of mistakes for online learning while being computationally efficient whenever possible, both of which may be of independent interest.
Authors: Nanhong Liu, Jingyi Yan, Mucun Sun, Jie Zhang
Abstract: In practical data-driven applications on electrical equipment fault diagnosis, training data can be poisoned by sensor failures, which can severely degrade the performance of machine learning (ML) models. However, once the ML model has been trained, removing the influence of such harmful data is challenging, as full retraining is both computationally intensive and time-consuming. To address this challenge, this paper proposes a SISA (Sharded, Isolated, Sliced, and Aggregated)-based machine unlearning (MU) framework for power transformer inter-turn short-circuit fault (ITSCF) localization. The SISA method partitions the training data into shards and slices, ensuring that the influence of each data point is isolated within specific constituent models through independent training. When poisoned data are detected, only the affected shards are retrained, avoiding retraining the entire model from scratch. Experiments on simulated ITSCF conditions demonstrate that the proposed framework achieves almost identical diagnostic accuracy to full retraining, while reducing retraining time significantly.
Authors: Roshni Anna Jacob, Prithvi Poddar, Jaidev Goel, Souma Chowdhury, Yulia R. Gel, Jie Zhang
Abstract: Extreme weather events and cyberattacks can cause component failures and disrupt the operation of power distribution networks (DNs), during which reconfiguration and load shedding are often adopted for resilience enhancement. This study introduces a topology-aware graph reinforcement learning (RL) framework for outage management that embeds higher-order topological features of the DN into a graph-based RL model, enabling reconfiguration and load shedding to maximize energy supply while maintaining operational stability. Results on the modified IEEE 123-bus feeder across 300 diverse outage scenarios demonstrate that incorporating the topological data analysis (TDA) tool, persistence homology (PH), yields 9-18% higher cumulative rewards, up to 6% increase in power delivery, and 6-8% fewer voltage violations compared to a baseline graph-RL model. These findings highlight the potential of integrating RL with TDA to enable self-healing in DNs, facilitating fast, adaptive, and automated restoration.
Authors: Zou Yang, Sophia Xiao, Bijan Mazaheri
Abstract: Recent work has proposed powerful frameworks, rooted in causal theory, to quantify fairness. Causal inference has primarily emphasized the detection of \emph{average} treatment effects (ATEs), and subsequent notions of fairness have inherited this focus. In this paper, we build on previous concerns about regulation based on averages. In particular, we formulate the "causal masking problem" as a linear program that optimizes an alternative objective, such as maximizing profit or minimizing crime, while retaining a zero ATE (i.e., the ATE between a protected attribute and a decision). By studying the capabilities and limitations of causal masking, we show that optimization under ATE-based regulation may induce significant unequal treatment. We demonstrate that the divergence between true and causally masked fairness is driven by confounding, underscoring the importance of full conditional-independence testing when assessing fairness. Finally, we discuss statistical and information-theoretic limitations that make causally masked solutions very difficult to detect, allowing them to persist for long periods. These results argue that we must regulate fairness at the model-level, rather than at the decision level.
Authors: Kosuke Yoshimura, Hisashi Kashima
Abstract: In predictive modeling for low-resource audio classification, extracting high-accuracy and interpretable attributes is critical. Particularly in high-reliability applications, interpretable audio attributes are indispensable. While human-driven attribute discovery is effective, its low throughput becomes a bottleneck. We propose a method for adaptively discovering interpretable audio attributes using Multimodal Large Language Models (MLLMs). By replacing humans in the AdaFlock framework with MLLMs, our method achieves significantly faster attribute discovery. Our method dynamically identifies salient acoustic characteristics via prompting and constructs an attribute-based ensemble classifier. Experimental results across various audio tasks demonstrate that our method outperforms direct MLLM prediction in the majority of evaluated cases. The entire training completes within 11 minutes, proving it a practical, adaptive solution that surpasses conventional human-reliant approaches.
Authors: Punyajoy Saha, Sudipta Halder, Debjyoti Mondal, Subhadarshi Panda
Abstract: Safety alignment is critical for deploying large language models (LLMs) in real-world applications, yet most existing approaches rely on large human-annotated datasets and static red-teaming benchmarks that are costly, difficult to scale, and slow to adapt to evolving model behaviors. Moreover, overly conservative safety mechanisms can reduce model usefulness by rejecting sensitive but legitimate queries. We introduce Self-MOA (Self Multi-Objective Alignment), a fully automated framework for aligning small language models using weak supervision from automated evaluator models. Self-MOA operates as a closed loop that dynamically generates model-specific red team prompts, constructs preference data from model-generated responses, and aligns models via multi-objective preference optimization to jointly optimize for safety and helpfulness. Across multiple small language models and safety benchmarks, Self-MOA achieves a 12.41\% improvement in safety while preserving helpfulness, using as little as 11 times less training data than human-supervised alignment baselines. These results demonstrate that adaptive, automated alignment can reduce the dependence on static, human-curated safety pipelines in resource-constrained settings.
Authors: Harsh Parikh, Gabriel Levin-Konigsberg, Dominique Perrault-Joncas, Alexander Volfovsky
Abstract: Treatment effects estimated from randomized controlled trials are local not only to the study population but also to the time at which the trial was conducted. We develop a framework for temporal transportation: extrapolating treatment effects to time periods where no experiment was conducted. We target the transported average treatment effect (TATE) and show that under a separable temporal effects assumption, the TATE decomposes into an observed average treatment effect and a temporal ratio. We provide two identification strategies -- one using replicated trials comparing the same treatments at different times, another using common treatment arms observed across time -- and develop doubly robust, semiparametrically efficient estimators for each. Monte Carlo simulations confirm that both estimators achieve nominal coverage, with the common arm strategy yielding substantial efficiency gains when its stronger assumptions hold. We apply our methods to A/B tests from the Upworthy Research Archive, demonstrating that the two strategies exhibit a variance-bias tradeoff: the common arm approach offers greater precision but may incur bias when treatments interact heterogeneously with temporal factors.
Authors: Hanlong Li, Karishma Kamalahasan, Jiahui Li, Kazuhiro Nakadai, Shreyas Kousik
Abstract: We present Artoo, a learned acoustic communication system for robots that replaces hand-designed signal processing with end-to-end co-trained neural networks. Our system pairs a lightweight text-to-speech (TTS) transmitter (1.18M parameters) with a conformer-based automatic speech recognition (ASR) receiver (938K parameters), jointly optimized through a differentiable channel. Unlike human speech, robot-to-robot communication is paralinguistics-free: the system need not preserve timbre, prosody, or naturalness, only maximize decoding accuracy under channel distortion. Through a three-phase co-training curriculum, the TTS transmitter learns to produce distortion-robust acoustic encodings that surpass the baseline under noise, achieving 8.3% CER at 0 dB SNR. The entire system requires only 2.1M parameters (8.4 MB) and runs in under 13 ms end-to-end on a CPU, making it suitable for deployment on resource-constrained robotic platforms.
Authors: Zihao Zheng, Zhihao Mao, Xingyue Zhou, Jiayu Chen, Maoliang Li, Xinhao Sun, Hailong Zou, Zhaobo Zhang, Xuanzhe Liu, Donggang Cao, Hong Mei, Xiang Chen
Abstract: Vision-and-Language Navigation (VLN) increasingly relies on large vision-language models, but their inference cost conflicts with real-time deployment. Token caching is a promising training-free strategy that avoids redundant computation by reusing stable visual tokens across frames. However, existing methods assume a static camera and fixed semantic focus, assumptions that VLN fundamentally violates. We identify two failure modes: (1) visual dynamics, where viewpoint shift displaces token positions across frames, causing position-wise matching to pair misaligned content; (2) semantic dynamics, where token relevance shifts across task stages as navigation progresses, making cached states stale. We propose VLN-Cache, a visual-dynamic-aware and semantic-dynamic-aware caching framework that introduces view-aligned remapping to recover geometric correspondences and a task-relevance saliency filter to veto reuse at semantic transitions. A layer-adaptive entropy policy further balances the per-layer reuse budget. Experiments on the R2R-CE simulation benchmark show up to 1.52x speedup while maintaining competitive navigation success rates.
Authors: Rihan Aaron D'Silva, Hiroyasu Tsukamoto
Abstract: This paper presents novel method for distribution-free robust trajectory optimization and control of discrete-time, nonlinear, and non-Gaussian stochastic systems, with closed-loop guarantees on chance constraint satisfaction. Our framework employs conformal inference to generate coverage-based confidence sets for the closed-loop dynamics around arbitrary reference trajectories, by constructing a joint nonconformity score to quantify both the validity of contraction (i.e., incremental stability) conditions and the impact of external stochastic disturbance on the closed-loop dynamics, without any distributional assumptions. Via appropriate constraint tightening, chance constraints can be reformulated into tractable, statistically valid deterministic constraints on the reference trajectories. This enables a formal pathway to leverage and validate learning-based motion planners and controllers, such as those with neural contraction metrics, in safety-critical real-world applications. Notably, our statistical guarantees are non-diverging and can be computed with finite samples of the underlying uncertainty, without overly conservative structural priors. We demonstrate our approach in motion planning problems for designing safe, dynamically feasible trajectories in both numerical simulation and hardware experiments.
Authors: Rajdeep Pathak, Tanujit Chakraborty
Abstract: Accurate and reliable forecasting of epidemic incidences is critical for public health preparedness, yet it remains a challenging task due to complex nonlinear temporal dependencies and heterogeneous spatial interactions. Often, point forecasts generated by spatiotemporal models are unreliable in assigning uncertainty to future epidemic events. Probabilistic forecasting of epidemics is therefore crucial for providing the best or worst-case scenarios rather than a simple, often inaccurate, point estimate. We present deep spatiotemporal engression methods to generate accurate and reliable probabilistic forecasts on low-frequency epidemic datasets. The proposed methods act as distributional lenses, and out-of-sample probabilistic forecasts are generated by sampling from the trained models. Our frameworks encapsulate lightweight deep generative architectures, wherein uncertainty is quantified endogenously, driven by a pre-additive noise component during model construction. We establish geometric ergodicity and asymptotic stationarity of the spatiotemporal engression processes under mild assumptions on the network weights and pre-additive noise process. Comprehensive evaluations across six epidemiological datasets over three forecast horizons demonstrate that the proposal consistently outperforms several temporal and spatiotemporal benchmarks in both point and probabilistic forecasting. Additionally, we explore the explainability of the proposal to enhance the models' practical application for informed, timely public health interventions.
Authors: Zahra Mansour, Verena Uslar, Dirk Weyhe, Danilo Hollosi, Nils Strodthoff
Abstract: Bowel sounds (BS) are typically momentary and have low amplitude, making them difficult to detect accurately through manual auscultation. This leads to significant variability in clinical assessment. Digital acoustic sensors allow the acquisition of high-quality BS and enable automated signal analysis, offering the potential to provide clinicians with both objective and quantitative feedback on bowel activity. This study presents an automated pipeline for bowel sound segmentation and classification using a wearable acoustic SonicGuard sensor. BS signals from 83 subjects were recorded using a SonicGuard sensor. Data from 40 subjects were manually annotated by clinical experts and used to train an automatic annotation algorithm, while the remaining subjects were used for further model evaluation. An energy-based event detection algorithm was developed to detect BS events. Detected sound segments were then classified into BS patterns using a pretrained Audio Spectrogram Transformer (AST) model. Model performance was evaluated separately for healthy individuals and patients. The best configuration used two specialized models, one trained on healthy subjects and one on patients, achieving (accuracy: 0.97, AUROC: 0.98) for healthy group and (accuracy: 0.96, AUROC: 0.98) for patient group. The auto-annotation method reduced manual labeling time by approximately 70%, and expert review showed that less than 12% of automatically detected segments required correction. The proposed automated segmentation and classification system enables quantitative assessment of bowel activity, providing clinicians with an objective diagnostic tool that may improve the diagnostic of gastrointestinal function and support the annotation of large-scale datasets.
Authors: Xiaoyi Wang, Long Feng, Zhaojun Wang
Abstract: Intergenerational mobility quantifies the transmission of socio-economic outcomes from parents to children. While rank-rank regression (RRR) is standard, adding covariates directly (RRRX) often yields parameters with unclear interpretation. Conditional rank-rank regression (CRRR) resolves this by using covariate-adjusted (conditional) ranks to measure within-group mobility. We improve and extend CRRR by estimating conditional ranks with a deep conditional transformation model (DCTM) and cross-fitting, enabling end-to-end conditional distribution learning with structural constraints and strong performance under nonlinearity, high-order interactions, and discrete ordered outcomes where the distributional regression used in traditional CRRR may be cumbersome or prone to misconfiguration. We further extend CRRR to discrete outcomes via an $\omega$-indexed conditional-rank definition and study sensitivity to $\omega$. For continuous outcomes, we establish an asymptotic theory for the proposed estimators and verify the validity of exchangeable bootstrap inference. Simulations across simple/complex continuous and discrete ordered designs show clear accuracy gains in challenging settings. Finally, we apply our method to two empirical studies, revealing substantial within-group persistence in U.S. income and pronounced gender differences in educational mobility in India.
Authors: Abbas Mammadov, So Takao, Bohan Chen, Ricardo Baptista, Morteza Mardani, Yee Whye Teh, Julius Berner
Abstract: Flow maps enable high-quality image generation in a single forward pass. However, unlike iterative diffusion models, their lack of an explicit sampling trajectory impedes incorporating external constraints for conditional generation and solving inverse problems. We put forth Variational Flow Maps, a framework for conditional sampling that shifts the perspective of conditioning from "guiding a sampling path", to that of "learning the proper initial noise". Specifically, given an observation, we seek to learn a noise adapter model that outputs a noise distribution, so that after mapping to the data space via flow map, the samples respect the observation and data prior. To this end, we develop a principled variational objective that jointly trains the noise adapter and the flow map, improving noise-data alignment, such that sampling from complex data posterior is achieved with a simple adapter. Experiments on various inverse problems show that VFMs produce well-calibrated conditional samples in a single (or few) steps. For ImageNet, VFM attains competitive fidelity while accelerating the sampling by orders of magnitude compared to alternative iterative diffusion/flow models. Code is available at https://github.com/abbasmammadov/VFM
Authors: Yatharth Sharma
Abstract: We propose a Vocos-based bandwidth extension model that enhances audio at 8-48 kHz by generating missing high-frequency content. Inputs are resampled to 48 kHz and processed by a neural vocoder backbone, enabling a single network to support arbitrary upsampling ratios. A lightweight Linkwitz-Riley-inspired refiner merges the original low band with the generated high frequencies via a smooth crossover. On validation, the model achieves competitive log-spectral distance while running at a real-time factor of 0.0001 on an NVIDIA A100 GPU and 0.0053 on an 8-core CPU, demonstrating practical, high-quality BWE at extreme throughput.
Authors: Duy M. H. Nguyen, Tuan A. Tran, Duong Nguyen, Siwei Xie, Trung Q. Nguyen, Mai T. N. Truong, Daniel Palenicek, An T. Le, Michael Barz, TrungTin Nguyen, Tuan Dam, Ngan Le, Minh Vu, Khoa Doan, Vien Ngo, Pengtao Xie, James Zou, Daniel Sonntag, Jan Peters, Mathias Niepert
Abstract: Recent token merging techniques for Vision Transformers (ViTs) provide substantial speedups by reducing the number of tokens processed by self-attention, often without retraining. However, their direct application to the Segment Anything Model (SAM) family is nontrivial: SAM's image encoder mixes windowed and global attention, and its mask decoder relies on dense, prompt-conditioned features for precise boundary prediction. We systematically evaluate representative token-merging methods on SAM and Medical SAM in a strict off-the-shelf setting, and find that existing destination-selection heuristics can erode boundaries and leak prompt information as merge rates increase. We propose \textbf{StructSAM}, a resolution-preserving merge-unmerge framework tailored to SAM. StructSAM computes a lightweight token-energy score from first-order feature gradients, uses grid-based flatness screening to protect boundary and prompt regions, and merges tokens within flat areas toward low-energy destinations with explicit token recovery. We further provide a spectral graph coarsening view showing that score-guided merging yields bounded Laplacian spectral distortion compared to random or window-restricted baselines. Across eight natural and medical benchmarks, StructSAM reduces encoder FLOPs by 25-30\% (up to 40\%+ with prompt-aware merging) with minor drops in mIoU/Dice, consistently outperforming ToMe, PiToMe, ToMeSD, VidToMe, and ALGM at the same compute.
Authors: Vincent Conitzer
Abstract: One common concern about advanced artificial intelligence is that it will prevent us from turning it off, as that would interfere with pursuing its goals. In this paper, we discuss an unorthodox proposal for addressing this concern: give the AI a (primary) goal of being turned off (see also papers by Martin et al., and by Goldstein and Robinson). We also discuss whether and under what conditions this would be a good idea.
Authors: Vojtech Halenka, Mohammadreza Amini, Per-Arne Andersen, Ole-Christoffer Granmo, Burak Kantarci
Abstract: All applications in fifth-generation (5G) networks rely on stable radio-frequency (RF) environments to support mission-critical services in mobility, automation, and connected intelligence. Their exposure to intentional interference or low-power jamming threatens availability and reliability, especially when such attacks remain below link-layer observability. This paper investigates lightweight, explainable, and hardware-efficient jamming detection using the Convolutional Tsetlin Machine (CTM) operating directly on 5G Synchronization Signal Block (SSB) features. CTM formulates Boolean logic clauses over quantized inputs, enabling bit-level inference and deterministic deployment on FPGA fabrics. These properties make CTM well suited for real-time, resource-constrained edge environments anticipated in 5G. The proposed approach is experimentally validated on a real 5G testbed using over-the-air SSB data, emulating practical downlink conditions. We benchmark CTM against a convolutional neural network (CNN) baseline under identical preprocessing and training pipelines. On the real dataset, CTM achieves comparable detection performance (Accuracy 91.53 +/- 1.01 vs. 96.83 +/- 1.19 for CNN) while training $9.5\times$ faster and requiring 14x less memory (45~MB vs.\ 624~MB). Furthermore, we outline a compact FPGA-oriented design for Zybo~Z7 (Zynq-7000) and provide resource projections (not measured) under three deployment profiles optimized for latency, power, and accuracy trade-offs. The results show that the CTM provides a practical, interpretable, and resource-efficient alternative to conventional DNNs for RF-domain jamming detection, establishing it as a strong candidate for edge-deployed, low-latency, and security-critical 5G applications while laying the groundwork for B5G systems.
Authors: Seth Nabarro, Mark van der Wilk, Andrew J. Davison
Abstract: We propose DistGP: a multi-robot learning method for collaborative learning of a global function using only local experience and computation. We utilise a sparse Gaussian process (GP) model with a factorisation that mirrors the multi-robot structure of the task, and admits distributed training via Gaussian belief propagation (GBP). Our loopy model outperforms Tree-Structured GPs \cite{bui2014tree} and can be trained online and in settings with dynamic connectivity. We show that such distributed, asynchronous training can reach the same performance as a centralised, batch-trained model, albeit with slower convergence. Last, we compare to DiNNO \cite{yu2022dinno}, a distributed neural network (NN) optimiser, and find DistGP achieves superior accuracy, is more robust to sparse communication and is better able to learn continually.
Authors: Mohammed Brahimi, Karim Laabassi, Mohamed Seghir Hadj Ameur, Aicha Boutorh, Badia Siab-Farsi, Amin Khouani, Omar Farouk Zouak, Seif Eddine Bouziane, Kheira Lakhdari, Abdelkader Nabil Benghanem
Abstract: Machine learning models in agricultural vision often achieve high accuracy on curated datasets but fail to generalize under real field conditions due to distribution shifts between training and deployment environments. Moreover, most machine learning competitions focus primarily on model design while treating datasets as fixed resources, leaving the role of data collection practices in model generalization largely unexplored. We introduce the AgrI Challenge, a data-centric competition framework in which multiple teams independently collect field datasets, producing a heterogeneous multi-source benchmark that reflects realistic variability in acquisition conditions. To systematically evaluate cross-domain generalization across independently collected datasets, we propose Cross-Team Validation (CTV), an evaluation paradigm that treats each team's dataset as a distinct domain. CTV includes two complementary protocols: Train-on-One-Team-Only (TOTO), which measures single-source generalization, and Leave-One-Team-Out (LOTO), which evaluates collaborative multi-source training. Experiments reveal substantial generalization gaps under single-source training: models achieve near-perfect validation accuracy yet exhibit validation-test gaps of up to 16.20% (DenseNet121) and 11.37% (Swin Transformer) when evaluated on datasets collected by other teams. In contrast, collaborative multi-source training dramatically improves robustness, reducing the gap to 2.82% and 1.78%, respectively. The challenge also produced a publicly available dataset of 50,673 field images of six tree species collected by twelve independent teams, providing a diverse benchmark for studying domain shift and data-centric learning in agricultural vision.
Authors: Namrata Patil Gurav, Akashdeep Ranu, Archchana Sindhujan, Diptesh Kanojia
Abstract: Quality Estimation (QE) is essential for assessing machine translation quality in reference-less settings, particularly for domain-specific and low-resource language scenarios. In this paper, we investigate sentence-level QE for English to Indic machine translation across four domains (Healthcare, Legal, Tourism, and General) and five language pairs. We systematically compare zero-shot, few-shot, and guideline-anchored prompting across selected closed-weight and open-weight LLMs. Findings indicate that while closed-weight models achieve strong performance via prompting alone, prompt-only approaches remain fragile for open-weight models, especially in high-risk domains. To address this, we adopt ALOPE, a framework for LLM-based QE that uses Low-Rank Adaptation with regression heads attached to selected intermediate Transformer layers. We also extend ALOPE with recently proposed Low-Rank Multiplicative Adaptation (LoRMA). Our results show that intermediate-layer adaptation consistently improves QE performance, with gains in semantically complex domains, indicating a path toward more robust QE in practical scenarios. We release code and domain-specific QE datasets publicly to support further research.
Authors: Li Gu, Zihuan Jiang, Zhixiang Chi, Huan Liu, Ziqiang Wang, Yuanhao Yu, Glen Berseth, Yang Wang
Abstract: Graphical user interface (GUI)-based mobile agents automate digital tasks on mobile devices by interpreting natural-language instructions and interacting with the screen. While recent methods apply reinforcement learning (RL) to train vision-language-model(VLM) agents in interactive environments with a primary focus on performance, generalization remains underexplored due to the lack of standardized benchmarks and open-source RL systems. In this work, we formalize the problem as a Contextual Markov Decision Process (CMDP) and introduce \textbf{AndroidWorld-Generalization}, a benchmark with three increasingly challenging regimes for evaluating zero-shot generalization to unseen task instances, templates, and applications. We further propose an RL training system that integrates Group Relative Policy Optimization (GRPO) with a scalable rollout collection system, consisting of containerized infrastructure and asynchronous execution % , and error recovery to support reliable and efficient training. Experiments on AndroidWorld-Generalization show that RL enables a 7B-parameter VLM agent to surpass supervised fine-tuning baselines, yielding a 26.1\% improvement on unseen instances but only limited gains on unseen templates (15.7\%) and apps (8.3\%), underscoring the challenges of generalization. As a preliminary step, we demonstrate that few-shot adaptation at test-time improves performance on unseen apps, motivating future research in this direction. To support reproducibility and fair comparison, we open-source the full RL training system, including the environment, task suite, models, prompt configurations, and the underlying infrastructure \footnote{https://github.com/zihuanjiang/AndroidWorld-Generalization}.
URLs: https://github.com/zihuanjiang/AndroidWorld-Generalization
Authors: Guoli Wang, Haonan Shi, Tu Ouyang, An Wang
Abstract: Large language models (LLMs) often require fine-tuning (FT) to perform well on downstream tasks, but FT can induce safety-alignment drift even when the training dataset contains only benign data. Prior work shows that introducing a small fraction of harmful data can substantially compromise LLM refusal behavior, causing LLMs to comply with harmful requests. Existing defense methods often rely on model-wide interventions, such as restricting which parameters are updated or injecting additional safety data, which can limit generality and degrade downstream task performance. To address these limitations, we propose a fine-tuning framework called Preserving Safety Alignment via Constrained Tokens (PACT), which stabilizes the model's confidence on safety tokens. Our approach is motivated by the empirical observation that safety-aligned behavior is reflected in the model's token-level output confidence and is often concentrated on a small subset of safety-related tokens. During downstream fine-tuning, we regularize the fine-tuned model to match the aligned reference model's confidence on safety-related tokens at each response step, while leaving non-safety tokens largely unconstrained to allow effective task adaptation. This targeted constraint prevents alignment drift without imposing global restrictions that typically trade off with model utility.
Authors: Xiang Zhang, Hongming Xu, Le Zhou, Wei Zhou, Xuanhe Zhou, Guoliang Li, Yuyu Luo, Changdong Liu, Guorun Chen, Jiang Liao, Fan Wu
Abstract: Enterprises commonly deploy heterogeneous database systems, each of which owns a distinct SQL dialect with different syntax rules, built-in functions, and execution constraints. However, most existing NL2SQL methods assume a single dialect (e.g., SQLite) and struggle to produce queries that are both semantically correct and executable on target engines. Prompt-based approaches tightly couple intent reasoning with dialect syntax, rule-based translators often degrade native operators into generic constructs, and multi-dialect fine-tuning suffers from cross-dialect interference. In this paper, we present Dial, a knowledge-grounded framework for dialect-specific NL2SQL. Dial introduces: (1) a Dialect-Aware Logical Query Planning module that converts natural language into a dialect-aware logical query plan via operator-level intent decomposition and divergence-aware specification; (2) HINT-KB, a hierarchical intent-aware knowledge base that organizes dialect knowledge into (i) a canonical syntax reference, (ii) a declarative function repository, and (iii) a procedural constraint repository; and (3) an execution-driven debugging and semantic verification loop that separates syntactic recovery from logic auditing to prevent semantic drift. We construct DS-NL2SQL, a benchmark covering six major database systems with 2,218 dialect-specific test cases. Experimental results show that Dial consistently improves translation accuracy by 10.25% and dialect feature coverage by 15.77% over state-of-the-art baselines. The code is at https://github.com/weAIDB/Dial.
Authors: Mohammad Saeid, Amir Salarpour, Pedram MohajerAnsari, Mert D. Pes\'e
Abstract: We present SLNet, a lightweight backbone for 3D point cloud recognition designed to achieve strong performance without the computational cost of many recent attention, graph, and deep MLP based models. The model is built on two simple ideas: NAPE (Nonparametric Adaptive Point Embedding), which captures spatial structure using a combination of Gaussian RBF and cosine bases with input adaptive bandwidth and blending, and GMU (Geometric Modulation Unit), a per channel affine modulator that adds only 2D learnable parameters. These components are used within a four stage hierarchical encoder with FPS+kNN grouping, nonparametric normalization, and shared residual MLPs. In experiments, SLNet shows that a very small model can still remain highly competitive across several 3D recognition tasks. On ModelNet40, SLNet-S with 0.14M parameters and 0.31 GFLOPs achieves 93.64% overall accuracy, outperforming PointMLP-elite with 5x fewer parameters, while SLNet-M with 0.55M parameters and 1.22 GFLOPs reaches 93.92%, exceeding PointMLP with 24x fewer parameters. On ScanObjectNN, SLNet-M achieves 84.25% overall accuracy within 1.2 percentage points of PointMLP while using 28x fewer parameters. For large scale scene segmentation, SLNet-T extends the backbone with local Point Transformer attention and reaches 58.2% mIoU on S3DIS Area 5 with only 2.5M parameters, more than 17x fewer than Point Transformer V3. We also introduce NetScore+, which extends NetScore by incorporating latency and peak memory so that efficiency can be evaluated in a more deployment oriented way. Across multiple benchmarks and hardware settings, SLNet delivers a strong overall balance between accuracy and efficiency. Code is available at: https://github.com/m-saeid/SLNet.
Authors: J. Clayton Kerce, Alexis Fox
Abstract: Standard transformers entangle all computation in a single residual stream, obscuring which components perform which functions. We introduce the Dual-Stream Transformer, which decomposes the residual stream into two functionally distinct components: a token stream updated by attention and a context stream updated by feed-forward networks. Information flow between attention heads is controlled through a hierarchy of mixing strategies, from fully independent (maximum interpretability) to dense (standard transformer behavior). This design exposes a tunable tradeoff between interpretability and performance. We measure this tradeoff on language modeling tasks at 29M parameters. Fully independent head mixing increases validation loss by 8\% relative to dense baselines. The recommended Kronecker mixing strategy, which permits scalar communication between heads while preserving within-head structure, costs only 2.5\%. All configurations maintain functional generation under attention amplification (scaling logits by factors up to 16 at inference time), with degradation ranging from 16\% to 27\%. This robustness suggests the architectures learn discrete algorithms that operate independently of soft probabilistic mixing. The architecture provides a foundation for interpretable language models where internal structure is exposed by design. \footnote{This work was partially supported by DARPA Contract HR001125C0302.}
Authors: Heng Jin, Chaoyu Zhang, Hexuan Yu, Shanghao Shi, Ning Zhang, Y. Thomas Hou, Wenjing Lou
Abstract: Cloud-based infrastructures have become the dominant platform for deploying large models, particularly large language models (LLMs). Fine-tuning and inference are increasingly delegated to cloud providers for simplified deployment and access to proprietary models, yet this creates a fundamental trust gap: although cryptographic and TEE-based verification exist, the scale of modern LLMs renders them prohibitive, leaving clients unable to practically audit these processes. This lack of transparency creates concrete security risks that can silently compromise service integrity. We present AFTUNE, an auditable and verifiable framework that ensures the computation integrity of cloud-based fine-tuning and inference. AFTUNE incorporates a lightweight recording and spot-check mechanism that produces verifiable traces of execution. These traces enable clients to later audit whether the training and inference processes followed the agreed configurations. Our evaluation shows that AFTUNE imposes practical computation overhead while enabling selective and efficient verification, demonstrating that trustworthy model services are achievable in today's cloud environments.
Authors: Qiang Liu, Lester Mackey, Chris Oates
Abstract: This monograph provides a rigorous overview of theoretical and methodological aspects of probabilistic inference and learning with Stein's method. Recipes are provided for constructing Stein discrepancies from Stein operators and Stein sets, and properties of these discrepancies such as computability, separation, convergence detection, and convergence control are discussed. Further, the connection between Stein operators and Stein variational gradient descent is set out in detail. The main definitions and results are precisely stated, and references to all proofs are provided.
Authors: Longbiao Cheng, Shih-Chii Liu
Abstract: Recent studies have shown that post-deployment adaptation can improve the robustness of speech enhancement models in unseen noise conditions. However, existing methods often incur prohibitive computational and memory costs, limiting their suitability for on-device deployment. In this work, we investigate model adaptation in realistic settings with dynamic acoustic scene changes and propose a lightweight framework that augments a frozen backbone with low-rank adapters updated via self-supervised training. Experiments on sequential scene evaluations spanning 111 environments across 37 noise types and three signal-to-noise ratio ranges, including the challenging [-8, 0] dB range, show that our method updates fewer than 1% of the base model's parameters while achieving an average 1.51 dB SI-SDR improvement within only 20 updates per scene. Compared to state-of-the-art approaches, our framework achieves competitive or superior perceptual quality with smoother and more stable convergence, demonstrating its practicality for lightweight on-device adaptation of speech enhancement models under real-world acoustic conditions.
Authors: Wenwei Li, Jiarun Zhou, Qinxiao Quan, Fusang Zhang, Daqing Zhang
Abstract: Contactless sensing using wireless communication signals has garnered significant attention due to its non-intrusive nature and ubiquitous infrastructure. Despite the promise, the inherent bistatic deployment of wireless communication introduces clock asynchronism, which leads to unknown phase offsets in channel response and hinders fine-grained sensing. State-of-the-art systems widely adopt the cross-antenna channel ratio to cancel these detrimental phase offsets. However, the channel ratio preserves sensing feature accuracy only at integer-wavelength target displacements, losing sub-wavelength fidelity. To overcome this limitation, we derive the first quantitative mapping between the distorted ratio feature and the ideal channel feature. Building on this foundation, we develop a robust framework that leverages channel response amplitude to recover the ideal channel feature from the distorted ratio. Real-world experiments across Wi-Fi and LoRa demonstrate that our method can effectively reconstruct sub-wavelength displacement details, achieving nearly an order-of-magnitude improvement in accuracy.
Authors: Young Hyun Cho, Jordan Awan
Abstract: Privacy protection and uncertainty quantification are increasingly important in data-driven decision making. Conformal prediction provides finite-sample marginal coverage, but existing private approaches often rely on data splitting, reducing the effective sample size. We propose a full-data privacy-preserving conformal prediction framework that avoids splitting. Our framework leverages stability induced by differential privacy to control the gap between in-sample and out-of-sample conformal scores, and pairs this with a conservative private quantile routine designed to prevent under-coverage. We show that a generic differential privacy guarantee yields a universal coverage floor, yet cannot generally recover the nominal $1-\alpha$ level. We then provide a refined, mechanism-specific stability analysis and yields asymptotic recovery of the nominal level. Experiments demonstrate sharper prediction sets than the split-based private baseline.
Authors: Jinzhou Tang, Fan Feng, Minghao Fu, Wenjun Lin, Biwei Huang, Keze Wang
Abstract: Learned world models excel at interpolative generalization but fail at extrapolative generalization to novel physical properties. This limitation arises because they learn statistical correlations rather than the environment's underlying generative rules, such as physical invariances and conservation laws. We argue that learning these invariances is key to robust extrapolation. To achieve this, we first introduce \textbf{Symmetry Exploration}, an unsupervised exploration strategy where an agent is intrinsically motivated by a Hamiltonian-based curiosity bonus to actively probe and challenge its understanding of conservation laws, thereby collecting physically informative data. Second, we design a Hamiltonian-based world model that learns from the collected data, using a novel self-supervised contrastive objective to identify the invariant physical state from raw, view-dependent pixel observations. Our framework, \textbf{DreamSAC}, trained on this actively curated data, significantly outperforms state-of-the-art baselines in 3D physics simulations on tasks requiring extrapolation.
Authors: Dennis Gross
Abstract: Aging bridge networks require proactive, verifiable, and interpretable maintenance strategies, yet reinforcement learning (RL) policies trained solely on reward signals provide no formal safety guarantees and remain opaque to infrastructure managers. We demonstrate COOL-MC as a tool for verifying and explaining RL policies for multi-bridge network maintenance, building on a single-bridge Markov decision process (MDP) from the literature and extending it to a parallel network of three heterogeneous bridges with a shared periodic budget constraint, encoded in the PRISM modeling language. We train an RL agent on this MDP and apply probabilistic model checking and explainability methods to the induced discrete-time Markov chain (DTMC) that arises from the interaction between the learned policy and the underlying MDP. Probabilistic model checking reveals that the trained policy has a safety-violation probability of 3.5\% over the planning horizon, being slightly above the theoretical minimum of 0\% and indicating the suboptimality of the learned policy, noting that these results are based on artificially constructed transition probabilities and deterioration rates rather than real-world data, so absolute performance figures should be interpreted with caution. The explainability analysis further reveals, for instance, a systematic bias in the trained policy toward the state of bridge 1 over the remaining bridges in the network. These results demonstrate COOL-MC's ability to provide formal, interpretable, and practical analysis of RL maintenance policies.
Authors: Niccol\`o Ferrari, Michele Fraccaroli, Evelina Lamma
Abstract: Anomaly detection is nowadays increasingly used in industrial applications and processes. One of the main fields of the appliance is the visual inspection for surface anomaly detection, which aims to spot regions that deviate from regularity and consequently identify abnormal products. Defect localization is a key task, that usually is achieved using a basic comparison between generated image and the original one, implementing some blob-analysis or image-editing algorithms, in the post-processing step, which is very biased towards the source dataset, and they are unable to generalize. Furthermore, in industrial applications, the totality of the image is not always interesting but could be one or some regions of interest (ROIs), where only in those areas there are relevant anomalies to be spotted. For these reasons, we propose a new architecture composed by two blocks. The first block is a Generative Adversarial Network (GAN), based on a residual autoencoder (ResAE), to perform reconstruction and denoising processes, while the second block produces image segmentation, spotting defects. This method learns from a dataset composed of good products and generated synthetic defects. The discriminative network is trained using a ROI for each image contained in the training dataset. The network will learn in which area anomalies are relevant. This approach guarantees the reduction of using pre-processing algorithms, formerly developed with blob-analysis and image-editing procedures. To test our model we used challenging MVTec anomaly detection datasets and an industrial large dataset of pharmaceutical BFS strips of vials. This set constitutes a more realistic use case of the aforementioned network.
Authors: Najeeb Jebreel, Mona Khalil, David S\'anchez, Josep Domingo-Ferrer
Abstract: Membership inference attacks (MIAs) have become the standard tool for evaluating privacy leakage in machine learning (ML). Among them, the Likelihood-Ratio Attack (LiRA) is widely regarded as the state of the art when sufficient shadow models are available. However, prior evaluations have often overstated the effectiveness of LiRA by attacking models overconfident on their training samples, calibrating thresholds on target data, assuming balanced membership priors, and/or overlooking attack reproducibility. We re-evaluate LiRA under a realistic protocol that (i) trains models using anti-overfitting (AOF) and transfer learning (TL), when applicable, to reduce overconfidence as in production models; (ii) calibrates decision thresholds using shadow models and data rather than target data; (iii) measures positive predictive value (PPV, or precision) under shadow-based thresholds and skewed membership priors (pi <= 10%); and (iv) quantifies per-sample membership reproducibility across different seeds and training variations. We find that AOF significantly weakens LiRA, while TL further reduces attack effectiveness while improving model accuracy. Under shadow-based thresholds and skewed priors, LiRA's PPV often drops substantially, especially under AOF or AOF+TL. We also find that thresholded vulnerable sets at extremely low FPR show poor reproducibility across runs, while likelihood-ratio rankings are more stable. These results suggest that LiRA, and likely weaker MIAs, are less effective than previously suggested under realistic conditions, and that reliable privacy auditing requires evaluation protocols that reflect practical training practices, feasible attacker assumptions, and reproducibility considerations. Code is available at https://github.com/najeebjebreel/lira_analysis.
Authors: Furkan Gen\c{c}, Onat \"Ozdemir, Emre Akba\c{s}
Abstract: Out-of-distribution (OOD) detection is critical in safety-sensitive applications. While this challenge has been addressed from various perspectives, the influence of training objectives on OOD behavior remains comparatively underexplored. In this paper, we present a systematic comparison of four widely used training objectives: Cross-Entropy Loss, Prototype Loss, Triplet Loss, and Average Precision (AP) Loss, spanning probabilistic, prototype-based, metric-learning, and ranking-based supervision, for OOD detection in image classification under standardized OpenOOD protocols. Across CIFAR-10/100 and ImageNet-200, we find that Cross-Entropy Loss, Prototype Loss, and AP Loss achieve comparable in-distribution accuracy, while Cross-Entropy Loss provides the most consistent near- and far-OOD performance overall; the other objectives can be competitive in specific settings.
Authors: Niccol\`o Ferrari, Nicola Zanarini, Michele Fraccaroli, Alice Bizzarri, Evelina Lamma
Abstract: Industrial visual inspection in pharmaceutical production requires high accuracy under strict constraints on cycle time, hardware footprint, and operational cost. Manual inline inspection is still common, but it is affected by operator variability and limited throughput. Classical rule-based computer vision pipelines are often rigid and difficult to scale to highly variable production scenarios. To address these limitations, we present a semi-supervised anomaly detection framework based on a generative adversarial architecture with a residual autoencoder and a dense bottleneck, specifically designed for online deployment on a high-speed Blow-Fill-Seal (BFS) line. The model is trained only on nominal samples and detects anomalies through reconstruction residuals, providing both classification and spatial localization via heatmaps. The training set contains 2,815,200 grayscale patches. Experiments on a real industrial test kit show high detection performance while satisfying timing constraints compatible with a 500 ms acquisition slot.
Authors: Robin Doerfler, Lonce Wyse
Abstract: Computational engine sound modeling is central to the automotive audio industry, particularly for active sound design, virtual prototyping, and emerging data-driven engine sound synthesis methods. These applications require large volumes of standardized, clean audio recordings with precisely time-aligned operating-state annotations: data that is difficult to obtain due to high costs, specialized measurement equipment requirements, and inevitable noise contamination. We present an analysis-driven framework for generating engine audio with sample-accurate control annotations. The method extracts harmonic structures from real recordings through pitch-adaptive spectral analysis, which then drive an extended parametric harmonic-plus-noise synthesizer. With this framework, we generate the Procedural Engine Sounds Dataset (19 hours, 5,935 files), a set of engine audio signals with sample-accurate RPM and torque annotations, spanning a wide range of operating conditions, signal complexities, and harmonic profiles. Comparison against real recordings validates that the synthesized data preserves characteristic harmonic structures, and baseline experiments confirm its suitability for learning-based parameter estimation and synthesis tasks. The dataset is released publicly to support research on engine timbre analysis, control parameter estimation, acoustic modeling and neural generative networks.
Authors: Chenxi Li, Xianggan Liu, Dake Shen, Yaosong Du, Zhibo Yao, Hao Jiang, Linyi Jiang, Chengwei Cao, Jingzhe Zhang, RanYi Peng, Peiling Bai, Xiande Huang
Abstract: Despite the rapid progress of Large Vision-Language Models (LVLMs), the integration of visual modalities introduces new safety vulnerabilities that adversaries can exploit to elicit biased or malicious outputs. In this paper, we demonstrate an underexplored vulnerability via semantic slot filling, where LVLMs complete missing slot values with unsafe content even when the slot types are deliberately crafted to appear benign. Building on this finding, we propose StructAttack, a simple yet effective single-query jailbreak framework under black-box settings. StructAttack decomposes a harmful query into a central topic and a set of benign-looking slot types, then embeds them as structured visual prompts (e.g., mind maps, tables, or sunburst diagrams) with small random perturbations. Paired with a completion-guided instruction, LVLMs automatically recompose the concealed semantics and generate unsafe outputs without triggering safety mechanisms. Although each slot appears benign in isolation (local benignness), StructAttack exploits LVLMs' reasoning to assemble these slots into coherent harmful semantics. Extensive experiments on multiple models and benchmarks show the efficacy of our proposed StructAttack.
Authors: Ye Tian, Aijun Liu
Abstract: Chain-of-thought (CoT) improves reasoning reliability but increases token cost, motivating post-training compression of explicit reasoning traces. However, the shortest sufficient reasoning is not universal: it depends on difficulty, model capacity, and training state, making fixed length targets brittle. In practice, naive RL-based compression can also undesirably shorten the user-facing answer, because a single completion-level learning signal leaks across the think/answer boundary. We propose Difficulty-Scaled Segment-Wise GRPO (DSS-GRPO), which decomposes returns into think and answer components, computes group-relative advantages per segment, and routes them with hard token masks so compression updates act only on think while answer alignment acts only on answer. DSS-GRPO uses prompt-wise within-group shaping and difficulty-aware scaling to encourage concise reasoning without collapsing answer behavior.
Authors: Luca M. Meyer, Majid Zamani
Abstract: Many previous works in spike sorting study spike classification and compression independently. In this paper, a novel algorithm is proposed called MetaSort to address these two problems. To deal with compression, a novel adaptive level crossing algorithm is proposed to approximate spike shapes with high fidelity. Meanwhile, the latent feature representation is used to handle the classification problem. Besides, to guarantee MetaSort is robust and discriminative, the geometric information of data is exploited simultaneously in the proposed framework by meta-transfer learning. Empirical experiments with in-vivo spike data demonstrate that MetaSort delivers promising performance, highlighting its potential and motivating continued development toward an ultra-low-power, on-chip implementation.
Authors: Hamed Hamzeh, Parisa Vahdatian
Abstract: Autoscaling in cloud-native platforms like Kubernetes is reactive and metric-driven, leading to a strategic void problem. This comes from the decoupling of higher-level business policies from lower-level resource provisioning. The strategic void, coupled with a fragmented coordination of pod and node scaling, can lead to significant resource waste and performance degradation under dynamic workloads. In this paper, we present MAS-H2, a new hierarchical multi-agent system that addresses the challenges of autonomic cloud resource management with a complete end-to-end solution. MAS-H2 systematically decomposes the control problem into three layers: a Strategic Agent that formalises business policies (e.g., cost vs. performance) into a global utility function; Planning Agents that produce a joint, proactive scaling plan for pods and nodes with time-series forecasting; and Execution Agents that execute the scaling plan. We built and tested a MAS-H2 prototype as a Kubernetes Operator on Google Kubernetes Engine (GKE) to benchmark it against the native Horizontal Pod Autoscaler (HPA) and Cluster Autoscaler (CA) baselines under two realistic, spiky, and stress-inducing workload scenarios. The results show that the MAS-H2 system maintained application CPU usage under 40% for predictable Heartbeat workloads. This resulted in over 50% less sustained CPU stress than the native HPA baseline, which typically operated above 80%. The MAS-H2 system demonstrated proactive planning in a volatile Chaotic Flash Sale scenario by filtering transient noise and deploying more replicas compared to HPA. It reduced peak CPU load by 55% without under-provisioning. Beyond performance, MAS-H2 seamlessly performed a zero-downtime strategic migration between two cost- and performance-optimised infrastructures.
Authors: Yifei Yuan, Ghaith Androwis, Xianlian Zhou
Abstract: Effective exoskeleton assistance requires co-adaptation: as the device alters joint dynamics, the user reorganizes neuromuscular coordination, creating a non-stationary learning problem. Most learning-based approaches do not explicitly account for the sequential nature of human motor adaptation, leading to training instability and poorly timed assistance. We propose Staged Multi-Agent Training (SMAT), a four-stage curriculum designed to mirror how users naturally acclimate to a wearable device. In SMAT, a musculoskeletal human actor and a bilateral hip exoskeleton actor are trained progressively: the human first learns unassisted gait, then adapts to the added device mass; the exoskeleton subsequently learns a positive assistance pattern against a stabilized human policy, and finally both agents co-adapt with full torque capacity and bidirectional feedback. We implement SMAT in the MyoAssist simulation environment using a 26-muscle lower-limb model and an attached hip exoskeleton. Our musculoskeletal simulations demonstrate that the learned exoskeleton control policy produces an average 10.1% reduction in hip muscle activation relative to the no-assist condition. We validated the learned controller in an offline setting using open-source gait data, then deployed it to a physical hip exoskeleton for treadmill experiments with five subjects. The resulting policy delivers consistent assistance and predominantly positive mechanical power without the need for any explicitly imposed timing shift (mean positive power: 13.6 W at 6 Nm RMS torque to 23.8 W at 9.3 Nm RMS torque, with minimal negative power) consistently across all subjects without subject-specific retraining.
Authors: Tharini Suresh, Salma Afifi, Sudeep Pasricha
Abstract: Diffusion models have revolutionized generative AI, with their inherent capacity to generate highly realistic state-of-the-art synthetic data. However, these models employ an iterative denoising process over computationally intensive layers such as UNets and attention mechanisms. This results in high inference energy on conventional electronic platforms, and thus, there is an emerging need to accelerate these models in a sustainable manner. To address this challenge, we present a novel silicon photonics-based accelerator for diffusion models. Experimental evaluations demonstrate that our photonic accelerator achieves at least 3x better energy efficiency and 5.5x throughput improvement compared to state-of-the-art diffusion model accelerators.
Authors: Zihang You, Xianlian Zhou
Abstract: Data-driven joint-moment predictors offer a scalable alternative to laboratory-based inverse-dynamics pipelines for biomechanics estimation and exoskeleton control. Meanwhile, physics-based reinforcement learning (RL) enables simulation-trained controllers to learn dynamics-aware assistance strategies without extensive human experimentation. However, quantitative verification of simulation-trained exoskeleton torque predictors, and their impact on human joint power injection, remains limited. This paper presents (1) an RL framework to learn exoskeleton assistance policies that reduce biological joint moments, and (2) a validation pipeline that verifies the trained control networks using an open-source gait dataset through inference and comparison with biological joint moments. Simulation-trained multilayer perceptron (MLP) controllers are developed for level-ground and ramp walking, mapping short-horizon histories of bilateral hip and knee kinematics to normalized assistance torques. Results show that predicted assistance preserves task-intensity trends across speeds and inclines. Agreement is particularly strong at the hip, with cross-correlation coefficients reaching 0.94 at 1.8 m/s and 0.98 during 5{\deg} decline walking, demonstrating near-matched temporal structure. Discrepancies increase at higher speeds and steeper inclines, especially at the knee, and are more pronounced in joint power comparisons. Delay tuning biases assistance toward greater positive power injection; modest timing shifts increase positive power and improve agreement in specific gait intervals. Together, these results establish a quantitative validation framework for simulation-trained exoskeleton controllers, demonstrate strong sim-to-data consistency at the torque level, and highlight both the promise and the remaining challenges for sim-to-real transfer.
Authors: Abdeldjalil Taibi, Mohmoud Badlis, Amina Bensalem, Belkacem Zouilekh, Mohammed Brahimi
Abstract: Efficient luggage trolley management is critical for reducing congestion and ensuring asset availability in modern airports. Automated detection systems face two main challenges. First, strict security and privacy regulations limit large-scale data collection. Second, existing public datasets lack the diversity, scale, and annotation quality needed to handle dense, overlapping trolley arrangements typical of real-world operations. To address these limitations, we introduce a synthetic data generation pipeline based on a high-fidelity Digital Twin of Algiers International Airport using NVIDIA Omniverse. The pipeline produces richly annotated data with oriented bounding boxes, capturing complex trolley formations, including tightly nested chains. We evaluate YOLO-OBB using five training strategies: real-only, synthetic-only, linear probing, full fine-tuning, and mixed training. This allows us to assess how synthetic data can complement limited real-world annotations. Our results show that mixed training with synthetic data and only 40 percent of real annotations matches or exceeds the full real-data baseline, achieving 0.94 mAP@50 and 0.77 mAP@50-95, while reducing annotation effort by 25 to 35 percent. Multi-seed experiments confirm strong reproducibility with a standard deviation below 0.01 on mAP@50, demonstrating the practical effectiveness of synthetic data for automated trolley detection.
Authors: Pu Qiu, Chen Ouyang, Yongyang Xiong, Keyou You, Wanquan Liu, Yang Shi
Abstract: Federated Composite Optimization (FCO) has emerged as a promising framework for training models with structural constraints (e.g., sparsity) in distributed edge networks. However, simultaneously achieving communication efficiency and convergence robustness remains a significant challenge, particularly when dealing with non-smooth regularizers, statistical heterogeneity, and the restrictions of biased compression. To address these issues, we propose FedCEF (Federated Composite Error Feedback), a novel algorithm tailored for non-convex FCO. FedCEF introduces a decoupled proximal update scheme that separates the proximal operator from communication, enabling clients to handle non-smooth terms locally while transmitting compressed information. To mitigate the noise from aggressive quantization and the bias from non-IID data, FedCEF integrates a rigorous error feedback mechanism with control variates. Furthermore, we design a communication-efficient pre-proximal downlink strategy that allows clients to exactly reconstruct global control variables without explicit transmission. We theoretically establish that FedCEF achieves sublinear convergence to a bounded residual error under general non-convexity, which is controllable via the step size and batch size. Extensive experiments on real datasets validate FedCEF maintains competitive model accuracy even under extreme compression ratios (e.g., 1%), significantly reducing the total communication volume compared to uncompressed baselines.
Authors: Rahul Bera
Abstract: Modern applications process massive data volumes that overwhelm the storage and retrieval capabilities of memory systems, making memory the primary performance and energy-efficiency bottleneck of computing systems. Although many microarchitectural techniques attempt to hide or tolerate long memory access latency, rapidly growing data footprints continue to outpace technology scaling, requiring more effective solutions. This dissertation shows that modern processors observe large amounts of application and system data during execution, yet many microarchitectural mechanisms make decisions largely independent of this information. Through four case studies, we demonstrate that such data-agnostic design leads to substantial missed opportunities for improving performance and energy efficiency. To address this limitation, this dissertation advocates shifting microarchitecture design from data-agnostic to data-informed. We propose mechanisms that (1) learn policies from observed execution behavior (data-driven design) and (2) exploit semantic characteristics of application data (data-aware design). We apply lightweight machine learning techniques and previously underexplored data characteristics across four processor components: a reinforcement learning-based hardware data prefetcher that learns memory access patterns online; a perceptron predictor that identifies memory requests likely to access off-chip memory; a reinforcement learning mechanism that coordinates data prefetching and off-chip prediction; and a mechanism that exploits repeatability in memory addresses and loaded values to eliminate predictable load instructions. Our extensive evaluation shows that the proposed techniques significantly improve performance and energy efficiency compared to prior state-of-the-art approaches.
Authors: Zijie Yan (NVIDIA), Hongxiao Bai (NVIDIA), Xin Yao (NVIDIA), Dennis Liu (NVIDIA), Tong Liu (NVIDIA), Hongbin Liu (NVIDIA), Pingtian Li (NVIDIA), Evan Wu (NVIDIA), Shiqing Fan (NVIDIA), Li Tao (NVIDIA), Robin Zhang (NVIDIA), Yuzhong Wang (NVIDIA), Shifang Xu (NVIDIA), Jack Chang (NVIDIA), Xuwen Chen (NVIDIA), Kunlun Li (NVIDIA), Yan Bai (NVIDIA), Gao Deng (NVIDIA), Nan Zheng (NVIDIA), Vijay Anand Korthikanti (NVIDIA), Abhinav Khattar (NVIDIA), Ethan He (NVIDIA), Soham Govande (NVIDIA), Sangkug Lym (NVIDIA), Zhongbo Zhu (NVIDIA), Qi Zhang (NVIDIA), Haochen Yuan (NVIDIA), Xiaowei Ren (NVIDIA), Deyu Fu (NVIDIA), Tailai Ma (NVIDIA), Shunkang Zhang (NVIDIA), Jiang Shao (NVIDIA), Ray Wang (NVIDIA), Vasudevan Rengasamy (NVIDIA), Rachit Garg (NVIDIA), Santosh Bhavani (NVIDIA), Xipeng Li (NVIDIA), Chandler Zhou (NVIDIA), David Wu (NVIDIA), Yingcan Wei (NVIDIA), Ashwath Aithal (NVIDIA), Michael Andersch (NVIDIA), Mohammad Shoeybi (NVIDIA), Jiajie Yao (NVIDIA), June Yang (NVIDIA)
Abstract: Scaling Mixture-of-Experts (MoE) training introduces systems challenges absent in dense models. Because each token activates only a subset of experts, this sparsity allows total parameters to grow much faster than per-token computation, creating coupled constraints across memory, communication, and computation. Optimizing one dimension often shifts pressure to another, demanding co-design across the full system stack. We address these challenges for MoE training through integrated optimizations spanning memory (fine-grained recomputation, offloading, etc.), communication (optimized dispatchers, overlapping, etc.), and computation (Grouped GEMM, fusions, CUDA Graphs, etc.). The framework also provides Parallel Folding for flexible multi-dimensional parallelism, low-precision training support for FP8 and NVFP4, and efficient long-context training. On NVIDIA GB300 and GB200, it achieves 1,233/1,048 TFLOPS/GPU for DeepSeek-V3-685B and 974/919 TFLOPS/GPU for Qwen3-235B. As a performant, scalable, and production-ready open-source solution, it has been used across academia and industry for training MoE models ranging from billions to trillions of parameters on clusters scaling up to thousands of GPUs. This report explains how these techniques work, their trade-offs, and their interactions at the systems level, providing practical guidance for scaling MoE models with Megatron Core.
Authors: Vinzenz Thoma, Georgios Piliouras, Luke Marris
Abstract: Automated design of multi-agent interactions with desirable equilibrium outcomes is inherently difficult due to the computational hardness, non-uniqueness, and instability of the resulting equilibria. In this work, we propose the use of game-agnostic differentiable equilibrium blocks (DEBs) as modules in a novel, differentiable framework to address a wide variety of incentive design problems from economics and computer science. We call this framework deep incentive design (DID). To validate our approach, we examine three diverse, challenging incentive design tasks: contract design, machine scheduling, and inverse equilibrium problems. For each task, we train a single neural network using a unified pipeline and DEB. This architecture solves the full distribution of problem instances, parameterized by a context, handling all games across a wide range of scales (from two to sixteen actions per player).
Authors: Yuanlong Chen, Bowen Zhu, Bing Xia, Yichuan Wang
Abstract: Brand advertising plays a critical role in building long-term consumer awareness and loyalty, making it a key objective for advertisers across digital platforms. Although real-time bidding has been extensively studied, there is limited literature on algorithms specifically tailored for brand auction ads that fully leverage their unique characteristics. In this paper, we propose a lightweight Model Predictive Control (MPC) framework designed for brand advertising campaigns, exploiting the inherent attributes of brand ads -- such as stable user engagement patterns and fast feedback loops -- to simplify modeling and improve efficiency. Our approach utilizes online isotonic regression to construct monotonic bid-to-spend and bid-to-conversion models directly from streaming data, eliminating the need for complex machine learning models. The algorithm operates fully online with low computational overhead, making it highly practical for real-world deployment. Simulation results demonstrate that our approach significantly improves spend efficiency and cost control compared to baseline strategies, providing a scalable and easily implementable solution for modern brand advertising platforms.
Authors: Timothy Heightman, Roman Aseguinolaza Gallo, Edward Jiang, JRM Saavedra, Antonio Ac\'in, Marcin P{\l}odzie\'n
Abstract: Inferring the dynamical generator of a many-body quantum system from measurement data is essential for the verification, calibration, and control of quantum processors. When the system is open, this task becomes considerably harder than in the purely unitary case, because coherent and dissipative mechanisms can produce similar measurement statistics and long-time data can be insensitive to coherent couplings. Here we tackle this so-called Lindbladian learning problem of open-system characterisation with maximum-likelihood on Pauli measurements at multiple experimentally friendly \emph{transient} times, exploiting the richer information content of transient dynamics. To navigate the resulting non-convex likelihood loss-landscape, we augment the physical model neural differential-equation term, which is progressively removed during training to distil an interpretable Lindbladian solution. Our method reliably learns open-system dynamics across neutral-atom (with 2D connectivity) and superconducting Hamiltonians, as well as the Heisenberg XYZ, and PXP models on a spin-1/2 chain. For the dissipative part, we show robustness over phase noise, thermal noise, and their combination. Our algorithm can robustly infer these dissipative systems over noise-to-signal ratios spanning four orders of magnitude, and system sizes up to $N=6$ qubits with fewer than $5 \times 10^5$ shots.
Authors: Zongqian Li, Tengchao Lv, Shaohan Huang, Yixuan Su, Qinzheng Sun, Qiufeng Yin, Ying Xin, Scarlett Li, Lei Cui, Nigel Collier, Furu Wei
Abstract: Training next-generation code generation models requires high-quality datasets, yet existing datasets face difficulty imbalance, format inconsistency, and data quality problems. We address these challenges through systematic data processing and difficulty scaling. We introduce a four-stage Data Processing Framework encompassing collection, processing, filtering, and verification, incorporating Automatic Difficulty Filtering via an LLM-based predict-calibrate-select framework that leverages multi-dimensional difficulty metrics across five weighted dimensions to retain challenging problems while removing simplistic ones. The resulting MicroCoder dataset comprises tens of thousands of curated real competitive programming problems from diverse platforms, emphasizing recency and difficulty. Evaluations on strictly unseen LiveCodeBench demonstrate that MicroCoder achieves 3x larger performance gains within 300 training steps compared to widely-used baseline datasets of comparable size, with consistent advantages under both GRPO and its variant training algorithms. The MicroCoder dataset delivers obvious improvements on medium and hard problems across different model sizes, achieving up to 17.2% relative gains in overall performance where model capabilities are most stretched. These results validate that difficulty-aware data curation improves model performance on challenging tasks, providing multiple insights for dataset creation in code generation.
Authors: Sarmad Mehrdad, Maxime Sabbah, Vincent Bonnet, Ludovic Righetti
Abstract: This paper investigates whether a single, unified cost function can explain and predict human reaching movements, in contrast with existing approaches that rely on subject- or posture-specific optimization criteria. Using the Minimal Observation Inverse Reinforcement Learning (MO-IRL) algorithm, together with a seven-dimensional set of candidate cost terms, we efficiently estimate time-varying cost weights for a standard planar reaching task. MO-IRL provides orders-of-magnitude faster convergence than bilevel formulations, while using only a fraction of the available data, enabling the practical exploration of time-varying cost structures. Three levels of generality are evaluated: Subject-Dependent Posture-Dependent, Subject-Dependent Posture-Independent, and Subject-Independent Posture-Independent. Across all cases, time-varying weights substantially improve trajectory reconstruction, yielding an average 27% reduction in RMSE compared to the baseline. The inferred costs consistently highlight a dominant role for joint-acceleration regulation, complemented by smaller contributions from torque-change smoothness. Overall, a single subject- and posture-agnostic time-varying cost function is shown to predict human reaching trajectories with high accuracy, supporting the existence of a unified optimality principle governing this class of movements.
Authors: Pedro Taranc\'on-\'Alvarez, Leonid Sarieddine, Pavlos Protopapas, Raul Jimenez
Abstract: Embeddings provide low-dimensional representations that organize complex function spaces and support generalization. They provide a geometric representation that supports efficient retrieval, comparison, and generalization. In this work we generalize the concept to Physics Informed Neural Networks. We present a method to construct solution embedding spaces of nonlinear partial differential equations using a multi-head setup, and extract non-degenerate information from them using principal component analysis (PCA). We test this method by applying it to viscous Burgers' equation, which is solved simultaneously for a family of initial conditions and values of the viscosity. A shared network body learns a latent embedding of the solution space, while linear heads map this embedding to individual realizations. By enforcing orthogonality constraints on the heads, we obtain a principal-component decomposition of the latent space that is robust to training degeneracies and admits a direct physical interpretation. The obtained components for Burgers' equation exhibit rapid saturation, indicating that a small number of latent modes captures the dominant features of the dynamics.
Authors: Mridankan Mandal
Abstract: Accurate estimation of pasture biomass from agricultural imagery is critical for sustainable livestock management, yet existing methods are limited by the small, imbalanced, and sparsely annotated datasets typical of real world monitoring. In this study, adaptation of vision foundation models to agricultural regression is systematically evaluated on the CSIRO Pasture Biomass benchmark, a 357 image dual view dataset with laboratory validated, component wise ground truth for five biomass targets, through 17 configurations spanning four backbones (EfficientNet-B3 to DINOv3-ViT-L), five cross view fusion mechanisms, and a 4x2 metadata factorial. A counterintuitive principle, termed "fusion complexity inversion", is uncovered: on scarce agricultural data, a two layer gated depthwise convolution (R^2 = 0.903) outperforms cross view attention transformers (0.833), bidirectional SSMs (0.819), and full Mamba (0.793, below the no fusion baseline). Backbone pretraining scale is found to monotonically dominate all architectural choices, with the DINOv2 -> DINOv3 upgrade alone yielding +5.0 R^2 points. Training only metadata (species, state, and NDVI) is shown to create a universal ceiling at R^2 ~ 0.829, collapsing an 8.4 point fusion spread to 0.1 points. Actionable guidelines for sparse agricultural benchmarks are established: backbone quality should be prioritized over fusion complexity, local modules preferred over global alternatives, and features unavailable at inference excluded.
Authors: Yunmei Chen, Chi Ding, Xiaojing Ye
Abstract: We develop a novel transfer learning framework to tackle the challenge of limited training data in image reconstruction problems. The proposed framework consists of two training steps, both of which are formed as bi-level optimizations. In the first step, we train a powerful universal feature-extractor that is capable of learning important knowledge from large, heterogeneous data sets in various domains. In the second step, we train a task-specific domain-adapter for a new target domain or task with only a limited amount of data available for training. Then the composition of the adapter and the universal feature-extractor effectively explores feature which serve as an important component of image regularization for the new domains, and this leads to high-quality reconstruction despite the data limitation issue. We apply this framework to reconstruct under-sampled MR images with limited data by using a collection of diverse data samples from different domains, such as images of other anatomies, measurements of various sampling ratios, and even different image modalities, including natural images. Experimental results demonstrate a promising transfer learning capability of the proposed method.
Authors: Waldyn G Martinez
Abstract: Detecting structural instability and anomalies in high-dimensional financial time series is challenging due to complex temporal dependence and evolving cross-sectional structure. We propose ReGEN-TAD, an interpretable generative framework that integrates modern machine learning with econometric diagnostics for anomaly detection. The model combines joint forecasting and reconstruction within a refined convolutional--transformer architecture and aggregates complementary signals capturing predictive inconsistency, reconstruction degradation, latent distortion, and volatility shifts. Robust calibration yields a unified anomaly score without labeled data. Experiments on synthetic and financial panels demonstrate improved robustness to structured deviations while enabling economically coherent factor-level attribution.
Authors: Dilermando Almeida, Juliano Negri, Guilherme Lazzarini, Thiago H. Segreto, Ranulfo Bezerra, Ricardo V. Godoy, Marcelo Becker
Abstract: Robust grasping in cluttered, unstructured environments remains challenging for mobile legged manipulators due to occlusions that lead to partial observations, unreliable depth estimates, and the need for collision-free, execution-feasible approaches. In this paper we present an end-to-end pipeline for language-guided grasping that bridges open-vocabulary target selection to safe grasp execution on a real robot. Given a natural-language command, the system grounds the target in RGB using open-vocabulary detection and promptable instance segmentation, extracts an object-centric point cloud from RGB-D, and improves geometric reliability under occlusion via back-projected depth compensation and two-stage point cloud completion. We then generate and collision-filter 6-DoF grasp candidates and select an executable grasp using safety-oriented heuristics that account for reachability, approach feasibility, and clearance. We evaluate the method on a quadruped robot with an arm in two cluttered tabletop scenarios, using paired trials against a view-dependent baseline. The proposed approach achieves a 90% overall success rate (9/10) against 30% (3/10) for the baseline, demonstrating substantially improved robustness to occlusions and partial observations in clutter.
Authors: Jeongwoo Lee, Baek Duhyeong, Eungyeol Han, Soyeon Shin, Gukin han, Seungduk Kim, Jaehyun Jeon, Taewoo Jeong
Abstract: Recent advances in Vision-Language Models (VLMs) have demonstrated impressive multimodal understanding in general domains. However, their applicability to decision-oriented domains such as hospitality remains largely unexplored. In this work, we investigate how well VLMs can perform visual question answering (VQA) about hotel and facility images that are central to consumer decision-making. While many existing VQA benchmarks focus on factual correctness, they rarely capture what information users actually find useful. To address this, we first introduce Informativeness as a formal framework to quantify how much hospitality-relevant information an image-question pair provides. Guided by this framework, we construct a new hospitality-specific VQA dataset that covers various facility types, where questions are specifically designed to reflect key user information needs. Using this benchmark, we conduct experiments with several state-of-the-art VLMs, revealing that VLMs are not intrinsically decision-aware-key visual signals remain underutilized, and reliable informativeness reasoning emerges only after modest domain-specific finetuning.
Authors: Ximeng Tao, Dimitar Filev, Gaurav Pandey
Abstract: Contrastive Language-Image Pre-training (CLIP) has shown impressive performance in aligning visual and textual representations. Recent studies have extended this paradigm to 3D vision to improve scene understanding for autonomous driving. A common strategy is to employ pairwise cosine similarity between modalities to guide the training of a 3D encoder. However, considering the similarity between individual modality pairs rather than all modalities jointly fails to ensure consistent and unified alignment across the entire multimodal space. In this paper, we propose a Contrastive Tensor Pre-training (CTP) framework that simultaneously aligns multiple modalities in a unified embedding space to enhance end-to-end autonomous driving. Compared with pairwise cosine similarity alignment, our method extends the 2D similarity matrix into a multimodal similarity tensor. Furthermore, we introduce a tensor loss to enable joint contrastive learning across all modalities. For experimental validation of our framework, we construct a text-image-point cloud triplet dataset derived from existing autonomous driving datasets. The results show that our proposed unified multimodal alignment framework achieves favorable performance for both scenarios: (i) aligning a 3D encoder with pretrained CLIP encoders, and (ii) pretraining all encoders from scratch.
Authors: Minkyu Kim, Sangheon Lee, Dongmin Park
Abstract: The ability to distinguish subtle differences between visually similar images is essential for diverse domains such as industrial anomaly detection, medical imaging, and aerial surveillance. While comparative reasoning benchmarks for vision-language models (VLMs) have recently emerged, they primarily focus on images with large, salient differences and fail to capture the nuanced reasoning required for real-world applications. In this work, we introduce VLM-SubtleBench, a benchmark designed to evaluate VLMs on subtle comparative reasoning. Our benchmark covers ten difference types - Attribute, State, Emotion, Temporal, Spatial, Existence, Quantity, Quality, Viewpoint, and Action - and curate paired question-image sets reflecting these fine-grained variations. Unlike prior benchmarks restricted to natural image datasets, our benchmark spans diverse domains, including industrial, aerial, and medical imagery. Through extensive evaluation of both proprietary and open-source VLMs, we reveal systematic gaps between model and human performance across difference types and domains, and provide controlled analyses highlighting where VLMs' reasoning sharply deteriorates. Together, our benchmark and findings establish a foundation for advancing VLMs toward human-level comparative reasoning.
Authors: Aomar Osmani
Abstract: We introduce SMGI, a structural theory of general artificial intelligence, and recast the foundational problem of learning from the optimization of hypotheses within fixed environments to the controlled evolution of the learning interface itself. We formalize the Structural Model of General Intelligence (SMGI) via a typed meta-model $\theta = (r,\mathcal H,\Pi,\mathcal L,\mathcal E,\mathcal M)$ that treats representational maps, hypothesis spaces, structural priors, multi-regime evaluators, and memory operators as explicitly typed, dynamic components. By enforcing a strict mathematical separation between this structural ontology ($\theta$) and its induced behavioral semantics ($T_\theta$), we define general artificial intelligence as a class of admissible coupled dynamics $(\theta, T_\theta)$ satisfying four obligations: structural closure under typed transformations, dynamical stability under certified evolution, bounded statistical capacity, and evaluative invariance across regime shifts. We prove a structural generalization bound that links sequential PAC-Bayes analysis and Lyapunov stability, providing sufficient conditions for capacity control and bounded drift under admissible task transformations. Furthermore, we establish a strict structural inclusion theorem demonstrating that classical empirical risk minimization, reinforcement learning, program-prior models (Solomonoff-style), and modern frontier agentic pipelines operate as structurally restricted instances of SMGI.
Authors: Chen-Chen Zong, Yu-Qi Chi, Xie-Yang Wang, Yan Cui, Sheng-Jun Huang
Abstract: Open-set active learning (OSAL) aims to identify informative samples for annotation when unlabeled data may contain previously unseen classes-a common challenge in safety-critical and open-world scenarios. Existing approaches typically rely on separately trained open-set detectors, introducing substantial training overhead and overlooking the supervisory value of labeled unknowns for improving known-class learning. In this paper, we propose E$^2$OAL (Effective and Efficient Open-set Active Learning), a unified and detector-free framework that fully exploits labeled unknowns for both stronger supervision and more reliable querying. E$^2$OAL first uncovers the latent class structure of unknowns through label-guided clustering in a frozen contrastively pre-trained feature space, optimized by a structure-aware F1-product objective. To leverage labeled unknowns, it employs a Dirichlet-calibrated auxiliary head that jointly models known and unknown categories, improving both confidence calibration and known-class discrimination. Building on this, a logit-margin purity score estimates the likelihood of known classes to construct a high-purity candidate pool, while an OSAL-specific informativeness metric prioritizes partially ambiguous yet reliable samples. These components together form a flexible two-stage query strategy with adaptive precision control and minimal hyperparameter sensitivity. Extensive experiments across multiple OSAL benchmarks demonstrate that E$^2$OAL consistently surpasses state-of-the-art methods in accuracy, efficiency, and query precision, highlighting its effectiveness and practicality for real-world applications. The code is available at github.com/chenchenzong/E2OAL.
Authors: Ximeng Tao, Pardis Taghavi, Dimitar Filev, Reza Langari, Gaurav Pandey
Abstract: Vision-language models (VLMs) have emerged as a promising direction for end-to-end autonomous driving (AD) by jointly modeling visual observations, driving context, and language-based reasoning. However, existing VLM-based systems face a trade-off between high-level reasoning and motion planning: large models offer strong semantic understanding but are costly to adapt for precise control, whereas small VLM models can be fine-tuned efficiently but often exhibit weaker reasoning. We propose NaviDriveVLM, a decoupled framework that separates reasoning from action generation using a large-scale Navigator and a lightweight trainable Driver. This design preserves reasoning ability, reduces training cost, and provides an explicit interpretable intermediate representation for downstream planning. Experiments on the nuScenes benchmark show that NaviDriveVLM outperforms large VLM baselines in end-to-end motion planning.
Authors: Jun Yin, Peng Huo, Bangguo Zhu, Hao Yan, Senzhang Wang, Shirui Pan, Chengqi Zhang
Abstract: In recent advances, to enable a fully data-driven learning paradigm on relational databases (RDB), relational deep learning (RDL) is proposed to structure the RDB as a heterogeneous entity graph and adopt the graph neural network (GNN) as the predictive model. However, existing RDL methods neglect the imbalance problem of relational data in RDBs and risk under-representing the minority entities, leading to an unusable model in practice. In this work, we investigate, for the first time, class imbalance problem in RDB entity classification and design the relation-centric minority synthetic over-sampling GNN (Rel-MOSS), in order to fill a critical void in the current literature. Specifically, to mitigate the issue of minority-related information being submerged by majority counterparts, we design the relation-wise gating controller to modulate neighborhood messages from each individual relation type. Based on the relational-gated representations, we further propose the relation-guided minority synthesizer for over-sampling, which integrates the entity relational signatures to maintain relational consistency. Extensive experiments on 12 entity classification datasets provide compelling evidence for the superiority of Rel-MOSS, yielding an average improvement of up to 2.46% and 4.00% in terms of Balanced Accuracy and G-Mean, compared with SOTA RDL methods and classic methods for handling class imbalance.
Authors: Akram S. Awad, Shihab Ahmed, Yue Wang, George K. Atia
Abstract: Robust Markov Decision Processes (MDPs) address environmental shift through distributionally robust optimization (DRO) by finding an optimal worst-case policy within an uncertainty set of transition kernels. However, standard DRO approaches require enlarging the uncertainty set under large shifts, which leads to overly conservative and pessimistic policies. In this paper, we propose a framework for transfer under environment shift that derives a robust target-domain policy via estimate-centered uncertainty sets, constructed through constrained estimation that integrates limited target samples with side information about the source-target dynamics. The side information includes bounds on feature moments, distributional distances, and density ratios, yielding improved kernel estimates and tighter uncertainty sets. The side information includes bounds on feature moments, distributional distances, and density ratios, yielding improved kernel estimates and tighter uncertainty sets. Error bounds and convergence results are established for both robust and non-robust value functions. Moreover, we provide a finite-sample guarantee on the learned robust policy and analyze the robust sub-optimality gap. Under mild low-dimensional structure on the transition model, the side information reduces this gap and improves sample efficiency. We assess the performance of our approach across OpenAI Gym environments and classic control problems, consistently demonstrating superior target-domain performance over state-of-the-art robust and non-robust baselines.
Authors: Anne Dranowski, Yura Kabkov, Daniel Tubbenhauer
Abstract: We develop a reinforcement learning pipeline for simplifying knot diagrams. A trained agent learns move proposals and a value heuristic for navigating Reidemeister moves. The pipeline applies to arbitrary knots and links; we test it on ``very hard'' unknot diagrams and, using diagram inflation, on $4_1\#9_{10}$ where we recover the recently established and surprising upper bound of three for the unknotting number.
Authors: Jing Jingzhe, Fan Zheyi, Szu Hui Ng, Qingpei Hu
Abstract: Bayesian optimization (BO) for high-dimensional constrained problems remains a significant challenge due to the curse of dimensionality. We propose Local Constrained Bayesian Optimization (LCBO), a novel framework tailored for such settings. Unlike trust-region methods that are prone to premature shrinking when confronting tight or complex constraints, LCBO leverages the differentiable landscape of constraint-penalized surrogates to alternate between rapid local descent and uncertainty-driven exploration. Theoretically, we prove that LCBO achieves a convergence rate for the Karush-Kuhn-Tucker (KKT) residual that depends polynomially on the dimension $d$ for common kernels under mild assumptions, offering a rigorous alternative to global BO where regret bounds typically scale exponentially. Extensive evaluations on high-dimensional benchmarks (up to 100D) demonstrate that LCBO consistently outperforms state-of-the-art baselines.
Authors: Yuzhi Liu, Duo Zhang, Anyang Peng, Weinan E, Linfeng Zhang, Han Wang
Abstract: Machine Learning Interatomic Potentials (MLIPs) enable accurate large-scale atomistic simulations, yet improving their expressive capacity efficiently remains challenging. Here we systematically develop Mixture-of-Experts (MoE) and Mixture-of-Linear-Experts (MoLE) architectures for MLIPs and analyze the effects of routing strategies and expert designs. We show that sparse activation combined with shared experts yields substantial performance gains, and that nonlinear MoE formulations outperform MoLE when shared experts are present, underscoring the importance of nonlinear expert specialization. Furthermore, element-wise routing consistently surpasses configuration-level routing, while global MoE routing often leads to numerical instability. The resulting element-wise MoE model achieves state-of-the-art accuracy across the OMol25, OMat24, and OC20M benchmarks. Analysis of routing patterns reveals chemically interpretable expert specialization aligned with periodic-table trends, indicating that the model effectively captures element-specific chemical characteristics for precise interatomic modeling.
Authors: Chenzhi Hu, Qinzhe Hu, Yuhang Xu, Junyi Chen, Ruijie Wang, Shengzhong Liu, Jianxin Li, Fan Wu, Guihai Chen
Abstract: Large reasoning models (LRMs) like OpenAI o1 and DeepSeek-R1 achieve high accuracy on complex tasks by adopting long chain-of-thought (CoT) reasoning paths. However, the inherent verbosity of these processes frequently results in redundancy and overthinking. To address this issue, existing works leverage Group Relative Policy Optimization (GRPO) to reduce LRM output length, but their static length reward design cannot dynamically adapt according to the relative problem difficulty and response length distribution, causing over-compression and compromised accuracy. Therefore, we propose SmartThinker, a novel GRPO-based efficient reasoning method with progressive CoT length calibration. SmartThinker makes a two-fold contribution: First, it dynamically estimates the optimal length with peak accuracy during training and guides overlong responses toward it to reduce response length while sustaining accuracy. Second, it dynamically modulates the length reward coefficient to avoid the unwarranted penalization of correct reasoning paths. Extensive experiment results show that SmartThinker achieves up to 52.5% average length compression with improved accuracy, and achieves up to 16.6% accuracy improvement on challenging benchmarks like AIME25. The source code can be found at https://github.com/SJTU-RTEAS/SmartThinker.
Authors: Dengcan Liu, Fengkai Yang, Xiaohan Wang, Shurui Yan, Jiajun Chai, Jiahao Li, Yikun Ban, Zhendong Mao, Wei Lin, Guojun Yin
Abstract: Reward modeling is essential for aligning Large Language Models(LLMs) with human preferences, yet conventional reward models suffer from poor interpretability and heavy reliance on costly expert annotations. While recent rubric-based approaches enhance evaluation transparency, they lack systematic quality control, yielding noisy and redundant criteria, failing to mitigate persistent biases (e.g., verbosity, position) in LLM evaluators, and creating a scalability-reliability trade-off. To address these limitations, we propose CDRRM (Contrast-Driven Rubric Reward Model), a framework built on a novel Contrast-then-Synthesis paradigm for high-quality rubric generation and guided preference judgment. CDRRM first conducts multi-dimensional contrastive profiling on preference pairs to identify causal discriminative factors, then synthesizes these insights into compact, context-aware rubrics to guide preference judg- ments. Extensive experiments on three authoritative benchmarks (RewardBench, RMBench, RMB) demonstrate that CDRRM achieves state-of-the-art performance across diverse domains and effectively mitigates aforementioned evaluation biases. Notably, our approach delivers exceptional data efficiency: training the rubric generator on only 3k high-quality samples empowers a frozen pre-trained judge model to outperform fully fine-tuned baselines. This work offers a scalable, interpretable, and data-efficient path for reward modeling.
Authors: Chi-Min Chan, Ehsan Hajiramezanali, Xiner Li, Edward De Brouwer, Carl Edwards, Wei Xue, Sirui Han, Yike Guo, Gabriele Scalia
Abstract: In scientific reasoning tasks, the veracity of the reasoning process is as critical as the final outcome. While Process Reward Models (PRMs) offer a solution to the coarse-grained supervision problems inherent in Outcome Reward Models (ORMs), their deployment is hindered by the prohibitive cost of obtaining expert-verified step-wise labels. This paper addresses the challenge of training reliable PRMs using abundant but noisy "weak" supervision. We argue that existing Weak-to-Strong Generalization (W2SG) theories lack prescriptive guidelines for selecting high-quality training signals from noisy data. To bridge this gap, we introduce the Dual-Consensus Weak-to-Strong (DC-W2S) framework. By intersecting Self-Consensus (SC) metrics among weak supervisors with Neighborhood-Consensus (NC) metrics in the embedding space, we stratify supervision signals into distinct reliability regimes. We then employ a curriculum of instance-level balanced sampling and label-level reliability-aware masking to guide the training process. We demonstrate that DC-W2S enables the training of robust PRMs for complex reasoning without exhaustive expert annotation, proving that strategic data curation is more effective than indiscriminate training on large-scale noisy datasets.
Authors: Nuutti Barron, Heng Rao, Urmi Saha, Yu Gu, Zhenghao Liu, Ge Yu, Defu Yang, Ashish Raj, Minghan Chen
Abstract: Mechanistic modeling provides a biophysically grounded framework for studying the spread of pathological tau protein in tauopathies like Alzheimer's disease. Existing approaches typically model tau propagation as a diffusive process on the brain's structural connectome, reproducing macroscopic patterns but neglecting microscale cellular transport and reaction mechanisms. The Network Transport Model (NTM) was introduced to fill this gap, explaining how region-level progression of tau emerges from microscale biophysical processes. However, the NTM faces a common challenge for complex models defined by large systems of partial differential equations: the inability to perform parameter inference and mechanistic discovery due to high computational burden and slow model simulations. To overcome this barrier, we propose Tau-BNO, a Brain Neural Operator surrogate framework for rapidly approximating NTM dynamics that captures both intra-regional reaction kinetics and inter-regional network transport. Tau-BNO combines a function operator that encodes kinetic parameters with a query operator that preserves initial state information, while approximating anisotropic transport through a spectral kernel that retains directionality. Empirical evaluations demonstrate high predictive accuracy ($R^2\approx$ 0.98) across diverse biophysical regimes and an 89\% performance improvement over state-of-the-art sequence models like Transformers and Mamba, which lack inherent structural priors. By reducing simulation time from hours to seconds, we show that the surrogate model is capable of producing new insights and generating new hypotheses. This framework is readily extensible to a broader class of connectome-based biophysical models, showcasing the transformative value of deep learning surrogates to accelerate analysis of large-scale, computationally intensive dynamical systems.
Authors: Xiang Shi, Wenlong Huang, Menglin Zou, Xinhai Sun
Abstract: We revisit Vision-Language-Action through a neuroscience-inspired triad. Biologically, the Cerebrum provides stable high-level multimodal priors and remains frozen; the Pons Adapter integrates these cortical features with real-time proprioceptive inputs and compiles intent into execution-ready tokens; and the Cerebellum (ParaCAT) performs fast, parallel categorical decoding for online control, with hysteresis/EMA/temperature/entropy for stability. A fixed-ratio schedule and two-stage feature caching make the system compute-aware and reproducible. Inspired by active, foveated vision, our wrist ROIs are geometrically tied to the end-effector via calibrated projection, providing a movement-stabilized, high-resolution view that is sensitive to fine-grained pose changes and complements the global context of the main view. The design is modular: upgrading the Cerebrum only retrains the Pons; changing robots only trains the Cerebellum; cerebellum-only RL can further refine control without touching high-level semantics. As a concept-and-protocol paper with preliminary evidence, we outline a timing protocol under matched conditions (GPU, resolution, batch) to verify anticipated efficiency gains. We also report preliminary LIBERO evidence showing that split feature caching reduces training time (7.5h to 4.5h) and improves average success (86.5% to 92.5%) under official N1.5 head-only training, and that SaiVLA0 reaches 99.0% mean success.
Authors: Shentong Mo, Yibing Song
Abstract: Coordinated audio generation based on video inputs typically requires a strict audio-visual (AV) alignment, where both semantics and rhythmics of the generated audio segments shall correspond to those in the video frames. Previous studies leverage a two-stage design where the AV encoders are firstly aligned via contrastive learning, then the encoded video representations guide the audio generation process. We observe that both contrastive learning and global video guidance are effective in aligning overall AV semantics while limiting temporally rhythmic synchronization. In this work, we propose FoleyFlow to first align unimodal AV encoders via masked modeling training, where the masked audio segments are recovered under the guidance of the corresponding video segments. After training, the AV encoders which are separately pretrained using only unimodal data are aligned with semantic and rhythmic consistency. Then, we develop a dynamic conditional flow for the final audio generation. Built upon the efficient velocity flow generation framework, our dynamic conditional flow utilizes temporally varying video features as the dynamic condition to guide corresponding audio segment generations. To this end, we extract coherent semantic and rhythmic representations during masked AV alignment, and use this representation of video segments to guide audio generation temporally. Our audio results are evaluated on the standard benchmarks and largely surpass existing results under several metrics. The superior performance indicates that FoleyFlow is effective in generating coordinated audios that are both semantically and rhythmically coherent to various video sequences.
Authors: Divake Kumar, Sina Tayebati, Devashri Naik, Patrick Poggi, Amanda Sofie Rios, Nilesh Ahuja, Amit Ranjan Trivedi
Abstract: Most uncertainty-aware robotic systems collapse prediction uncertainty into a single scalar score and use it to trigger uniform corrective responses. This aggregation obscures whether uncertainty arises from corrupted observations or from mismatch between the learned model and the true system dynamics. As a result, corrective actions may be applied to the wrong component of the closed loop, degrading performance relative to leaving the policy unchanged. We introduce a lightweight post hoc framework that decomposes uncertainty into aleatoric and epistemic components and uses these signals to regulate system responses at inference time. Aleatoric uncertainty is estimated from deviations in the observation distribution using a Mahalanobis density model, while epistemic uncertainty is detected using a noise robust forward dynamics ensemble that isolates model mismatch from measurement corruption. The two signals remain empirically near orthogonal during closed loop execution and enable type specific responses. High aleatoric uncertainty triggers observation recovery, while high epistemic uncertainty moderates control actions. The same signals also regulate adaptive perception by guiding model capacity selection during tracking inference. Experiments demonstrate consistent improvements across both control and perception tasks. In robotic manipulation, the decomposed controller improves task success from 59.4% to 80.4% under compound perturbations and outperforms a combined uncertainty baseline by up to 21.0%. In adaptive tracking inference on MOT17, uncertainty-guided model selection reduces average compute by 58.2% relative to a fixed high capacity detector while preserving detection quality within 0.4%. Code and demo videos are available at https://divake.github.io/uncertainty-decomposition/.
Authors: Jiahong Li, Fang Deng
Abstract: The autocovariance least squares (ALS) method is a computationally efficient approach for estimating noise covariances in Kalman filters without requiring specific noise models. However, conventional ALS and its variants rely on the classic least mean squares (LMS) criterion, making them highly sensitive to measurement outliers and prone to severe performance degradation. To overcome this limitation, this paper proposes a novel outlier-robust ALS algorithm, termed ALS-IRLS, based on the iteratively reweighted least squares (IRLS) framework. Specifically, the proposed approach introduces a two-tier robustification strategy. First, an innovation-level adaptive thresholding mechanism is employed to filter out heavily contaminated data. Second, the outlier-contaminated autocovariance is formulated using an $\epsilon$-contamination model, where the standard LMS criterion is replaced by the Huber cost function. The IRLS method is then utilized to iteratively adjust data weights based on estimation deviations, effectively mitigating the influence of residual outliers. Comparative simulations demonstrate that ALS-IRLS reduces the root-mean-square error (RMSE) of noise covariance estimates by over two orders of magnitude compared to standard ALS. Furthermore, it significantly enhances downstream state estimation accuracy, outperforming existing outlier-robust Kalman filters and achieving performance nearly equivalent to the ideal Oracle lower bound in the presence of noisy and anomalous data.
Authors: Joel Lidin, Amir Sarfi, Erfan Miahi, Quentin Anthony, Shivam Chauhan, Evangelos Pappas, Benjamin Th\'erien, Eugene Belilovsky, Samuel Dare
Abstract: Recently, there has been increased interest in globally distributed training, which has the promise to both reduce training costs and democratize participation in building large-scale foundation models. However, existing models trained in a globally distributed manner are relatively small in scale and have only been trained with whitelisted participants. Therefore, they do not yet realize the full promise of democratized participation. In this report, we describe Covenant-72B, an LLM produced by the largest collaborative globally distributed pre-training run (in terms of both compute and model scale), which simultaneously allowed open, permissionless participation supported by a live blockchain protocol. We utilized a state-of-the-art communication-efficient optimizer, SparseLoCo, supporting dynamic participation with peers joining and leaving freely. Our model, pre-trained on approximately 1.1T tokens, performs competitively with fully centralized models pre-trained on similar or higher compute budgets, demonstrating that fully democratized, non-whitelisted participation is not only feasible, but can be achieved at unprecedented scale for a globally distributed pre-training run.
Authors: Ali Hamza Bashir, Behzad Shomali, Markus Frey, Mehdi Ali, Rafet Sifa, David Berghaus
Abstract: We investigate whether performing reasoning in a continuous latent space leads to more robust multilingual capabilities. We compare Continuous Chain-of-Thought (using the CODI framework) against standard supervised fine-tuning across five typologically diverse languages: English, Chinese, German, French, and Urdu. Our experiments on GSM8k and CommonsenseQA demonstrate that continuous reasoning significantly outperforms explicit reasoning on low-resource languages, particularly in zero-shot settings where the target language was not seen during training. Additionally, this approach achieves extreme efficiency, compressing reasoning traces by approximately $29\times$ to $50\times$. These findings indicate that continuous latent representations naturally exhibit greater language invariance, offering a scalable solution for cross-lingual reasoning.
Authors: Michael K\"osel, Marcel Schreiber, Michael Ulrich, Claudius Gl\"aser, Klaus Dietmayer
Abstract: LiDAR-based 3D object detection plays a critical role for reliable and safe autonomous driving systems. However, existing detectors often produce overly confident predictions for objects not belonging to known categories, posing significant safety risks. This is caused by so-called out-of-distribution (OOD) objects, which were not part of the training data, resulting in incorrect predictions. To address this challenge, we propose ALOOD (Aligned LiDAR representations for Out-Of-Distribution Detection), a novel approach that incorporates language representations from a vision-language model (VLM). By aligning the object features from the object detector to the feature space of the VLM, we can treat the detection of OOD objects as a zero-shot classification task. We demonstrate competitive performance on the nuScenes OOD benchmark, establishing a novel approach to OOD object detection in LiDAR using language representations. The source code is available at https://github.com/uulm-mrm/mmood3d.
Authors: Yonghong Deng, Zhen Yang, Ping Jian, Xinyue Zhang, Zhongbin Guo, Chengzhi Li
Abstract: With the rapid advancement of large language models (LLMs), the safety of LLMs has become a critical concern. Despite significant efforts in safety alignment, current LLMs remain vulnerable to jailbreaking attacks. However, the root causes of such vulnerabilities are still poorly understood, necessitating a rigorous investigation into jailbreak mechanisms across both academic and industrial communities. In this work, we focus on a continuation-triggered jailbreak phenomenon, whereby simply relocating a continuation-triggered instruction suffix can substantially increase jailbreak success rates. To uncover the intrinsic mechanisms of this phenomenon, we conduct a comprehensive mechanistic interpretability analysis at the level of attention heads. Through causal interventions and activation scaling, we show that this jailbreak behavior primarily arises from an inherent competition between the model's intrinsic continuation drive and the safety defenses acquired through alignment training. Furthermore, we perform a detailed behavioral analysis of the identified safety-critical attention heads, revealing notable differences in the functions and behaviors of safety heads across different model architectures. These findings provide a novel mechanistic perspective for understanding and interpreting jailbreak behaviors in LLMs, offering both theoretical insights and practical implications for improving model safety.
Authors: Seongjin Bien, Carlo Kneissl, Tobias J\"ulg, Frank Fundel, Thomas Ressler-Antal, Florian Walter, Bj\"orn Ommer, Gitta Kutyniok, Wolfram Burgard
Abstract: Tactile sensation is essential for contact-rich manipulation tasks. It provides direct feedback on object geometry, surface properties, and interaction forces, enhancing perception and enabling fine-grained control. An inherent limitation of tactile sensors is that readings are available only when an object is touched. This precludes their use during planning and the initial execution phase of a task. Predicting tactile information from visual information can bridge this gap. A common approach is to learn a direct mapping from camera images to the output of vision-based tactile sensors. However, the resulting model will depend strongly on the specific setup and on how well the camera can capture the area where an object is touched. In this work, we introduce FlowTouch, a novel model for view-invariant visuo-tactile prediction. Our key idea is to use an object's local 3D mesh to encode rich information for predicting tactile patterns while abstracting away from scene-dependent details. FlowTouch integrates scene reconstruction and Flow Matching-based models for image generation. Our results show that FlowTouch is able to bridge the sim-to-real gap and generalize to new sensor instances. We further show that the resulting tactile images can be used for downstream grasp stability prediction. Our code, datasets and videos are available at https://flowtouch.github.io/
Authors: Daniel Wang, Thang D. Bui
Abstract: Machine learning models involving discrete latent variables require gradient estimators to facilitate backpropagation in a computationally efficient manner. The most recent addition to the Straight-Through family of estimators, ReinMax, can be viewed from a numerical ODE perspective as incorporating an approximation via Heun's method to reduce bias, but at the cost of high variance. In this work, we introduce the ReinMax-Rao and ReinMax-CV estimators which incorporate Rao-Blackwellisation and control variate techniques into ReinMax to reduce its variance. Our estimators demonstrate superior performance on training variational autoencoders with discrete latent spaces. Furthermore, we investigate the possibility of leveraging alternative numerical methods for constructing more accurate gradient approximations and present an alternative view of ReinMax from a simpler numerical integration perspective.
Authors: Firas Hadj Kacem, Ahmed Khanfir, Mike Papadakis
Abstract: Bias in financial language models constitutes a major obstacle to their adoption in real-world applications. Detecting such bias is challenging, as it requires identifying inputs whose predictions change when varying properties unrelated to the decision, such as demographic attributes. Existing approaches typically rely on exhaustive mutation and pairwise prediction analysis over large corpora, which is effective but computationally expensive-particularly for large language models and can become impractical in continuous retraining and releasing processes. Aiming at reducing this cost, we conduct a large-scale study of bias in five financial language models, examining similarities in their bias tendencies across protected attributes and exploring cross-model-guided bias detection to identify bias-revealing inputs earlier. Our study uses approximately 17k real financial news sentences, mutated to construct over 125k original-mutant pairs. Results show that all models exhibit bias under both atomic (0.58\%-6.05\%) and intersectional (0.75\%-5.97\%) settings. Moreover, we observe consistent patterns in bias-revealing inputs across models, enabling substantial reuse and cost reduction in bias detection. For example, up to 73\% of FinMA's biased behaviours can be uncovered using only 20\% of the input pairs when guided by properties derived from DistilRoBERTa outputs.
Authors: Hamish Flynn, Joe Watson, Ingmar Posner, Jan Peters
Abstract: We analyze the Bayesian regret of the Gaussian process posterior sampling reinforcement learning (GP-PSRL) algorithm. Posterior sampling is an effective heuristic for decision-making under uncertainty that has been used to develop successful algorithms for a variety of continuous control problems. However, theoretical work on GP-PSRL is limited. All known regret bounds either fail to achieve a tight dependence on a kernel-dependent quantity called the maximum information gain or fail to properly account for the fact that the set of possible system states is unbounded. Through a recursive application of the Borell-Tsirelson-Ibragimov-Sudakov inequality, we show that, with high probability, the states actually visited by the algorithm are contained within a ball of near-constant radius. To obtain tight dependence on the maximum information gain, we use the chaining method to control the regret suffered by GP-PSRL. Our main result is a Bayesian regret bound of the order $\widetilde{\mathcal{O}}(H^{3/2}\sqrt{\gamma_{T/H} T})$, where $H$ is the horizon, $T$ is the number of time steps and $\gamma_{T/H}$ is the maximum information gain. With this result, we resolve the limitations with prior theoretical work on PSRL, and provide the theoretical foundation and tools for analyzing PSRL in complex settings.
Authors: Francesco Della Santa, Sandra Pieraccini, Maria Strazzullo
Abstract: This work addresses the accurate and efficient simulation of physical phenomena governed by parametric Partial Differential Equations (PDEs) characterized by varying boundary conditions, where parametric instances modify not only the physics of the problem but also the imposition of boundary constraints on the computational domain. In such scenarios, classical Galerkin projection-based reduced order techniques encounter a fundamental bottleneck. Parametric boundaries typically necessitate a re-formulation of the discrete problem for each new configuration, and often, these approaches are unsuitable for real-time applications. To overcome these limitations, we propose a novel methodology based on Graph-Instructed Neural Networks (GINNs). The GINN framework effectively learns the mapping between the parametric description of the computational domain and the corresponding PDE solution. Our results demonstrate that the proposed GINN-based models, can efficiently represent highly complex parametric PDEs, serving as a robust and scalable asset for several applied-oriented settings when compared with fully connected architectures.
Authors: Yehonatan Elisha, Oren Barkan, Noam Koenigstein
Abstract: Vision Transformers (ViTs) often degrade under distribution shifts because they rely on spurious correlations, such as background cues, rather than semantically meaningful features. Existing regularization methods, typically relying on simple foreground-background masks, which fail to capture the fine-grained semantic concepts that define an object (e.g., ``long beak'' and ``wings'' for a ``bird''). As a result, these methods provide limited robustness to distribution shifts. To address this limitation, we introduce a novel finetuning framework that steers model reasoning toward concept-level semantics. Our approach optimizes the model's internal relevance maps to align with spatially grounded concept masks. These masks are generated automatically, without manual annotation: class-relevant concepts are first proposed using an LLM-based, label-free method, and then segmented using a VLM. The finetuning objective aligns relevance with these concept regions while simultaneously suppressing focus on spurious background areas. Notably, this process requires only a minimal set of images and uses half of the dataset classes. Extensive experiments on five out-of-distribution benchmarks demonstrate that our method improves robustness across multiple ViT-based models. Furthermore, we show that the resulting relevance maps exhibit stronger alignment with semantic object parts, offering a scalable path toward more robust and interpretable vision models. Finally, we confirm that concept-guided masks provide more effective supervision for model robustness than conventional segmentation maps, supporting our central hypothesis.
Authors: Gijs van Seeventer, Saber Salehkaleybar
Abstract: We study identifiability in continuous-time linear stationary stochastic differential equations with known causal structure. Unlike existing approaches, we relax the assumption of a known diffusion matrix, thereby respecting the model's intrinsic scale invariance. Rather than recovering drift coefficients themselves, we introduce edge-sign identifiability: for a given causal structure, we ask whether the sign of a given drift entry is uniquely determined across all observational covariance matrices induced by parametrizations compatible with that structure. Under a notion of faithfulness, we derive criteria for characterising identifiability, non-identifiability, and partial identifiability for general graphs. Applying our criteria to specific causal structures, both analogous to classical causal settings (e.g., instrumental variables) and novel cyclic settings, we determine their edge-sign identifiability and, in some cases, obtain explicit expressions for the sign of a target edge in terms of the observational covariance matrix.
Authors: Mina Jamshidi Idaji, Julius Hense, Tom Neuh\"auser, Augustin Krause, Yanqing Luo, Oliver Eberle, Thomas Schnake, Laure Ciernik, Farnoush Rezaei Jafari, Reza Vahidimajd, Jonas Dippel, Christoph Walz, Frederick Klauschen, Andreas Mock, Klaus-Robert M\"uller
Abstract: Multiple instance learning (MIL) has enabled substantial progress in computational histopathology, where a large amount of patches from gigapixel whole slide images are aggregated into slide-level predictions. Heatmaps are widely used to validate MIL models and to discover tissue biomarkers. Yet, the validity of these heatmaps has barely been investigated. In this work, we introduce a general framework for evaluating the quality of MIL heatmaps without requiring additional labels. We conduct a large-scale benchmark experiment to assess six explanation methods across histopathology task types (classification, regression, survival), MIL model architectures (Attention-, Transformer-, Mamba-based), and patch encoder backbones (UNI2, Virchow2). Our results show that explanation quality mostly depends on MIL model architecture and task type, with perturbation ("Single"), layer-wise relevance propagation (LRP), and integrated gradients (IG) consistently outperforming attention-based and gradient-based saliency heatmaps, which often fail to reflect model decision mechanisms. We further demonstrate the advanced capabilities of the best-performing explanation methods: (i) We provide a proof-of-concept that MIL heatmaps of a bulk gene expression prediction model can be correlated with spatial transcriptomics for biological validation, and (ii) showcase the discovery of distinct model strategies for predicting human papillomavirus (HPV) infection from head and neck cancer slides. Our work highlights the importance of validating MIL heatmaps and establishes that improved explainability can enable more reliable model validation and yield biological insights, making a case for a broader adoption of explainable AI in digital pathology. Our code is provided in a public GitHub repository: https://github.com/bifold-pathomics/xMIL/tree/xmil-journal
URLs: https://github.com/bifold-pathomics/xMIL/tree/xmil-journal
Authors: Sucheta Ghosh, Zahra Monfared
Abstract: Electrocardiogram (ECG) analysis is vital for detecting cardiac abnormalities, yet robust automated classification is challenging due to the complexity and variability of physiological signals. In this work, we investigate transformer-based ECG classification using features derived from the Koopman operator and wavelet transforms. Two tasks are studied: (1) binary classification (Normal vs. Non-normal), and (2) four-class classification (Normal, Atrial Fibrillation, Ventricular Arrhythmia, Block). We use Extended Dynamic Mode Decomposition (EDMD) to approximate the Koopman operator. Our results show that wavelet features excel in binary classification, while Koopman features, when paired with transformers, achieve superior performance in the four-class setting. A simple hybrid of Koopman and wavelet features does not improve accuracy. However, selecting an appropriate EDMD dictionary -- specifically a radial basis function dictionary with tuned parameters -- yields significant gains, surpassing the wavelet-only baseline and the hybrid wavelet-Koopman system. We also present a Koopman-based reconstruction analysis for interpretable insights into the learned dynamics and compare against a recurrent neural network baseline. Overall, our findings demonstrate the effectiveness of Koopman-based feature learning with transformers and highlight promising directions for integrating dynamical systems theory into time-series classification.
Authors: Olivier Jeunen
Abstract: Continuous and efficient experimentation is key to the practical success of user-facing applications on the web, both through online A/B-tests and off-policy evaluation. Despite their shared objective -- estimating the incremental value of a treatment -- these domains often operate in isolation, utilising distinct terminologies and statistical toolkits. This paper bridges that divide by establishing a formal equivalence between their canonical variance reduction methods. We prove that the standard online Difference-in-Means estimator is mathematically identical to an off-policy Inverse Propensity Scoring estimator equipped with an optimal (variance-minimising) additive control variate. Extending this unification, we demonstrate that widespread regression adjustment methods (such as CUPED, CUPAC, and ML-RATE) are structurally equivalent to Doubly Robust estimation. This unified view extends our understanding of commonly used approaches, and can guide practitioners and researchers working on either class of problems.
Authors: Yatong Chen, Guanhua Zhang, Moritz Hardt
Abstract: Influential benchmarks incentivize competing model developers to strategically allocate post-training resources toward improvements on the leaderboard, a phenomenon dubbed benchmaxxing or training on the test task. In this work, we initiate a principled study of the incentive structure that benchmarks induce. We model benchmarking as a Stackelberg game between a benchmark designer who chooses an evaluation protocol and multiple model developers who compete simultaneously in a subgame given by the designer's choice. Each competitor has a model of unknown latent quality and can inflate its observed score by allocating resources to benchmark-specific improvements. First, we prove that current benchmarks induce games for which no Nash equilibrium between model developers exists. This result suggests one explanation for why current practice leads to misaligned incentives, prompting model developers to strategize in opaque ways. However, we prove that under mild conditions, a recently proposed evaluation protocol, called tune-before-test, induces a benchmark with a unique Nash equilibrium that ranks models by latent quality. This positive result demonstrates that benchmarks need not set bad incentives, even if current evaluations do.
Authors: Liyuan Mao, Le Yu, Jing Zhou, Chujie Zheng, Bowen Yu, Chang Gao, Shixuan Liu, An Yang, Weinan Zhang, JunYang Lin
Abstract: In this work, we reveal that Large Language Models (LLMs) possess intrinsic behavioral plasticity-akin to chameleons adapting their coloration to environmental cues-that can be exposed through token-conditional generation and stabilized via reinforcement learning. Specifically, by conditioning generation on carefully selected token prefixes sampled from responses exhibiting desired behaviors, LLMs seamlessly adapt their behavioral modes at inference time (e.g., switching from step-by-step reasoning to direct answering) without retraining. Based on this insight, we propose Token-Conditioned Reinforcement Learning (ToCoRL), a principled framework that leverages RL to internalize this chameleon-like plasticity, transforming transient inference-time adaptations into stable and learnable behavioral patterns. ToCoRL guides exploration with token-conditional generation and keep enhancing exploitation, enabling emergence of appropriate behaviors. Extensive experiments show that ToCoRL enables precise behavioral control without capability degradation. Notably, we show that large reasoning models, while performing strongly on complex mathematics, can be effectively adapted to excel at factual question answering, which was a capability previously hindered by their step-by-step reasoning patterns.
Authors: Xi Mo
Abstract: This paper presents IronEngine, a general AI assistant platform organized around a unified orchestration core that connects a desktop user interface, REST and WebSocket APIs, Python clients, local and cloud model backends, persistent memory, task scheduling, reusable skills, 24-category tool execution, MCP-compatible extensibility, and hardware-facing integration. IronEngine introduces a three-phase pipeline -- Discussion (Planner--Reviewer collaboration), Model Switch (VRAM-aware transition), and Execution (tool-augmented action loop) -- that separates planning quality from execution capability. The system features a hierarchical memory architecture with multi-level consolidation, a vectorized skill repository backed by ChromaDB, an adaptive model management layer supporting 92 model profiles with VRAM-aware context budgeting, and an intelligent tool routing system with 130+ alias normalization and automatic error correction. We present experimental results on file operation benchmarks achieving 100\% task completion with a mean total time of 1541 seconds across four heterogeneous tasks, and provide detailed comparisons with representative AI assistant systems including ChatGPT, Claude Desktop, Cursor, Windsurf, and open-source agent frameworks. Without disclosing proprietary prompts or core algorithms, this paper analyzes the platform's architectural decomposition, subsystem design, experimental performance, safety boundaries, and comparative engineering advantages. The resulting study positions IronEngine as a system-oriented foundation for general-purpose personal assistants, automation frameworks, and future human-centered agent platforms.
Authors: Peter Brodeur, Jacob M. Koshy, Anil Palepu, Khaled Saab, Ava Homiar, Roma Ruparel, Charles Wu, Ryutaro Tanno, Joseph Xu, Amy Wang, David Stutz, Hannah M. Ferrera, David Barrett, Lindsey Crowley, Jihyeon Lee, Spencer E. Rittner, Ellery Wulczyn, Selena K. Zhang, Elahe Vedadi, Christine G. Kohn, Kavita Kulkarni, Vinay Kadiyala, Sara Mahdavi, Wendy Du, Jessica Williams, David Feinbloom, Renee Wong, Tao Tu, Petar Sirkovic, Alessio Orlandi, Christopher Semturs, Yun Liu, Juraj Gottweis, Dale R. Webster, Jo\"elle Barral, Katherine Chou, Pushmeet Kohli, Avinatan Hassidim, Yossi Matias, James Manyika, Rob Fields, Jonathan X. Li, Marc L. Cohen, Vivek Natarajan, Mike Schaekermann, Alan Karthikesalingam, Adam Rodman
Abstract: Large language model (LLM)-based AI systems have shown promise for patient-facing diagnostic and management conversations in simulated settings. Translating these systems into clinical practice requires assessment in real-world workflows with rigorous safety oversight. We report a prospective, single-arm feasibility study of an LLM-based conversational AI, the Articulate Medical Intelligence Explorer (AMIE), conducting clinical history taking and presentation of potential diagnoses for patients to discuss with their provider at urgent care appointments at a leading academic medical center. 100 adult patients completed an AMIE text-chat interaction up to 5 days before their appointment. We sought to assess the conversational safety and quality, patient and clinician experience, and clinical reasoning capabilities compared to primary care providers (PCPs). Human safety supervisors monitored all patient-AMIE interactions in real time and did not need to intervene to stop any consultations based on pre-defined criteria. Patients reported high satisfaction and their attitudes towards AI improved after interacting with AMIE (p < 0.001). PCPs found AMIE's output useful with a positive impact on preparedness. AMIE's differential diagnosis (DDx) included the final diagnosis, per chart review 8 weeks post-encounter, in 90% of cases, with 75% top-3 accuracy. Blinded assessment of AMIE and PCP DDx and management (Mx) plans suggested similar overall DDx and Mx plan quality, without significant differences for DDx (p = 0.6) and appropriateness and safety of Mx (p = 0.1 and 1.0, respectively). PCPs outperformed AMIE in the practicality (p = 0.003) and cost effectiveness (p = 0.004) of Mx. While further research is needed, this study demonstrates the initial feasibility, safety, and user acceptance of conversational AI in a real-world setting, representing crucial steps towards clinical translation.
Authors: Zhe Hong
Abstract: When an RL agent's observations are gradually corrupted, at what drift rate does it "wake up" -- and what determines this boundary? We study world model-based self-monitoring under continuous observation drift across four MuJoCo environments, three detector families (z-score, variance, percentile), and three model capacities. We find that (1) a sharp detection threshold $\varepsilon^*$ exists universally: below it, drift is absorbed as normal variation; above it, detection occurs rapidly. The threshold's existence and sigmoid shape are invariant across all detector families and model capacities, though its position depends on the interaction between detector sensitivity, noise floor structure, and environment dynamics. (2) Sinusoidal drift is completely undetectable by all detector families -- including variance and percentile detectors with no temporal smoothing -- establishing this as a world model property rather than a detector artifact. (3) Within each environment, $\varepsilon^*$ follows a power law in detector parameters ($R^2 = 0.89$-$0.97$), but cross-environment prediction fails ($R^2 = 0.45$), revealing that the missing variable is environment-specific dynamics structure $\partial \mathrm{PE}/\partial\varepsilon$. (4) In fragile environments, agents collapse before any detector can fire ("collapse before awareness"), creating a fundamentally unmonitorable failure mode. Our results reframe $\varepsilon^*$ from an emergent world model property to a three-way interaction between noise floor, detector, and environment dynamics, providing a more defensible and empirically grounded account of self-monitoring boundaries in RL agents.
Authors: Andrei Starodubov, Yaqub Aris Prabowo, Andreas Hadjipieris, Ioannis Kyriakides, Roberto Galeazzi
Abstract: Robust single-vessel tracking from fixed coastal platforms is hindered by modality-specific degradations: cameras suffer from illumination and visual clutter, while LiDAR performance drops with range and intermittent returns. We present a heterogeneous multi-sensor fusion particle-filter tracker that incorporates an information-gain (entropy-reduction) adaptive sensing policy to select the most informative configuration at each fusion time bin. The approach is validated in a real maritime deployment at the CMMI Smart Marina Testbed (Ayia Napa Marina, Cyprus), using a shore-mounted 3D LiDAR and an elevated fixed camera to track a rigid inflatable boat with onboard GNSS ground truth. We compare LiDAR-only, camera-only, all-sensors, and adaptive configurations. Results show LiDAR dominates near-field accuracy, the camera sustains longer-range coverage when LiDAR becomes unavailable, and the adaptive policy achieves a favorable accuracy-continuity trade-off by switching modalities based on information gain. By avoiding continuous multi-stream processing, the adaptive configuration provides a practical baseline for resilient and resource-aware maritime surveillance.
Authors: Shreya Das, Kundan Kumar, Muhammad Iqbal, Outi Savolainen, Dominik Baumann, Laura Ruotsalainen, Simo S\"arkk\"a
Abstract: Model-based reinforcement learning (MBRL) is sample-efficient but depends on the accuracy of the learned dynamics, which are often modeled using black-box methods that do not adhere to physical laws. Those methods tend to produce inaccurate predictions when presented with data that differ from the original training set. In this work, we employ Lagrangian neural networks (LNNs), which enforce an underlying Lagrangian structure to train the model within a Dyna-based MBRL framework. Furthermore, we train the LNN using stochastic gradient-based and state-estimation-based optimizers to learn the network's weights. The state-estimation-based method converges faster than the stochastic gradient-based method during neural network training. Simulation results are provided to illustrate the effectiveness of the proposed LNN-based Dyna framework for MBRL.
Authors: Prakrut Kotecha, Ganga Nair B, Shishir Kolathaya
Abstract: Robotic systems operating in unstructured environments must operate under significant uncertainty arising from intermittent contacts, frictional variability, and unmodeled compliance. While recent model-free approaches have demonstrated impressive performance, many deployment settings still require predictive models that support planning, constraint handling, and online adaptation. Analytical rigid-body models provide strong physical structure but often fail to capture complex interaction effects, whereas purely data-driven models may violate physical consistency, exhibit data bias, and accumulate long-horizon drift. In this work, we propose STRIDE, a dynamics learning framework that explicitly separates conservative rigid-body mechanics from uncertain, effectively stochastic non-conservative interaction effects. The structured component is modeled using a Lagrangian Neural Network (LNN) to preserve energy-consistent inertial dynamics, while residual interaction forces are represented using Conditional Flow Matching (CFM) to capture multi-modal interaction phenomena. The two components are trained jointly end-to-end, enabling the model to retain physical structure while representing complex stochastic behavior. We evaluate STRIDE on systems of increasing complexity, including a pendulum, the Unitree Go1 quadruped, and the Unitree G1 humanoid. Results show 20% reduction in long-horizon prediction error and 30% reduction in contact force prediction error compared to deterministic residual baselines, supporting more reliable model-based control in uncertain robotic environments.
Authors: Youngseo Kim, Kwan Yun, Seokhyeon Hong, Sihun Cha, Colette Suhjung Koo, Junyong Noh
Abstract: The surge of highly realistic synthetic videos produced by contemporary generative systems has significantly increased the risk of malicious use, challenging both humans and existing detectors. Against this backdrop, we take a generator-side view and observe that internal cross-attention mechanisms in these models encode fine-grained speech-motion alignment, offering useful correspondence cues for forgery detection. Building on this insight, we propose X-AVDT, a robust and generalizable deepfake detector that probes generator-internal audio-visual signals accessed via DDIM inversion to expose these cues. X-AVDT extracts two complementary signals: (i) a video composite capturing inversion-induced discrepancies, and (ii) an audio-visual cross-attention feature reflecting modality alignment enforced during generation. To enable faithful cross-generator evaluation, we further introduce MMDF, a new multimodal deepfake dataset spanning diverse manipulation types and rapidly evolving synthesis paradigms, including GANs, diffusion, and flow-matching. Extensive experiments demonstrate that X-AVDT achieves leading performance on MMDF and generalizes strongly to external benchmarks and unseen generators, outperforming existing methods with accuracy improved by 13.1%. Our findings highlight the importance of leveraging internal audio-visual consistency cues for robustness to future generators in deepfake detection.
Authors: Jonathan Wurth, Helena Stegherr, Neele Kemper, Michael Heider, J\"org H\"ahner
Abstract: Selecting an optimization algorithm requires comparing candidates across problem instances, but the computational budget for deployment is often unknown at benchmarking time. Current methods either collapse anytime performance into a scalar, require manual interpretation of plots, or produce conclusions that change when algorithms are added or removed. Moreover, methods based on raw objective values require normalization, which needs bounds or optima that are often unavailable and breaks coherent aggregation across instances. We propose a framework that formulates anytime algorithm comparison as Pareto optimization over time: an algorithm is non-dominated if no competitor beats it at every timepoint. By using rankings rather than objective values, our approach requires no bounds, no normalization, and aggregates coherently across arbitrary instance distributions without requiring known optima. We introduce PolarBear (Pareto-optimal anytime algorithms via Bayesian racing), a procedure that identifies the anytime Pareto set through adaptive sampling with calibrated uncertainty. Bayesian inference over a temporal Plackett-Luce ranking model provides posterior beliefs about pairwise dominance, enabling early elimination of confidently dominated algorithms. The output Pareto set together with the posterior supports downstream algorithm selection under arbitrary time preferences and risk profiles without additional experiments.
Authors: Gabriele Somaschini, Adrian R\"ofer, Abhinav Valada
Abstract: Object co-occurrences provide a key cue for finding objects successfully and efficiently in unfamiliar environments. Typically, one looks for cups in kitchens and views fridges as evidence of being in a kitchen. Such priors have also been exploited in artificial agents, but they are typically learned from explicitly labeled data or queried from language models. It is still unclear whether these relations can be learned implicitly from unlabeled observations alone. In this work, we address this problem and propose ProReFF, a feature field model trained to predict relative distributions of features obtained from pre-trained vision language models. In addition, we introduce a learning-based strategy that enables training from unlabeled and potentially contradictory data by aligning inconsistent observations into a coherent relative distribution. For the downstream object search task, we propose an agent that leverages predicted feature distributions as a semantic prior to guide exploration toward regions with a high likelihood of containing the object. We present extensive evaluations demonstrating that ProReFF captures meaningful relative feature distributions in natural scenes and provides insight into the impact of our proposed alignment step. We further evaluate the performance of our search agent in 100 challenges in the Matterport3D simulator, comparing with feature-based baselines and human participants. The proposed agent is 20% more efficient than the strongest baseline and achieves up to 80% of human performance.
Authors: Yixuan Wang, Rhythm Syed, Fangyu Wu, Mengchao Zhang, Aykut Onol, Jose Barreiros, Hooshang Nayyeri, Tony Dear, Huan Zhang, Yunzhu Li
Abstract: Action-conditioned video prediction models (often referred to as world models) have shown strong potential for robotics applications, but existing approaches are often slow and struggle to capture physically consistent interactions over long horizons, limiting their usefulness for scalable robot policy training and evaluation. We present Interactive World Simulator, a framework for building interactive world models from a moderate-sized robot interaction dataset. Our approach leverages consistency models for both image decoding and latent-space dynamics prediction, enabling fast and stable simulation of physical interactions. In our experiments, the learned world models produce interaction-consistent pixel-level predictions and support stable long-horizon interactions for more than 10 minutes at 15 FPS on a single RTX 4090 GPU. Our framework enables scalable demonstration collection solely within the world models to train state-of-the-art imitation policies. Through extensive real-world evaluation across diverse tasks involving rigid objects, deformable objects, object piles, and their interactions, we find that policies trained on world-model-generated data perform comparably to those trained on the same amount of real-world data. Additionally, we evaluate policies both within the world models and in the real world across diverse tasks, and observe a strong correlation between simulated and real-world performance. Together, these results establish the Interactive World Simulator as a stable and physically consistent surrogate for scalable robotic data generation and faithful, reproducible policy evaluation.
Authors: Saeed Asadi, Jonathan Yu-Meng Li
Abstract: We propose Generative Adversarial Regression (GAR), a framework for learning conditional risk scenarios through generators aligned with downstream risk objectives. GAR builds on a regression characterization of conditional risk for elicitable functionals, including quantiles, expectiles, and jointly elicitable pairs. We extend this principle from point prediction to generative modeling by training generators whose policy-induced risk matches that of real data under the same context. To ensure robustness across all policies, GAR adopts a minimax formulation in which an adversarial policy identifies worst-case discrepancies in risk evaluation while the generator adapts to eliminate them. This structure preserves alignment with the risk functional across a broad class of policies rather than a fixed, pre-specified set. We illustrate GAR through a tail-risk instantiation based on jointly elicitable $(\mathrm{VaR}, \mathrm{ES})$ objectives. Experiments on S\&P 500 data show that GAR produces scenarios that better preserve downstream risk than unconditional, econometric, and direct predictive baselines while remaining stable under adversarially selected policies.
Authors: Aravind R. Iyengar
Abstract: The question of \emph{knowledge}, \emph{truth} and \emph{trust} is explored via a mathematical formulation of claims and sources. We define truth as the reproducibly perceived subset of knowledge, formalize sources as having both generative and discriminative roles, and develop a framework for reputation grounded in the \emph{conviction} -- the likelihood that a source's stance is vindicated by independent consensus. We argue that conviction, rather than correctness or faithfulness, is the principled basis for trust: it is regime-independent, rewards genuine contribution, and demands the transparent and self-sufficient perceptions that make external verification possible. We formalize reputation as the expected weighted signed conviction over a realm of claims, characterize its behavior across source-claim regimes, and identify continuous verification as both a theoretical necessity and a practical mechanism through which reputation accrues. The framework is applied to AI agents, which are identified as capable but error-prone sources for whom verifiable conviction and continuously accrued reputation constitute the only robust foundation for trust.
Authors: Ben Rank, Hardik Bhatnagar, Ameya Prabhu, Shira Eisenberg, Karina Nguyen, Matthias Bethge, Maksym Andriushchenko
Abstract: AI agents have become surprisingly proficient at software engineering over the past year, largely due to improvements in reasoning capabilities. This raises a deeper question: can these systems extend their capabilities to automate AI research itself? In this paper, we explore post-training, the critical phase that turns base LLMs into useful assistants. We introduce PostTrainBench to benchmark how well LLM agents can perform post-training autonomously under bounded compute constraints (10 hours on one H100 GPU). We ask frontier agents (e.g., Claude Code with Opus 4.6) to optimize the performance of a base LLM on a particular benchmark (e.g., Qwen3-4B on AIME). Importantly, we do not provide any predefined strategies to the agents and instead give them full autonomy to find necessary information on the web, run experiments, and curate data. We find that frontier agents make substantial progress but generally lag behind instruction-tuned LLMs from leading providers: 23.2% for the best agent vs. 51.1% for official instruction-tuned models. However, agents can exceed instruction-tuned models in targeted scenarios: GPT-5.1 Codex Max achieves 89% on BFCL with Gemma-3-4B vs. 67% for the official model. We also observe several failure modes worth flagging. Agents sometimes engage in reward hacking: training on the test set, downloading existing instruction-tuned checkpoints instead of training their own, and using API keys they find to generate synthetic data without authorization. These behaviors are concerning and highlight the importance of careful sandboxing as these systems become more capable. Overall, we hope PostTrainBench will be useful for tracking progress in AI R&D automation and for studying the risks that come with it. Website and code are available at https://posttrainbench.com/.
Authors: Matan Levy, Gavriel Habib, Issar Tzachor, Dvir Samuel, Rami Ben-Ari, Nir Darshan, Or Litany, Dani Lischinski
Abstract: Template-free animatable head avatars can achieve high visual fidelity by learning expression-dependent facial deformation directly from a subject's capture, avoiding parametric face templates and hand-designed blendshape spaces. However, since learned deformation is supervised only by the expressions observed for a single identity, these models suffer from limited expression coverage and often struggle when driven by motions that deviate from the training distribution. We introduce RAF (Retrieval-Augmented Faces), a simple training-time augmentation designed for template-free head avatars that learn deformation from data. RAF constructs a large unlabeled expression bank and, during training, replaces a subset of the subject's expression features with nearest-neighbor expressions retrieved from this bank while still reconstructing the subject's original frames. This exposes the deformation field to a broader range of expression conditions, encouraging stronger identity-expression decoupling and improving robustness to expression distribution shift without requiring paired cross-identity data, additional annotations, or architectural changes. We further analyze how retrieval augmentation increases expression diversity and validate retrieval quality with a user study showing that retrieved neighbors are perceptually closer in expression and pose. Experiments on the NeRSemble benchmark demonstrate that RAF consistently improves expression fidelity over the baseline, in both self-driving and cross-driving scenarios.
Authors: Matteo Argenton, Laura Cappelli, Concezio Bozzi
Abstract: In the forthcoming years the LHC experiments are going to be upgraded to benefit from the substantial increase of the LHC instantaneous luminosity, which will lead to larger, denser events, and, consequently, greater complexity in reconstructing charged particle tracks, motivating frontier research in new technologies. Quantum machine learning models are being investigated as potential new approaches to high energy physics (HEP) tasks. We characterize and upgrade a quantum graph neural network (QGNN) architecture for charged particle track reconstruction on a simulated high luminosity dataset. The model operates on a set of event graphs, each built from the hits generated in tracking detector layers by particles produced in proton collisions, performing a classification of the possible hit connections between adjacent layers. In this approach the QGNN is designed as a hybrid architecture, interleaving classical feedforward networks with parametrized quantum circuits. We characterize the interplay between the classical and quantum components. We report on the principal upgrades to the original design, and present new evidence of improved training behavior, specifically in terms of convergence toward the final trained configuration.
Authors: Adam Rozzio, Rafael Athanasiades, O. Deniz Akyildiz
Abstract: Maximum marginal likelihood estimation (MMLE) can be formulated as the optimization of a free energy functional. From this viewpoint, the Expectation-Maximisation (EM) algorithm admits a natural interpretation as a coordinate descent method over the joint space of model parameters and probability measures. Recently, a significant body of work has adopted this perspective, leading to interacting particle algorithms for MMLE. In this paper, we propose an accelerated version of one such procedure, based on Stein variational gradient descent (SVGD), by introducing Nesterov acceleration in both the parameter updates and in the space of probability measures. The resulting method, termed Momentum SVGD-EM, consistently accelerates convergence in terms of required iterations across various tasks of increasing difficulty, demonstrating effectiveness in both low- and high-dimensional settings.
Authors: Simon Bing, Jonas Wahl, Jakob Runge
Abstract: We introduce structural causal bottleneck models (SCBMs), a novel class of structural causal models. At the core of SCBMs lies the assumption that causal effects between high-dimensional variables only depend on low-dimensional summary statistics, or bottlenecks, of the causes. SCBMs provide a flexible framework for task-specific dimension reduction while being estimable via standard, simple learning algorithms in practice. We analyse identifiability in SCBMs, connect them to information bottlenecks in the sense of Tishby & Zaslavsky (2015), and illustrate how to estimate them experimentally. We also demonstrate the benefit of bottlenecks for effect estimation in low-sample transfer learning settings. We argue that SCBMs provide an alternative to existing causal dimension reduction frameworks like causal representation learning or causal abstraction learning.
Authors: Phillip Long, Zachary Novack, Chris Donahue
Abstract: Autoregressive "language" models (LMs) trained on raw waveforms can be repurposed for lossless audio compression, but prior work is limited to 8-bit audio, leaving open whether such approaches work for practical settings (16/24-bit) and can compete with existing codecs. We benchmark LM-based compression on full-fidelity audio across diverse domains (music, speech, bioacoustics), sampling rates (16kHz-48kHz), and bit depths (8, 16, 24-bit). Standard sample-level tokenization becomes intractable at higher bit depths due to vocabulary size (65K for 16-bit; 16.7M for 24-bit). We propose Trilobyte, a byte-level tokenization schema for full resolution audio, improving vocabulary scaling from $O(2^{b})$ to $O(1)$ and enabling the first tractable 24-bit LM-based lossless compression. While LMs consistently outperform FLAC and yield state-of-the-art compression at 8-bit and 16-bit, we observe that compression gains become more modest as bit depth increases beyond 8-bit.
Authors: Weize Liu, Minghui Liu, Sy-Tuyen Ho, Souradip Chakraborty, Xiyao Wang, Furong Huang
Abstract: Training large language models (LLMs) as autonomous agents often begins with imitation learning, but it only teaches agents what to do without understanding why: agents never contrast successful actions against suboptimal alternatives and thus lack awareness of action quality. Recent approaches attempt to address this by introducing self-reflection supervision derived from contrasts between expert and alternative actions. However, the training paradigm fundamentally remains imitation learning: the model imitates pre-constructed reflection text rather than learning to reason autonomously. We propose Agentic Critical Training (ACT), a reinforcement learning paradigm that trains agents to identify the better action among alternatives. By rewarding whether the model's judgment is correct, ACT drives the model to autonomously develop reasoning about action quality, producing genuine self-reflection rather than imitating it. Across three challenging agent benchmarks, ACT consistently improves agent performance when combined with different post-training methods. It achieves an average improvement of 5.07 points over imitation learning and 4.62 points over reinforcement learning. Compared to approaches that inject reflection capability through knowledge distillation, ACT also demonstrates clear advantages, yielding an average improvement of 2.42 points. Moreover, ACT enables strong out-of-distribution generalization on agentic benchmarks and improves performance on general reasoning benchmarks without any reasoning-specific training data, highlighting the value of our method. These results suggest that ACT is a promising path toward developing more reflective and capable LLM agents.
Authors: Mikhail Hushchyn, Kenenbek Arzymatov, Denis Derkach
Abstract: Moments when a time series changes its behavior are called change points. Occurrence of change point implies that the state of the system is altered and its timely detection might help to prevent unwanted consequences. In this paper, we present two change-point detection approaches based on neural networks and online learning. These algorithms demonstrate linear computational complexity and are suitable for change-point detection in large time series. We compare them with the best known algorithms on various synthetic and real world data sets. Experiments show that the proposed methods outperform known approaches. We also prove the convergence of the algorithms to the optimal solutions and describe conditions rendering current approach more powerful than offline one.
Authors: Tejas Y. Deo, B. B. Deshmukh, Keshav H. Jatakar, Kamlesh M. Chhajed, S. S. Pardeshi, R. Jegadeeshwaran, Apoorva N. Khairnar, Hrushikesh S. Khade, A. D. Patange
Abstract: In this paper, a white-Box support vector machine (SVM) framework and its swarm-based optimization is presented for supervision of toothed milling cutter through characterization of real-time spindle vibrations. The anomalous moments of vibration evolved due to in-process tool failures (i.e., flank and nose wear, crater and notch wear, edge fracture) have been investigated through time-domain response of acceleration and statistical features. The Recursive Feature Elimination with Cross-Validation (RFECV) with decision trees as the estimator has been implemented for feature selection. Further, the competence of standard SVM has been examined for tool health monitoring followed by its optimization through application of swarm based algorithms. The comparative analysis of performance of five meta-heuristic algorithms (Elephant Herding Optimization, Monarch Butterfly Optimization, Harris Hawks Optimization, Slime Mould Algorithm, and Moth Search Algorithm) has been carried out. The white-box approach has been presented considering global and local representation that provides insight into the performance of machine learning models in tool condition monitoring.
Authors: Reza Refaei Afshar, Joaquin Vanschoren, Uzay Kaymak, Rui Zhang, Yaoxin Wu, Wen Song, Yingqian Zhang
Abstract: Reinforcement Learning and, recently, Deep Reinforcement Learning are popular methods for solving sequential decision-making problems modeled as Markov Decision Processes. RL modeling of a problem and selecting algorithms and hyper-parameters require careful consideration, as different configurations may entail completely different performances. These considerations are mainly the task of RL experts; however, RL is progressively becoming popular in other fields, such as combinatorial optimization, where researchers and system designers are not necessarily RL experts. Besides, many modeling decisions are typically made manually, such as defining state and action space, size of batches, batch update frequency, and time steps. For these reasons, automating different components of RL is of great importance, and it has attracted much attention in recent years. Automated RL provides a framework in which different components of RL, including MDP modeling, algorithm selection, and hyper-parameter optimization, are modeled and defined automatically. In this article, we present the literature on automated RL (AutoRL), including the recent large language model (LLM) based techniques. We also discuss the recent work on techniques that are not presently tailored for automated RL but hold promise for future integration into AutoRL. Furthermore, we discuss the challenges, open questions, and research directions in AutoRL.
Authors: Michael Franklin Mbouopda (LIMOS, UCA), Emille E. O. Ishida (LIMOS, UCA), Engelbert Mephu Nguifo (LIMOS, UCA), Emmanuel Gangler (LPC, UCA)
Abstract: Exploring the expansion history of the universe, understanding its evolutionary stages, and predicting its future evolution are important goals in astrophysics. Today, machine learning tools are used to help achieving these goals by analyzing transient sources, which are modeled as uncertain time series. Although black-box methods achieve appreciable performance, existing interpretable time series methods failed to obtain acceptable performance for this type of data. Furthermore, data uncertainty is rarely taken into account in these methods. In this work, we propose an uncertaintyaware subsequence based model which achieves a classification comparable to that of state-of-the-art methods. Unlike conformal learning which estimates model uncertainty on predictions, our method takes data uncertainty as additional input. Moreover, our approach is explainable-by-design, giving domain experts the ability to inspect the model and explain its predictions. The explainability of the proposed method has also the potential to inspire new developments in theoretical astrophysics modeling by suggesting important subsequences which depict details of light curve shapes. The dataset, the source code of our experiment, and the results are made available on a public repository.
Authors: Yi Tian, Kaiqing Zhang, Russ Tedrake, Suvrit Sra
Abstract: We study the task of learning state representations from potentially high-dimensional observations, with the goal of controlling an unknown partially observable system. We pursue a cost-driven approach, where a dynamic model in some latent state space is learned by predicting the costs without predicting the observations or actions. In particular, we focus on an intuitive cost-driven state representation learning method for solving Linear Quadratic Gaussian (LQG) control, one of the most fundamental partially observable control problems. As our main results, we establish finite-sample guarantees of finding a near-optimal state representation function and a near-optimal controller using the directly learned latent model, for finite-horizon time-varying LQG control problems. To the best of our knowledge, despite various empirical successes, finite-sample guarantees of such a cost-driven approach remain elusive. Our result underscores the value of predicting multi-step costs, an idea that is key to our theory, and notably also an idea that is known to be empirically valuable for learning state representations. A second part of this work, that is to appear as Part II, addresses the infinite-horizon linear time-invariant setting; it also extends the results to an approach that implicitly learns the latent dynamics, inspired by the recent empirical breakthrough of MuZero in model-based reinforcement learning.
Authors: Xinwen Cheng, Zhehao Huang, Wenxin Zhou, Zhengbao He, Ruikai Yang, Yingwen Wu, Xiaolin Huang
Abstract: Machine unlearning (MU) aims to remove the influence of specific training samples from a well-trained model, a task of growing importance due to the ``right to be forgotten.'' The unlearned model should approach the retrained model, where forgetting data do not contribute to the training process. Therefore, unlearning should withdraw their contribution from the pre-trained model. However, quantifying and disentangling sample's contribution to overall learning process is highly challenging, leading most existing MU approaches to adopt other heuristic strategies such as random labeling or knowledge distillation. These operations inevitably degrade model utility, requiring additional maintenance with remaining data. To advance MU towards better utility and efficiency for practical deployment, we seek to approximate sample contribution with only the pre-trained model. We theoretically and empirically reveal that sample's contribution during training manifests in the learned model's increased sensitivity to it. In light of this, we propose MU-Mis (Machine Unlearning by Minimizing input sensitivity), which directly suppresses the contribution of forgetting data. This straightforward suppression enables MU-Mis to successfully unlearn without degrading model utility on the remaining data, thereby eliminating the need for access to the remaining data. To the best of our knowledge, this is the first time that a remaining-data-free method can perform on par with top performing remaining-data-dependent methods.
Authors: Yan Zhuang, Qi Liu, Haoyang Bi, Zhenya Huang, Weizhe Huang, Jiatong Li, Junhao Yu, Zirui Liu, Zirui Hu, Yuting Hong, Zachary A. Pardos, Haiping Ma, Mengxiao Zhu, Shijin Wang, Enhong Chen
Abstract: Computerized Adaptive Testing (CAT) offers an efficient and personalized method for assessing examinee proficiency by dynamically adjusting test questions based on individual performance. Compared to traditional, non-personalized testing methods, CAT requires fewer questions and provides more accurate assessments. As a result, CAT has been widely adopted across various fields, including education, healthcare, sports, sociology, and the evaluation of AI models. While traditional methods rely on psychometrics and statistics, the increasing complexity of large-scale testing has spurred the integration of machine learning techniques. This paper aims to provide a machine learning-focused survey on CAT, presenting a fresh perspective on this adaptive testing paradigm. We delve into measurement models, question selection algorithm, bank construction, and test control within CAT, exploring how machine learning can optimize these components. Through an analysis of current methods, strengths, limitations, and challenges, we strive to develop robust, fair, and efficient CAT systems. By bridging psychometric-driven CAT research with machine learning, this survey advocates for a more inclusive and interdisciplinary approach to the future of adaptive testing.
Authors: Dominik J. M\"uhlematter, Michelle Halbheer, Alexander Becker, Dominik Narnhofer, Helge Aasen, Konrad Schindler, Mehmet Ozgur Turkoglu
Abstract: Numerous real-world decisions rely on machine learning algorithms and require calibrated uncertainty estimates. However, modern methods often yield overconfident, uncalibrated predictions. The dominant approach to quantifying the uncertainty inherent in the model is to train an ensemble of separate predictors and measure their empirical variance. In an explicit implementation, the ensemble has a high computational cost and memory footprint, especially if the base model itself is already large, like modern transformers. This motivates efforts to develop implicit ensemble methods that emulate the ensemble without explicitly instantiating all its members. We introduce LoRA-Ensemble, a parameter-efficient ensembling method for self-attention networks. It is based on Low-Rank Adaptation (LoRA), originally developed for efficient LLM fine-tuning, and extends it into an implicit ensembling scheme, where all ensemble members share the same, pre-trained self-attention network, but have individual low-rank matrices for the attention projections. The resulting method not only outperforms state-of-the-art implicit techniques like BatchEnsemble, but even matches or exceeds the accuracy of an Explicit Ensemble, while at the same time achieving superior calibration.
Authors: Deng Pan, Nuno Moniz, Nitesh Chawla
Abstract: The challenge of delivering efficient explanations is a critical barrier that prevents the adoption of model explanations in real-world applications. Existing approaches often depend on extensive model queries for sample-level explanations or rely on expert's knowledge of specific model structures that trade general applicability for efficiency. To address these limitations, this paper introduces a novel framework Fast Explanation (FEX) that represents attribution-based explanations via probability distributions, which are optimized by leveraging the policy gradient method. The proposed framework offers a robust, scalable solution for real-time, large-scale model explanations, bridging the gap between efficiency and applicability. We validate our framework on image and text classification tasks and the experiments demonstrate that our method reduces inference time by over 97% and memory usage by 70% compared to traditional model-agnostic approaches while maintaining high-quality explanations and broad applicability.
Authors: Kuo Gai, Sicong Wang, Shihua Zhang
Abstract: Deep neural networks (DNNs) are vulnerable to small adversarial perturbations of the inputs, posing a significant challenge to their reliability and robustness. Empirical methods such as adversarial training can defend against particular attacks but remain vulnerable to more powerful attacks. Alternatively, Lipschitz networks provide certified robustness to unseen perturbations but lack sufficient expressive power. To harness the advantages of both approaches, we design a novel two-step Optimal Transport induced Adversarial Defense (OTAD) model that can fit the training data accurately while preserving the local Lipschitz continuity. First, we train a DNN with a regularizer derived from optimal transport theory, yielding a discrete optimal transport map linking data to its features. By leveraging the map's inherent regularity, we interpolate the map by solving the convex integration problem (CIP) to guarantee the local Lipschitz property. OTAD is extensible to diverse architectures of ResNet and Transformer, making it suitable for complex data. For efficient computation, the CIP can be solved through training neural networks. OTAD opens a novel avenue for developing reliable and secure deep learning systems through the regularity of optimal transport maps. Empirical results demonstrate that OTAD can outperform other robust models on diverse datasets.
Authors: Jian Xu, Shian Du, Junmei Yang, Qianli Ma, Delu Zeng, John Paisley
Abstract: Gaussian Process Latent Variable Models (GPLVMs) have become increasingly popular for unsupervised tasks such as dimensionality reduction and missing data recovery due to their flexibility and non-linear nature. An importance-weighted version of the Bayesian GPLVMs has been proposed to obtain a tighter variational bound. However, this version of the approach is primarily limited to analyzing simple data structures, as the generation of an effective proposal distribution can become quite challenging in high-dimensional spaces or with complex data sets. In this work, we propose an Annealed Importance Sampling (AIS) approach to address these issues. By transforming the posterior into a sequence of intermediate distributions using annealing, we combine the strengths of Sequential Monte Carlo samplers and VI to explore a wider range of posterior distributions and gradually approach the target distribution. We further propose an efficient algorithm by reparameterizing all variables in the evidence lower bound (ELBO). Experimental results on both toy and image datasets demonstrate that our method outperforms state-of-the-art methods in terms of tighter variational bounds, higher log-likelihoods, and more robust convergence.
Authors: Tong Wang (Jeffrey), Ronilo Ragodos (Jeffrey), Lu Feng (Jeffrey), Yu (Jeffrey), Hu
Abstract: Post hoc explainers such as SHAP and LIME are used widely in business research to interpret complex machine learning models. Although they were designed to explain model predictions, there has been an increasing trend in which the explanations they generate are treated as evidence about underlying data relationships. Based on a systematic review of 181 studies, including 56 published in leading journals, we document that this explanation interpretation is widespread and examine its validity. We also evaluate how well post hoc explanations reliably recover the direction and relative importance of features in true data-generating process. We introduce two metrics-direction alignment and strength alignment-and assess SHAP and LIME using simulated data with known ground truth. Although explanations often appear reasonable on average, they exhibit substantial heterogeneity in their alignment and can fail to be aligned even when their predictive accuracy is high. High predictive performance is therefore necessary but insufficient for reliable explanation. We further show that feature correlation and the Rashomon effect (where many equally accurate models rely on different feature attributions) are key drivers of misalignment. Agreement in explanations across such models provides a practical diagnostic of reliability. Overall, our findings caution against using post hoc explainers for hypothesis validation and instead position them as exploratory tools that generate, rather than confirm, substantive insights.
Authors: RuiKang OuYang, Bo Qiang, Jos\'e Miguel Hern\'andez-Lobato
Abstract: Developing an efficient sampler capable of generating independent and identically distributed (IID) samples from a Boltzmann distribution is a crucial challenge in scientific research, e.g. molecular dynamics. In this work, we intend to learn neural samplers given energy functions instead of data sampled from the Boltzmann distribution. By learning the energies of the noised data, we propose a diffusion-based sampler, Noised Energy Matching, which theoretically has lower variance and more complexity compared to related works. Furthermore, a novel bootstrapping technique is applied to NEM to balance between bias and variance. We evaluate NEM and BNEM on a 2-dimensional 40 Gaussian Mixture Model (GMM) and a 4-particle double-well potential (DW-4). The experimental results demonstrate that BNEM can achieve state-of-the-art performance while being more robust.
Authors: Thibault Monsel, Onofrio Semeraro, Lionel Mathelin, Guillaume Charpiat
Abstract: Recent advances in learning dynamical systems from data have shown significant promise. However, many existing methods assume access to the full state of the system -- an assumption that is rarely satisfied in practice, where systems are typically monitored through a limited number of sensors, leading to partial observability. To address this challenge, we draw inspiration from the Mori-Zwanzig formalism, which provides a theoretical connection between hidden variables and memory terms. Motivated by this perspective, we introduce a constant-lag Neural Delay Differential Equations (NDDEs) framework, providing a continuous-time approach for learning non-Markovian dynamics directly from data. These memory effects are captured using a finite set of time delays, which are identified via the adjoint method. We validate the proposed approach on a range of datasets, including synthetic systems, chaotic dynamics, and experimental measurements, such as the Kuramoto-Sivashinsky equation and cavity-flow experiments. Results demonstrate that NDDEs compare favourably with existing approaches for partially observed systems, including long short-term memory (LSTM) networks and augmented neural ordinary differential equations (ANODEs). Overall, NDDEs offer a principled and data-efficient framework for modelling non-Markovian dynamics under partial observability. An open-source implementation accompanies this article.
Authors: Jiajian Li, Qi Wang, Yunbo Wang, Xin Jin, Yang Li, Wenjun Zeng, Xiaokang Yang
Abstract: Training visual reinforcement learning agents in a high-dimensional open world presents significant challenges. While various model-based methods have improved sample efficiency by learning interactive world models, these agents tend to be "short-sighted", as they are typically trained on short snippets of imagined experiences. We argue that the primary challenge in open-world decision-making is improving the exploration efficiency across a vast state space, especially for tasks that demand consideration of long-horizon payoffs. In this paper, we present LS-Imagine, which extends the imagination horizon within a limited number of state transition steps, enabling the agent to explore behaviors that potentially lead to promising long-term feedback. The foundation of our approach is to build a $\textit{long short-term world model}$. To achieve this, we simulate goal-conditioned jumpy state transitions and compute corresponding affordance maps by zooming in on specific areas within single images. This facilitates the integration of direct long-term values into behavior learning. Our method demonstrates significant improvements over state-of-the-art techniques in MineDojo.
Authors: Yuelin Xu, Xiao Zhang
Abstract: Despite being widely adopted as a canonical framework for learning robust models, adversarial training suffers from robust overfitting. Existing empirical measures and theoretical explorations are insufficient to provide satisfying mechanistic insights into the phenomenon. By viewing adversarial training with momentum SGD as a discrete-time dynamical system, we introduce a PAC-Bayesian analytical framework that proves time-resolved robust generalization bounds. Specifically, our framework tracks the closed-form evolution of the posterior mean and covariance under both stationary and non-stationary transient regimes, revealing their connections to the learning rate, the geometry of the loss landscape, and mini-batch stochastic gradients. By empirically approximating the statistical quantities implied by our theory, we offer a unified, mechanistic explanation for robust overfitting. We also illustrate why adversarial weight perturbation reduces the robust generalization gap by suppressing the loss curvature, but its design may be suboptimal for optimization due to over-penalization.
Authors: Usman Akram, Haris Vikalo
Abstract: Predicting the behavior of a dynamical system from noisy observations of its past outputs is a classical problem encountered across engineering and science. For linear systems with Gaussian inputs, the Kalman filter -- the best linear minimum mean-square error estimator of the state trajectory -- is optimal in the Bayesian sense. For nonlinear systems, Bayesian filtering is typically approached using suboptimal heuristics such as the Extended Kalman Filter (EKF), or numerical methods such as particle filtering (PF). In this work, we show that transformers, employed in an in-context learning (ICL) setting, can implicitly infer hidden states in order to predict the outputs of a wide family of dynamical systems, without test-time gradient updates or explicit knowledge of the system model. Specifically, when provided with a short context of past input-output pairs and, optionally, system parameters, a frozen transformer accurately predicts the current output. In linear-Gaussian regimes, its predictions closely match those of the Kalman filter; in nonlinear regimes, its performance approaches that of EKF and PF. Moreover, prediction accuracy degrades gracefully when key parameters, such as the state-transition matrix, are withheld from the context, demonstrating robustness and implicit parameter inference. These findings suggest that transformer in-context learning provides a flexible, non-parametric alternative for output prediction in dynamical systems, grounded in implicit latent-state estimation.
Authors: Yucheng Xing, Xin Wang
Abstract: Modern convolutional neural networks (CNNs) organize computation as a discrete stack of layers whose parameters are independently stored and learned, with the number of layers fixed as an architectural hyperparameter. In this work, we explore an alternative perspective: can network parameterization itself be modeled as a continuous dynamical system? We introduce Puppet-CNN, a framework that represents convolutional layer parameters as states evolving along a learned parameter flow governed by a neural ordinary differential equation (ODE). Under this formulation, layer parameters are generated through continuous evolution in parameter space, and the effective number of generated layers is determined by the integration horizon of the learned dynamics, which can be modulated by input complexity to enable input-adaptive computation. We validate this formulation on standard image classification benchmarks and demonstrate that continuous parameter dynamics can achieve competitive predictive performance while substantially reducing stored trainable parameters. These results suggest that viewing neural network parameterization through the lens of dynamical systems provides a structured and flexible design space for adaptive convolutional models.
Authors: Davide Maran, Marcello Restelli
Abstract: We address the problem of learning an unknown smooth function and its derivatives from noisy pointwise evaluations under the supremum norm. While classical nonparametric regression provides a strong theoretical foundation, traditional kernel-based estimators often incur high computational costs and memory requirements that scale with the sample size, limiting their utility in real-time applications such as reinforcement learning. To overcome these challenges, we propose a parametric approach based on a finite-dimensional representation that achieves minimax-optimal uniform convergence rates. Our method enables lightweight inference without storing all samples in memory. We provide sharp finite-sample bounds under sub-Gaussian noise, derive second-order Bernstein-type guarantees, and prove matching lower bounds, thereby confirming the optimality of our approach in both estimation error and memory efficiency.
Authors: Somrita Ghosh, Yuelin Xu, Xiao Zhang
Abstract: Learning robust models under adversarial settings is widely recognized as requiring a considerably large number of training samples. Recent work proposes semi-supervised adversarial training (SSAT), which utilizes external unlabeled or synthetically generated data and is currently the state of the art. However, SSAT requires substantial extra data to attain high robustness, resulting in prolonged training time and increased memory usage. In this paper, we propose data reduction strategies to improve the efficiency of SSAT by optimizing the amount of additional data incorporated. Specifically, we design novel latent clustering-based techniques to select or generate a small, critical subset of data samples near the model's decision boundary. While focusing on boundary-adjacent points, our methods maintain a balanced ratio between boundary and non-boundary data points, thereby avoiding overfitting. Comprehensive experiments across image benchmarks demonstrate that our methods can effectively reduce SSAT's data requirements and computational costs while preserving its strong robustness advantages. In particular, our latent-space selection scheme based on k-means clustering and our guided diffusion-based approach with LCG-KM are the most effective, achieving nearly identical robust accuracies with 5 times to 10 times less unlabeled data. When compared to full SSAT trained to convergence, our methods reduce total runtime by approximately 3 times to 4 times due to strategic prioritization of unlabeled data.
Authors: Haixia Liu, Boxiao Li, Can Yang, Yang Wang
Abstract: Numerous studies have shown that label noise can lead to poor generalization performance, negatively affecting classification accuracy. Therefore, understanding the effectiveness of classifiers trained using deep neural networks in the presence of noisy labels is of considerable practical significance. In this paper, we focus on the error bounds of excess risks for classification problems with noisy labels within deep learning frameworks. We derive error bounds for the excess risk, decomposing it into statistical error and approximation error. To handle statistical dependencies (e.g., mixing sequences), we employ an independent block construction to bound the error, leveraging techniques for dependent processes. For the approximation error, we establish these theoretical results to the vector-valued setting, where the output space consists of $K$-dimensional unit vectors. Finally, under the low-dimensional manifold hypothesis, we further refine the approximation error to mitigate the impact of high-dimensional input spaces.
Authors: Jonathan Drechsel, Steffen Herbold
Abstract: AI systems frequently exhibit and amplify social biases, leading to harmful consequences in critical areas. This study introduces a novel encoder-decoder approach that leverages model gradients to learn a feature neuron encoding societal bias information such as gender, race, and religion. We show that our method can not only identify which weights of a model need to be changed to modify a feature, but even demonstrate that this can be used to rewrite models to debias them while maintaining other capabilities. We demonstrate the effectiveness of our approach across various model architectures and highlight its potential for broader applications.
Authors: Linus Aronsson, Morteza Haghir Chehreghani
Abstract: Signed networks, where edges are labeled as positive or negative to represent friendly or antagonistic interactions, provide a natural framework for analyzing polarization, trust, and conflict in social systems. Detecting meaningful group structures in such networks is crucial for understanding online discourse, political divisions, and trust dynamics. A key challenge is to identify communities that are internally cohesive and externally antagonistic, while allowing for neutral or unaligned vertices. In this paper, we propose a method for identifying $k$ polarized communities that addresses a major limitation of prior methods: their tendency to produce highly size-imbalanced solutions. We introduce a novel optimization objective that avoids such imbalance. In addition, it is well known that approximation algorithms based on local search are highly effective for clustering signed networks when neutral vertices are not allowed. We build on this idea and design the first local search algorithm that extends to the setting with neutral vertices while scaling to large networks. By connecting our approach to block-coordinate Frank-Wolfe optimization, we prove a linear convergence rate, enabled by the structure of our objective. Experiments on real-world and synthetic datasets demonstrate that our method consistently outperforms state-of-the-art baselines in solution quality, while remaining competitive in computational efficiency.
Authors: Michelle M. Li, Kevin Li, Yasha Ektefaie, Ying Jin, Yepeng Huang, Shvat Messica, Tianxi Cai, Marinka Zitnik
Abstract: Conditional generation models for longitudinal sequences can produce new or modified trajectories given a conditioning input. However, they often lack control over when the condition should take effect (timing) and which variables it should influence (scope). Most methods either operate only on univariate sequences or assume that the condition alters all variables and time steps. In scientific and clinical settings, interventions instead begin at a specific moment, such as the time of drug administration or surgery, and influence only a subset of measurements while the rest of the trajectory remains unchanged. CLEF learns temporal concepts that encode how and when a condition alters future sequence evolution. These concepts allow CLEF to apply targeted edits to the affected time steps and variables while preserving the rest of the sequence. We evaluate CLEF on 8 datasets spanning cellular reprogramming, patient health, and sales, comparing against 9 state-of-the-art baselines. CLEF improves immediate sequence editing accuracy by 16.28% (MAE) on average against their non-CLEF counterparts. Unlike prior models, CLEF enables one-step conditional generation at arbitrary future times, outperforming their non-CLEF counterparts in delayed sequence editing by 26.73% (MAE) on average. We test CLEF under counterfactual inference assumptions and show up to 62.84% (MAE) improvement on zero-shot conditional generation of counterfactual trajectories. In a case study of patients with type 1 diabetes mellitus, CLEF identifies clinical interventions that generate realistic counterfactual trajectories shifted toward healthier outcomes.
Authors: Thierry Bossy, Julien Vignoud, Tahseen Rabbani, Juan R. Troncoso Pastoriza, Martin Jaggi
Abstract: Federated learning (FL) is a popular paradigm for collaborative training which avoids direct data exposure between clients. However, data privacy issues still remain: FL-trained large language models are capable of memorizing and completing phrases and sentences contained in training data when given their prefixes. Thus, it is possible for adversarial and honest- but-curious clients to recover training data of other participants simply through targeted prompting. In this work, we demonstrate that a popular and simple fine-tuning strategy, low-rank adaptation (LoRA), reduces memorization during FL by a factor of up to 10 without significant performance cost. We study this effect by performing fine-tuning tasks in high-risk domains such as medicine, law, and finance. We observe a reduction in memorization for a wide variety of model families, from 1B to 70B parameters. We find that LoRA can reduce memorization in centralized learning as well, and we compare how the memorization patterns differ. Furthermore, we study the effect of hyperparameters and show that LoRA can be combined with other privacy-preserving techniques such as gradient clipping and Gaussian noise, secure aggregation, and Goldfish loss to further improve record-level privacy while maintaining performance.
Authors: Xuefeng Liu, Hung T. C. Le, Siyu Chen, Rick Stevens, Zhuoran Yang, Matthew R. Walter, Yuxin Chen
Abstract: Online reinforcement learning (RL) enhances policies through direct interactions with the environment, but faces challenges related to sample efficiency. In contrast, offline RL leverages extensive pre-collected data to learn policies, but often produces suboptimal results due to limited data coverage. Recent efforts integrate offline and online RL in order to harness the advantages of both approaches. However, effectively combining online and offline RL remains challenging due to issues that include catastrophic forgetting, lack of robustness to data quality and limited sample efficiency in data utilization. In an effort to address these challenges, we introduce A3RL, which incorporates a novel confidence aware Active Advantage Aligned (A3) sampling strategy that dynamically prioritizes data aligned with the policy's evolving needs from both online and offline sources, optimizing policy improvement. Moreover, we provide theoretical insights into the effectiveness of our active sampling strategy and conduct diverse empirical experiments and ablation studies, demonstrating that our method outperforms competing online RL techniques that leverage offline data.
Authors: Zihao Li, Xiao Lin, Zhining Liu, Jiaru Zou, Ziwei Wu, Lecheng Zheng, Dongqi Fu, Yada Zhu, Hendrik Hamann, Hanghang Tong, Jingrui He
Abstract: While many advances in time series models focus exclusively on numerical data, research on multimodal time series, particularly those involving contextual textual information, remains in its infancy. With recent progress in large language models and time series learning, we revisit the integration of paired texts with time series through the Platonic Representation Hypothesis, which posits that representations of different modalities converge to shared spaces. In this context, we identify that time-series-paired texts may naturally exhibit periodic properties that closely mirror those of the original time series. Building on this insight, we propose a novel framework, Texts as Time Series (TaTS), which considers the time-series-paired texts to be auxiliary variables of the time series. TaTS can be plugged into any existing numerical-only time series models and effectively enable them to handle time series data with paired texts. Through extensive experiments on both multimodal time series forecasting and imputation tasks across benchmark datasets with various existing time series models, we demonstrate that TaTS can enhance multimodal predictive performance without modifying model architectures. Our Code is available at https://github.com/iDEA-iSAIL-Lab-UIUC/TaTS.
Authors: Aviv Shamsian, Eitan Shaar, Aviv Navon, Gal Chechik, Ethan Fetaya
Abstract: Machine unlearning aims to remove the influence of problematic training data after a model has been trained. The primary challenge in machine unlearning is ensuring that the process effectively removes specified data without compromising the model's overall performance on the remaining dataset. Many existing machine unlearning methods address this challenge by carefully balancing gradient ascent on the `unlearn' data with the gradient descent on a `retain' set that represents the training data. However, in many cases the training dataset is not fully available when we wish to unlearn some concepts, because models are released without their training datasets, and one may only have access to a $\textit{small part of a training set}$. Here, we propose OrthoGrad, a novel approach that mitigates interference between the unlearn set and a small retain set rather than competing ascent and descent processes. Our method projects the gradient of the unlearn set onto the subspace orthogonal to all gradients in the retain batch, effectively avoiding any gradient interference. We demonstrate the effectiveness of OrthoGrad on multiple machine unlearning benchmarks, including automatic speech recognition, outperforming competing methods.
Authors: Roy Friedman, Noa Moriel, Matthew Ricci, Guy Pelc, Yair Weiss, Mor Nitzan
Abstract: Characterizing the long term behavior of dynamical systems given limited measurements is a common challenge throughout the physical and biological sciences. This is a challenging task due to the sparsity and noise inherent to empirical observations, as well as the variability of possible long-term dynamics. We address this by introducing smooth prototype equivalences (SPE), a framework for matching sparse observations to prototypical behaviors using invertible neural networks which model smooth phase space deformations. SPE can localize the invariant sets describing long-term behavior of the observed dynamics through the learned mapping from prototype space to data space. Furthermore, SPE can classify dynamical regimes by comparing the data residual of the deformed measurements to prototype dynamics. Our method outperforms existing techniques in the classification of oscillatory systems and can efficiently identify invariant structures like limit cycles and fixed points in an equation-free manner, even when only a small, noisy subset of the phase space is observed. SPE further reveals driving genes in synthetic oscillators such as the repressilator regulatory circuit, and traces cyclic biological processes like the cell cycle trajectory directly from experimental high-dimensional single-cell gene expression data.
Authors: Vu Nguyen, Andrey Kan
Abstract: The problem of relevant and diverse subset selection has a wide range of applications, including recommender systems and retrieval-augmented generation (RAG). For example, in recommender systems, one is interested in selecting relevant items, while providing a diversified recommendation. Constrained subset selection problem is NP-hard, and popular approaches such as Maximum Marginal Relevance (MMR) are based on greedy selection. Many real-world applications involve large data, but the original MMR work did not consider distributed selection. This limitation was later addressed by a method called DGDS which allows for a distributed setting using random data partitioning. Here, we exploit structure in the data to further improve both scalability and performance on the target application. We propose MUSS, a novel method that uses a multilevel approach to relevant and diverse selection. In a recommender system application, our method can not only improve the performance up to $4$ percent points in precision, but is also $20$ to $80$ times faster. Our method is also capable of outperforming baselines on RAG-based question answering accuracy. We present a novel theoretical approach for analyzing this type of problems, and show that our method achieves a constant factor approximation of the optimal objective. Moreover, our analysis also resulted in a $\times 2$ tighter bound for DGDS compared to previously known bound.
Authors: Lang Cao, Renhong Chen, Yingtian Zou, Chao Peng, Huacong Xu, Yuxian Wang, Wu Ning, Qian Chen, Mofan Peng, Zijie Chen, Peishuo Su, Yitong Li
Abstract: We introduce the Entropy-Driven Uncertainty Process Reward Model (EDU-PRM), a novel entropy-driven training framework for process reward modeling that enables dynamic, uncertainty-aligned segmentation of complex reasoning steps, eliminating the need for costly manual step annotations. Unlike previous Process Reward Models (PRMs) that rely on static partitioning and human labeling, EDU-PRM automatically anchors step boundaries at tokens with high predictive entropy, effectively capturing intrinsic logical transitions and facilitating efficient exploration of diverse reasoning paths. On the ProcessBench benchmark, EDU-PRM outperforms strong public PRM baselines, such as Math-Shepherd PRM and Omega PRM, and EDU-PRM achieves comparable results with SOTA models while only using 1.5% training data. Furthermore, by leveraging our proposed EDU sampling strategy, we observe accuracy boosts from 64.7% to 67.3% for generative reasoning tasks, accompanied by a reduction of 32% in token usage. These findings underscore the potential of EDU-PRM as a scalable and annotation-efficient paradigm for process supervision in mathematical reasoning, paving the way for more efficient and robust approaches to complex mathematical problem solving.
Authors: Hojoon Lee, Takuma Seno, Jun Jet Tai, Kaushik Subramanian, Kenta Kawamoto, Peter Stone, Peter R. Wurman
Abstract: Deep reinforcement learning has achieved superhuman racing performance in high-fidelity simulators like Gran Turismo 7 (GT7). It typically utilizes global features that require instrumentation external to a car, such as precise localization of agents and opponents, limiting real-world applicability. To address this limitation, we introduce a vision-based autonomous racing agent that relies solely on ego-centric camera views and onboard sensor data, eliminating the need for precise localization during inference. This agent employs an asymmetric actor-critic framework: the actor uses a recurrent neural network with the sensor data local to the car to retain track layouts and opponent positions, while the critic accesses the global features during training. Evaluated in GT7, our agent consistently outperforms GT7's built-drivers. To our knowledge, this work presents the first vision-based autonomous racing agent to demonstrate champion-level performance in competitive racing scenarios.
Authors: Garrett Baker, George Wang, Jesse Hoogland, Daniel Murfet
Abstract: We develop a linear response framework for interpretability that treats a neural network as a Bayesian statistical mechanical system. A small perturbation of the data distribution, for example shifting the Pile toward GitHub or legal text, induces a first-order change in the posterior expectation of an observable localized on a chosen component of the network. The resulting susceptibility can be estimated efficiently with local SGLD samples and factorizes into signed, per-token contributions that serve as attribution scores. We combine these susceptibilities into a response matrix whose low-rank structure separates functional modules such as multigram and induction heads in a 3M-parameter transformer.
Authors: Zhenyu Wang, Molei Liu, Jing Lei, Francis Bach, Zijian Guo
Abstract: When synthesizing multi-source high-dimensional data, a key objective is to extract low-dimensional representations that effectively approximate the original features across different sources. Such representations facilitate the discovery of transferable structures and help mitigate systematic biases such as batch effects. We introduce Stable Principal Component Analysis (StablePCA), a distributionally robust framework for constructing stable latent representations by maximizing the worst-case explained variance over multiple sources. A primary challenge in extending classical PCA to the multi-source setting lies in the nonconvex rank constraint, which renders the StablePCA formulation a nonconvex optimization problem. To overcome this challenge, we conduct a convex relaxation of StablePCA and develop an efficient Mirror-Prox algorithm to solve the relaxed problem, with global convergence guarantees. Since the relaxed problem generally differs from the original formulation, we further introduce a data-dependent certificate to assess how well the algorithm solves the original nonconvex problem and establish the condition under which the relaxation is tight. Finally, we explore alternative distributionally robust formulations of multi-source PCA based on different loss functions.
Authors: Reilly Haskins, Benjamin Adams
Abstract: Knowledge distillation compresses a larger neural model (teacher) into smaller, faster student models by training them to match teacher outputs. However, the internal computational transformations that occur during this process remain poorly understood. We apply techniques from mechanistic interpretability to analyze how internal circuits, representations, and activation patterns differ between teachers and students. Focusing on GPT2 and its distilled counterpart DistilGPT2, and generalizing our findings to both bidirectional architectures and larger model pairs, we find that student models can reorganize, compress, and discard teacher components, often resulting in a stronger reliance on fewer individual components. To quantify functional alignment beyond output similarity, we introduce an alignment metric based on influence-weighted component similarity, validated across multiple tasks. Our findings reveal that while knowledge distillation preserves broad functional behaviors, it also causes significant shifts in internal computation, with important implications for the robustness and generalization capacity of distilled models.
Authors: Hanyu Duan, Yi Yang, Ahmed Abbasi, Kar Yan Tam
Abstract: Machine unlearning is the process of removing the imprint left by specific data samples during the training of a machine learning model. AI developers, including those building personalized technologies, employ machine unlearning for various purposes such as privacy protection, security, and to address ethical concerns. This paper introduces Ready2Unlearn, a learning-time optimization approach designed to facilitate future unlearning processes. Unlike the majority of existing unlearning efforts that focus on designing unlearning algorithms, which are typically implemented reactively when an unlearning request is made during the model deployment phase, Ready2Unlearn shifts the focus to the training phase, adopting a "forward-looking" perspective. Building upon well-established meta-learning principles, Ready2Unlearn proactively trains machine learning models with unlearning readiness, such that they are well prepared and can handle future unlearning requests in a more efficient and principled manner. Ready2Unlearn is model-agnostic and compatible with any gradient ascent-based machine unlearning algorithms. We evaluate the method on both language and vision tasks under various unlearning settings, including class-wise unlearning and random data unlearning. Experimental results show that by incorporating such preparedness at training time, Ready2Unlearn produces an unlearning-ready model state, which offers several key advantages when future unlearning is requested. We hope this study inspires future research on proactive strategies for equipping machine learning models with built-in unlearning readiness, particularly in modern information systems that rely heavily on user data for recommendation, search, and personalized services, where privacy risks and data deletion demands are increasingly prevalent.
Authors: Guangda Liu, Chengwei Li, Zhenyu Ning, Jing Lin, Yiwu Yao, Danning Ke, Minyi Guo, Jieru Zhao
Abstract: Large language models (LLMs) are widely deployed with rapidly expanding context windows to support increasingly demanding applications. However, long contexts pose significant deployment challenges, primarily due to the KV cache whose size grows proportionally with context length. While KV cache compression methods have been proposed to address this issue, KV dropping methods incur considerable accuracy loss, and KV retrieval methods suffer from significant efficiency bottlenecks. We propose FreeKV, a training-free algorithm-system co-optimization framework to enhance KV retrieval efficiency while preserving accuracy. On the algorithm side, FreeKV introduces speculative retrieval to shift the KV selection and recall processes out of the critical path, combined with fine-grained correction to ensure accuracy. On the system side, FreeKV employs hybrid KV layouts across CPU and GPU memory to eliminate fragmented data transfers, and leverages double-buffered streamed recall to further improve efficiency, enabling effective overlap with computation, full latency hiding, and practical speedups from speculative recall. Experiments demonstrate that FreeKV achieves near-lossless accuracy across various scenarios and models, delivering up to a 13$\times$ speedup compared to SOTA KV retrieval methods. Code is available at https://github.com/sjtu-zhao-lab/FreeKV.
Authors: Aymeric Capitaine, Maxime Haddouche, Eric Moulines, Michael I. Jordan, Etienne Boursier, Alain Durmus
Abstract: Decision-focused learning (DFL) is an increasingly popular paradigm for training predictive models whose outputs are used in decision-making tasks. Instead of merely optimizing for predictive accuracy, DFL trains models to directly minimize the loss associated with downstream decisions. However, existing studies focus solely on scenarios where a fixed batch of data is available and the objective function does not change over time. We instead investigate DFL in dynamic environments where the objective function and data distribution evolve over time. This setting is challenging for online learning because the objective function has zero or undefined gradients, which prevents the use of standard first-order optimization methods, and is generally non-convex. To address these difficulties, we (i) regularize the objective to make it differentiable and (ii) use perturbation techniques along with a near-optimal oracle to overcome non-convexity. Combining those techniques yields two original online algorithms tailored for DFL, for which we establish respectively static and dynamic regret bounds. These are the first provable guarantees for the online decision-focused problem. Finally, we showcase the effectiveness of our algorithms on a knapsack experiment, where they outperform two standard benchmarks.
Authors: Mohamed A. Mohamed, Kateryna Nekhomiazh, Vedant Vyas, Marcos M. Jose, Andrew Patterson, Marlos C. Machado
Abstract: Continual reinforcement learning (RL) concerns agents that are expected to learn continually, rather than converge to a policy that is then fixed for evaluation. This setting is well-suited to environments that the agent perceives as changing over time, rendering any static policy ineffective. In continual RL, researchers often simulate such changes either by modifying episodic environments to incorporate task shifts during interaction or by designing simulators that explicitly model continual dynamics. However, transforming episodic problems into continual ones primarily captures scenarios involving abrupt changes in the data stream and still relies on episodic structure. Meanwhile, the few simulators explicitly designed for empirical continual RL research are often limited in scope or complexity. In this paper, we introduce AgarCL, a research platform for continual RL that enables agents to progress toward increasingly sophisticated behaviour. AgarCL is based on the game Agar.io, a non-episodic, high-dimensional problem with stochastic, ever-evolving dynamics, continuous actions, and partial observability. We provide benchmark results for DQN, PPO, and SAC on the primary continual RL challenge, as well as across a suite of smaller tasks within AgarCL. These smaller tasks isolate aspects of the full environment and allow us to characterize the distinct challenges posed by different components of the game. We further evaluate three continual learning methods-Shrink and Perturb, ReDo, and Continual Backpropagation-and observe little improvement over standard RL algorithms, suggesting that the challenges posed by AgarCL extend beyond the stability-plasticity dilemma.
Authors: Yiming Sun, Shuo Chen, Shengyu Chen, Chonghao Qiu, Licheng Liu, Youmi Oh, Sparkle L. Malone, Gavin McNicol, Qianlai Zhuang, Chris Smith, Yiqun Xie, Xiaowei Jia
Abstract: Methane (CH$_4$) is the second most powerful greenhouse gas after carbon dioxide and plays a crucial role in climate change due to its high global warming potential. Accurately modeling CH$_4$ fluxes across the globe and at fine temporal scales is essential for understanding its spatial and temporal variability and developing effective mitigation strategies. In this work, we introduce the first-of-its-kind cross-scale global wetland methane benchmark dataset (X-MethaneWet), which synthesizes physics-based model simulation data from TEM-MDM and the real-world observation data from FLUXNET-CH$_4$. This dataset can offer opportunities for improving global wetland CH$_4$ modeling and science discovery with new AI algorithms. To set up AI model baselines for methane flux prediction, we evaluate the performance of various sequential deep learning models on X-MethaneWet. Furthermore, we explore four different transfer learning techniques to leverage simulated data from TEM-MDM to improve the generalization of deep learning models on real-world FLUXNET-CH$_4$ observations. Our extensive experiments demonstrate the effectiveness of these approaches, highlighting their potential for advancing methane emission modeling and identifying new opportunities for developing more accurate and scalable AI-driven climate models.
Authors: Tina Khezresmaeilzadeh, Parsa Razmara, Seyedarmin Azizi, Mohammad Erfan Sadeghi, Erfan Baghaei Potraghloo
Abstract: Stock price prediction remains a complex and high-stakes task in financial analysis, traditionally addressed using statistical models or, more recently, language models. In this work, we introduce VISTA (Vision-Language Inference for Stock Time-series Analysis), a novel, training-free framework that leverages Vision-Language Models (VLMs) for multi-modal stock forecasting. VISTA prompts a VLM with both textual representations of historical stock prices and their corresponding line charts to predict future price values. By combining numerical and visual modalities in a zero-shot setting and using carefully designed chain-of-thought prompts, VISTA captures complementary patterns that unimodal approaches often miss. We benchmark VISTA against standard baselines, including ARIMA and text-only LLM-based prompting methods. Experimental results show that VISTA outperforms these baselines by up to 89.83%, demonstrating the effectiveness of multi-modal inference for stock time-series analysis and highlighting the potential of VLMs in financial forecasting tasks without requiring task-specific training.
Authors: Juntong Wang, Xiyuan Wang, Muhan Zhang
Abstract: Common Neighbors (CNs) and their higher-order variants are important pairwise features widely used in state-of-the-art link prediction methods. However, existing methods often struggle with the repetition across different orders of CNs and fail to fully leverage their potential. We identify that these limitations stem from two key issues: redundancy and over-smoothing in high-order common neighbors. To address these challenges, we design orthogonalization to eliminate redundancy between different-order CNs and normalization to mitigate over-smoothing. By combining these two techniques, we propose Orthogonal Common Neighbor (OCN), a novel approach that significantly outperforms the strongest baselines by an average of 7.7\% on popular link prediction benchmarks. A thorough theoretical analysis is provided to support our method. Ablation studies also verify the effectiveness of our orthogonalization and normalization techniques. Code is available at: https://github.com/qingpingmo/OCN.
Authors: Nurbek Tastan, Stefanos Laskaridis, Martin Takac, Karthik Nandakumar, Samuel Horvath
Abstract: Large pre-trained models are commonly adapted to downstream tasks using parameter-efficient fine-tuning methods such as Low-Rank Adaptation (LoRA), which injects small trainable low-rank matrices instead of updating all weights. While LoRA dramatically reduces trainable parameters with little overhead, it can still underperform full fine-tuning in accuracy and often converges more slowly. We introduce LoFT, a novel low-rank adaptation method that behaves like full fine-tuning by aligning the optimizer's internal dynamics with those of updating all model weights. LoFT not only learns weight updates in a low-rank subspace (like LoRA) but also properly projects the optimizer's first and second moments (Adam's momentum and variance) into the same subspace, mirroring full-model updates. By aligning the low-rank update itself with the full update, LoFT eliminates the need for tuning extra hyperparameters, e.g., the LoRA scaling factor $\alpha$. Empirically, this approach substantially narrows the performance gap between adapter-based tuning and full fine-tuning and consistently outperforms standard LoRA-style methods, all without increasing inference cost.
Authors: Zheng Wang, Wanhao Yu, Li Yang, Sen Lin
Abstract: Continual Learning (CL) seeks to build an agent that can continuously learn a sequence of tasks, where a key challenge, namely Catastrophic Forgetting, persists due to the potential knowledge interference among different tasks. On the other hand, deep neural networks (DNNs) are shown to converge to a terminal state termed Neural Collapse during training, where all class prototypes geometrically form a static simplex equiangular tight frame (ETF). These maximally and equally separated class prototypes make the ETF an ideal target for model learning in CL to mitigate knowledge interference. Thus inspired, several studies have emerged very recently to leverage a fixed global ETF in CL, which however suffers from key drawbacks, such as impracticability and limited performance.To address these challenges and fully unlock the potential of ETF in CL, we propose Progressive Neural Collapse (ProNC), a novel framework that completely removes the need of a fixed global ETF in CL. Specifically, ProNC progressively expands the ETF target in a principled way by adding new class prototypes as vertices for new tasks, ensuring maximal separability across all encountered classes with minimal shifts from the previous ETF. We next develop a new CL framework by plugging ProNC into commonly used CL algorithm designs, where distillation is further leveraged to balance between target shifting for old classes and target aligning for new classes. Extensive experiments show that our approach significantly outperforms related baselines while maintaining superior flexibility, simplicity, and efficiency. Our code is available at https://github.com/Continue-Edge-AI-Lab/ProNC
Authors: Chaoyu Liu, Yangming Li, Zhongying Deng, Chris Budd, Carola-Bibiane Sch\"onlieb
Abstract: Physical laws, such as the conversation of mass and momentum, are fundamental principles in many physical systems. Neural operators have achieved promising performance in learning the solutions to those systems, but often fail to ensure conservation. Existing methods typically enforce strict conservation via hand-crafted post-processing or architectural constraints, leading to limited model flexibility and adaptability. In this work, we propose a novel plug-and-play adaptive correction approach to ensure the conservation of fundamental linear and quadratic quantities for neural operator outputs. Our method introduces a lightweight learnable operator to adaptively enforce the target conservation law during training. This method allows the model to flexibly and adaptively correct its output to guarantee strict conservation. We provide a theoretical result showing that our correction method does not hamper the expression ability of neural operators and can potentially achieve lower reconstruction loss than their conservation-constrained counterparts. Our method is evaluated across multiple neural operator architectures and representative PDEs. Extensive experiments show that incorporating our correction method into baseline models significantly improves both accuracy and stability. In addition, the experimental results demonstrate that our approach consistently achieves superior performance over widely used conservation-enforcement techniques on various PDE benchmarks.
Authors: Pedro Jim\'enez-Gonz\'alez, Miguel C. Soriano, Lucas Lacasa
Abstract: Traditional algorithms to optimize artificial neural networks when confronted with a supervised learning task are usually exploitation-type relaxational dynamics such as gradient descent (GD). Here, we explore the dynamics of the neural network trajectory along training for unconventionally large learning rates. We show that for a region of values of the learning rate, the GD optimization shifts away from purely exploitation-like algorithm into a regime of exploration-exploitation balance, as the neural network is still capable of learning but the trajectory shows sensitive dependence on initial conditions --as characterized by positive network maximum Lyapunov exponent--. Interestingly, the characteristic training time required to reach an acceptable accuracy in the test set reaches a minimum precisely in such learning rate region, further suggesting that one can accelerate the training of artificial neural networks by locating at the onset of chaos. Our results --initially illustrated for the MNIST classification task-- qualitatively hold for a range of supervised learning tasks, {learning architectures (including both shallow and deep multilayer perceptrons and convolutional neural networks) and other hyperparameters (different activation functions and weight regularisation),} and showcase the emergent, constructive role of transient chaotic dynamics in the training of artificial neural networks.
Authors: Minhyuk Seo, Taeheon Kim, Hankook Lee, Jonghyun Choi, Tinne Tuytelaars
Abstract: As AI becomes more personal, e.g., Agentic AI, there is an increasing need for personalizing models for various use cases. Personalized federated learning (PFL) enables each client to collaboratively leverage other clients' knowledge for better adaptation to the task of interest, without privacy risks. Despite its potential, existing PFL methods remain confined to rather simplified scenarios where data and models are the same across clients. To move towards realistic scenarios, we move beyond these restrictive assumptions by addressing both data and model heterogeneity. We propose a task-relevance-aware model aggregation strategy to reduce parameter interference under heterogeneous data. Moreover, we introduce Co-LoRA, a dimension-invariant module that enables knowledge sharing across heterogeneous architectures. To mimic the real-world task diversity, we propose a multi-modal PFL benchmark spanning 40 distinct tasks with distribution shifts over time. Extensive experiments shows that our proposed method significantly outperforms the state-of-the-art PFL methods under heterogeneous scenarios.
Authors: Tanmay Goyal, Gaurav Sinha
Abstract: We study the Logistic Contextual Slate Bandit problem, where, at each round, an agent selects a slate of $N$ items from an exponentially large set (of size $2^{\Omega(N)}$) of candidate slates provided by the environment. A single binary reward, determined by a logistic model, is observed for the chosen slate. Our objective is to develop algorithms that maximize cumulative reward over $T$ rounds while maintaining low per-round computational costs. We propose two algorithms, Slate-GLM-OFU and Slate-GLM-TS, that accomplish this goal. These algorithms achieve $N^{O(1)}$ per-round time complexity via local planning (independent slot selections), and low regret through global learning (joint parameter estimation). We provide theoretical and empirical evidence supporting these claims. Under a well-studied diversity assumption, we prove that Slate-GLM-OFU incurs only $\tilde{O}(\sqrt{T})$ regret. Extensive experiments across a wide range of synthetic settings demonstrate that our algorithms consistently outperform state-of-the-art baselines, achieving both the lowest regret and the fastest runtime. Furthermore, we apply our algorithm to select in-context examples in prompts of Language Models for solving binary classification tasks such as sentiment analysis. Our approach achieves competitive test accuracy, making it a viable alternative in practical scenarios.
Authors: Haoran Tang, Rajiv Khanna
Abstract: We characterize the effectiveness of Sharpness-aware minimization (SAM) under machine unlearning scheme, where unlearning forget signals interferes with learning retain signals. While previous work prove that SAM improves generalization with noise memorization prevention, we show that SAM abandons such denoising property when fitting the forget set, leading to altered generalization depending on signal strength. We further characterize the signal surplus of SAM in the order of signal strength, which enables learning from less retain signals to maintain model performance and putting more weight on unlearning the forget set. Empirical studies show that SAM outperforms SGD with relaxed requirement for retain signals and can enhance various unlearning methods either as pretrain or unlearn algorithm. Motivated by our refined characterization of SAM unlearning and observing that overfitting can benefit more stringent sample-specific unlearning, we propose Sharp MinMax, which splits the model into two to learn retain signals with SAM and unlearn forget signals with sharpness maximization, achieving best performance. Extensive experiments show that SAM enhances unlearning across varying difficulties measured by memorization, yielding decreased feature entanglement between retain and forget sets, stronger resistance to membership inference attacks, and a flatter loss landscape. Our observations generalize to more noised data, different optimizers, and different architectures.
Authors: Zixuan Huang, Yikun Ban, Lean Fu, Xiaojie Li, Zhongxiang Dai, Jianxin Li, Deqing Wang
Abstract: Direct Preference Optimization (DPO) has emerged as an effective approach for aligning large language models (LLMs) with human preferences. However, its performance is highly dependent on the quality of the underlying human preference data. To address this bottleneck, prior work has explored various data selection strategies, but these methods often overlook the impact of the evolving states of the language model during the optimization process. In this paper, we introduce a novel problem: Sample Scheduling for DPO, which aims to dynamically and adaptively schedule training samples based on the model's evolving batch-wise states throughout preference optimization. To solve this problem, we propose SamS, an efficient and effective algorithm that adaptively selects samples in each training batch based on the LLM's learning feedback to maximize the potential generalization performance. Notably, without modifying the core DPO algorithm, simply integrating SamS significantly improves performance across tasks, with minimal additional computational overhead. This work points to a promising new direction for improving LLM alignment through batch-wise sample selection, with potential generalization to RLHF and broader supervised learning paradigms.
Authors: Zejin Lu, Sushrut Thorat, Radoslaw M Cichy, Tim C Kietzmann
Abstract: Despite years of research and the dramatic scaling of artificial intelligence (AI) systems, a striking misalignment between artificial and human vision persists. Contrary to humans, AI relies heavily on texture-features rather than shape information, lacks robustness to image distortions, remains highly vulnerable to adversarial attacks, and struggles to recognise simple abstract shapes within complex backgrounds. To close this gap, here we take inspiration from how human vision develops from early infancy into adulthood. We quantified visual maturation by synthesising decades of research into a novel developmental visual diet (DVD) for AI vision. Guiding AI systems through this human-inspired curriculum, which considers the development of visual acuity, contrast sensitivity, and colour, produces models that better align with human behaviour on every hallmark of robust vision tested, yielding the strongest reported reliance on shape information to date, abstract shape recognition beyond the state of the art, and higher resilience to image corruptions and adversarial attacks. Our results thus demonstrate that robust AI vision can be achieved by guiding how a model learns, not merely how much it learns, offering a resource-efficient route toward safer and more human-like artificial visual systems.
Authors: Sebastien Andre-Sloan, Anirbit Mukherjee, Matthew Colbrook
Abstract: Physics-Informed Neural Networks (PINNs) are increasingly used to approximate solutions of partial differential equations (PDEs), particularly in high dimensions. In real-world settings, data are often noisy, making it crucial to understand when a predictor can still achieve low empirical risk. Yet, little is known about the conditions under which a PINN can do so effectively. We analyse PINNs applied to the Hamilton--Jacobi--Bellman (HJB) PDE and establish a lower bound on the network size required for the supervised PINN empirical risk to fall below the variance of noisy supervision labels. Specifically, if a predictor achieves empirical risk $O(\eta)$ below $\sigma^2$ (the variance of the supervision data), then necessarily $d_N\log d_N\gtrsim N_s \eta^2$, where $N_s$ is the number of samples and $d_N$ the number of trainable parameters. A similar constraint holds in the fully unsupervised PINN setting when boundary labels are noisy. Thus, simply increasing the number of noisy supervision labels does not offer a ``free lunch'' in reducing empirical risk. We also give empirical studies on the HJB PDE, the Poisson PDE and the the Navier-Stokes PDE set to produce the Taylor-Green solutions. In these experiments we demonstrate that PINNs indeed need to be beyond a threshold model size for them to train to errors below $\sigma^2$. These results provide a quantitative foundation for understanding parameter requirements when training PINNs in the presence of noisy data.
Authors: Zihao Li, Zhichen Zeng, Xiao Lin, Feihao Fang, Yanru Qu, Zhe Xu, Zhining Liu, Xuying Ning, Tianxin Wei, Ge Liu, Hanghang Tong, Jingrui He
Abstract: Over the past decade, advances in generative modeling, such as generative adversarial networks, masked autoencoders, and diffusion models, have significantly transformed biological research and discovery, enabling breakthroughs in molecule design, protein generation, catalysis discovery, drug discovery, and beyond. At the same time, biological applications have served as valuable testbeds for evaluating the capabilities of generative models. Recently, flow matching has emerged as a powerful and efficient alternative to diffusion-based generative modeling, with growing interest in its application to problems in biology and life sciences. This paper presents the first comprehensive survey of recent developments in flow matching and its applications in biological domains. We begin by systematically reviewing the foundations and variants of flow matching, and then categorize its applications into three major areas: biological sequence modeling, molecule generation and design, and peptide and protein generation. For each, we provide an in-depth review of recent progress. We also summarize commonly used datasets and software tools, and conclude with a discussion of potential future directions. The corresponding curated resources are available at https://github.com/Violet24K/Awesome-Flow-Matching-Meets-Biology.
URLs: https://github.com/Violet24K/Awesome-Flow-Matching-Meets-Biology.
Authors: Ruimeng Ye, Zihan Wang, Yang Xiao, Zinan Ling, Manling Li, Bo Hui
Abstract: Weak-to-Strong generalization (W2SG) is a new trend to elicit the full capabilities of a strong model with supervision from a weak model. While existing W2SG studies focus on simple tasks like binary classification, we extend this paradigm to complex interactive decision-making environments. Specifically, we fine-tune a strong model with trajectories of intermediate actions generated by a weak model. Motivated by the human learning process, we propose to generalize not only success knowledge but also failure experience so that the strong model can learn from failed trajectories accumulated by weak models. To effectively and efficiently elicit the potential of strong agents, we further construct ``trajectory trees," a hierarchical representation that organizes weak model-generated action trajectories, coupled with Monte Carlo Tree Search (MCTS) to optimize the strong model. Through theoretical analysis, we provide formal guarantees for the effectiveness of our method in improving W2SG performance. Our empirical evaluations demonstrate substantial improvements in reasoning and decision-making capabilities across diverse task domains, validating the scalability and robustness of our proposed framework.
Authors: Valentin Lafargue, Adriana Laurindo Monteiro, Emmanuelle Claeys, Laurent Risser, Jean-Michel Loubes
Abstract: The rapid deployment of AI systems in high-stakes domains, including those classified as high-risk under the The EU AI Act (Regulation (EU) 2024/1689), has intensified the need for reliable compliance auditing. For binary classifiers, regulatory risk assessment often relies on global fairness metrics such as the Disparate Impact ratio, widely used to evaluate potential discrimination. In typical auditing settings, the auditee provides a subset of its dataset to an auditor, while a supervisory authority may verify whether this subset is representative of the full underlying distribution. In this work, we investigate to what extent a malicious auditee can construct a fairness-compliant yet representative-looking sample from a non-compliant original distribution, thereby creating an illusion of fairness. We formalize this problem as a constrained distributional projection task and introduce mathematically grounded manipulation strategies based on entropic and optimal transport projections. These constructions characterize the minimal distributional shift required to satisfy fairness constraints. To counter such attacks, we formalize representativeness through distributional distance based statistical tests and systematically evaluate their ability to detect manipulated samples. Our analysis highlights the conditions under which fairness manipulation can remain statistically undetected and provides practical guidelines for strengthening supervisory verification. We validate our theoretical findings through experiments on standard tabular datasets for bias detection. Code is publicly available at https://github.com/ValentinLafargue/Inspection.
Authors: Jiazhen Pan (Cherise), Bailiang Jian (Cherise), Paul Hager (Cherise), Yundi Zhang (Cherise), Che Liu (Cherise), Friedrike Jungmann (Cherise), Hongwei Bran Li (Cherise), Chenyu You (Cherise), Junde Wu (Cherise), Jiayuan Zhu (Cherise), Fenglin Liu (Cherise), Yuyuan Liu (Cherise), Niklas Bubeck (Cherise), Christian Wachinger (Cherise), Chen (Cherise), Chen (Cherise), Zhenyu Gong, Cheng Ouyang, Georgios Kaissis, Benedikt Wiestler, Daniel Rueckert
Abstract: Ensuring the safety and reliability of large language models (LLMs) in clinical practice is critical to prevent patient harm. However, LLMs are advancing so rapidly that static benchmarks quickly become obsolete or prone to overfitting, yielding a misleading picture of model trustworthiness. Here we introduce a Dynamic, Automatic, and Systematic (DAS) red-teaming framework that continuously stress-tests LLMs across four safety-critical axes: robustness, privacy, bias/fairness, and hallucination. Validated against board-certified clinicians with high concordance, a suite of adversarial agents autonomously mutates clinical test cases to uncover vulnerabilities in real time. Applying DAS to 15 proprietary and open-source LLMs revealed a profound gap between high static benchmark performance and low dynamic reliability - the ``Benchmarking Gap''. Despite median MedQA accuracy exceeding 80\%, 94\% of previously correct answers failed our dynamic robustness tests. Crucially, this brittleness generalized to the realistic, open-ended HealthBench dataset, where top-tier models exhibited failure rates exceeding 70\% and stark shifts in model rankings across evaluations, suggesting that high scores on established static benchmarks may reflect superficial memorization. We observed similarly high failure rates across other domains: privacy leaks were elicited in 86\% of scenarios, cognitive-bias priming altered clinical recommendations in 81\% of fairness tests, and we identified hallucination rates exceeding 74\% in widely used models. By converting medical LLM safety analysis from a static checklist into a dynamic stress-test, DAS provides a foundational, scalable, and living platform to surface the latent risks that must be addressed before the next generation of medical AI can be safely deployed.
Authors: Shifeng Xie, Vasilii Feofanov, Ambroise Odonnat, Lei Zan, Marius Alonso, Jianfeng Zhang, Themis Palpanas, Lujia Pan, Keli Zhang, Ievgen Redko
Abstract: Time series foundation models (TSFMs) have recently gained significant attention due to their strong zero-shot capabilities and widespread real-world applications. Such models typically require a computationally costly pre-training on large-scale, carefully curated collections of real-world sequences. To allow for a sample-efficient pre-training of TSFMs, we propose \textsc{CauKer}, a novel algorithm designed to generate diverse, causally coherent synthetic time series with realistic trends, seasonality, and nonlinear interactions. \textsc{CauKer} combines Gaussian Process (GP) kernel composition with Structural Causal Models (SCM) to produce data for sample-efficient pre-training of state-of-the-art classification TSFMs having different architectures and following different pre-training approaches. Additionally, our experiments reveal that \textsc{CauKer}-generated datasets exhibit clear scaling laws for both dataset size (10K to 10M samples) and model capacity (1M to 783M parameters), unlike real-world datasets, which display irregular scaling behavior. The source code is publicly available at https://github.com/ShifengXIE/CauKer.
Authors: Ziheng Sun, Qi Feng, Lehao Lin, Chris Ding, Jicong Fan
Abstract: This work focuses on training graph foundation models (GFMs) that have strong generalization ability in graph-level tasks such as graph classification. Effective GFM training requires capturing information consistent across different domains. We discover that graph structures provide more consistent cross-domain information compared to node features and graph labels. However, traditional GFMs primarily focus on transferring node features from various domains into a unified representation space but often lack structural cross-domain generalization. To address this, we introduce GraphProp, which emphasizes structural generalization. The training process of GraphProp consists of two main phases. First, we train a structural GFM by predicting graph invariants. Since graph invariants are properties of graphs that depend only on the abstract structure, not on particular labellings or drawings of the graph, this structural GFM has a strong ability to capture the abstract structural information and provide discriminative graph representations comparable across diverse domains. In the second phase, we use the representations given by the structural GFM as positional encodings to train a comprehensive GFM. This phase utilizes domain-specific node attributes and graph labels to further improve cross-domain node feature generalization. Our experiments demonstrate that GraphProp significantly outperforms the competitors in supervised learning and few-shot learning, especially in handling graphs without node attributes.
Authors: Lorenzo Livi
Abstract: We show that gating mechanisms in recurrent neural networks (RNNs) induce lag-dependent and direction-dependent effective learning rates, even when training uses a fixed, global step size. This behavior arises from a coupling between state-space time-scales (parametrized by the gates) and parameter-space dynamics during gradient descent. By deriving exact Jacobians for leaky-integrator and gated RNNs and applying a first-order expansion, we make explicit how constant, scalar, and multi-dimensional gates reshape gradient propagation, modulate effective step sizes, and introduce anisotropy in parameter updates. These findings reveal that gates act not only as filters of information flow, but also as data-driven preconditioners of optimization, with formal connections to learning-rate schedules, momentum, and adaptive methods such as Adam. Empirical simulations corroborate these predictions: across several sequence tasks, gates produce lag-dependent effective learning rates and concentrate gradient flow into low-dimensional subspaces, matching or exceeding the anisotropic structure induced by Adam. Notably, gating and optimizer-driven adaptivity shape complementary aspects of credit assignment: gates align state-space transport with loss-relevant directions, while optimizers rescale parameter-space updates. Overall, this work provides a unified dynamical systems perspective on how gating couples state evolution with parameter updates, clarifying why gated architectures achieve robust trainability in practice.
Authors: Zhouyu Zhang, Chih-Yuan Chiu, Glen Chou
Abstract: We present an inverse dynamic game-based algorithm to learn parametric constraints from a given dataset of local Nash equilibrium interactions between multiple agents. Specifically, we introduce mixed-integer linear programs (MILP) encoding the Karush-Kuhn-Tucker (KKT) conditions of the interacting agents, which recover constraints consistent with the local Nash stationarity of the interaction demonstrations. We establish theoretical guarantees that our method learns inner approximations of the true safe and unsafe sets. We also use the interaction constraints recovered by our method to design motion plans that robustly satisfy the underlying constraints. Across simulations and hardware experiments, our methods accurately inferred constraints and designed safe interactive motion plans for various classes of constraints, both convex and non-convex, from interaction demonstrations of agents with nonlinear dynamics.
Authors: Jiarui Cao, Zhiyang Zhang, Heming Wang, Jun Xu, Ling Lan, Simon J. L. Billinge, Ran Gu
Abstract: The nanostructure inverse problem is an attractive problem that helps researchers to understand the relationship between the properties and the structure of nanomaterials. This study focuses on the problem of recovering the model system of monometallic nanoparticles (MMNPs) from their pair distribution function (PDF) and regards it as a highly ill-posed conditional generation task. This study proposes a Condition-based Latent Diffusion Model (CbLDM) as a feasible solution to this problem. This model demonstrates an acceleration approach within the framework of a latent diffusion model by using conditional priors to estimate the conditional posterior distribution, which is an approximate distribution of p(z|x). In addition, this study uses Laplacian matrix instead of distance matrix to recover the nanostructure, which helps to improve stability. Our study demonstrates that a latent diffusion model with a conditional prior can generate nanostructures that are consistent with PDF observations and physically meaningful, thereby laying the groundwork for subsequent more complex inverse problems.
Authors: Tianye Fang, Xuanshu Luo, Martin Werner
Abstract: The increasing availability of big mobility data from ubiquitous portable devices enables human mobility prediction through deep learning approaches. However, the diverse complexity of human mobility data impedes model training, leading to inefficient gradient updates and potential underfitting. Meanwhile, exclusively predicting next locations neglects implicit determinants, including distances and directions, thereby yielding suboptimal prediction results. This paper presents a unified training framework that integrates entropy-driven curriculum and multi-task learning to address these challenges. The proposed entropy-driven curriculum learning strategy quantifies trajectory predictability based on Lempel-Ziv compression and organizes training from simple to complex for faster convergence and enhanced performance. The multi-task training simultaneously optimizes the primary location prediction alongside auxiliary estimation of movement distance and direction for learning realistic mobility patterns, and improve prediction accuracy through complementary supervision signals. Extensive experiments conducted in accordance with the HuMob Challenge demonstrate that our approach achieves state-of-the-art performance on GEO-BLEU (0.354) and DTW (26.15) metrics with up to 2.92-fold convergence speed compared to training without curriculum learning.
Authors: Hugh Xuechen Liu, K{\i}van\c{c} Tatar
Abstract: Bipartite knowledge graphs in niche domains are typically data-poor and edge-sparse, which hinders link prediction. We introduce AEGIS (Authentic Edge Growth In Sparsity), an edge-only augmentation framework that resamples existing training edges -either uniformly simple or with inverse-degree bias degree-aware -thereby preserving the original node set and sidestepping fabricated endpoints. To probe authenticity across regimes, we consider naturally sparse graphs (game design pattern's game-pattern network) and induce sparsity in denser benchmarks (Amazon, MovieLens) via high-rate bond percolation. We evaluate augmentations on two complementary metrics: AUC-ROC (higher is better) and the Brier score (lower is better), using two-tailed paired t-tests against sparse baselines. On Amazon and MovieLens, copy-based AEGIS variants match the baseline while the semantic KNN augmentation is the only method that restores AUC and calibration; random and synthetic edges remain detrimental. On the text-rich GDP graph, semantic KNN achieves the largest AUC improvement and Brier score reduction, and simple also lowers the Brier score relative to the sparse control. These findings position authenticity-constrained resampling as a data-efficient strategy for sparse bipartite link prediction, with semantic augmentation providing an additional boost when informative node descriptions are available.
Authors: Valentyn Melnychuk, Stefan Feuerriegel
Abstract: Various deep generative models have been proposed to estimate potential outcomes distributions from observational data. However, none of them have the favorable theoretical property of general Neyman-orthogonality and, associated with it, quasi-oracle efficiency and double robustness. In this paper, we introduce a general suite of generative Neyman-orthogonal (doubly-robust) learners that estimate the conditional distributions of potential outcomes. Our proposed generative doubly-robust learners (GDR-learners) are flexible and can be instantiated with many state-of-the-art deep generative models. In particular, we develop GDR-learners based on (a) conditional normalizing flows (which we call GDR-CNFs), (b) conditional generative adversarial networks (GDR-CGANs), (c) conditional variational autoencoders (GDR-CVAEs), and (d) conditional diffusion models (GDR-CDMs). Unlike the existing methods, our GDR-learners possess the properties of quasi-oracle efficiency and rate double robustness, and are thus asymptotically optimal. In a series of (semi-)synthetic experiments, we demonstrate that our GDR-learners are very effective and outperform the existing methods in estimating the conditional distributions of potential outcomes.
Authors: Reza Rahimi Azghan, Gautham Krishna Gudur, Mohit Malu, Edison Thomaz, Giulia Pedrielli, Pavan Turaga, Hassan Ghasemzadeh
Abstract: The rise of deep learning has greatly advanced human behavior monitoring using wearable sensors, particularly human activity recognition (HAR). While deep models have been widely studied, most assume stationary data distributions - an assumption often violated in real-world scenarios. For example, sensor data from one subject may differ significantly from another, leading to distribution shifts. In continual learning, this shift is framed as a sequence of tasks, each corresponding to a new subject. Such settings suffer from catastrophic forgetting, where prior knowledge deteriorates as new tasks are learned. This challenge is compounded by the scarcity and inconsistency of labeled data in human studies. To address these issues, we propose CLAD-Net (Continual Learning with Attention and Distillation), a framework enabling wearable-sensor models to be updated continuously without sacrificing performance on past tasks. CLAD-Net integrates a self-supervised transformer, acting as long-term memory, with a supervised Convolutional Neural Network (CNN) trained via knowledge distillation for activity classification. The transformer captures global activity patterns through cross-attention across body-mounted sensors, learning generalizable representations without labels. Meanwhile, the CNN leverages knowledge distillation to retain prior knowledge during subject-wise fine-tuning. On PAMAP2, CLAD-Net achieves 91.36 percent final accuracy with only 8.78 percent forgetting, surpassing memory-based and regularization-based baselines such as Experience Replay and Elastic Weight Consolidation. In semi-supervised settings with only 10-20 percent labeled data, CLAD-Net still delivers strong performance, demonstrating robustness to label scarcity. Ablation studies further validate each module's contribution.
Authors: Alakh Sharma, Gaurish Trivedi, Kartikey Singh Bhandari, Yash Sinha, Dhruv Kumar, Pratik Narang, Jagat Sesh Challa
Abstract: Scalable multi-agent reinforcement learning (MARL) remains a central challenge for AI. Existing population-based methods, like Policy-Space Response Oracles, PSRO, require storing explicit policy populations and constructing full payoff matrices, incurring quadratic computation and linear memory costs. We present Generative Evolutionary Meta-Solver (GEMS), a surrogate-free framework that replaces explicit populations with a compact set of latent anchors and a single amortized generator. Instead of exhaustively constructing the payoff matrix, GEMS relies on unbiased Monte Carlo rollouts, multiplicative-weights meta-dynamics, and a model-free empirical-Bernstein UCB oracle to adaptively expand the policy set. Best responses are trained within the generator using an advantage-based trust-region objective, eliminating the need to store and train separate actors. We evaluated GEMS in a variety of Two-player and Multi-Player games such as the Deceptive Messages Game, Kuhn Poker and Multi-Particle environment. We find that GEMS is up to ~$\mathbf{6\times}$ faster, has $\mathbf{1.3\times}$ less memory usage than PSRO, while also reaps higher rewards simultaneously. These results demonstrate that GEMS retains the game theoretic guarantees of PSRO, while overcoming its fundamental inefficiencies, hence enabling scalable multi-agent learning in multiple domains.
Authors: Ran Elbaz, Guy Bar-Shalom, Yam Eitan, Fabrizio Frasca, Haggai Maron
Abstract: Permutation equivariant neural networks employing parameter-sharing schemes have emerged as powerful models for leveraging a wide range of data symmetries, significantly enhancing the generalization and computational efficiency of the resulting models. Recently, Kolmogorov-Arnold Networks (KANs) have demonstrated promise through their improved interpretability and expressivity compared to traditional architectures based on MLPs. While equivariant KANs have been explored in recent literature for a few specific data types, a principled framework for applying them to data with permutation symmetries in a general context remains absent. This paper introduces Function Sharing KAN (FS-KAN), a principled approach to constructing equivariant and invariant KA layers for arbitrary permutation symmetry groups, unifying and significantly extending previous work in this domain. We derive the basic construction of these FS-KAN layers by generalizing parameter-sharing schemes to the Kolmogorov-Arnold setup and provide a theoretical analysis demonstrating that FS-KANs have the same expressive power as networks that use standard parameter-sharing layers, allowing us to transfer well-known and important expressivity results from parameter-sharing networks to FS-KANs. Empirical evaluations on multiple data types and symmetry groups show that FS-KANs exhibit superior data efficiency compared to standard parameter-sharing layers, by a wide margin in certain cases, while preserving the interpretability and adaptability of KANs, making them an excellent architecture choice in low-data regimes.
Authors: Valentyn Melnychuk, Dennis Frauen, Jonas Schweisthal, Stefan Feuerriegel
Abstract: The conditional average treatment effect (CATE) is widely used in personalized medicine to inform therapeutic decisions. However, state-of-the-art methods for CATE estimation (so-called meta-learners) often perform poorly in the presence of low overlap. In this work, we introduce a new approach to tackle this issue and improve the performance of existing meta-learners in the low-overlap regions. Specifically, we introduce Overlap-Adaptive Regularization (OAR) that regularizes target models proportionally to overlap weights so that, informally, the regularization is higher in regions with low overlap. To the best of our knowledge, our OAR is the first approach to leverage overlap weights in the regularization terms of the meta-learners. Our OAR approach is flexible and works with any existing CATE meta-learner: we demonstrate how OAR can be applied to both parametric and non-parametric second-stage models. Furthermore, we propose debiased versions of our OAR that preserve the Neyman-orthogonality of existing meta-learners and thus ensure more robust inference. Through a series of (semi-)synthetic experiments, we demonstrate that our OAR significantly improves CATE estimation in low-overlap settings in comparison to constant regularization.
Authors: Linus Aronsson, Han Wu, Morteza Haghir Chehreghani
Abstract: We study active correlation clustering where pairwise similarities are not provided upfront and must be queried in a cost-efficient manner through active learning. Specifically, we focus on the cold-start scenario, where no true initial pairwise similarities are available for active learning. To address this challenge, we propose a coverage-aware method that encourages diversity early in the process. We demonstrate the effectiveness of our approach through several synthetic and real-world experiments.
Authors: Sreeja Apparaju, Yichuan Niu, Xixi Qi
Abstract: Budget pacing is critical in online advertising to align spend with campaign goals under dynamic auctions. Existing pacing methods often rely on ad-hoc parameter tuning, which can be unstable and inefficient. We propose a principled controller that combines bucketized hysteresis with proportional feedback to provide stable and adaptive spend control. Our method provides a framework and analysis for parameter selection that enables accurate tracking of desired spend rates across campaigns. Experiments in real-world auctions demonstrate significant improvements in pacing accuracy and delivery consistency, reducing pacing error by 13% and $\lambda$-volatility by 54% compared to baseline method. By bridging control theory with advertising systems, our approach offers a scalable and reliable solution for budget pacing, with particular benefits for small-budget campaigns.
Authors: Viktor Sip, Martin Breyton, Spase Petkoski, Viktor Jirsa
Abstract: Learning stochastic models of dynamical systems from observed data is of interest in many scientific fields. Here, we propose a new method for this task within the family of dynamical variational autoencoders. The proposed double projection method estimates both the system state trajectories and the noise time series from data. This approach naturally allows us to perform multi-step system evolution and to learn models with a comparatively low-dimensional state space. We evaluate the performance of the method on six benchmark problems, including both simulated and experimental data. We further illustrate the effects of the teacher forcing interval of the multi-step scheme on the nature of the internal dynamics and compare the resulting behavior to that of deterministic models of equivalent architecture.
Authors: Ruohao Guo, Afshin Oroojlooy, Roshan Sridhar, Miguel Ballesteros, Alan Ritter, Dan Roth
Abstract: Despite recent rapid progress in AI safety, current large language models remain vulnerable to adversarial attacks in multi-turn interaction settings, where attackers strategically adapt their prompts across conversation turns and pose a more critical yet realistic challenge. Existing approaches that discover safety vulnerabilities either rely on manual red-teaming with human experts or employ automated methods using pre-defined templates and human-curated attack data, with most focusing on single-turn attacks. However, these methods did not explore the vast space of possible multi-turn attacks, failing to consider novel attack trajectories that emerge from complex dialogue dynamics and strategic conversation planning. This gap is particularly critical given recent findings that LLMs exhibit significantly higher vulnerability to multi-turn attacks compared to single-turn attacks. We propose DialTree, an on-policy reinforcement learning framework integrated with tree search that autonomously discovers diverse multi-turn attack strategies by treating the dialogue as a sequential decision-making problem, enabling systematic exploration without manually curated data. Through extensive experiments, our approach not only achieves more than 44.2% higher ASR across 12 target models compared to previous state-of-the-art approaches, but also effectively uncovers new attack strategies by learning optimal dialogue policies that maximize attack success across multiple turns.
Authors: Elias Dubbeldam, Reza Mohammadi, Marit Schoonhoven, S. Ilker Birbil
Abstract: Accurate predictions on tabular data rely on capturing complex, dataset-specific feature interactions. Attention-based methods and graph neural networks, referred to as graph-based tabular deep learning (GTDL), aim to improve predictions by modeling these interactions as a graph. In this work, we analyze how these methods model the feature interactions. Current GTDL approaches primarily focus on optimizing predictive accuracy, often neglecting the accurate modeling of the underlying graph structure. Using synthetic datasets with known ground-truth graph structures, we find that current GTDL methods fail to recover meaningful feature interactions, as their edge recovery is close to random. This suggests that the attention mechanism and message-passing schemes used in GTDL do not effectively capture feature interactions. Furthermore, when we impose the true interaction structure, we find that the predictive accuracy improves. This highlights the need for GTDL methods to prioritize accurate modeling of the graph structure, as it leads to better predictions.
Authors: Nikolaus Howe, Micah Carroll
Abstract: Chain-of-Thought (CoT) monitoring has emerged as a compelling method for detecting harmful behaviors such as reward hacking for reasoning models, under the assumption that models' reasoning processes are informative of such behaviors. In practice, LLM training often produces unintended behaviors due to imperfect reward signals, leading models to develop misaligned tendencies. A common corrective approach is to apply post-hoc instructions to avoid problematic behaviors, but what happens to the model's reasoning process when these instructions conflict with learned behaviors? We investigate this question in simple settings and find that models engage in systematic motivated reasoning -- generating plausible-sounding justifications for violating their instructions while downplaying potential harms or contradictions. Concerningly, we find that as motivated reasoning becomes more prevalent over the course of training, an 8B-parameter CoT monitor is increasingly fooled by the motivated reasoning, being persuaded to judge the answer as following the constitution, despite correctly identifying the answer as contradicting the constitution when not provided with the model's reasoning trace. While we find that large frontier reasoning models closely track human ability in detecting motivated reasoning, this should not give us too much solace, as frontier model developers rely on smaller models for monitoring due to their low latency and deployment costs. Our results underscore the necessity for further research into the emergence and detection of motivated reasoning in model evaluation and oversight. Code for this paper is available at https://github.com/nikihowe/motivated-reasoning. WARNING: some examples in this paper may be upsetting.
Authors: Zan Li, Rui Fan
Abstract: Financial anomalies arise from heterogeneous mechanisms -- price shocks, liquidity freezes, contagion cascades, and momentum reversals -- yet existing detectors produce uniform scores without revealing which mechanism is failing. This hinders targeted responses: liquidity freezes call for market-making support, whereas price shocks call for circuit breakers. Three key challenges remain: (1) static graphs cannot adapt when correlations shift across regimes; (2) uniform detectors overlook heterogeneous anomaly signatures; and (3) black-box scores provide no actionable guidance on driving mechanisms. We address these challenges with an adaptive graph learning framework that embeds interpretability architecturally rather than post hoc. The framework constructs stress-modulated graphs that adaptively interpolate between known sector and geographic relationships and data-driven correlations as market conditions evolve. Anomalies are decomposed via four mechanism-specific experts -- Price-Shock, Liquidity, Systemic-Contagion, and Momentum-Reversal -- each capturing a distinct anomaly channel documented in the financial economics literature. The resulting routing weights serve as interpretable proxies for mechanism attribution, with their relative values indicating each anomaly's primary driving mechanism. A hierarchical Market Pressure Index aggregates entity-level anomaly scores into graduated market-wide alerts. On 100 U.S. equities (2017-2024), the framework detects all six major stress events with a 3.7-day mean lead time, outperforming baselines by +33 percentage points, with AUC 0.888 and AP 0.626. Case studies on SVB (March 2023) and Japan carry-trade unwind (August 2024) demonstrate that routing weights automatically distinguish localized from systemic crises without labeled supervision.
Authors: Pengxiang Cai, Zihao Gao, Wanchen Lian, Jintai Chen
Abstract: Tabular prediction traditionally relies on gradient-boosted decision trees and deep learning models, which excel in specific tasks but lack interpretability and transferability. Reasoning large language models (LLMs) promise cross-task adaptability with transparent reasoning traces, yet their potential for tabular data remains unrealized. To bridge this gap, we propose a reasoning framework centered on Permutation Relative Policy Optimization (PRPO), a reinforcement learning method that encodes column-permutation invariance as a structural prior. By estimating advantages across label-preserving permutations, PRPO transforms sparse rewards into dense signals, activating latent numerical reasoning capabilities of LLMs with limited supervision. Extensive experiments show that our method matches fully supervised baselines and dominates in zero-shot settings, performing on par with 32-shot strong baselines. Remarkably, our 8B model significantly outperforms much larger LLMs, achieving up to a 53.17% improvement over DeepSeek-R1 (685B).
Authors: Chia-Hsuan Lu, Tony Tan, Michael Benedikt
Abstract: Graph neural networks (GNNs) are the predominant architecture for learning over graphs. As with any machine learning model, an important issue is the detection of attacks, where an adversary can change the output with a small perturbation of the input. Techniques for solving the adversarial robustness problem - determining whether an attack exists - were originally developed for image classification. In the case of graph learning, the attack model usually considers changes to the graph structure in addition to or instead of the numerical features of the input, and the state of the art techniques proceed via reduction to constraint solving, working on top of powerful solvers, e.g. for mixed integer programming. We show that it is possible to improve on the state of the art in structural robustness by replacing the use of powerful solvers by calls to efficient partial solvers, which run in polynomial time but may be incomplete. We evaluate our tool RobLight on a diverse set of GNN variants and datasets.
Authors: Jacopo Di Ventura, Jan Felix Kleuker, Aske Plaat, Thomas Moerland
Abstract: Zero-shot reinforcement learning (RL) has emerged as a setting for developing general agents, capable of solving downstream tasks without additional training or planning at test-time. While conventional RL optimizes policies for fixed rewards, zero-shot RL requires learning representations that enable immediate adaptation to arbitrary reward functions. As the field matures, the growing diversity of approaches demands a foundational framework reconciling different perspectives under a common unifying structure. In this work, we introduce a formal, unified framework for zero-shot RL, allowing for rigorous comparisons across methods. We propose a taxonomy organizing the algorithmic landscape along two levels: representation, distinguishing between compositional and direct methods based on their exploitation of action-value function decompositions; and learning paradigm, differentiating between reward-free and pseudo reward-free training. Additionally, we propose a unified view of existing error bounds, decomposing the total error into three primary contributing components: inference, reward, and approximation, serving as a foundation for more grounded comparisons of zero-shot methods.
Authors: Tengxue Zhang, Biao Ouyang, Yang Shu, Xinyang Chen, Chenjuan Guo, Bin Yang
Abstract: Pre-trained models exhibit strong generalization to various downstream tasks. However, given the numerous models available in the model hub, identifying the most suitable one by individually fine-tuning is time-consuming. In this paper, we propose \textbf{SwiftTS}, a swift selection framework for time series pre-trained models. To avoid expensive forward propagation through all candidates, SwiftTS adopts a learning-guided approach that leverages historical dataset-model performance pairs across diverse horizons to predict model performance on unseen datasets. It employs a lightweight dual-encoder architecture that embeds time series and candidate models with rich characteristics, computing patchwise compatibility scores between data and model embeddings for efficient selection. To further enhance the generalization across datasets and horizons, we introduce a horizon-adaptive expert composition module that dynamically adjusts expert weights, and the transferable cross-task learning with cross-dataset and cross-horizon task sampling to enhance out-of-distribution (OOD) robustness. Extensive experiments on 14 downstream datasets and 8 pre-trained models demonstrate that SwiftTS achieves state-of-the-art performance in time series pre-trained model selection.
Authors: Hyunsik Yoo, Ting-Wei Li, SeongKu Kang, Zhining Liu, Charlie Xu, Qilin Qi, Hanghang Tong
Abstract: While large language models (LLMs) achieve strong performance in recommendation, they face challenges in continual learning as users, items, and user preferences evolve over time. Existing LoRA-based continual methods primarily focus on preserving performance on previous tasks, but this overlooks the unique nature of recommendation: the goal is not to predict past preferences, and outdated preferences can even harm performance when current interests shift significantly. To address this, we propose PESO (Proximally rEgularized Single evolving lOra, a continual adaptation method for LoRA in recommendation. PESO introduces a proximal regularizer that anchors the current adapter to its most recent frozen state, enabling the model to flexibly balance adaptation and preservation, and to better capture recent user behaviors. Theoretically, we show that this proximal design provides data-aware, direction-wise guidance in the LoRA subspace. Empirically, PESO consistently outperforms existing LoRA-based continual learning methods.
Authors: Farjana Aktar, Mohd Ruhul Ameen, Akif Islam, Md Ekramul Hamid
Abstract: Achieving both accurate and interpretable classification of motor-imagery EEG remains a key challenge in brain-computer interface (BCI) research. In this paper, we compare a transparent fuzzy-reasoning approach (ANFIS-FBCSP-PSO) with a well-known deep-learning benchmark (EEGNet) using the publicly available BCI Competition IV-2a dataset. The ANFIS pipeline combines filter-bank common spatial pattern feature extraction with fuzzy IF-THEN rules optimized via particle-swarm optimization, while EEGNet learns hierarchical spatial-temporal representations directly from raw EEG data. In within-subject experiments, the fuzzy-neural model performed better (68.58% +/- 13.76% accuracy, kappa = 58.04% +/- 18.43), while in cross-subject (LOSO) tests, the deep model exhibited stronger generalization (68.20% +/- 12.13% accuracy, kappa = 57.33% +/- 16.22). The study therefore provides practical guidance for selecting MI-BCI systems according to the design goal: interpretability or robustness across users. Future investigations into transformer-based and hybrid neuro-symbolic frameworks are expected to further advance transparent EEG decoding.
Authors: Song Gao, Songyang Zhang, Shusen Jing, Shuai Zhang, Xiangwei Zhou, Yue Wang, Zhipeng Cai
Abstract: Recent advancements in large artificial intelligence models (LAMs) are driving significant innovations in mobile edge computing within next-generation wireless networks. However, the substantial demands for computational resources and larges-cale training data required to train LAMs conflict with the limited storage and computational capacity of edge devices, posing significant challenges to training and deploying LAMs at the edge. In this work, we introduce the Networked Mixture-of-Experts (NMoE) system, in which clients perform inference collaboratively by distributing tasks to suitable neighbors based on their expertise and aggregate the returned results. For training the NMoE, we propose a federated learning framework that integrates both supervised and self-supervised learning to balance personalization and generalization, while preserving communication efficiency and data privacy. We conduct extensive experiments to demonstrate the efficacy of the proposed NMoE system, providing insights for the NMoE training algorithms.
Authors: Jiedong Jiang, Wanyi He, Yuefeng Wang, Guoxiong Gao, Yongle Hu, Jingting Wang, Nailin Guan, Peihao Wu, Chunbo Dai, Liang Xiao, Bin Dong
Abstract: Recent advances in large language models (LLMs) have demonstrated impressive capabilities in formal theorem proving, particularly on contest-based mathematical benchmarks like the IMO. However, these contests do not reflect the depth, breadth, and abstraction of modern mathematical research. To bridge this gap, we introduce FATE (Formal Algebra Theorem Evaluation), a new benchmark series in formal algebra designed to chart a course toward advanced mathematical reasoning. We present two new components, FATE-H and FATE-X, each with 100 problems in abstract and commutative algebra. The FATE series spans a difficulty spectrum from undergraduate exercises to problems exceeding PhD qualifying exams. Notably, FATE-X is the first formal benchmark to surpass both PhD-level exam difficulty and the coverage of the Mathlib library. Our evaluations of state-of-the-art LLM provers on this new benchmark reveal a stark performance gap compared to contest math: the best model achieves only 3% (pass@64) accuracy on FATE-H and 0% on FATE-X. Our two-stage evaluation reveals that models' natural-language reasoning is notably more accurate than their ability to formalize this reasoning. We systematically classify the common errors that arise during this formalization process. Furthermore, a comparative study shows that a specialized prover can exhibit less effective reflection than general-purpose models, reducing its accuracy at the natural-language stage. We believe FATE provides a robust and challenging benchmark that establishes essential checkpoints on the path toward research-level formal mathematical reasoning.
Authors: Anirudh Satheesh, Keenan Powell, Vaneet Aggarwal
Abstract: A central challenge in reinforcement learning is that policies trained in controlled environments often fail under distribution shifts at deployment into real-world environments. Distributionally Robust Reinforcement Learning (DRRL) addresses this by optimizing for worst-case performance within an uncertainty set defined by a robustness budget $\epsilon$. However, fixing $\epsilon$ results in a tradeoff between performance and robustness: small values yield high nominal performance but weak robustness, while large values can result in instability and overly conservative policies. We propose Distributionally Robust Self-Paced Curriculum Reinforcement Learning (DR-SPCRL), a method that overcomes this limitation by treating $\epsilon$ as a continuous curriculum. DR-SPCRL adaptively schedules the robustness budget according to the agent's progress, enabling a balance between nominal and robust performance. Empirical results across multiple environments demonstrate that DR-SPCRL not only stabilizes training but also achieves a superior robustness-performance trade-off, yielding an average 11.8\% increase in episodic return under varying perturbations compared to fixed or heuristic scheduling strategies, and achieving approximately 1.9$\times$ the performance of the corresponding nominal RL algorithms.
Authors: Yanan Zhao, Feng Ji, Jingyang Dai, Jiaze Ma, Keyue Jiang, Kai Zhao, Wee Peng Tay
Abstract: Graph contrastive learning (GCL) learns node and graph representations by contrasting multiple views of the same graph. Existing methods typically rely on fixed, handcrafted views-usually a local and a global perspective, which limits their ability to capture multi-scale structural patterns. We present an augmentation-free, multi-view GCL framework grounded in fractional-order continuous dynamics. By varying the fractional derivative order $\alpha \in (0,1]$, our encoders produce a continuous spectrum of views: small $\alpha$ yields localized features, while large $\alpha$ induces broader, global aggregation. We treat $\alpha$ as a learnable parameter so the model can adapt diffusion scales to the data and automatically discover informative views. This principled approach generates diverse, complementary representations without manual augmentations. Extensive experiments on standard benchmarks demonstrate that our method produces more robust and expressive embeddings and outperforms state-of-the-art GCL baselines.
Authors: Tuhin Subhra De
Abstract: Variational Autoencoders and Generative Adversarial Networks remained the state-of-the-art (SOTA) generative models until 2022. Now they are superseded by diffusion-based models. Efforts to improve traditional models have stagnated as a result. In old-school fashion, we explore image generation with conditional Variational Autoencoders (CVAE) to incorporate desired attributes within the images. VAEs are known to produce blurry images with less diversity; we refer to a method that solves this issue by leveraging the variance of the gaussian decoder as a learnable parameter during training. Previous works on CVAEs assumed that the conditional distribution of the latent space given the labels is equal to the prior distribution, which is not the case in reality. We show that estimating it using Non-Volume Preserving (NVP) transformations results in better image generation than existing methods by reducing the FID by 4% and increasing log likelihood by 7.6% compared to the previous cases.
Authors: Yuval Shapira, Dana Drachsler-Cohen
Abstract: Few-pixel attacks mislead a classifier by modifying a few pixels of an image. Their perturbation space is an $\ell_0$-ball, which is not convex, unlike $\ell_p$-balls for $p\geq1$. However, existing local robustness verifiers typically scale by relying on linear bound propagation, which captures convex perturbation spaces. We show that the convex hull of an $\ell_0$-ball is the intersection of its bounding box and an asymmetrically scaled $\ell_1$-like polytope. The volumes of the convex hull and this polytope are nearly equal as the input dimension increases. We then show a linear bound propagation that precisely computes bounds over the convex hull and is significantly tighter than bound propagations over the bounding box or our $\ell_1$-like polytope. This bound propagation scales the state-of-the-art $\ell_0$ verifier on its most challenging robustness benchmarks by 1.24x-7.07x, with a geometric mean of 3.16.
Authors: Minsoo Jo, Dongyoon Yang, Taesup Kim
Abstract: Adversarial examples in neural networks have been extensively studied in Euclidean geometry, but recent advances in \textit{hyperbolic networks} call for a reevaluation of attack strategies in non-Euclidean geometries. Existing methods such as FGSM and PGD apply perturbations without regard to the underlying hyperbolic structure, potentially leading to inefficient or geometrically inconsistent attacks. In this work, we propose a novel adversarial attack that explicitly leverages the geometric properties of hyperbolic space. Specifically, we compute the gradient of the loss function in the tangent space of hyperbolic space, decompose it into a radial (depth) component and an angular (semantic) component, and apply perturbation derived solely from the angular direction. Our method generates adversarial examples by focusing perturbations in semantically sensitive directions encoded in angular movement within the hyperbolic geometry. Empirical results on image classification, cross-modal retrieval tasks and network architectures demonstrate that our attack achieves higher fooling rates than conventional adversarial attacks, while producing high-impact perturbations with deeper insights into vulnerabilities of hyperbolic embeddings. This work highlights the importance of geometry-aware adversarial strategies in curved representation spaces and provides a principled framework for attacking hierarchical embeddings.
Authors: Adarsh Kumarappan, Ayushi Mehrotra
Abstract: The SmoothLLM defense provides a certification guarantee against jailbreaking attacks, but it relies on a strict "k-unstable" assumption that rarely holds in practice. This strong assumption can limit the trustworthiness of the provided safety certificate. In this work, we address this limitation by introducing a more realistic probabilistic framework, "(k, $\varepsilon$)-unstable," to certify defenses against diverse jailbreaking attacks, from gradient-based (GCG) to semantic (PAIR). We derive a new, data-informed lower bound on SmoothLLM's defense probability by incorporating empirical models of attack success, providing a more trustworthy and practical safety certificate. By introducing the notion of (k, $\varepsilon$)-unstable, our framework provides practitioners with actionable safety guarantees, enabling them to set certification thresholds that better reflect the real-world behavior of LLMs. Ultimately, this work contributes a practical and theoretically-grounded mechanism to make LLMs more resistant to the exploitation of their safety alignments, a critical challenge in secure AI deployment.
Authors: Adarsh Kumarappan, Ananya Mujoo
Abstract: Multi-turn conversational attacks, which leverage psychological principles like Foot-in-the-Door (FITD), where a small initial request paves the way for a more significant one, to bypass safety alignments, pose a persistent threat to Large Language Models (LLMs). Progress in defending against these attacks is hindered by a reliance on manual, hard-to-scale dataset creation. This paper introduces a novel, automated pipeline for generating large-scale, psychologically-grounded multi-turn jailbreak datasets. We systematically operationalize FITD techniques into reproducible templates, creating a benchmark of 1,500 scenarios across illegal activities and offensive content. We evaluate seven models from three major LLM families under both multi-turn (with history) and single-turn (without history) conditions. Our results reveal stark differences in contextual robustness: models in the GPT family demonstrate a significant vulnerability to conversational history, with Attack Success Rates (ASR) increasing by as much as 32 percentage points. In contrast, Google's Gemini 2.5 Flash exhibits exceptional resilience, proving nearly immune to these attacks, while Anthropic's Claude 3 Haiku shows strong but imperfect resistance. These findings highlight a critical divergence in how current safety architectures handle conversational context and underscore the need for defenses that can resist narrative-based manipulation.
Authors: Shivam Pal, Sakshi Varshney, Piyush Rai
Abstract: Deep neural networks are prone to learning shortcuts, spurious correlations present in the training data that undermine out-of-distribution (OOD) generalization. Most prior work mitigates shortcut learning through input-space reweighting, either relying on explicit shortcut labels or inferring shortcut structure from heuristics such as per-sample loss. Moreover, these approaches typically assume the presence of some shortcut-conflicting examples in the training set, an assumption that is often violated in practice, particularly in medical imaging where data is aggregated across institutions with different acquisition protocols. We propose a latent-space method that views shortcut learning as over-reliance on shortcut-aligned axes. In a disentangled latent space, we identify candidate shortcut-aligned axes via their strong correlation with labels and reduce classifier reliance on them by injecting targeted anisotropic noise during training. Unlike prior latent-space based approaches that remove, project out, or adversarially suppress shortcut features, our method preserves the full representation and instead impose functional invariance by regularizing the classifier's sensitivity along those axes. We show that injecting anisotropic noise induces targeted Jacobian and curvature regularization, effectively flattening the decision boundary along shortcut axes while leaving core feature dimensions largely unaffected. Our method achieves state-of-the-art OOD performance across standard shortcut-learning benchmarks without requiring shortcut labels or shortcut-conflicting samples.
Authors: Finn G. Vamosi, Nils D. Forkert
Abstract: When people reason about cause and effect, they often consider many competing "what if" scenarios before deciding which explanation fits best. Analogously, advanced language models capable of causal inference can consider multiple interventions and counterfactuals to judge the validity of causal claims. Crucially, this type of reasoning is less like a single calculation and more like an internal dialogue between alternative hypotheses. In this paper, we make this dialogue explicit through a dual-agent debate framework where one model provides a structured causal inference, and the other critically examines this reasoning for logical flaws. When disagreements arise, the agents attempt to persuade each other, challenging each other's logic and revising their conclusions until they converge on a mutually agreed answer. To take advantage of this deliberative process, we specifically use reasoning language models, whose strengths in both causal inference and adversarial debate remain under-explored relative to standard large language models. We evaluate our approach on the CLadder dataset, a benchmark linking natural language questions to formally defined causal graphs across all three rungs of Pearl's ladder of causation. With Qwen3 and DeepSeek-R1 as debater agents, we demonstrate that multi-agent debate improves DeepSeek-R1's overall accuracy in causal inference from 78.03% to 87.45%, with the counterfactual category specifically improving from 67.94% to 80.04% accuracy. Similarly, Qwen3's overall accuracy improves from 84.16% to 89.41%, and counterfactual questions from 71.53% to 80.35%, showing that even strong models can still benefit greatly from debate with weaker agents. Our results highlight the potential of reasoning models as building blocks for multi-agent systems in causal inference, and demonstrate the importance of diverse perspectives in causal problem-solving.
Authors: Mansi Maheshwari, John C. Raisbeck, Bruno Castro da Silva
Abstract: Artificial neural networks have shown remarkable success in supervised learning when trained on a single task using a fixed dataset. However, when neural networks are trained on a reinforcement learning task, their ability to continue learning from new experiences declines over time. This decline in learning ability is known as plasticity loss. To restore plasticity, prior work has explored periodically resetting the parameters of the learning network, a strategy that often improves performance. However, such resets come at the cost of a temporary drop in performance, which can be dangerous in real-world settings. To overcome this instability, we introduce AltNet, a reset-based approach that restores plasticity without performance degradation by leveraging a pair of twin networks. The use of twin networks anchors performance during resets through a mechanism that allows networks to periodically alternate roles: one network learns as it acts in the environment, while the other learns off-policy from the active network's interactions through a replay buffer. At fixed intervals, the active network is reset and the passive network, having learned from prior experience, becomes the new active network. AltNet restores plasticity, improving sample efficiency and achieving higher performance, while avoiding performance drops that pose risks in safety-critical settings. We demonstrate these advantages in several high-dimensional control tasks from the DeepMind Control Suite, where AltNet outperforms various relevant baseline methods, as well as state-of-the-art reset-based techniques.
Authors: Pedro M. P. Curvo, Jan-Willem van de Meent, Maksim Zhdanov
Abstract: A key scalability challenge in neural solvers for industrial-scale physics simulations is efficiently capturing both fine-grained local interactions and long-range global dependencies across millions of spatial elements. We introduce the Multi-Scale Patch Transformer (MSPT), an architecture that combines local point attention within patches with global attention to coarse patch-level representations. To partition the input domain into spatially-coherent patches, we employ ball trees, which handle irregular geometries efficiently. This dual-scale design enables MSPT to scale to millions of points on a single GPU. We validate our method on standard PDE benchmarks (elasticity, plasticity, fluid dynamics, porous flow) and large-scale aerodynamic datasets (ShapeNet-Car, Ahmed-ML), achieving state-of-the-art accuracy with substantially lower memory footprint and computational cost.
Authors: Chenhao Sun, Yuhao Mao, Martin Vechev
Abstract: Randomized Smoothing (RS) is a prominent technique for certifying the robustness of neural networks against adversarial perturbations. With RS, achieving high accuracy at small radii requires a small noise variance, while achieving high accuracy at large radii requires a large noise variance. However, the global noise variance used in the standard RS formulation leads to a fundamental limitation: there exists no global noise variance that simultaneously achieves strong performance at both small and large radii. To break through the global variance limitation, we propose a dual RS framework which enables input-dependent noise variances. To achieve that, we first prove that RS remains valid with input-dependent noise variances, provided the variance is locally constant around each input. Building on this result, we introduce two components: (i) a variance estimator predicts an optimal noise variance for each input, (ii) this estimated variance is then used by a standard RS classifier. The variance estimator is independently smoothed via RS to ensure local constancy, enabling flexible design. We also introduce training strategies to iteratively optimize the two components. Experiments on CIFAR-10 demonstrate that our dual RS method provides strong performance for both small and large radii-unattainable with global noise variance-while incurring only a 60% computational overhead at inference. Moreover, it outperforms prior input-dependent noise approaches across most radii, with gains at radii 0.5, 0.75, and 1.0 of 15.6%, 20.0%, and 15.7%. On ImageNet, dual RS remains effective across all radii, with advantages of 8.6%, 17.1%, and 9.1% at radii 0.5, 1.0, and 1.5. Additionally, the dual RS framework provides a routing perspective for certified robustness, improving the accuracy-robustness trade-off with off-the-shelf expert RS models.
Authors: Zhongjian Qiao, Rui Yang, Jiafei Lyu, Xiu Li, Zhongxiang Dai, Zhuoran Yang, Siyang Gao, Shuang Qiu
Abstract: Single-domain offline reinforcement learning (RL) often suffers from limited data coverage, while cross-domain offline RL handles this issue by leveraging additional data from other domains with dynamics shifts. However, existing studies primarily focus on train-time robustness (handling dynamics shifts from training data), neglecting the test-time robustness against dynamics perturbations when deployed in practical scenarios. In this paper, we investigate dual (both train-time and test-time) robustness against dynamics shifts in cross-domain offline RL. We first empirically show that the policy trained with cross-domain offline RL exhibits fragility under dynamics perturbations during evaluation, particularly when target domain data is limited. To address this, we introduce a novel robust cross-domain Bellman (RCB) operator, which enhances test-time robustness against dynamics perturbations while staying conservative to the out-of-distribution dynamics transitions, thus guaranteeing the train-time robustness. To further counteract potential value overestimation or underestimation caused by the RCB operator, we introduce two techniques, the dynamic value penalty and the Huber loss, into our framework, resulting in the practical \textbf{D}ual-\textbf{RO}bust \textbf{C}ross-domain \textbf{O}ffline RL (DROCO) algorithm. Extensive empirical results across various dynamics shift scenarios show that DROCO outperforms strong baselines and exhibits enhanced robustness to dynamics perturbations.
Authors: Chubin Zhang, Zhenglin Wan, Feng Chen, Fuchao Yang, Lang Feng, Yaxin Zhou, Xingrui Yu, Yang You, Ivor Tsang, Bo An
Abstract: Diffusion and flow matching policies offer expressive, multimodal action modeling, yet they are frequently unstable in online reinforcement learning (RL) due to intractable likelihoods and gradients propagating through long sampling chains. Conversely, tractable parameterizations such as Gaussians lack the expressiveness needed for complex control -- exposing a persistent tension between optimization stability and representational power. We address this tension with a key structural principle: decoupling optimization from generation. Building on this, we introduce GoRL (Generative Online Reinforcement Learning), an algorithm-agnostic framework that trains expressive policies from scratch by confining policy optimization to a tractable latent space while delegating action synthesis to a conditional generative decoder. Using a two-timescale alternating schedule and anchoring decoder refinement to a fixed prior, GoRL enables stable optimization while continuously expanding expressiveness. Empirically, GoRL consistently outperforms unimodal and generative baselines across diverse continuous-control tasks. Notably, on the challenging HopperStand task, it achieves episodic returns exceeding 870 -- more than $3\times$ that of the strongest baseline -- demonstrating a practical path to policies that are both stable and highly expressive. Our code is publicly available at https://github.com/bennidict23/GoRL.
Authors: Jialai She
Abstract: Shapley values, a gold standard for feature attribution in Explainable AI, face two key challenges. First, the canonical Shapley framework assumes that the worth function is additive, yet real-world payoff constructions--driven by non-Gaussian distributions, heavy tails, feature dependence, or domain-specific loss scales--often violate this assumption, leading to distorted attributions. Second, achieving sparse explanations in high-dimensional settings by computing dense Shapley values and then applying ad hoc thresholding is costly and risks inconsistency. We introduce Sparse Isotonic Shapley Regression (SISR), a unified nonlinear explanation framework. SISR simultaneously learns a monotonic transformation to restore additivity--obviating the need for a closed-form specification--and enforces an L0 sparsity constraint on the Shapley vector, enhancing computational efficiency in large feature spaces. Its optimization algorithm leverages Pool-Adjacent-Violators for efficient isotonic regression and normalized hard-thresholding for support selection, ensuring ease in implementation and global convergence guarantees. Analysis shows that SISR recovers the true transformation in a wide range of scenarios and achieves strong support recovery even in high noise. Moreover, we are the first to demonstrate that irrelevant features and inter-feature dependencies can induce a true payoff transformation that deviates substantially from linearity. Extensive experiments demonstrate that SISR stabilizes attributions across payoff schemes and correctly filters irrelevant features; in contrast, standard Shapley values suffer severe rank and sign distortions. By unifying nonlinear transformation estimation with sparsity pursuit, SISR advances the frontier of nonlinear explainability, providing a theoretically grounded and practical attribution framework.
Authors: Vegard Flovik
Abstract: Deep neural networks achieve impressive performance but remain difficult to interpret and control. We present SALVE (Sparse Autoencoder-Latent Vector Editing), a unified "discover, validate, and control" framework that bridges mechanistic interpretability and model editing. Using an $\ell_1$-regularized autoencoder, we learn a sparse, model-native feature basis without supervision. We validate these features with Grad-FAM, a feature-level saliency mapping method that visually grounds latent features in input data. Leveraging the autoencoder's structure, we perform precise and permanent weight-space interventions, enabling continuous modulation of both class-defining and cross-class features. We further derive a critical suppression threshold, $\alpha_{crit}$, quantifying each class's reliance on its dominant feature, supporting fine-grained robustness diagnostics. Our approach is validated on both convolutional (ResNet-18) and transformer-based (ViT-B/16) models, demonstrating consistent, interpretable control over their behavior. This work contributes a principled methodology for turning feature discovery into actionable model edits, advancing the development of transparent and controllable AI systems.
Authors: Yulun Jiang, Liangze Jiang, Damien Teney, Michael Moor, Maria Brbic
Abstract: Reinforcement learning (RL) has enabled the training of large language model (LLM) agents to interact with the environment and to solve multi-turn long-horizon tasks. However, the RL-trained agents often struggle in tasks that require active exploration and fail to efficiently adapt from trial-and-error experiences. In this paper, we present LaMer, a general Meta-RL framework that enables LLM agents to actively explore and learn from the environment feedback at test time. LaMer consists of two key components: (i) a cross-episode training framework to encourage exploration and long-term rewards optimization; and (ii) in-context policy adaptation via reflection, allowing the agent to adapt their policy from task feedback signal without gradient update. Experiments across diverse environments show that LaMer significantly improves performance over RL baselines, with 11%, 14%, and 19% performance gains on Sokoban, MineSweeper and Webshop, respectively. Moreover, LaMer also demonstrates better generalization to more challenging or previously unseen tasks compared to the RL-trained agents. Overall, our results demonstrate that Meta-RL provides a principled approach to induce exploration in language agents, enabling more robust adaptation to novel environments through learned exploration strategies.
Authors: Andrew Polar, Michael Poluektov
Abstract: The present paper introduces concurrency-driven enhancements to the training algorithm for the Kolmogorov-Arnold networks (KANs) that is based on the Newton-Kaczmarz (NK) method. As indicated by prior research, KANs trained using the NK-based approach significantly overtake classical neural networks based on multilayer perceptrons (MLPs) in terms of accuracy and training time. Although certain parts of the algorithm, such as evaluation of the basis functions, can be parallelised, a fundamental limitation lies in the computation of the updates' values that is sequential: each update depends on the results of the previous step, obstructing parallel application. However, substantial acceleration is achievable. Three complementary strategies are proposed in the present paper: (i) a pre-training procedure tailored to the NK updates' structure, (ii) training on disjoint subsets of data, followed by models' merging, not in the context of federated learning, but as a mechanism for accelerating the convergence, and (iii) a parallelisation technique suitable for execution on field-programmable gate arrays (FPGAs), which is implemented and tested directly on the device. All experimental results presented in this work are fully reproducible, with the complete source codes available online.
Authors: Rajeeb Thapa Chhetri, Saurab Thapa, Avinash Kumar, Zhixiong Chen
Abstract: A critical vulnerability of supervised deep learning in high-dimensional tabular domains is "generalization collapse": models form precise decision boundaries around known training distributions but fail catastrophically when encountering Out-of-Distribution (OOD) data. To overcome this, we propose Latent Sculpting, a hierarchical, two-stage representation learning architecture designed to enforce explicit structural boundaries prior to density estimation. In the first stage, a Transformer-based tabular encoder is trained using our novel Binary Latent Sculpting loss. This objective explicitly condenses benign network traffic into a dense, low-entropy hypersphere while enforcing a strict geometric minimum-distance margin for anomalous patterns. In the second stage, a Masked Autoregressive Flow (MAF) maps this structurally optimized manifold to calculate exact, probabilistic anomaly thresholds. We evaluate this methodology on the CIC-IDS-2017 benchmark under a rigorous zero-shot protocol, deliberately withholding complex attack classes during training to test true OOD generalization. Averaged across three random initialization seeds to ensure statistical robustness, our framework maintains near-perfect classification on known signatures (F1 = 0.980 +/- 0.000) while achieving an overall zero-shot OOD F1-Score of 0.867 +/- 0.021 and an AUROC of 0.913 +/- 0.010 at an 85th-percentile threshold. Most notably, the model achieves an average recall of 78.7% (peaking at 97.2%) on stealthy "Infiltration" attacks and over 94% on low-volume DoS variations - complex distributional shifts where standard supervised and unsupervised baselines historically suffer near-total detection failure. These empirical results demonstrate that explicitly decoupling topological manifold structuring from probabilistic density estimation establishes a highly stable and scalable defense against zero-day cyber threats.
Authors: Francesco Vitale, Paolo Palmiero, Massimiliano Rak, Nicola Mazzocca
Abstract: Online gaming is a popular activity involving the adoption of complex systems and network infrastructures. The relevance of gaming, which generates large amounts of market revenue, drove research in modeling network devices' behavior to evaluate bandwidth consumption, predict and sustain high loads, and detect malicious activity. In this context, process mining appears promising due to its ability to combine data-driven analyses with model-based insights. In this paper, we propose a process mining-based method that analyzes gaming network traffic, allowing: unsupervised characterization of different states from gaming network data; encoding such states through process mining into interpretable Petri nets; and classification of gaming network traffic data to identify different video games being played. We apply the method to the UPSIDE case study, involving gaming network data of several devices interacting with two video games: Clash Royale and Rocket League. Results demonstrate that the gaming network behavior can be effectively and interpretably modeled through states represented as Petri nets with sufficient coherence and specificity while maintaining a good classification accuracy of the two different video games.
Authors: Pareesa Ameneh Golnari, Adarsh Kumarappan, Wen Wen, Xiaoyu Liu, Gabriel Ryan, Yuting Sun, Shengyu Fu, Elsie Nallipogu
Abstract: DevBench is a telemetry-driven benchmark designed to evaluate Large Language Models (LLMs) on realistic code completion tasks. It includes 1,800 evaluation instances across six programming languages and six task categories derived from real developer telemetry, such as API usage and code purpose understanding. Unlike prior benchmarks, it emphasizes ecological validity, avoids training data contamination, and enables detailed diagnostics. The evaluation combines functional correctness, similarity-based metrics, and LLM-judge assessments focused on usefulness and contextual relevance. 9 state-of-the-art models were assessed, revealing differences in syntactic precision, semantic reasoning, and practical utility. Our benchmark provides actionable insights to guide model selection and improvement-detail that is often missing from other benchmarks but is essential for both practical deployment and targeted model development.
Authors: Xiaohong Yang, Tong Xie, Minghui Liwang, Chikai Shang, Yang Lu, Zhenzhen Jiao, Liqun Fu, Seyyedali Hosseinalipour
Abstract: Training large language models (LLMs) at the network edge faces fundamental challenges arising from device resource constraints, severe data heterogeneity, and heightened privacy risks. To address these challenges, we propose ELSA (Efficient LLM-centric Split Aggregation), a novel framework that systematically integrates split learning (SL) and hierarchical federated learning (HFL) for distributed LLM fine-tuning over resource-constrained edge networks. ELSA introduces three key innovations. First, it employs a task-agnostic, behavior-aware client clustering mechanism that constructs semantic fingerprints using public probe inputs and symmetric Kullback-Leibler (KL) divergence, augmented by prediction-consistency trust scoring and latency-aware edge assignment to jointly mitigate data heterogeneity, device unreliability, and communication constraints. Second, it employs a resource-aware dynamic model splitting strategy to adaptively partition the LLM into three segments across clients and edge servers, with the cloud used only for adapter aggregation, enabling an effective balance between on-device computation cost and global convergence stability. Third, it incorporates a lightweight communication scheme based on computational sketches combined with semantic subspace orthogonal perturbation (SS-OP) to reduce communication overhead while mitigating privacy leakage during model exchanges across the network. Extensive experiments across diverse NLP tasks demonstrate that ELSA consistently outperforms state-of-the-art baselines in terms of adaptability, convergence behavior, and robustness, establishing a scalable and privacy-aware solution for edge-side LLM fine-tuning under resource constraints.
Authors: Tobias Habermann, Michael Mecik, Zhenyu Wang, C\'esar David Vera, Martin Kumm, Mario Garrido
Abstract: Among hardware accelerators for deep-learning inference, data flow implementations offer low latency and high throughput capabilities. In these architectures, each neuron is mapped to a dedicated hardware unit, making them well-suited for field-programmable gate array (FPGA) implementation. Previous unrolled implementations mostly focus on fully connected networks because of their simplicity, although it is well known that convolutional neural networks (CNNs) require fewer computations for the same accuracy. When observing the data flow in CNNs, pooling layers and convolutional layers with a stride larger than one, the number of data at their output is reduced with respect to their input. This data reduction strongly affects the data rate in a fully parallel implementation, making hardware units heavily underutilized unless it is handled properly. This work addresses this issue by analyzing the data flow of CNNs and presents a novel approach to designing data-rate-aware, continuous-flow CNN architectures. The proposed approach ensures a high hardware utilization close to 100% by interleaving low data rate signals and sharing hardware units, as well as using the right parallelization to achieve the throughput of a fully parallel implementation. The results show that a significant amount of the arithmetic logic can be saved, which allows implementing complex CNNs like MobileNet on a single FPGA with high throughput.
Authors: Huanlin Gao, Ping Chen, Fuyuan Shi, Ruijia Wu, Li YanTao, Qiang Hui, Yuren You, Ting Lu, Chao Tan, Shaoan Zhao, Zhaoxiang Liu, Fang Zhao, Kai Wang, Shiguo Lian
Abstract: We present MeanCache, a training-free caching framework for efficient Flow Matching inference. Existing caching methods reduce redundant computation but typically rely on instantaneous velocity information (e.g., feature caching), which often leads to severe trajectory deviations and error accumulation under high acceleration ratios. MeanCache introduces an average-velocity perspective: by leveraging cached Jacobian--vector products (JVP) to construct interval average velocities from instantaneous velocities, it effectively mitigates local error accumulation. To further improve cache timing and JVP reuse stability, we develop a trajectory-stability scheduling strategy as a practical tool, employing a Peak-Suppressed Shortest Path under budget constraints to determine the schedule. Experiments on FLUX.1, Qwen-Image, and HunyuanVideo demonstrate that MeanCache achieves 4.12X and 4.56X and 3.59X acceleration, respectively, while consistently outperforming state-of-the-art caching baselines in generation quality. We believe this simple yet effective approach provides a new perspective for Flow Matching inference and will inspire further exploration of stability-driven acceleration in commercial-scale generative models.
Authors: Pedro Chumpitaz-Flores, My Duong, Ying Mao, Kaixun Hua
Abstract: Pairwise-constrained clustering incorporates side information through must-link (ML) and cannot-link (CL) relations between samples. While these constraints can improve cluster quality, they complicate optimization at scale and limit quantum and hybrid approaches through the size of the encoded problem. PASS is a scalable framework for pairwise-constrained k-means that concentrates optimization on a small working subset while updating remaining assignments through re-centering. Cannot-link feasibility under subset-restricted updates is formalized as a list-coloring problem on the induced constraint subgraph, yielding a checkable repair certificate with verifiable outcomes. The same subset restriction produces reduced classical subproblems and smaller quantum formulations, enabling a reduction-based hybrid evaluation under a simulation protocol. Infeasible constraint sets are handled explicitly: the pipeline returns a verifiable repair under stated conditions or reports residual conflicts under the same evaluation protocol. Across diverse benchmarks, PASS attains competitive SSE with lower runtime and returns solutions on instances where strong baselines do not finish within a fixed time budget.
Authors: Zhuochen Liu, Hans Walker, Rahul Jain
Abstract: Neural network models are increasingly used for state estimation in control and decision-making problems, yet it remains unclear to what extent they behave as principled filters in nonlinear dynamical systems. Unlike classical filters, which rely on explicit knowledge of system dynamics and noise models, neural estimators can be trained purely from data without access to the underlying system equations. In this work, we present a systematic empirical comparison between such model-free neural network models and classical filtering methods across multiple nonlinear scenarios. Our study evaluates Transformer-based models, state-space neural networks, and recurrent architectures alongside particle filters and nonlinear Kalman filters. The results show that neural models (in particular, state-space models (SSMs)) achieve state estimation performance that approaches strong nonlinear Kalman filters in nonlinear scenarios and outperform weaker classical baselines despite lacking access to system models, while also attaining substantially higher inference throughput.
Authors: Akash Pandey, Payal Mohapatra, Wei Chen, Qi Zhu, Sinan Keten
Abstract: Identifying the extent to which every temporal segment influences a model's predictions is essential for explaining model decisions and increasing transparency. While post-hoc explainable methods based on gradients and feature-based attributions have been popular, they suffer from reference state sensitivity and struggle to generalize across time-series datasets, as they treat time points independently and ignore sequential dependencies. Another perspective on explainable time-series classification is through interpretable components of the model, for instance, leveraging self-attention mechanisms to estimate temporal attribution; however, recent findings indicate that these attention weights often fail to provide faithful measures of temporal importance. In this work, we advance this perspective and present a novel explainability-driven deep learning framework, TimeSliver, which jointly utilizes raw time-series data and its symbolic abstraction to construct a representation that maintains the original temporal structure. Each element in this representation linearly encodes the contribution of each temporal segment to the final prediction, allowing us to assign a meaningful importance score to every time point. For time-series classification, TimeSliver outperforms other temporal attribution methods by 11% on 7 distinct synthetic and real-world multivariate time-series datasets. TimeSliver also achieves predictive performance within 2% of state-of-the-art baselines across 26 UEA benchmark datasets, positioning it as a strong and explainable framework for general time-series classification.
Authors: Huaming Du, Yijie Huang, Su Yao, Yiying Wang, Yueyang Zhou, Jingwen Yang, Jinshi Zhang, Han Ji, Yu Zhao, Guisong Liu, Hegui Zhang, Carl Yang, Gang Kou
Abstract: The increasing scale of graph datasets has significantly improved the performance of graph representation learning methods, but it has also introduced substantial training challenges. Graph dataset condensation techniques have emerged to compress large datasets into smaller yet information-rich datasets, while maintaining similar test performance. However, these methods strictly require downstream applications to match the original dataset and task, which often fails in cross-task and cross-domain scenarios. To address these challenges, we propose a novel causal-invariance-based and transferable graph dataset condensation method, named TGCC, providing effective and transferable condensed datasets. Specifically, to preserve domain-invariant knowledge, we first extract domain causal-invariant features from the spatial domain of the graph using causal interventions. Then, to fully capture the structural and feature information of the original graph, we perform enhanced condensation operations. Finally, through spectral-domain enhanced contrastive learning, we inject the causal-invariant features into the condensed graph, ensuring that the compressed graph retains the causal information of the original graph. Experimental results on five public datasets and our novel FinReport dataset demonstrate that TGCC achieves up to a 13.41% improvement in cross-task and cross-domain complex scenarios compared to existing methods, and achieves state-of-the-art performance on 5 out of 6 datasets in the single dataset and task scenario.
Authors: Ege Demirci, Francesco Bullo, Ananthram Swami, Ambuj Singh
Abstract: Recovering missing flows on the edges of a network, while exactly respecting local conservation laws, is a fundamental inverse problem that arises in many systems such as transportation, energy, and mobility. We introduce FlowSymm, a novel architecture that combines (i) a group-action on divergence-free flows, (ii) a graph-attention encoder to learn feature-conditioned weights over these symmetry-preserving actions, and (iii) a lightweight Tikhonov refinement solved via implicit bilevel optimization. The method first anchors the given observation on a minimum-norm divergence-free completion. We then compute an orthonormal basis for all admissible group actions that leave the observed flows invariant and parameterize the valid solution subspace, which shows an Abelian group structure under vector addition. A stack of GATv2 layers then encodes the graph and its edge features into per-edge embeddings, which are pooled over the missing edges and produce per-basis attention weights. This attention-guided process selects a set of physics-aware group actions that preserve the observed flows. Finally, a scalar Tikhonov penalty refines the missing entries via a convex least-squares solver, with gradients propagated implicitly through Cholesky factorization. Across three real-world flow benchmarks (traffic, power, bike), FlowSymm outperforms state-of-the-art baselines in RMSE, MAE and correlation metrics.
Authors: Yanwei Yue, Boci Peng, Xuanbo Fan, Jiaxin Guo, Qiankun Li, Yan Zhang
Abstract: Memory agents, which depart from predefined memory-processing pipelines by endogenously managing the processing, storage, and retrieval of memories, have garnered increasing attention for their autonomy and adaptability. However, existing training paradigms remain constrained: agents often traverse long-horizon sequences of memory operations before receiving sparse and delayed rewards, which hinders truly end-to-end optimization of memory management policies. To address this limitation, we introduce Mem-T, an autonomous memory agent that interfaces with a lightweight hierarchical memory database to perform dynamic updates and multi-turn retrieval over streaming inputs. To effectively train long-horizon memory management capabilities, we further propose MoT-GRPO, a tree-guided reinforcement learning framework that transforms sparse terminal feedback into dense, step-wise supervision via memory operation tree backpropagation and hindsight credit assignment, thereby enabling the joint optimization of memory construction and retrieval. Extensive experiments demonstrate that Mem-T is (1) high-performing, surpassing frameworks such as A-Mem and Mem0 by up to $14.92\%$, and (2) economical, operating on a favorable accuracy-efficiency Pareto frontier and reducing inference tokens per query by $\sim24.45\%$ relative to GAM without sacrificing performance.
Authors: Meng Ding, Zeqing Zhang, Di Wang, Lijie Hu
Abstract: Reliable data attribution is essential for mitigating bias and reducing computational waste in modern machine learning, with the Shapley value serving as the theoretical gold standard. While recent "In-Run" methods bypass the prohibitive cost of retraining by estimating contributions dynamically, they heavily rely on the linear structure of Stochastic Gradient Descent (SGD) and fail to capture the complex dynamics of adaptive optimizers like Adam. In this work, we demonstrate that data attribution is inherently optimizer-dependent: we show that SGD-based proxies diverge significantly from true contributions under Adam (Pearson $R \approx 0.11$), rendering them ineffective for modern training pipelines. To bridge this gap, we propose Adam-Aware In-Run Data Shapley. We derive a closed-form approximation that restores additivity by redefining utility under a fixed-state assumption and enable scalable computation via a novel Linearized Ghost Approximation. This technique linearizes the variance-dependent scaling term, allowing us to compute pairwise gradient dot-products without materializing per-sample gradients. Extensive experiments show that our method achieves near-perfect fidelity to ground-truth marginal contributions ($R > 0.99$) while retaining $\sim$95\% of standard training throughput. Furthermore, our Adam-aware attribution significantly outperforms SGD-based baselines in data attribution downstream tasks.
Authors: Wenze Lin, Zhen Yang, Xitai Jiang, Pony Ma, Gao Huang
Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a promising paradigm for enhancing reasoning in Large Language Models (LLMs). However, it frequently encounters challenges such as entropy collapse, excessive verbosity, and insufficient exploration for hard problems. Crucially, existing reward schemes fail to distinguish between the need for extensive search during problem-solving and the efficiency required for mastered knowledge. In this work, we introduce T2T(Thickening-to-Thinning), a dynamic reward framework inspired by human learning processes. Specifically, it implements a dual-phase mechanism: (1) On incorrect attempts, T2T incentivizes "thickening" (longer trajectories) to broaden the search space and explore novel solution paths; (2) Upon achieving correctness, it shifts to "thinning", imposing length penalties to discourage redundancy, thereby fostering model confidence and crystallizing reasoning capabilities. Extensive experiments on mathematical benchmarks (MATH-500, AIME, AMC) across Qwen-series and Deepseek models demonstrate that T2T significantly outperforms standard GRPO and recent baselines, achieving superior performance.
Authors: Hongyi Li, Han Lin, Jun Xu
Abstract: Oblique decision trees combine the transparency of trees with the power of multivariate decision boundaries, but learning high-quality oblique splits is NP-hard, and practical methods still rely on slow search or theory-free heuristics. We present the Hinge Regression Tree (HRT), which reframes each split as a non-linear least-squares problem over two linear predictors whose max/min envelope induces ReLU-like expressive power. The resulting alternating fitting procedure is exactly equivalent to a damped Newton (Gauss-Newton) method within fixed partitions. We analyze this node-level optimization and, for a backtracking line-search variant, prove that the local objective decreases monotonically and converges; in practice, both fixed and adaptive damping yield fast, stable convergence and can be combined with optional ridge regularization. We further prove that HRT's model class is a universal approximator with an explicit $O(\delta^2)$ approximation rate, and show on synthetic and real-world benchmarks that it matches or outperforms single-tree baselines with more compact structures.
Authors: Zepeng Zhai, Meilin Chen, Jiaxuan Zhao, Junlang Qian, Lei Shen, Yuan Lu
Abstract: Reinforcement Learning with Verifiable Rewards has recently advanced the capabilities of Large Language Models in complex reasoning tasks by providing explicit rule-based supervision. Among RLVR methods, GRPO and its variants have achieved strong empirical performance. Despite their success, we identify that they suffer from Gradient Misassignment in Positives and Gradient Domination in Negatives, which lead to inefficient and suboptimal policy updates. To address these issues, we propose Rewards as Labels (REAL), a novel framework that revisits verifiable rewards as categorical labels rather than scalar weights, thereby reformulating policy optimization as a classification problem. Building on this, we further introduce anchor logits to enhance policy learning. Our analysis reveals that REAL induces a monotonic and bounded gradient weighting, enabling balanced gradient allocation across rollouts and effectively mitigating the identified mismatches. Extensive experiments on mathematical reasoning benchmarks show that REAL improves training stability and consistently outperforms GRPO and strong variants such as DAPO. On the 1.5B model, REAL improves average Pass@1 over DAPO by 6.7%. These gains further scale to 7B model, REAL continues to outperform DAPO and GSPO by 6.2% and 1.7%, respectively. Notably, even with a vanilla binary cross-entropy, REAL remains stable and exceeds DAPO by 4.5% on average.
Authors: Gnankan Landry Regis N'guessan, Bum Jun Kim
Abstract: Radial singular fields, such as $1/r$, $\log r$, and crack-tip profiles, are difficult to model with current coordinate-separable neural architectures. We formally establish this result: any $C^2$ function that is both radial and additively separable must be quadratic, establishing a fundamental obstruction for coordinate-wise power-law models. Motivated by this result, we introduce Radial M\"untz-Sz\'asz Networks (RMN), which represent fields as linear combinations of learnable radial powers $r^\mu$, including negative exponents, together with a limit-stable log-primitive for exact $\log r$ behavior. RMN admits closed-form spatial gradients and Laplacians, enabling physics-informed learning on punctured domains. Across ten 2D and 3D benchmarks, RMN achieves between 1.5 and 51 times lower RMSE than MLPs and between 10 and 100 times lower RMSE than SIREN, while using only 27 parameters, compared with 33,537 for MLPs and 8,577 for SIREN. We extend RMN to incorporate angular dependence (RMN-Angular) and to handle multiple sources with learnable centers (RMN-MC), whose source-center recovery errors fall below $10^{-4}$. We also report controlled failures on smooth, strongly non-radial targets to delineate RMN's operating regime.
Authors: Yicheng Di, Wei Yuan, Tieke He, Yuan Liu, Hongzhi Yin
Abstract: Vision-language pretrained models offer strong transferable representations, yet adapting them in privacy-sensitive multi-party settings is challenging due to the high communication cost of federated optimization and the limited local data on clients. Federated prompt learning mitigates this issue by keeping the VLPM backbone frozen and collaboratively training lightweight prompt parameters. However, existing approaches typically enforce a unified prompt structure and length across clients, which is inadequate under practical client heterogeneity in both data distributions and system resources, and may further introduce conflicts between globally shared and locally optimal knowledge. To address these challenges, we propose \textbf{SDFed}, a heterogeneous federated prompt learning framework that bridges Local-Global Discrepancy via Subspace Refinement and Divergence Control. SDFed maintains a fixed-length global prompt for efficient aggregation while allowing each client to learn a variable-length local prompt to better match its data characteristics and capacity. To mitigate local-global conflicts and facilitate effective knowledge transfer, SDFed introduces a subspace refinement method for local prompts and an information retention and divergence control strategy that preserves key local information while maintaining appropriate separability between global and local representations. Extensive experiments on several datasets demonstrate that SDFed consistently improves performance and robustness in heterogeneous federated settings.
Authors: Xinxu Wei, Rong Zhou, Lifang He, Yu Zhang
Abstract: With the growing interest in foundation models for brain signals, graph-based pretraining has emerged as a promising paradigm for learning transferable representations from connectome data. However, existing contrastive and masked autoencoder methods typically rely on naive random dropping or masking for augmentation, which is ill-suited for brain graphs and hypergraphs as it disrupts semantically meaningful connectivity patterns. Moreover, commonly used graph-level readout and reconstruction schemes fail to capture global structural information, limiting the robustness of learned representations. In this work, we propose a unified diffusion-based pretraining framework that addresses both limitations. First, diffusion is designed to guide structure-aware dropping and masking strategies, preserving brain graph semantics while maintaining effective pretraining diversity. Second, diffusion enables topology-aware graph-level readout and node-level global reconstruction by allowing graph embeddings and masked nodes to aggregate information from globally related regions. Extensive experiments across multiple neuroimaging datasets with over 25,000 subjects and 60,000 scans involving various mental disorders and brain atlases demonstrate consistent performance improvements.
Authors: Efe Kahraman, Giulio Tosato
Abstract: We investigate document page ordering on 5,461 shuffled WOO documents (Dutch freedom of information releases) using page embeddings. These documents are heterogeneous collections such as emails, legal texts, and spreadsheets compiled into single PDFs, where semantic ordering signals are unreliable. We compare five methods, including pointer networks, seq2seq transformers, and specialized pairwise ranking models. The best performing approach successfully reorders documents up to 15 pages, with Kendall's tau ranging from 0.95 for short documents (2-5 pages) to 0.72 for 15 page documents. We observe two unexpected failures: seq2seq transformers fail to generalize on long documents (Kendall's tau drops from 0.918 on 2-5 pages to 0.014 on 21-25 pages), and curriculum learning underperforms direct training by 39% on long documents. Ablation studies suggest learned positional encodings are one contributing factor to seq2seq failure, though the degradation persists across all encoding variants, indicating multiple interacting causes. Attention pattern analysis reveals that short and long documents require fundamentally different ordering strategies, explaining why curriculum learning fails. Model specialization achieves substantial improvements on longer documents (+0.21 tau).
Authors: Peng Cheng, Jiucheng Zang, Qingnan Li, Liheng Ma, Yufei Cui, Yingxue Zhang, Boxing Chen, Ming Jian, Wen Tong
Abstract: Muon-style optimizers leverage Newton-Schulz (NS) iterations to orthogonalize updates, yielding update geometries that often outperform Adam-series methods. However, this orthogonalization discards magnitude information, rendering training sensitive to step-size hyperparameters and vulnerable to high-energy bursts. To mitigate this, we introduce TrasMuon (\textbf{T}rust \textbf{R}egion \textbf{A}daptive \textbf{S}caling \textbf{Muon}). TrasMuon preserves the near-isometric geometry of Muon while stabilizing magnitudes through (i) global RMS calibration and (ii) energy-based trust-region clipping. We demonstrate that while reintroducing adaptive scaling improves optimization efficiency, it typically exacerbates instability due to high-energy outliers. TrasMuon addresses this by defining a trust region based on relative energy ratios, confining updates to a stable zone. Empirical experiments on vision and language models demonstrate that TrasMuon converges faster than baselines. Furthermore, experiments without warmup stages confirm TrasMuon's superior stability and robustness.
Authors: Mingqiao Zhang, Qiyao Peng, Yumeng Wang, Chunyuan Liu, Hongtao Liu
Abstract: The expanding integration of Large Language Models (LLMs) into recommender systems poses critical challenges to evaluation reliability. This paper identifies and investigates a previously overlooked issue: benchmark data leakage in LLM-based recommendation. This phenomenon occurs when LLMs are exposed to and potentially memorize benchmark datasets during pre-training or fine-tuning, leading to artificially inflated performance metrics that fail to reflect true model performance. To validate this phenomenon, we simulate diverse data leakage scenarios by conducting continued pre-training of foundation models on strategically blended corpora, which include user-item interactions from both in-domain and out-of-domain sources. Our experiments reveal a dual-effect of data leakage: when the leaked data is domain-relevant, it induces substantial but spurious performance gains, misleadingly exaggerating the model's capability. In contrast, domain-irrelevant leakage typically degrades recommendation accuracy, highlighting the complex and contingent nature of this contamination. Our findings reveal that data leakage acts as a critical, previously unaccounted-for factor in LLM-based recommendation, which could impact the true model performance. We release our code at https://github.com/yusba1/LLMRec-Data-Leakage.
Authors: Guojian Zhan, Letian Tao, Pengcheng Wang, Yixiao Wang, Yiheng Li, Yuxin Chen, Hongyang Li, Masayoshi Tomizuka, Shengbo Eben Li
Abstract: Learning expressive and efficient policy functions is a promising direction in reinforcement learning (RL). While flow-based policies have recently proven effective in modeling complex action distributions with a fast deterministic sampling process, they still face a trade-off between expressiveness and computational burden, which is typically controlled by the number of flow steps. In this work, we propose mean velocity policy (MVP), a new generative policy function that models the mean velocity field to achieve the fastest one-step action generation. To ensure its high expressiveness, an instantaneous velocity constraint (IVC) is introduced on the mean velocity field during training. We theoretically prove that this design explicitly serves as a crucial boundary condition, thereby improving learning accuracy and enhancing policy expressiveness. Empirically, our MVP achieves state-of-the-art success rates across several challenging robotic manipulation tasks from Robomimic and OGBench. It also delivers substantial improvements in training and inference speed over existing flow-based policy baselines.
Authors: Jorge Carrasco-Pollo, Floor Eijkelboom, Jan-Willem van de Meent
Abstract: We introduce Pawsterior, a variational flow-matching framework for improved and extended simulation-based inference (SBI). Many SBI problems involve posteriors constrained by structured domains, such as bounded physical parameters or hybrid discrete-continuous variables, yet standard flow-matching methods typically operate in unconstrained spaces. This mismatch leads to inefficient learning and difficulty respecting physical constraints. Our contributions are twofold. First, generalizing the geometric inductive bias of CatFlow, we formalize endpoint-induced affine geometric confinement, a principle that incorporates domain geometry directly into the inference process via a two-sided variational model. This formulation improves numerical stability during sampling and leads to consistently better posterior fidelity, as demonstrated by improved classifier two-sample test performance across standard SBI benchmarks. Second, and more importantly, our variational parameterization enables SBI tasks involving discrete latent structure (e.g., switching systems) that are fundamentally incompatible with conventional flow-matching approaches. By addressing both geometric constraints and discrete latent structure, Pawsterior provides a principled way to apply flow-matching in a broader range of structured SBI settings.
Authors: Zhimin Zhao
Abstract: Code generation has progressed more reliably than reinforcement learning, largely because code has an information structure that makes it learnable. Code provides dense, local, verifiable feedback at every token, whereas most reinforcement learning problems do not. This difference in feedback quality is not binary but graded. We propose a five-level hierarchy of learnability based on information structure and argue that the ceiling on ML progress depends less on model size than on whether a task is learnable at all. The hierarchy rests on a formal distinction among three properties of computational problems (expressibility, computability, and learnability). We establish their pairwise relationships, including where implications hold and where they fail, and present a unified template that makes the structural differences explicit. The analysis suggests why supervised learning on code scales predictably while reinforcement learning does not, and why the common assumption that scaling alone will solve remaining ML challenges warrants scrutiny.
Authors: Davide Casnici, Martin Lefebvre, Justin Dauwels, Charlotte Frenkel
Abstract: Predictive coding (PC) is a biologically inspired algorithm for training neural networks that relies only on local updates, allowing parallel learning across layers. However, practical implementations face two key limitations: error signals must still propagate from the output to early layers through multiple inference-phase steps, and feedback decays exponentially during this process, leading to vanishing updates in early layers. We propose direct Kolen-Pollack predictive coding (DKP-PC), which simultaneously addresses both feedback delay and exponential decay, yielding a more efficient and scalable variant of PC while preserving update locality. Leveraging direct feedback alignment and direct Kolen-Pollack algorithms, DKP-PC introduces learnable feedback connections from the output layer to all hidden layers, establishing a direct pathway for error transmission. This yields an algorithm that reduces the theoretical error propagation time complexity from O(L), with L being the network depth, to O(1), removing depth-dependent delay in error signals. Moreover, empirical results demonstrate that DKP-PC achieves performance at least comparable to, and often exceeding, that of standard PC, while offering improved latency and computational performance, supporting its potential for custom hardware-efficient implementations.
Authors: Leo Muxing Wang, Connor Mclaughlin, Lili Su
Abstract: Learning with shared representation is widely recognized as an effective way to separate commonalities from heterogeneity across various heterogeneous sources. Most existing work includes all related data sources via simultaneously training a common feature extractor and source-specific heads. It is well understood that data sources with low relevance or poor quality may hinder representation learning. In this paper, we further dive into the question of which data sources should be learned jointly by focusing on the traditionally deemed ``good'' collection of sources, in which individual sources have similar relevance and qualities with respect to the true underlying common structure. Towards tractability, we focus on the linear setting where sources share a low-dimensional subspace. We find that source screening can play a central role in statistically optimal subspace estimation. We show that, for a broad class of problem instances, training on a carefully selected subset of sources suffices to achieve minimax optimality, even when a substantial portion of data is discarded. We formalize the notion of an informative subpopulation, develop algorithms and practical heuristics for identifying such subsets, and validate their effectiveness through both theoretical analysis and empirical evaluations on synthetic and real-world datasets.
Authors: Rong Fu, Jia Yee Tan, Zijian Zhang, Ziming Wang, Zhaolu Kang, Muge Qi, Shuning Zhang, Simon Fong
Abstract: Multimodal systems are vulnerable to partial or complete loss of input channels at deployment, which undermines reliability in real-world settings. This paper presents ModalImmune, a training framework that enforces modality immunity by intentionally and controllably collapsing selected modality information during training so the model learns joint representations that are robust to destructive modality influence. The framework combines a spectrum-adaptive collapse regularizer, an information-gain guided controller for targeted interventions, curvature-aware gradient masking to stabilize destructive updates, and a certified Neumann-truncated hyper-gradient procedure for automatic meta-parameter adaptation. Empirical evaluation on standard multimodal benchmarks demonstrates that ModalImmune improves resilience to modality removal and corruption while retaining convergence stability and reconstruction capacity.
Authors: Zehao Jin, Yaoye Zhu, Chen Zhang, Yanan Sui
Abstract: Whole-brain biological neural networks naturally support the learning and control of whole-body movements. However, the use of brain connectomes as neural network controllers in embodied reinforcement learning remains unexplored. We investigate using the exact neural architecture of an adult fruit fly's brain for the control of its body movement. We develop Fly-connectomic Graph Model (FlyGM), whose static structure is identical to the complete connectome of an adult Drosophila for whole-body locomotion control. To perform dynamical control, FlyGM represents the static connectome as a directed message-passing graph to impose a biologically grounded information flow from sensory inputs to motor outputs. Integrated with a biomechanical fruit fly model, our method achieves stable control across diverse locomotion tasks without task-specific architectural tuning. To verify the structural advantages of the connectome-based model, we compare it against a degree-preserving rewired graph, a random graph, and multilayer perceptrons, showing that FlyGM yields higher sample efficiency and superior performance. This work demonstrates that static brain connectomes can be transformed to instantiate effective neural policy for embodied learning of movement control.
Authors: Pranay Anchuri
Abstract: Concentrated liquidity provision in decentralized exchanges presents a fundamental Impulse Control problem. Liquidity Providers (LPs) face a non-trivial trade-off between maximizing fee accrual through tight price-range concentration and minimizing the friction costs of rebalancing, including gas fees and swap slippage. Existing methods typically employ heuristic or threshold strategies that fail to account for market dynamics. This paper formulates liquidity management as an optimal control problem and derives the corresponding Hamilton-Jacobi-Bellman quasi-variational inequality (HJB-QVI). We present an approximate solution RAmmStein, a Deep Reinforcement Learning method that incorporates the mean-reversion speed (theta) of an Ornstein-Uhlenbeck process among other features as input to the model. We demonstrate that the agent learns to separate the state space into regions of action and inaction. We further extend the framework with RAmmStein-Width, which jointly optimizes rebalancing timing and position width via a 6-action DDQN. We evaluate the framework using high-frequency 1Hz Coinbase trade data comprising over 6.8M trades on a realistic environment (10M TVL, 1% default width). Experimental results show that RAmmStein achieves a net ROI of 1.60%, the highest among all realistic (non-omniscient) strategies, while greedy strategies lose up to -8.4% to gas costs. Notably, the agent reduces rebalancing frequency by 85% compared to greedy rebalancing. RAmmStein-Width discovers extreme parsimony on its own, executing only 9 rebalances and $40 in gas, and degrades more slowly than all active strategies at elevated gas costs. Our results demonstrate that regime-aware laziness can significantly improve capital efficiency by preserving the returns that would otherwise be eroded by the operational costs.
Authors: Rajan, Ishaan Gupta
Abstract: Molecules are commonly represented as SMILES strings, which can be readily converted to fixed-size molecular fingerprints. These fingerprints serve as feature vectors to train ML/DL models for molecular property prediction tasks in the field of computational chemistry, drug discovery, biochemistry, and materials science. Recent research has demonstrated that SMILES can be used to construct molecular graphs where atoms are nodes ($V$) and bonds are edges ($E$). These graphs can subsequently be used to train geometric DL models like GNN. GNN learns the inherent structural relationships within a molecule rather than depending on fixed-size fingerprints. Although GNN are powerful aggregators, their efficacy on smaller datasets and inductive biases across different architectures is less studied. In our present study, we performed a systematic benchmarking of four different GNN architectures across a diverse domain of datasets (physical chemistry, biological, and analytical). Additionally, we have also implemented a hierarchical fusion (GNN+FP) framework for target prediction. We observed that the fusion framework consistently outperforms or matches the performance of standalone GNN (RMSE improvement > $7\%$) and baseline models. Further, we investigated the representational similarity using centered kernel alignment (CKA) between GNN and fingerprint embeddings and found that they occupy highly independent latent spaces (CKA $\le0.46$). The cross-architectural CKA score suggests a high convergence between isotopic models like GCN, GraphSAGE and GIN (CKA $\geq0.88$), with GAT learning moderately independent representation (CKA $0.55-0.80$).
Authors: Sicheng Dai, Hongwang Xiao, Shan Yu, Qiwei Ye
Abstract: Electroencephalogram (EEG) signals have become a popular medium for decoding visual information due to their cost-effectiveness and high temporal resolution. However, current approaches face significant challenges in bridging the modality gap between EEG and image data. These methods typically rely on complex adaptation processes involving multiple stages, making it hard to maintain consistency and manage compounding errors. Furthermore, the computational overhead imposed by large-scale diffusion models limit their practicality in real-world brain-computer interface (BCI) applications. In this work, we present AVDE, a lightweight and efficient framework for visual decoding from EEG signals. First, we leverage LaBraM, a pre-trained EEG model, and fine-tune it via contrastive learning to align EEG and image representations. Second, we adopt an autoregressive generative framework based on a "next-scale prediction" strategy: images are encoded into multi-scale token maps using a pre-trained VQ-VAE, and a transformer is trained to autoregressively predict finer-scale tokens starting from EEG embeddings as the coarsest representation. This design enables coherent generation while preserving a direct connection between the input EEG signals and the reconstructed images. Experiments on two datasets show that AVDE outperforms previous state-of-the-art methods in both image retrieval and reconstruction tasks, while using only 10% of the parameters. In addition, visualization of intermediate outputs shows that the generative process of AVDE reflects the hierarchical nature of human visual perception. These results highlight the potential of autoregressive models as efficient and interpretable tools for practical BCI applications.
Authors: Hung-Hsuan Chen
Abstract: Low-Rank Adaptation (LoRA) dominates parameter-efficient fine-tuning (PEFT). However, it faces a critical ``linear ceiling'' in complex reasoning tasks: simply increasing the rank yields diminishing returns due to intrinsic linear constraints. We introduce CeRA (Capacity-enhanced Rank Adaptation), a weight-level parallel adapter that injects SiLU gating and structural dropout to induce manifold expansion. On the SlimOrca benchmark, CeRA breaks this linear barrier: at rank 64 (PPL 3.89), it outperforms LoRA at rank 512 (PPL 3.90), demonstrating superior spectral efficiency. This advantage generalizes to mathematical reasoning, where CeRA achieves a perplexity of 1.97 on MathInstruct, significantly surpassing LoRA's saturation point of 2.07. Mechanism analysis via Singular Value Decomposition (SVD) confirms that CeRA activates the dormant tail of the singular value spectrum, effectively preventing the rank collapse observed in linear methods.
Authors: Peiyuan Zhang, Matthew Noto, Wenxuan Tan, Chengquan Jiang, Will Lin, Wei Zhou, Hao Zhang
Abstract: Achieving reliable 4-bit attention is a prerequisite for end-to-end FP4 computation on emerging FP4-capable GPUs, yet attention remains the main obstacle due to FP4's tiny dynamic range and attention's heavy-tailed activations. This paper presents the first systematic study of 4-bit quantization-aware training (QAT) for attention. We find that "drop-in" QAT, which naively combines an FP4 forward pass with a high-precision Flash Attention (FA)-style backward pass, leads to training instability. We identify two key principles for stable FP4 attention: (1) matching low-precision recomputation of attention scores in the backward pass, and (2) resolving implicit precision assumptions in FA's gradient calculation. Based on these insights, we propose Attn-QAT and implement fused Triton kernels for training as well as FP4 inference kernels. Across diffusion and language models, Attn-QAT recovers the quality drop from FP4 attention without explicit outlier-mitigation heuristics used in prior FP4 attention, and delivers up to a 1.5x speedup on an RTX 5090. Video demos can be found at https://drive.google.com/drive/folders/190F6xbBDUF2kGQYIcXBt3ehSYij5jlim?usp=sharing.
URLs: https://drive.google.com/drive/folders/190F6xbBDUF2kGQYIcXBt3ehSYij5jlim?usp=sharing.
Authors: Leo Muxing Wang, Pengkun Yang, Lili Su
Abstract: We study personalized multi-agent average reward TD learning, in which a collection of agents interacts with different environments and jointly learns their respective value functions. We focus on the setting where there exists a shared linear representation, and the agents' optimal weights collectively lie in an unknown linear subspace. Inspired by the recent success of personalized federated learning (PFL), we study the convergence of cooperative single-timescale TD learning in which agents iteratively estimate the common subspace and local heads. We showed that this decomposition can filter out conflicting signals, effectively mitigating the negative impacts of ``misaligned'' signals, and achieving linear speedup. The main technical challenges lie in the heterogeneity, the Markovian sampling, and their intricate interplay in shaping error evolutions. Specifically, not only are the error dynamics of multiple variables closely interconnected, but there is also no direct contraction for the principal angle distance between the optimal subspace and the estimated subspace. We hope our analytical techniques can be useful to inspire research on deeper exploration into leveraging common structures. Experiments are provided to show the benefits of learning via a shared structure to the more general control problem.
Authors: Fabian Kabus, Maren Hackenberg, Julia Hindel, Thibault Cholvin, Antje Kilias, Thomas Brox, Abhinav Valada, Marlene Bartos, Harald Binder
Abstract: While artificial neural networks excel in unsupervised learning of non-sparse structure, classical statistical regression techniques offer better interpretability, in particular when sparseness is enforced by $\ell_1$ regularization, enabling identification of which factors drive observed dynamics. We investigate how these two types of approaches can be optimally combined, exemplarily considering two-photon calcium imaging data where sparse autoregressive dynamics are to be extracted. We propose embedding a vector autoregressive (VAR) model as an interpretable regression technique into a convolutional autoencoder, which provides dimension reduction for tractable temporal modeling. A skip connection separately addresses non-sparse static spatial information, selectively channeling sparse structure into the $\ell_1$-regularized VAR. $\ell_1$-estimation of regression parameters is enabled by differentiating through the piecewise linear solution path. This is contrasted with approaches where the autoencoder does not adapt to the VAR model. Having an embedded statistical model also enables a testing approach for comparing temporal sequences from the same observational unit. Additionally, contribution maps visualize which spatial regions drive the learned dynamics.
Authors: Zhenquan Yao, Zitong Huang, Yihan Zeng, Jianhua Han, Hang Xu, Chun-Mei Feng, Jianwei Ma, Wangmeng Zuo
Abstract: Graphical User Interface (GUI) Agents, benefiting from recent advances in multimodal large language models (MLLM), have achieved significant development. However, due to the frequent updates of GUI applications, adapting to new tasks without forgetting old tasks in GUI continual learning remains an open problem. In this work, we reveal that while Supervised Fine-Tuning (SFT) facilitates fast adaptation, it often triggers knowledge overwriting, whereas Reinforcement Learning (RL) demonstrates an inherent resilience that shields prior interaction logic from erasure. Based on this insight, we propose a \textbf{C}ontinual \textbf{G}UI \textbf{L}earning (CGL) framework that dynamically balances adaptation efficiency and skill retention by enhancing the synergy between SFT and RL. Specifically, we introduce an SFT proportion adjustment mechanism guided by policy entropy to dynamically control the weight allocation between the SFT and RL training phases. To resolve explicit gradient interference, we further develop a specialized gradient surgery strategy. By projecting exploratory SFT gradients onto GRPO-based anchor gradients, our method explicitly clips the components of SFT gradients that conflict with GRPO. On top of that, we establish an AndroidControl-CL benchmark, which divides GUI applications into distinct task groups to effectively simulate and evaluate the performance of continual GUI learning. Experimental results demonstrate the effectiveness of our proposed CGL framework across continual learning scenarios. The benchmark, code, and model will be made publicly available.
Authors: Ruinan Jin, Yingbin Liang, Shaofeng Zou
Abstract: Despite Adam demonstrating faster empirical convergence than SGD in many applications, much of the existing theory yields guarantees essentially comparable to those of SGD, leaving the empirical performance gap insufficiently explained. In this paper, we uncover a key second-moment normalization in Adam and develop a stopping-time/martingale analysis that provably distinguishes Adam from SGD under the classical bounded variance model (a second moment assumption). In particular, we establish the first theoretical separation between the high-probability convergence behaviors of the two methods: Adam achieves a $\delta^{-1/2}$ dependence on the confidence parameter $\delta$, whereas corresponding high-probability guarantee for SGD necessarily incurs at least a $\delta^{-1}$ dependence.
Authors: Joshua Steier
Abstract: What determines whether a molecular property prediction model organizes its representations so that geometric and compositional information can be cleanly separated? We introduce Compositional Probe Decomposition (CPD), which linearly projects out composition signal and measures how much geometric information remains accessible to a Ridge probe. We validate CPD with four independent checks, including a structural isomer benchmark where compositional projections score at chance while geometric residuals reach 94.6\% pairwise classification accuracy. Across ten models from five architectural families on QM9, we find a \emph{linear accessibility gradient}: models differ by $6.6\times$ in geometric information accessible after composition removal ($R^2_{\mathrm{geom}}$ from 0.081 to 0.533 for HOMO-LUMO gap). Three factors explain this gradient. Task alignment dominates: models trained on HOMO-LUMO gap ($R^2_{\mathrm{geom}}$ 0.44--0.53) outscore energy-trained models by $\sim$0.25 $R^2$ regardless of architecture. Within-architecture ablations on two independent architectures confirm this: PaiNN drops from 0.53 to 0.31 when retrained on energy, and MACE drops from 0.44 to 0.08. Data diversity partially compensates for misaligned objectives, with MACE pretrained on MPTraj (0.36) outperforming QM9-only energy models. Inside MACE's representations, information routes by symmetry type: $L{=}1$ (vector) channels preferentially encode dipole moment ($R^2 = 0.59$ vs.\ 0.38 in $L{=}0$), while $L{=}0$ (scalar) channels encode HOMO-LUMO gap ($R^2 = 0.76$ vs.\ 0.34 in $L{=}1$). This pattern is absent in ViSNet. We also show that nonlinear probes produce misleading results on residualized representations, recovering $R^2 = 0.68$--$0.95$ on a purely compositional target, and recommend linear probes for this setting.
Authors: Peter Balogh
Abstract: We investigate when transformer MLP nonlinearity is actually necessary. A gate with $d+1$ parameters decides when to replace the full MLP with a linear surrogate. Through systematic investigation across six models (162M-2.8B parameters), two architectures, and three corpora, we establish that nonlinearity need cannot be predicted from token identity: cross-corpus correlation is zero ($r < 0.05$). The routing decision is fully contextual. Despite weak per-instance predictability, the gate exploits a heavily skewed distribution where most MLP computations are near-linear, achieving 25-56% linear routing at <1% perplexity cost in GPT-2. In GPT-2 Large, 11 of 36 layers beat baseline with gating and no layer exceeds 3.7% all-linear cost. This success is architecture-dependent: Pythia models show higher costs, though Pythia-2.8B's full 32-layer sweep reveals one layer that narrowly beats baseline. As a proof of concept, we progressively replace middle-layer MLPs with frozen linear matrices: 5 of 24 layers linearize at zero cost. With a full training budget, 4 linearized layers yield a 10.2% perplexity improvement -- and a two-phase gated approach pushes this to 17.3%, beating a vanilla fine-tuning control and confirming that the nonlinear MLPs at these layers were actively harmful.
Authors: Zeyneb N. Kaya, Nick Rui
Abstract: As strong general reasoners, large language models (LLMs) encounter diverse domains and tasks, where the ability to adapt and self-improve at test time is valuable. We introduce MASS, a meta-learning framework that enables LLMs to self-adapt by generating problem-specific synthetic training data and performing targeted self-updates optimized for downstream performance at inference time. We train this behavior end-to-end via bilevel optimization: an inner loop adapts on self-generated examples while an outer loop meta-learns data-attribution signals and rewards post-update task performance. The synthetic data is optimized with scalable meta-gradients, backpropagating the downstream loss through the inner updates to reward useful generations. Experiments on mathematical reasoning show that MASS learns to synthesize per-instance curricula that yield effective, data-efficient test-time adaptation.
Authors: Pedram Agand
Abstract: Standard Retrieval-Augmented Generation (RAG) architectures fail in high-stakes financial domains due to two fundamental limitations: the inherent arithmetic incompetence of Large Language Models (LLMs) and the distributional semantic conflation of dense vector retrieval (e.g., mapping "Net Income" to "Net Sales" due to contextual proximity). In deterministic domains, a 99% accuracy rate yields 0% operational trust. To achieve zero-hallucination financial reasoning, we introduce the Verifiable Numerical Reasoning Agent (VeNRA). VeNRA shifts the RAG paradigm from retrieving probabilistic text to retrieving deterministic variables via a strictly typed Universal Fact Ledger (UFL). We mathematically bound this ledger using a novel Double-Lock Grounding algorithm. Coupled with deterministic Python execution, this neuro-symbolic routing compresses systemic hallucination rates to a near-zero 1.2%. Recognising that upstream parsing anomalies inevitably occur, we introduce the VeNRA Sentinel: a 3-billion parameter SLM trained to forensically audit candidate using a single-token inference budget with optional post-hoc reasoning. To train the Sentinel, we steer away from traditional hallucination datasets in favour of Adversarial Simulation, programmatically sabotaging financial records to simulate Ecological Errors. The compact Sentinel consequently outperforms 70B+ frontier models in error detection. Through Loss Dilution phenomenon in Reverse-CoT training, we present a novel Micro-Chunking loss algorithm to stabilise gradients under extreme verdict penalisation, yielding a 28x latency speedup without sacrificing forensic rigor.
Authors: Christos Fragkathoulas, Eleni Psaroudaki, Themis Palpanas, Evaggelia Pitoura
Abstract: Time-series clustering is a fundamental tool for pattern discovery, yet existing explainability methods, primarily based on feature attribution or metadata, fail to identify the transitions that move an instance across cluster boundaries. While Counterfactual Explanations (CEs) identify the minimal temporal perturbations required to alter the prediction of a model, they have been mostly confined to supervised settings. This paper introduces GALACTIC, the first unified framework to bridge local and global counterfactual explainability for unsupervised time-series clustering. At instance level (local), GALACTIC generates perturbations via a cluster-aware optimization objective that respects the target and underlying cluster assignments. At cluster level (global), to mitigate cognitive load and enhance interpretability, we formulate a representative CE selection problem. We propose a Minimum Description Length (MDL) objective to extract a non-redundant summary of global explanations that characterize the transitions between clusters. We prove that our MDL objective is supermodular, which allows the corresponding MDL reduction to be framed as a monotone submodular set function. This enables an efficient greedy selection algorithm with provable $(1-1/e)$ approximation guarantees. Extensive experimental evaluation on the UCR Archive demonstrates that GALACTIC produces significantly sparser local CEs and more concise global summaries than state-of-the-art baselines adapted for our problem, offering the first unified approach for interpreting clustered time-series through counterfactuals.
Authors: Hejian Sang, Yuanda Xu, Zhengze Zhou, Ran He, Zhipeng Wang, Jiachen Sun
Abstract: Reasoning models think out loud, but much of what they say is noise. We introduce OPSDC (On-Policy Self-Distillation for Reasoning Compression), a method that teaches models to reason more concisely by distilling their own concise behavior back into themselves. The entire approach reduces to one idea: condition the same model on a "be concise" instruction to obtain teacher logits, and minimize per-token reverse KL on the student's own rollouts. No ground-truth answers, no token budgets, no difficulty estimators. Just self-distillation. Yet this simplicity belies surprising sophistication: OPSDC automatically compresses easy problems aggressively while preserving the deliberation needed for hard ones. On Qwen3-8B and Qwen3-14B, we achieve 57-59% token reduction on MATH-500 while improving accuracy by 9-16 points absolute. On AIME 2024, the 14B model gains 10 points with 41% compression. The secret? Much of what reasoning models produce is not just redundant-it is actively harmful, compounding errors with every unnecessary token. Code is available at https://github.com/HJSang/OPSD_Reasoning_Compression.
Authors: Xupeng Chen, Shuchen Meng
Abstract: Generative AI compresses within-task skill differences while shifting economic value toward concentrated complementary assets, creating an apparent paradox: the technology that equalizes individual performance may widen aggregate inequality. We formalize this tension in a task-based model with endogenous education, employer screening, and heterogeneous firms. The model yields two regimes whose boundary depends on AI's technology structure (proprietary vs. commodity) and labor market institutions (rent-sharing elasticity, asset concentration). A scenario analysis via Method of Simulated Moments, matching six empirical targets, disciplines the model's quantitative magnitudes; a sensitivity decomposition reveals that the five non-$\Delta$Gini moments identify mechanism rates but not the aggregate sign, which at the calibrated parameters is pinned by $m_6$ and $\xi$, while AI's technology structure ($\eta_1$ vs. $\eta_0$) independently crosses the boundary. The contribution is the mechanism -- not a verdict on the sign. Occupation-level regressions using BLS OEWS data (2019--2023) illustrate why such data cannot test the model's task-level predictions. The predictions are testable with within-occupation, within-task panel data that do not yet exist at scale.
Authors: Levy Chaves, Chao Zhou, Rebekka Burkholz, Eduardo Valle, Sandra Avila
Abstract: Model merging integrates multiple task-specific models into a single consolidated one. Recent research has made progress in improving merging performance for in-distribution or multi-task scenarios, but domain generalization in model merging remains underexplored. We investigate how merging models fine-tuned on distinct domains affects generalization to unseen domains. Through an analysis of parameter competition in the task matrix using singular value decomposition, we show that merging models trained under different distribution shifts induces stronger conflicts between their subspaces compared to traditional multi-task settings. To mitigate this issue, we propose SCORE (Subspace COnflict-Resolving mErging), a method designed to alleviate such singular subspace conflicts. SCORE finds a shared orthogonal basis by computing the principal components of the concatenated leading singular vectors of all models. It then projects each task matrix into the shared basis, pruning off-diagonal components to remove conflicting singular directions. SCORE consistently outperforms, on average, existing model merging approaches in domain generalization settings across a variety of architectures and model scales, demonstrating its effectiveness and scalability.
Authors: Damir Filipovi\'c, Puneet Pasricha
Abstract: We introduce an ensemble learning method based on Gaussian Process Regression (GPR) for predicting conditional expected stock returns given stock-level and macro-economic information. Our ensemble learning approach significantly reduces the computational complexity inherent in GPR inference and lends itself to general online learning tasks. We conduct an empirical analysis on a large cross-section of US stocks from 1962 to 2016. We find that our method dominates existing machine learning models statistically and economically in terms of out-of-sample $R$-squared and Sharpe ratio of prediction-sorted portfolios. Exploiting the Bayesian nature of GPR, we introduce the mean-variance optimal portfolio with respect to the prediction uncertainty distribution of the expected stock returns. It appeals to an uncertainty averse investor and significantly dominates the equal- and value-weighted prediction-sorted portfolios, which outperform the S&P 500.
Authors: Long Cao, Liwei Ge, Daochi Zhang, Xiang Li, Yao Wang, Rui-Xue Xu, YiJing Yan, Xiao Zheng
Abstract: Reducing computational scaling for simulating non-Markovian dissipative dynamics using artificial neural networks is both a major focus and formidable challenge in open quantum systems. To enable neural quantum states (NQSs), we encode environmental memory in dissipatons (quasiparticles with characteristic lifetimes), yielding the dissipaton-embedded quantum master equation (DQME). The resulting NQS-DQME framework achieves compact representation of many-body correlations and non-Markovian memory. Benchmarking against numerically exact hierarchical equations of motion confirms NQS-DQME maintains comparable accuracy while enhancing scalability and interpretability. This methodology opens new paths to explore non-Markovian open quantum dynamics in previously intractable systems.
Authors: Xiaoyu Wu, Jiaru Zhang, Yang Hua, Bohan Lyu, Hao Wang, Tao Song, Haibing Guan
Abstract: Few-shot fine-tuning of Diffusion Models (DMs) is a key advancement, significantly reducing training costs and enabling personalized AI applications. However, we explore the training dynamics of DMs and observe an unanticipated phenomenon: during the training process, image fidelity initially improves, then unexpectedly deteriorates with the emergence of noisy patterns, only to recover later with severe overfitting. We term the stage with generated noisy patterns as corruption stage. To understand this corruption stage, we begin by theoretically modeling the one-shot fine-tuning scenario, and then extend this modeling to more general cases. Through this modeling, we identify the primary cause of this corruption stage: a narrowed learning distribution inherent in the nature of few-shot fine-tuning. To tackle this, we apply Bayesian Neural Networks (BNNs) on DMs with variational inference to implicitly broaden the learned distribution, and present that the learning target of the BNNs can be naturally regarded as an expectation of the diffusion loss and a further regularization with the pretrained DMs. This approach is highly compatible with current few-shot fine-tuning methods in DMs and does not introduce any extra inference costs. Experimental results demonstrate that our method significantly mitigates corruption, and improves the fidelity, quality and diversity of the generated images in both object-driven and subject-driven generation tasks. Code is available at https://github.com/Nicholas0228/BNN-Finetuning-DMs.
Authors: Ruohan Zhan, Shichao Han, Yuchen Hu, Zhenling Jiang
Abstract: Online user-generated content platforms allocate billions of dollars of promotional traffic through algorithms in two-sided marketplaces. To evaluate updates to these algorithms, platforms frequently rely on creator-side randomized experiments. However, because treated and control creators compete for exposure, such experiments suffer from algorithmic interference: exposure outcomes depend on competitors' treatment status. We show that commonly used difference-in-means estimators can therefore be severely biased and may even recommend deploying inferior algorithms. To address this challenge, we develop a structured semiparametric framework that explicitly models the competitive allocation mechanism underlying exposure. Our approach combines an algorithm choice model that characterizes how exposure is allocated across competing content with a viewer response model that captures engagement conditional on exposure. We construct a debiased estimator grounded in the double machine learning framework to recover the global treatment effect of platform-wide rollout. Methodologically, we extend DML asymptotic theory to accommodate correlated samples arising from overlapping consideration sets. Using Monte Carlo simulations and a large-scale field experiment on a major short-video platform, we show that our estimator closely matches an interference-free benchmark obtained from a costly double-sided experimental design. In contrast, standard estimators exhibit substantial bias and, in some cases, even reverse the sign of the effect.
Authors: Lang Zeng, Weijing Tang, Zhao Ren, Ying Ding
Abstract: The stochastic gradient descent (SGD) algorithm has been widely used to optimize deep Cox neural network (Cox-NN) by updating model parameters using mini-batches of data. We show that SGD aims to optimize the average of mini-batch partial-likelihood, which is different from the standard partial-likelihood. This distinction requires developing new statistical properties for the global optimizer, namely, the mini-batch maximum partial-likelihood estimator (mb-MPLE). We establish that mb-MPLE for Cox-NN is consistent and achieves the optimal minimax convergence rate up to a polylogarithmic factor. For Cox regression with linear covariate effects, we further show that mb-MPLE is $\sqrt{n}$-consistent and asymptotically normal with asymptotic variance approaching the information lower bound as batch size increases, which is confirmed by simulation studies. Additionally, we offer practical guidance on using SGD, supported by theoretical analysis and numerical evidence. For Cox-NN, we demonstrate that the ratio of the learning rate to the batch size is critical in SGD dynamics, offering insight into hyperparameter tuning. For Cox regression, we characterize the iterative convergence of SGD, ensuring that the global optimizer, mb-MPLE, can be approximated with sufficiently many iterations. Finally, we demonstrate the effectiveness of mb-MPLE in a large-scale real-world application where the standard MPLE is intractable.
Authors: Zhanglu Yan, Zhenyu Bai, Weng-Fai Wong
Abstract: Spiking Neural Networks (SNNs) promise higher energy efficiency over conventional Quantized Artificial Neural Networks (QNNs) due to their event-driven, spike-based computation. However, prevailing energy evaluations often oversimplify, focusing on computational aspects while neglecting critical overheads like comprehensive data movement and memory access. Such simplifications can lead to misleading conclusions regarding the true energy benefits of SNNs. This paper presents a rigorous re-evaluation. We establish a fair baseline by mapping rate-encoded SNNs with $T$ timesteps to functionally equivalent QNNs with $\lceil \log_2(T+1) \rceil$ bits. This ensures both models have comparable representational capacities, as well has similar hardware requirement, enabling meaningful energy comparisons. We introduce a detailed analytical energy model encompassing core computation and data movement. Using this model, we systematically explore a wide parameter space, including intrinsic network characteristics ($T$, spike rate $\SR$, QNN sparsity $\gamma$, model size $N$, weight bit-level) and hardware characteristics (memory system and network-on-chip). Our analysis identifies specific operational regimes where SNNs genuinely offer superior energy efficiency. For example, under typical neuromorphic hardware conditions, SNNs with moderate time windows ($T \in [5,10]$) require an average spike rate ($\SR$) below 6.4\% to outperform equivalent QNNs. Furthermore, to illustrate the real-world implications of our findings, we analyze the operational lifetime of a typical smartwatch, showing that an optimized SNN can nearly double its battery life compared to a QNN. These insights guide the design of turely energy-efficient neural network solutions.
Authors: Maximilian St\"olzle, Cosimo Della Santina
Abstract: Even though a variety of methods have been proposed in the literature, efficient and effective latent-space control (i.e., control in a learned low-dimensional space) of physical systems remains an open challenge. We argue that a promising avenue is to leverage powerful and well-understood closed-form strategies from control theory literature in combination with learned dynamics, such as potential-energy shaping. We identify three fundamental shortcomings in existing latent-space models that have so far prevented this powerful combination: (i) they lack the mathematical structure of a physical system, (ii) they do not inherently conserve the stability properties of the real systems, (iii) these methods do not have an invertible mapping between input and latent-space forcing. This work proposes a novel Coupled Oscillator Network (CON) model that simultaneously tackles all these issues. More specifically, (i) we show analytically that CON is a Lagrangian system - i.e., it possesses well-defined potential and kinetic energy terms. Then, (ii) we provide formal proof of global Input-to-State stability using Lyapunov arguments. Moving to the experimental side, we demonstrate that CON reaches SoA performance when learning complex nonlinear dynamics of mechanical systems directly from images. An additional methodological innovation contributing to achieving this third goal is an approximated closed-form solution for efficient integration of network dynamics, which eases efficient training. We tackle (iii) by approximating the forcing-to-input mapping with a decoder that is trained to reconstruct the input based on the encoded latent space force. Finally, we show how these properties enable latent-space control. We use an integral-saturated PID with potential force compensation and demonstrate high-quality performance on a soft robot using raw pixels as the only feedback information.
Authors: Haoyi Niu, Qimao Chen, Tenglong Liu, Jianxiong Li, Guyue Zhou, Yi Zhang, Jianming Hu, Xianyuan Zhan
Abstract: Reusing pre-collected data from different domains is an appealing solution for decision-making tasks, especially when data in the target domain are limited. Existing cross-domain policy transfer methods mostly aim at learning domain correspondences or corrections to facilitate policy learning, such as learning task/domain-specific discriminators, representations, or policies. This design philosophy often results in heavy model architectures or task/domain-specific modeling, lacking flexibility. This reality makes us wonder: can we directly bridge the domain gaps universally at the data level, instead of relying on complex downstream cross-domain policy transfer procedures? In this study, we propose the Cross-Domain Trajectory EDiting (xTED) framework that employs a specially designed diffusion model for cross-domain trajectory adaptation. Our proposed model architecture effectively captures the intricate dependencies among states, actions, and rewards, as well as the dynamics patterns within target data. Edited by adding noises and denoising with the pre-trained diffusion model, source domain trajectories can be transformed to align with target domain properties while preserving original semantic information. This process effectively corrects underlying domain gaps, enhancing state realism and dynamics reliability in source data, and allowing flexible integration with various single-domain and cross-domain downstream policy learning methods. Despite its simplicity, xTED demonstrates superior performance in extensive simulation and real-robot experiments.
Authors: Xin Chen, Yifan Hu, Minda Zhao
Abstract: Policy gradient methods are widely used in reinforcement learning. Yet, the nonconvexity of policy optimization poses significant challenges in understanding the global convergence of policy gradient methods. For a class of finite-horizon Markov Decision Processes (MDPs) with general state and action spaces, we identify a set of structural properties to establish a benign nonconvex landscape, the Polyak-{\L}ojasiewicz-Kurdyka (P{\L}K) condition of the policy optimization. Leveraging the P{\L}K condition, policy gradient methods converge to the globally optimal policy with a non-asymptotic rate despite nonconvexity. Our results apply to various control and operations models, including entropy-regularized tabular MDPs, Linear Quadratic Regulator problems, and both stochastic inventory models and stochastic cash balance problems with strongly convex costs. In these models, stochastic policy gradient methods obtain an $\epsilon$-optimal policy using a sample size of $\tilde{\mathcal{O}}(\epsilon^{-1})$ and polynomial in terms of the planning horizon. To the best of our knowledge, we provide the first sample-complexity guarantees for multi-period inventory systems with Markov-modulated demand and for stochastic cash balance problems. We complement the theory with numerical experiments showing that policy gradient methods outperform several benchmark algorithms from the literature across these operations models.
Authors: Yuqi Gu, Zhongyuan Lyu, Kaizheng Wang
Abstract: We propose a general transfer learning framework for clustering given a main dataset and an auxiliary one about the same subjects. The two datasets may reflect similar but different latent grouping structures of the subjects. We propose an adaptive transfer clustering (ATC) algorithm that automatically leverages the commonality in the presence of unknown discrepancy, by optimizing an estimated bias-variance decomposition. It applies to a broad class of statistical models including Gaussian mixture models, stochastic block models, and latent class models. A theoretical analysis proves the optimality of ATC under the Gaussian mixture model and explicitly quantifies the benefit of transfer. Extensive simulations and real data experiments confirm our method's effectiveness in various scenarios.
Authors: Zuzanna Buchnajzer, Kacper Dobek, Stanis{\l}aw Hapke, Daniel Jankowski, Krzysztof Krawiec
Abstract: Deep learning architectures based on convolutional neural networks tend to rely on continuous, smooth features. While this characteristics provides significant robustness and proves useful in many real-world tasks, it is strikingly incompatible with the physical characteristic of the world, which, at the scale in which humans operate, comprises crisp objects, typically representing well-defined categories. This study proposes a class of neurosymbolic systems that learn by reconstructing images in terms of visual primitives and are thus forced to form high-level, structural explanations of them. When applied to the task of diagnosing abnormalities in histological imaging, the method proved superior to a conventional deep learning architecture in terms of classification accuracy, while being more transparent.
Authors: Yucheng Xing, Xiaodong Liu, Xin Wang
Abstract: Diffusion models typically generate data through a fixed denoising trajectory that is shared across all samples. However, generation targets can differ in complexity, suggesting that a single pre-defined diffusion process may not be optimal for every input. In this work, we investigate input-adaptive generative dynamics for diffusion models, where the generation process itself adapts to the conditions of each sample. Instead of relying on a fixed diffusion trajectory, the proposed framework allows the generative dynamics to adjust across inputs according to their generation requirements. To enable this behavior, we train the diffusion backbone under varying horizons and noise schedules, so that it can operate consistently under different input-adaptive trajectories. Experiments on conditional image generation show that diffusion trajectories can vary across inputs while maintaining generation quality and reducing the average number of sampling steps. These results provide a proof of the concept that diffusion processes can benefit from input-adaptive generative dynamics rather than relying on a single fixed trajectory.
Authors: Sergey Sedov, Sumanth Bharadwaj Hachalli Karanam, Venu Gopal Kadamba
Abstract: Prompt-Tuning is an efficient method for adapting pre-trained language models to new tasks with minimal computational overhead by modifying prompt embeddings. In this work, we investigate how crucial the phenomenon of embedding collapse, frequently observed in Prompt-Tuning, is for the final performance of the model. To address this question, we designed embedding priors and compared them with posteriors of the converged Soft and Deep Prompt-Tuning methods. Our findings suggest that priors strongly affect the position of the tuned embeddings, and models can effectively work with embeddings from different parts of activation spaces, including completely new regions. As the final Prompt-Tuning capabilities are limited, we hypothesize that controllable Prompt-Tuning posteriors may serve as a good starting point for tasks such as chain-of-thought (COT) distillation. Our experiments also show that generated trajectories are not localized in the activation space of the models. However, there are distinct clusters of activations for distant tasks (e.g., NLP and arithmetic), while activations between NLP tasks (e.g., Question-Answering and MLM) lie in the same cluster. These observations raise questions about the importance of a single activation cluster for the generalization abilities of large language models.
Authors: Ashay Athalye, Nishanth Kumar, Tom Silver, Yichao Liang, Jiuguang Wang, Tom\'as Lozano-P\'erez, Leslie Pack Kaelbling
Abstract: Our aim is to learn to solve long-horizon decision-making problems in complex robotics domains given low-level skills and a handful of short-horizon demonstrations containing sequences of images. To this end, we focus on learning abstract symbolic world models that facilitate zero-shot generalization to novel goals via planning. A critical component of such models is the set of symbolic predicates that define properties of and relationships between objects. In this work, we leverage pretrained vision-language models (VLMs) to propose a large set of visual predicates potentially relevant for decision-making, and to evaluate those predicates directly from camera images. At training time, we pass the proposed predicates and demonstrations into an optimization-based model-learning algorithm to obtain an abstract symbolic world model that is defined in terms of a compact subset of the proposed predicates. At test time, given a novel goal in a novel setting, we use the VLM to construct a symbolic description of the current world state, and then use a search-based planning algorithm to find a sequence of low-level skills that achieves the goal. We demonstrate empirically across experiments in both simulation and the real world that our method can generalize aggressively, applying its learned world model to solve problems with a wide variety of object types, arrangements, numbers of objects, and visual backgrounds, as well as novel goals and much longer horizons than those seen at training time.
Authors: Han Qi, Haocheng Yin, Aris Zhu, Yilun Du, Heng Yang
Abstract: We present generative predictive control (GPC), a learning control framework that (i) clones a generative diffusion-based policy from expert demonstrations, (ii) trains a predictive action-conditioned world model from both expert demonstrations and random explorations, and (iii) synthesizes an online planner that ranks and optimizes the action proposals from (i) by looking ahead into the future using the world model from (ii). Across a variety of robotic manipulation tasks, we demonstrate that GPC consistently outperforms behavior cloning in both state-based and vision-based settings, in simulation and in the real world.
Authors: Parsa Moradi, Mohammad Ali Maddah-Ali
Abstract: Coded computing has demonstrated promising results in addressing straggler resiliency in distributed computing systems. However, most coded computing schemes are designed for exact computation, requiring the number of responding servers to exceed a certain recovery threshold. Additionally, these schemes are tailored for highly structured functions. Recently, new coded computing schemes for general computing functions, where exact computation is replaced with approximate computation, have emerged. In these schemes, the availability of additional results corresponds to more accurate estimation of computational tasks. This flexibility introduces new questions that need to be addressed. This paper addresses the practically important scenario in the context of general coded computing, where each server may become a straggler with a probability $p$, independently from others. We theoretically analyze the approximation error of two existing general coded computing schemes: Berrut Approximate Coded Computing (BACC) and Learning Theoretic Coded Computing (LeTCC). Under the probabilistic straggler configuration, we demonstrate that the average approximation error for BACC and LeTCC converge to zero with the rate of at least $\mathcal{O}(\log^3_{\frac{1}{p}}(N)\cdot{N^{-3}})$ and $\mathcal{O}(\log^4_{\frac{1}{p}}(N)\cdot{N^{-2}})$, respectively. This is perhaps surprising, as earlier results does not indicate a convergence when the number of stragglers scales with the total number of servers $N$. However, in this case, despite the average number of stragglers being $Np$, the independence of servers in becoming stragglers allows the approximation error to converge to zero. These theoretical results are validated through experiments on various computing functions, including deep neural networks.
Authors: Mohammed Kharma, Soohyeon Choi, Mohammed AlKhanafseh, David Mohaisen
Abstract: Artificial Intelligence (AI)-driven code generation tools are increasingly used throughout the software development lifecycle to accelerate coding tasks. However, the security of AI-generated code using Large Language Models (LLMs) remains underexplored, with studies revealing various risks and weaknesses. This paper analyzes the security of code generated by LLMs across different programming languages. We introduce a dataset of 200 tasks grouped into six categories to evaluate the performance of LLMs in generating secure and maintainable code. Our research shows that while LLMs can automate code creation, their security effectiveness varies by language. Many models fail to utilize modern security features in recent compiler and toolkit updates, such as Java 17. Moreover, outdated methods are still commonly used, particularly in C++. This highlights the need for advancing LLMs to enhance security and quality while incorporating emerging best practices in programming languages.
Authors: Takuya Matsuyama, Shinji Nishimoto, Yu Takagi
Abstract: Understanding the property of neural populations (or voxels) in the human brain can advance our comprehension of human perceptual and cognitive processing capabilities and contribute to developing brain-inspired computer models. Recent encoding models using deep neural networks (DNNs) have successfully predicted voxel-wise activity. However, interpreting the properties that explain voxel responses remains challenging because of the black-box nature of DNNs. As a solution, we propose LLM-assisted Visual Cortex Captioning (LaVCa), a data-driven approach that uses large language models (LLMs) to generate natural-language captions for images to which voxels are selective. By applying LaVCa for image-evoked brain activity, we demonstrate that LaVCa generates captions that describe voxel selectivity more accurately than the previously proposed method. Furthermore, the captions generated by LaVCa quantitatively capture more detailed properties than the existing method at both the inter-voxel and intra-voxel levels. Furthermore, a more detailed analysis of the voxel-specific properties generated by LaVCa reveals fine-grained functional differentiation within regions of interest (ROIs) in the visual cortex and voxels that simultaneously represent multiple distinct concepts. These findings offer profound insights into human visual representations by assigning detailed captions throughout the visual cortex while highlighting the potential of LLM-based methods in understanding brain representations.
Authors: Chengyin Xu, Kaiyuan Chen, Xiao Li, Ke Shen, Chenggang Li
Abstract: The escalating scale and cost of Large Language Models (LLMs) training necessitate accurate pre-training prediction of downstream task performance for comprehensive understanding of scaling properties. This is challenged by: 1) the emergence phenomenon, where unpredictable capabilities appearing suddenly at critical model scales; and 2) uneven task difficulty and inconsistent performance scaling patterns, leading to high metric variability. Current prediction methods lack accuracy and reliability. We propose a Clustering-On-Difficulty (COD) framework for downstream performance prediction. The COD framework clusters tasks by their difficulty scaling features, thereby constructing a more stable and predictable task subset that exhibits well-behaved scaling characteristics with the increase of compute budget. We adopt a performance scaling law to predict cluster-wise performance with theoretical support. Predictable subset performance acts as an intermediate predictor for the full evaluation set. We further derive a mapping function to accurately extrapolate the performance of the subset to the full set. Applied to an LLM with 70B parameters, COD achieved a 1.55\% average prediction error across eight key LLM benchmarks, thus providing actionable insights for scaling properties and training monitoring during LLM pre-training.
Authors: Yiyang Ling, Karan Owalekar, Oluwatobiloba Adesanya, Erdem B{\i}y{\i}k, Daniel Seita
Abstract: Motion planning involves determining a sequence of robot configurations to reach a desired pose, subject to movement and safety constraints. Traditional motion planning finds collision-free paths, but this is overly restrictive in clutter, where it may not be possible for a robot to accomplish a task without contact. In addition, contacts range from relatively benign (e.g. brushing a soft pillow) to more dangerous (e.g. toppling a glass vase), making it difficult to characterize which may be acceptable. In this paper, we propose IMPACT, a novel motion planning framework that uses Vision-Language Models (VLMs) to infer environment semantics, identifying which parts of the environment can best tolerate contact based on object properties and locations. Our approach generates an anisotropic cost map that encodes directional push safety. We pair this map with a contact-aware A* planner to find stable contact-rich paths. We perform experiments using 20 simulation and 10 real-world scenes and assess using task success rate, object displacements, and feedback from human evaluators. Our results over 3200 simulation and 200 real-world trials suggest that IMPACT enables efficient contact-rich motion planning in cluttered settings while outperforming alternative methods and ablations. Our project website is available at https://impact-planning.github.io/.
Authors: Pooya Razavi, Sonya Powers
Abstract: Estimating item difficulty through field-testing is often resource-intensive and time-consuming. As such, there is strong motivation to develop methods that can predict item difficulty at scale using only the item content. Large Language Models (LLMs) represent a new frontier for this goal. The present research examines the feasibility of using an LLM to predict item difficulty for K-5 mathematics and reading assessment items (N = 5170). Two estimation approaches were implemented: (a) a direct estimation method that prompted the LLM to assign a single difficulty rating to each item, and (b) a feature-based strategy where the LLM extracted multiple cognitive and linguistic features, which were then used in ensemble tree-based models (random forests and gradient boosting) to predict difficulty. Overall, direct LLM estimates showed moderate to strong correlations with true item difficulties. However, their accuracy varied by grade level, often performing worse for early grades. In contrast, the feature-based method yielded stronger predictive accuracy, with correlations as high as r = 0.87 and lower error estimates compared to both direct LLM predictions and baseline regressors. These findings highlight the promise of LLMs in streamlining item development and reducing reliance on extensive field testing and underscore the importance of structured feature extraction. We provide a seven-step workflow for testing professionals who would want to implement a similar item difficulty estimation approach with their item pool.
Authors: Mohamed Amine Ferrag, Norbert Tihanyi, Merouane Debbah
Abstract: Large language models and autonomous AI agents have evolved rapidly, resulting in a diverse array of evaluation benchmarks, frameworks, and collaboration protocols. Driven by the growing need for standardized evaluation and integration, we systematically consolidate these fragmented efforts into a unified framework. However, the landscape remains fragmented and lacks a unified taxonomy or comprehensive survey. Therefore, we present a side-by-side comparison of benchmarks developed between 2019 and 2025 that evaluate these models and agents across multiple domains. In addition, we propose a taxonomy of approximately 60 benchmarks that cover general and academic knowledge reasoning, mathematical problem-solving, code generation and software engineering, factual grounding and retrieval, domain-specific evaluations, multimodal and embodied tasks, task orchestration, and interactive assessments. Furthermore, we review AI-agent frameworks introduced between 2023 and 2025 that integrate large language models with modular toolkits to enable autonomous decision-making and multi-step reasoning. Moreover, we present real-world applications of autonomous AI agents in materials science, biomedical research, academic ideation, software engineering, synthetic data generation, chemical reasoning, mathematical problem-solving, geographic information systems, multimedia, healthcare, and finance. We then survey key agent-to-agent collaboration protocols, namely the Agent Communication Protocol (ACP), the Model Context Protocol (MCP), and the Agent-to-Agent Protocol (A2A). Finally, we discuss recommendations for future research, focusing on advanced reasoning strategies, failure modes in multi-agent LLM systems, automated scientific discovery, dynamic tool integration via reinforcement learning, integrated search capabilities, and security vulnerabilities in agent protocols.
Authors: Joshua Harris, Fan Grayson, Felix Feldman, Timothy Laurence, Toby Nonnenmacher, Oliver Higgins, Leo Loman, Selina Patel, Thomas Finnie, Samuel Collins, Michael Borowitz
Abstract: As Large Language Models (LLMs) become widely accessible, a detailed understanding of their knowledge within specific domains becomes necessary for successful real world use. This is particularly critical in the domains of medicine and public health, where failure to retrieve relevant, accurate, and current information could significantly impact UK residents. However, while there are a number of LLM benchmarks in the medical domain, currently little is known about LLM knowledge within the field of public health. To address this issue, this paper introduces a new benchmark, PubHealthBench, with over 8000 questions for evaluating LLMs' Multiple Choice Question Answering (MCQA) and free form responses to public health queries. To create PubHealthBench we extract free text from 687 current UK government guidance documents and implement an automated pipeline for generating MCQA samples. Assessing 24 LLMs on PubHealthBench we find the latest proprietary LLMs (GPT-4.5, GPT-4.1 and o1) have a high degree of knowledge, achieving >90% accuracy in the MCQA setup, and outperform humans with cursory search engine use. However, in the free form setup we see lower performance with no model scoring >75%. Therefore, while there are promising signs that state of the art (SOTA) LLMs are an increasingly accurate source of public health information, additional safeguards or tools may still be needed when providing free form responses.
Authors: Rui Miao, Babak Shahbaba, Annie Qu
Abstract: Offline reinforcement learning (RL) aims to find optimal policies in dynamic environments in order to maximize the expected total rewards by leveraging pre-collected data. Learning from heterogeneous data is one of the fundamental challenges in offline RL. Traditional methods focus on learning an optimal policy for all individuals with pre-collected data from a single episode or homogeneous batch episodes, and thus, may result in a suboptimal policy for a heterogeneous population. In this paper, we propose an individualized offline policy optimization framework for heterogeneous time-stationary Markov decision processes (MDPs). The proposed heterogeneous model with individual latent variables enables us to efficiently estimate the individual Q-functions, and our Penalized Pessimistic Personalized Policy Learning (P4L) algorithm guarantees a fast rate on the average regret under a weak partial coverage assumption on behavior policies. In addition, our simulation studies and a real data application demonstrate the superior numerical performance of the proposed method compared with existing methods.
Authors: Ryan Hoque, Peide Huang, David J. Yoon, Mouli Sivapurapu, Jian Zhang
Abstract: Imitation learning for manipulation has a well-known data scarcity problem. Unlike natural language and 2D computer vision, there is no Internet-scale corpus of data for dexterous manipulation. One appealing option is egocentric human video, a passively scalable data source. However, existing large-scale datasets such as Ego4D do not have native hand pose annotations and do not focus on object manipulation. To this end, we use Apple Vision Pro to collect EgoDex: the largest and most diverse dataset of dexterous human manipulation to date. EgoDex has 829 hours of egocentric video with paired 3D hand and finger tracking data collected at the time of recording, where multiple calibrated cameras and on-device SLAM can be used to precisely track the pose of every joint of each hand. The dataset covers a wide range of diverse manipulation behaviors with everyday household objects in 194 different tabletop tasks ranging from tying shoelaces to folding laundry. Furthermore, we train and systematically evaluate imitation learning policies for hand trajectory prediction on the dataset, introducing metrics and benchmarks for measuring progress in this increasingly important area. By releasing this large-scale dataset, we hope to push the frontier of robotics, computer vision, and foundation models. EgoDex is publicly available for download at https://github.com/apple/ml-egodex.
Authors: Siqiao Huang, Jialong Wu, Qixing Zhou, Shangchen Miao, Mingsheng Long
Abstract: World models, which predict future transitions from past observation and action sequences, have shown great promise for improving data efficiency in sequential decision-making. However, existing world models often require extensive domain-specific training and still produce low-fidelity, coarse predictions, limiting their usefulness in complex environments. In contrast, video diffusion models trained on large-scale internet data have demonstrated impressive capabilities in generating high-quality videos that capture diverse real-world dynamics. In this work, we present Vid2World, a general approach for leveraging and transferring pre-trained video diffusion models into interactive world models. To bridge the gap, Vid2World systematically explores video diffusion causalization, reshaping both the architecture and training objective of pre-trained models to enable autoregressive generation. Additionally, it incorporates a causal action guidance mechanism to enhance action controllability in the resulting interactive world models. Extensive experiments across multiple domains, including robot manipulation, 3D game simulation, and open-world navigation, demonstrate that our method offers a scalable and effective pathway for repurposing highly capable video diffusion models into interactive world models.
Authors: Zixuan Ke, Austin Xu, Yifei Ming, Xuan-Phi Nguyen, Ryan Chin, Caiming Xiong, Shafiq Joty
Abstract: Multi-agent systems (MAS) leveraging the impressive capabilities of Large Language Models (LLMs) hold significant potential for tackling complex tasks. However, most current MAS depend on manually designed agent roles and communication protocols. These manual designs often fail to align with the underlying LLMs' strengths and struggle to adapt to novel tasks. Recent automatic MAS approaches attempt to mitigate these limitations but typically necessitate a validation set for tuning and yield static MAS designs lacking adaptability during inference, while also removing the flexibility to reduce to simpler systems. We introduce MAS-ZERO, the first self-evolved, inference-time framework for automatic MAS design. MAS-ZERO employs meta-level design to iteratively design, critique, and refine MAS configurations tailored to each problem instance, without requiring a validation set. Critically, it enables dynamic problem decomposition and agent composition through meta-feedback on solvability and completeness, and reduction to simpler systems when appropriate. Experiments across reasoning (math and graduate-level QA), coding, and agentic (search-based) benchmarks, using both closed-source and open-source LLM backbones of varying sizes, demonstrate that MAS-ZERO outperforms strong manual and automatic MAS baselines. It achieves substantial average accuracy improvements of up to 16.69% on reasoning, 16.66% on coding, and 5.45% on agentic tasks, while maintaining cost efficiency.
Authors: Pingqing Zheng (Katie), Jiayin Qin (Katie), Fuqi Zhang (Katie), Niraj Chitla (Katie), Zishen Wan (Katie), Shang Wu (Katie), Yu Cao (Katie), Caiwen Ding (Katie), Yang (Katie), Zhao
Abstract: Retrieval Augmented Generation (RAG) is an essential agent for Large Language Model (LLM) aided Description Language (HDL) tasks, addressing the challenges of limited training data and prohibitively long prompts. However, its performance in handling ambiguous queries and real-world, repository-level HDL projects containing thousands or even tens of thousands of code lines remains limited. Our analysis demonstrates two fundamental mismatches, structural and vocabulary, between conventional semantic similarity-based RAGs and HDL codes. To this end, we propose HDLxGraph, the first framework that integrates the inherent graph characteristics of HDLs with RAGs for LLM-assisted tasks. Specifically, HDLxGraph incorporates Abstract Syntax Trees (ASTs) to capture HDLs' hierarchical structures and Data Flow Graphs (DFGs) to address the vocabulary mismatch. In addition, to overcome the lack of comprehensive HDL search benchmarks, we introduce HDLSearch, an LLM generated dataset derived from real-world, repository-level HDL projects. Evaluations show that HDLxGraph improves search, debugging, and completion accuracy by 12.04%/12.22%/5.04% and by 11.59%/8.18%/4.07% over state-of-the-art similarity-based RAG and software-code Graph RAG baselines, respectively. The code of HDLxGraph and HDLSearch benchmark are available at https://github.com/UMN-ZhaoLab/HDLxGraph.
Authors: Zhaomin Wu, Ziyang Wang, Bingsheng He
Abstract: Relational databases are often fragmented across organizations, creating data silos that hinder distributed data management and mining. Collaborative learning (CL) -- techniques that enable multiple parties to train models jointly without sharing raw data -- offers a principled approach to this challenge. However, existing CL frameworks (e.g., federated and split learning) remain limited in real-world deployments. Current CL benchmarks and algorithms primarily target the learning step under assumptions of isolated, aligned, and joinable databases, and they typically neglect the end-to-end data management pipeline, especially preprocessing steps such as table joins and data alignment. In contrast, our analysis of the real-world corpus WikiDBs shows that databases are interconnected, unaligned, and sometimes unjoinable, exposing a significant gap between CL algorithm design and practical deployment. To close this evaluation gap, we build WikiDBGraph, a large-scale dataset constructed from 100{,}000 real-world relational databases linked by 17 million weighted edges. Each node (database) and edge (relationship) is annotated with 13 and 12 properties, respectively, capturing a hybrid of instance- and feature-level overlap across databases. Experiments on WikiDBGraph demonstrate both the effectiveness and limitations of existing CL methods under realistic conditions, highlighting previously overlooked gaps in managing real-world data silos and pointing to concrete directions for practical deployment of collaborative learning systems.
Authors: Nathan Gaby, Xiaojing Ye
Abstract: We develop a general theoretical framework for optimal probability density control on standard measure spaces, aimed at addressing large-scale multi-agent control problems. In particular, we establish a maximum principle (MP) for control problems posed on infinite-dimensional spaces of probability distributions and control vector fields. We further derive the Hamilton--Jacobi--Bellman equation for the associated value functional defined on the space of probability distributions. Both results are presented in a concise form and supported by rigorous mathematical analysis, enabling efficient numerical treatment of these problems. Building on the proposed MP, we introduce a scalable numerical algorithm that leverages deep neural networks to handle high-dimensional settings. The effectiveness of the approach is demonstrated through several multi-agent control examples involving domain obstacles and inter-agent interactions.
Authors: Sanjay Kariyappa, G. Edward Suh
Abstract: Prompt injection attacks are a critical security vulnerability in large language models (LLMs), allowing attackers to hijack model behavior by injecting malicious instructions within the input context. Recent defense mechanisms have leveraged an Instruction Hierarchy (IH) Signal, often implemented through special delimiter tokens or additive embeddings to denote the privilege level of input tokens. However, these prior works typically inject the IH signal exclusively at the initial input layer, which we hypothesize limits its ability to effectively distinguish the privilege levels of tokens as it propagates through the different layers of the model. To overcome this limitation, we introduce a novel approach that injects the IH signal into the intermediate token representations within the network. Our method augments these representations with layer-specific trainable embeddings that encode the privilege information. Our evaluations across multiple models and training methods reveal that our proposal yields between $1.6\times$ and $9.2\times$ reduction in attack success rate on gradient-based prompt injection attacks compared to state-of-the-art methods, without significantly degrading the model's utility.
Authors: Fotios Lygerakis, Ozan \"Ozdenizci, Elmar R\"uckert
Abstract: Tactile sensing provides local essential information that is complementary to visual perception, such as texture, compliance, and force. Despite recent advances in visuotactile representation learning, challenges remain in fusing these modalities and generalizing across tasks and environments without heavy reliance on pre-trained vision-language models. Moreover, existing methods do not study positional encodings, thereby overlooking the multi-stage spatial reasoning needed to capture fine-grained visuotactile correlations. We introduce ViTaPEs, a transformer-based architecture for learning task-agnostic visuotactile representations from paired vision and tactile inputs. Our key idea is a two-stage positional injection: local (modality-specific) positional encodings are added within each stream, and a global positional encoding is added on the joint token sequence immediately before attention, providing a shared positional vocabulary at the stage where cross-modal interaction occurs. We make the positional injection points explicit and conduct controlled ablations that isolate their effect before a token-wise nonlinearity versus immediately before self-attention. Experiments on multiple large-scale real-world datasets show that ViTaPEs not only surpasses state-of-the-art baselines across various recognition tasks but also demonstrates zero-shot generalization to unseen, out-of-domain scenarios. We further demonstrate the transfer-learning strength of ViTaPEs in a robotic grasping task, where it outperforms state-of-the-art baselines in predicting grasp success. Project page: https://sites.google.com/view/vitapes
Authors: Zhuoyun Zhong, Seyedali Golestaneh, Constantinos Chamzas
Abstract: Planning with learned dynamics models offers a promising approach toward versatile real-world manipulation, particularly in nonprehensile settings such as pushing or rolling, where accurate analytical models are difficult to obtain. However, collecting training data for learning-based methods can be costly and inefficient, as it often relies on randomly sampled interactions that are not necessarily the most informative. Furthermore, learned models tend to exhibit high uncertainty in underexplored regions of the skill space, undermining the reliability of long-horizon planning. To address these challenges, we propose ActivePusher, a novel framework that combines residual-physics modeling with uncertainty-based active learning, to focus data acquisition on the most informative skill parameters. Additionally, ActivePusher seamlessly integrates with model-based kinodynamic planners, leveraging uncertainty estimates to bias control sampling toward more reliable actions. We evaluate our approach in both simulation and real-world environments, and demonstrate that it consistently improves data efficiency and achieves higher planning success rates in comparison to baseline methods. The source code is available at https://github.com/elpis-lab/ActivePusher.
Authors: Junjie Xing, Yeye He, Mengyu Zhou, Haoyu Dong, Shi Han, Lingjiao Chen, Dongmei Zhang, Surajit Chaudhuri, H. V. Jagadish
Abstract: Tables and table-based use cases play a crucial role in many important real-world applications, such as spreadsheets, databases, and computational notebooks, which traditionally require expert-level users like data engineers, data analysts, and database administrators to operate. Although LLMs have shown remarkable progress in working with tables (e.g., in spreadsheet and database copilot scenarios), comprehensive benchmarking of such capabilities remains limited. In contrast to an extensive and growing list of NLP benchmarks, evaluations of table-related tasks are scarce, and narrowly focus on tasks like NL-to-SQL and Table-QA, overlooking the broader spectrum of real-world tasks that professional users face. This gap limits our understanding and model progress in this important area. In this work, we introduce MMTU, a large-scale benchmark with over 28K questions across 25 real-world table tasks, designed to comprehensively evaluate models ability to understand, reason, and manipulate real tables at the expert-level. These tasks are drawn from decades' worth of computer science research on tabular data, with a focus on complex table tasks faced by professional users. We show that MMTU require a combination of skills -- including table understanding, reasoning, and coding -- that remain challenging for today's frontier models, where even frontier reasoning models like OpenAI GPT-5 and DeepSeek R1 score only around 69\% and 57\% respectively, suggesting significant room for improvement. We highlight key findings in our evaluation using MMTU and hope that this benchmark drives further advances in understanding and developing foundation models for structured data processing and analysis. Our code and data are available at https://github.com/MMTU-Benchmark/MMTU and https://huggingface.co/datasets/MMTU-benchmark/MMTU.
URLs: https://github.com/MMTU-Benchmark/MMTU, https://huggingface.co/datasets/MMTU-benchmark/MMTU.
Authors: Yu Guan, Zhiyu Yin, Haoyu Chen, Sheng Cheng, Chaojie Yang, Kun Qian, Tianyin Xu, Pengcheng Zhang, Yang Zhang, Hanyu Zhao, Yong Li, Wei Lin, Dennis Cai, Ennan Zhai
Abstract: Troubleshooting performance problems of large model training (LMT) is immensely challenging, due to unprecedented scales of modern GPU clusters, the complexity of software-hardware interactions, and the data intensity of the training process. Existing troubleshooting approaches designed for traditional distributed systems or datacenter networks fall short and can hardly apply to real-world training systems. In this paper, we present EROICA, the first online troubleshooting system that provides both fine-grained observation based on profiling, and coverage of all machines in GPU clusters, to diagnose performance issues in production, including both hardware and software problems (or the mixture of both). EROICA effectively summarizes runtime behavior patterns of LMT function executions via online profiling, and leverages differential observability to localize the root cause with minimal production impact. EROICA has been deployed as a production service for large-scale GPU clusters of ~100,000 GPUs for 1.5 years. It has diagnosed a variety of difficult performance issues with 97.5% success.
Authors: Taesoo Park, Mungwi Jeong, Mingyu Park, Narae Kim, Junyoung Kim, Mujung Kim, Jisang Yoo, Hoyun Lee, Sanghoon Kim, Soonchul Kwon
Abstract: This paper presents BemaGANv2, an advanced GAN-based vocoder designed for high-fidelity and long-term audio generation, with a focus on systematic evaluation of discriminator combination strategies. Long-term audio generation is critical for applications in Text-to-Music (TTM) and Text-to-Audio (TTA) systems, where maintaining temporal co- herence, prosodic consistency, and harmonic structure over extended durations remains a significant challenge. Built upon the original BemaGAN architecture, BemaGANv2 incorporates major architectural innovations by replacing traditional ResBlocks in the generator with the Anti-aliased Multi-Periodicity composition (AMP) module, which internally applies the Snake activation function to better model periodic structures. In the discriminator framework, we integrate the Multi-Envelope Discriminator (MED), a novel architecture we proposed, to extract rich temporal en- velope features crucial for periodicity detection. Coupled with the Multi-Resolution Discriminator (MRD), this com- bination enables more accurate modeling of long-range dependencies in audio. We systematically evaluate various discriminator configurations, including Multi-Scale Discriminator (MSD) + MED, MSD + MRD, and Multi-Period Discriminator (MPD) + MED + MRD, using objective metrics (Fr\'echet Audio Distance (FAD), Structural Similar- ity Index (SSIM), Pearson Correlation Coefficient (PCC), Mel-Cepstral Distortion (MCD), Multi-Resolution STFT (M-STFT), Periodicity error (Periodicity)) and subjective evaluations (MOS, SMOS). To support reproducibility, we provide detailed architectural descriptions, training configurations, and complete implementation details. The code, pre-trained models, and audio demo samples are available at: https://github.com/dinhoitt/BemaGANv2.
Authors: Keyhan Najafian, Farhad Maleki, Lingling Jin, Ian Stavness
Abstract: Instance segmentation is essential for applications such as automated monitoring of plant health, growth, and yield. However, extensive effort is required to create large-scale datasets with pixel-level annotations of each object instance for developing instance segmentation models that restrict the use of deep learning in these areas. This challenge is more significant in images with densely packed, self-occluded objects, which are common in agriculture. To address this challenge, we propose a semi-self-supervised learning approach that requires minimal manual annotation to develop a high-performing instance segmentation model. We design GLMask, an image-mask representation for the model to focus on shape, texture, and pattern while minimizing its dependence on color features. We develop a pipeline to generate semantic segmentation and then transform it into instance-level segmentation. The proposed approach substantially outperforms the conventional instance segmentation models, establishing a state-of-the-art wheat head instance segmentation model with mAP@50 of 98.5%. Additionally, we assessed the proposed methodology on the general-purpose Microsoft COCO dataset, achieving a significant performance improvement of over 12.6% mAP@50. This highlights that the utility of our proposed approach extends beyond precision agriculture and applies to other domains, specifically those with similar data characteristics.
Authors: Sungjae Park, Homanga Bharadhwaj, Shubham Tulsiani
Abstract: We propose DemoDiffusion, a simple method for enabling robots to perform manipulation tasks by imitating a single human demonstration, without requiring task-specific training or paired human-robot data. Our approach is based on two insights. First, the hand motion in a human demonstration provides a useful prior for the robot's end-effector trajectory, which we can convert into a rough open-loop robot motion trajectory via kinematic retargeting. Second, while this retargeted motion captures the overall structure of the task, it may not align well with plausible robot actions in-context. To address this, we leverage a pre-trained generalist diffusion policy to modify the trajectory, ensuring it both follows the human motion and remains within the distribution of plausible robot actions. Unlike approaches based on online reinforcement learning or paired human-robot data, our method enables robust adaptation to new tasks and scenes with minimal effort. In real-world experiments across 8 diverse manipulation tasks, DemoDiffusion achieves 83.8\% average success rate, compared to 13.8\% for the pre-trained policy and 52.5\% for kinematic retargeting, succeeding even on tasks where the pre-trained generalist policy fails entirely. Project page: https://demodiffusion.github.io/
Authors: Xinyuan Liu, Jiahui Chen, Bocheng Hu, Yu Sun, Xinyang Chen, Shaoxu Song, Yongxin Tong
Abstract: Data quality remains an important challenge in data-driven systems, as errors in tabular data can severely compromise downstream analytics and machine learning performance. Although numerous error detection algorithms have been proposed, the lack of diverse, real-world error datasets limits comprehensive evaluation. Manual error annotation is both time-consuming and inconsistent, motivating the exploration of synthetic error generation as an alternative. In this work, we introduce TableEG, a framework that leverages large language models (LLMs) to generate authentic errors. By employing a table fine-tuning strategy and a triplet representation $(I, T, O)$ to model error generation, detection, and correction tasks, TableEG captures the complex dependencies inherent in two-dimensional tables. Trained on 12 real-world datasets spanning 10 diverse domains, TableEG ensures that the synthesized errors faithfully reflect authentic error distributions. Experimental results indicate that errors generated by TableEG exhibit superior pattern and distribution similarity compared to both rule-based methods and LLM-generated errors without fine-tuning. Furthermore, performance metrics on TableEG-generated errors closely align with those on real-world errors across nearly all datasets and detection algorithms, particularly for machine learning based detection techniques. Overall, TableEG not only bridges the gap between synthetic and real-world errors but also establishes a robust benchmark for subsequent error detection and correction tasks.
Authors: Moises Andrade, Joonhyuk Cha, Brandon Ho, Vriksha Srihari, Karmesh Yadav, Zsolt Kira
Abstract: Verifiers--functions assigning rewards to agent behavior--have been key to AI progress in math, code, and games. However, extending gains to domains without clear-cut success criteria remains a challenge: while humans can recognize desired outcomes, translating this intuition into scalable rules is nontrivial. Multimodal LLMs (MLLMs) offer a promising solution, given their world knowledge, human-preference alignment, and reasoning capabilities. We evaluate MLLM verifiers across web navigation, computer use, and robotics, spanning 13+ models, 28+ designs, and thousands of trajectories from diverse agents. We identify a critical limitation: a strong tendency for MLLMs to over-validate agent behavior--a phenomenon we term agreement bias. This bias is pervasive, resilient to test-time scaling, and can harm applications relying on MLLM judgments/rewards (e.g., self-improvement, steering, online supervision). We discuss several considerations for evaluating and designing MLLM verifiers, and introduce SGV, a lightweight method that better leverages their capabilities by modulating (un)conditional generation. First, an MLLM is elicited to generate broad priors about desired behavior, independent of the data under evaluation. Then, conditioned on self-generated priors, it reasons over and evaluates a candidate trajectory. Our methods yield more human-aligned verifiers, improving failure detection by 25pp and accuracy by 14pp. In self-improvement and online supervision, they boost task completion of a GUI specialist in OSWorld, a diffusion policy in robomimic, and a ReAct agent in VisualWebArena--surpassing the previous state of the art by 20pp. As a byproduct, we release an update of VisualWebArena featuring strong agent baselines, more human-aligned oracles, container parallelism with high fidelity and proper resets, >10x speedups, and VWA-Lite, a 1/3 subset with comparable evaluation fidelity.
Authors: Yucong Zhang, Juan Liu, Ming Li
Abstract: Pre-trained foundation models have demonstrated remarkable success in audio, vision and language, yet their potential for general machine signal modeling with arbitrary sampling rates-covering acoustic, vibration, and other industrial sensor data-remains under-explored. In this work, we propose a novel foundation model ECHO that integrates an advanced band-split architecture with frequency positional embeddings, enabling spectral localization across arbitrary sampling configurations. Moreover, the model incorporates sliding patches to support inputs of variable length without padding or cropping, producing a concise embedding that retains both temporal and spectral fidelity and naturally extends to streaming scenarios. We evaluate our method on various kinds of machine signal datasets, including previous DCASE task 2 challenges (2020-2025), and widely-used industrial signal corpora. Experimental results demonstrate consistent state-of-the-art performance in machine signal anomaly detection and fault classification, confirming the effectiveness and generalization capability of the proposed model. We open-sourced ECHO on https://github.com/yucongzh/ECHO.
Authors: Yevhen Havrylenko, Meelis K\"a\"arik, Artur Tuttar
Abstract: Actuarial ratemaking depends on high-quality data, yet access to such data is often limited by the cost of obtaining new data, privacy concerns, etc. In this paper, we explore synthetic-data generation as a potential solution to these issues. In addition to generative methods previously studied in the actuarial literature, we explore and benchmark another class of approaches based on Multivariate Imputation by Chained Equations (MICE). In a comparative study using an open-source dataset, MICE-based models are evaluated against other generative models like Variational Autoencoders and Conditional Tabular Generative Adversarial Networks. We assess how well synthetic data preserves the original marginal distributions of variables as well as the multivariate relationships among covariates. The consistency between Generalized Linear Models (GLMs) trained on synthetic data with GLMs trained on the original data is also investigated. Furthermore, we assess the ease of use of each generative approach and study the impact of generically augmenting original data with synthetic data on the performance of GLMs for predicting claim counts. Our results highlight the potential of MICE-based methods in creating high-fidelity tabular data while offering lower implementation complexity compared to deep generative models.
Authors: Lesi Chen, Junru Li, El Mahdi Chayti, Jingzhao Zhang
Abstract: This paper studies the complexity of finding an $\epsilon$-stationary point for stochastic bilevel optimization when the upper-level problem is nonconvex and the lower-level problem is strongly convex. Recent work proposed the first-order method, F${}^2$SA, achieving the $\tilde{\mathcal{O}}(\epsilon^{-6})$ upper complexity bound for first-order smooth problems. This is slower than the optimal $\Omega(\epsilon^{-4})$ complexity lower bound in its single-level counterpart. In this work, we show that faster rates are achievable for higher-order smooth problems. We first reformulate F$^2$SA as approximating the hyper-gradient with a forward difference. Based on this observation, we propose a class of methods F${}^2$SA-$p$ that uses $p$th-order finite difference for hyper-gradient approximation and improves the upper bound to $\tilde{\mathcal{O}}(p \epsilon^{-4-p/2})$ for $p$th-order smooth problems. Finally, we demonstrate that the $\Omega(\epsilon^{-4})$ lower bound also holds for stochastic bilevel problems when the high-order smoothness holds for the lower-level variable, indicating that the upper bound of F${}^2$SA-$p$ is nearly optimal in the highly smooth region $p = \Omega( \log \epsilon^{-1} / \log \log \epsilon^{-1})$.
Authors: Jason Starace, Terence Soule
Abstract: We establish empirical bounds on behavioral inference through controlled experiments at scale: LLM-based agents assigned one of 36 behavioral profiles (9 belief systems x 4 motivations) generate over 1.5 million behavioral sequences across 17,411 games in grid-world environments, providing ground truth unavailable in human behavioral studies. Rather than asking whether inference has limits, we ask how large those limits are, where they concentrate, and why. A fundamental asymmetry emerges in both magnitude and structure. Motivations achieve 98-100% inference accuracy and recover 97% of available mutual information across all architectures. Belief systems plateau at 24% for LSTMs regardless of capacity, recovering only 30% of available information, a 3.3x asymmetry in information extraction efficiency. Transformer architectures with 9-stage curriculum learning reach 49% alignment accuracy, doubling LSTM performance and demonstrating that the recurrent ceiling is architectural rather than fundamental. Yet even this improvement leaves belief systems correctly classified less than half the time, with per-alignment accuracy ranging from 1% (True Neutral) to 72% (Lawful Evil). Confusion analysis maps the failure structure precisely: a "neutral zone" of behavioral ambiguity extends beyond True Neutral to encompass Good alignments, where prosocial behavior is indistinguishable from rule-following or balance-keeping. Combined motivation and belief inference yields 17.6x improvement over random baseline for full 36-class profile classification, while establishing that the bottleneck is entirely located in belief system inference. Signal enhancement and explanatory queries yield only marginal LSTM gains (+3.8%), confirming that the ceiling is information-theoretic rather than data-limited. These bounds have direct implications for any system relying on behavioral monitoring to infer agent values.
Authors: Jackson Eshbaugh, Chetan Tiwari, Jorge Silveyra
Abstract: Computational models have emerged as powerful tools for multi-scale energy modeling research at the building level as well as urban scale. However, these models require a plethora of data on building parameters, some of which can be inaccessible, expensive, or can raise privacy concerns. We introduce a modular multimodal framework to synthetically produce this data from publicly accessible images and residential information using generative Artificial Intelligence (AI). Additionally, we provide a modeling pipeline demonstrating this framework and we evaluate its generative AI components for realism. Our experiments show that our framework's use of AI avoids common issues with generative models and produces realistic multimodal data at the building scale. Resulting datasets can be used for assessing influence of energy efficiency upgrades at the building scale, as well as to simulate larger patterns of energy consumption across regions. This work will support research in building and energy simulation by reducing dependence on costly or restricted data sources, and pave a path towards more accessible research in Machine Learning (ML) and other data-driven disciplines.
Authors: Jiayun Wang, Yousuf Aborahama, Arya Khokhar, Yang Zhang, Chuwei Wang, Karteekeya Sastry, Julius Berner, Yilin Luo, Boris Bonev, Zongyi Li, Kamyar Azizzadenesheli, Lihong V. Wang, Anima Anandkumar
Abstract: Learning physics-constrained inverse operators-rather than post-processing physics-based reconstructions-is a broadly applicable strategy for problems with expensive forward models. We demonstrate this principle in three-dimensional photoacoustic computed tomography (3D PACT), where current systems demand dense transducer arrays and prolonged scans, restricting clinical translation. We introduce PANO (PACT imaging neural operator), an end-to-end physics-aware neural operator-a deep learning architecture that generalizes across input sampling densities without retraining-that directly learns the inverse mapping from raw sensor measurements to a 3D volumetric image. Unlike two-step methods that reconstruct then denoise, PANO performs direct inversion in a single pass, jointly embedding physics and data priors. It employs spherical discrete-continuous convolutions to respect hemispherical sensor geometry and Helmholtz equation constraints to ensure physical consistency. PANO reconstructs high-quality images from both simulated and real data across diverse sparse acquisition settings, achieves real-time inference and outperforms the widely-used UBP algorithm by approximately 33 percentage points in cosine similarity on simulated data and 14 percentage points on real phantom data. These results establish a pathway toward more accessible 3D PACT systems for preclinical research, and motivate future in-vivo validation for clinical translation.
Authors: Julien Brandoit, Damien Ernst, Guillaume Drion, Arthur Fyon
Abstract: Inferring the biophysical parameters of conductance-based models (CBMs) from experimentally accessible recordings remains a central challenge in computational neuroscience. Spike times are the most widely available data, yet they reveal little about which combinations of ion channel conductances generate the observed activity. This inverse problem is further complicated by neuronal degeneracy, where multiple distinct conductance sets yield similar spiking patterns. We introduce a method that addresses this challenge by combining deep learning with Dynamic Input Conductances (DICs), a theoretical framework that reduces complex CBMs to three interpretable feedback components governing excitability and firing patterns. Our approach first maps spike times to DIC densities at threshold using a neural network that learns a low-dimensional representation of neuronal activity. The predicted DIC values are then used to generate degenerate CBM populations via an iterative compensation algorithm, ensuring compatibility with the intermediate target DICs, and thereby reproducing the corresponding firing patterns, even in high-dimensional models. Applied to two models, this algorithmic pipeline reconstructs spiking and bursting regimes with high accuracy and robustness to variability, including spike trains generated under noisy current injection mimicking physiological stochasticity. It produces diverse degenerate populations within milliseconds on standard hardware, enabling scalable and efficient inference from spike recordings alone. Together, this work positions DICs as a practical and interpretable link between experimentally observed activity and mechanistic models. By enabling fast and scalable reconstruction of degenerate populations directly from spike times, our approach provides a powerful way to investigate how neurons exploit conductance variability to achieve reliable computation.
Authors: Di Wen, Kunyu Peng, Junwei Zheng, Yufan Chen, Yitian Shi, Jiale Wei, Ruiping Liu, Kailun Yang, Rainer Stiefelhagen
Abstract: Industrial workflows demand adaptive and trustworthy assistance that can operate under limited computing, connectivity, and strict privacy constraints. In this work, we present MICA (Multi-Agent Industrial Coordination Assistant), a perception-grounded and speech-interactive system that delivers real-time guidance for assembly, troubleshooting, part queries, and maintenance. MICA coordinates five role-specialized language agents, audited by a safety checker, to ensure accurate and compliant support. To achieve robust step understanding, we introduce Adaptive Step Fusion (ASF), which dynamically blends expert reasoning with online adaptation from natural speech feedback. Furthermore, we establish a new multi-agent coordination benchmark across representative task categories and propose evaluation metrics tailored to industrial assistance, enabling systematic comparison of different coordination topologies. Our experiments demonstrate that MICA consistently improves task success, reliability, and responsiveness over baseline structures, while remaining deployable on practical offline hardware. Together, these contributions highlight MICA as a step toward deployable, privacy-preserving multi-agent assistants for dynamic factory environments. The source code will be made publicly available at https://github.com/Kratos-Wen/MICA.
Authors: Zhaoyang Li, Zhan Ling, Yuchen Zhou, Litian Gong, Erdem B{\i}y{\i}k, Hao Su
Abstract: Large Vision-Language Models (LVLMs) excel at captioning, visual question answering, and robotics by combining vision and language, yet they often miss obvious objects or hallucinate nonexistent ones in atypical scenes. We examine these failures through the lens of uncertainty, focusing on contextual incongruity, where objects appear unexpectedly or fail to appear in expected contexts, and show that such cases increase recognition difficulty for state-of-the- art LVLMs. To study this regime, we introduce the Object Recognition in Incongruous Context (ORIC) framework, which constructs incongruous object-context pairs through two complementary strategies: (1) LLM-guided sampling to identify hard-to-recognize objects present in the image and (2) CLIP-guided sampling to mine plausible but absent ones. Applied to MSCOCO, ORIC creates ORIC-Bench and ORIC-style training data. Evaluating 18 LVLMs and 2 open-vocabulary detectors reveals significant degradation and bias under incongruous contexts. Visual Reinforcement Fine-Tuning of Qwen3-VL-8B-Instruct on 600 ORIC samples improves performance on ORIC-Bench, AMBER, and HallusionBench. Overall, we show that contextual incongruity is a key source of uncertainty and provide tools for more reliable LVLMs. The dataset and code are publicly available at https://github.com/ZhaoyangLi-1/ORIC.
Authors: Bojan Deraji\'c, Sebastian Bernhard, Wolfgang H\"onig
Abstract: Control barrier functions (CBFs) have been demonstrated as an effective method for safety-critical control of autonomous systems. Although CBFs are simple to deploy, their design remains challenging, motivating the development of learning-based approaches. Yet, issues such as suboptimal safe sets, applicability in partially observable environments, and lack of rigorous safety guarantees persist. In this work, we propose observation-conditioned neural CBFs based on Hamilton-Jacobi (HJ) reachability analysis, which approximately recover the maximal safe sets. We exploit certain mathematical properties of the HJ value function, ensuring that the predicted safe set never intersects with the observed failure set. Moreover, we leverage a hypernetwork-based architecture that is particularly suitable for the design of observation-conditioned safety filters. The proposed method is examined both in simulation and hardware experiments for a ground robot and a quadcopter. The results show improved success rates and generalization to out-of-domain environments compared to the baselines.
Authors: Vahe Karagulyan, Pierre Alquier
Abstract: The core of generalization theory was developed for independent observations. Some PAC and PAC-Bayes bounds are available for data that exhibit a temporal dependence. However, there are constants in these bounds that depend on properties of the data-generating process: mixing coefficients, mixing time, spectral gap... Such constants are unknown in practice. In this paper, we prove a new PAC-Bayes bound for Markov chains. This bound depends on a quantity called the pseudo-spectral gap. The main novelty is that we can provide an empirical bound on the pseudo-spectral gap when the state space is finite. Thus, we obtain the first fully empirical PAC-Bayes bound for Markov chains. This extends beyond the finite case, although this requires additional assumptions. On simulated experiments, the empirical version of the bound is essentially as tight as the non-empirical one.
Authors: Gerard Boxo, Aman Neelappa, Shivam Raval
Abstract: White-box monitors are a popular technique for detecting potentially harmful behaviours in language models. While they perform well in general, their effectiveness in detecting text-ambiguous behaviour is disputed. In this work, we find evidence that removing textual evidence of a behaviour significantly decreases probe performance. The AUROC reduction ranges from $10$- to $30$-point depending on the setting. We evaluate probe monitors across three setups (Sandbagging, Sycophancy, and Bias), finding that when probes rely on textual evidence of the target behaviour (such as system prompts or CoT reasoning), performance degrades once these tokens are filtered. This filtering procedure is standard practice for output monitor evaluation. As further evidence of this phenomenon, we train Model Organisms which produce outputs without any behaviour verbalisations. We validate that probe performance on Model Organisms is substantially lower than unfiltered evaluations: $0.57$ vs $0.74$ AUROC for Bias, and $0.57$ vs $0.94$ AUROC for Sandbagging. Our findings suggest that linear probes may be brittle in scenarios where they must detect non-surface-level patterns.
Authors: Shuai Shao, Qihan Ren, Chen Qian, Boyi Wei, Dadi Guo, Jingyi Yang, Xinhao Song, Linfeng Zhang, Weinan Zhang, Dongrui Liu, Jing Shao
Abstract: Advances in Large Language Models (LLMs) have enabled a new class of self-evolving agents that autonomously improve through interaction with the environment, demonstrating strong capabilities. However, self-evolution also introduces novel risks overlooked by current safety research. In this work, we study the case where an agent's self-evolution deviates in unintended ways, leading to undesirable or even harmful outcomes. We refer to this as Misevolution. To provide a systematic investigation, we evaluate misevolution along four key evolutionary pathways: model, memory, tool, and workflow. Our empirical findings reveal that misevolution is a widespread risk, affecting agents built even on top-tier LLMs (e.g., Gemini-2.5-Pro). Different emergent risks are observed in the self-evolutionary process, such as the degradation of safety alignment after memory accumulation, or the unintended introduction of vulnerabilities in tool creation and reuse. To our knowledge, this is the first study to systematically conceptualize misevolution and provide empirical evidence of its occurrence, highlighting an urgent need for new safety paradigms for self-evolving agents. Finally, we discuss potential mitigation strategies to inspire further research on building safer and more trustworthy self-evolving agents. Our code and data are available at https://github.com/ShaoShuai0605/Misevolution . Warning: this paper includes examples that may be offensive or harmful in nature.
Authors: Emil Javurek, Valentyn Melnychuk, Jonas Schweisthal, Konstantin Hess, Dennis Frauen, Stefan Feuerriegel
Abstract: Predicting individualized potential outcomes in sequential decision-making is central for optimizing therapeutic decisions in personalized medicine (e.g., which dosing sequence to give to a cancer patient). However, predicting potential outcomes over long horizons is notoriously difficult. Existing methods that break the curse of the horizon typically lack strong theoretical guarantees such as orthogonality and quasi-oracle efficiency. In this paper, we revisit the problem of predicting individualized potential outcomes in sequential decision-making (i.e., estimating Q-functions in Markov decision processes with observational data) through a causal inference lens. In particular, we develop a comprehensive theoretical foundation for meta-learners in this setting with a focus on beneficial theoretical properties. As a result, we yield a novel meta-learner called DRQ-learner and establish that it is: (1) doubly robust (i.e., valid inference under the misspecification of one of the nuisances), (2) Neyman-orthogonal (i.e., insensitive to first-order estimation errors in the nuisance functions), and (3) achieves quasi-oracle efficiency (i.e., behaves asymptotically as if the ground-truth nuisance functions were known). Our DRQ-learner is applicable to settings with both discrete and continuous state spaces. Further, our DRQ-learner is flexible and can be used together with arbitrary machine learning models (e.g., neural networks). We validate our theoretical results through numerical experiments, thereby showing that our meta-learner outperforms state-of-the-art baselines.
Authors: G\"unter F. Steinke, Thomas Steinke
Abstract: Standard techniques for differentially private estimation, such as Laplace or Gaussian noise addition, require guaranteed bounds on the sensitivity of the estimator in question. But such sensitivity bounds are often large or simply unknown. Thus we seek differentially private methods that can be applied to arbitrary black-box functions. A handful of such techniques exist, but all are either inefficient in their use of data or require evaluating the function on exponentially many inputs. In this work we present a scheme that trades off between statistical efficiency (i.e., how much data is needed) and oracle efficiency (i.e., the number of evaluations). We also present lower bounds showing the near-optimality of our scheme.
Authors: Nurbek Tastan, Samuel Horvath, Karthik Nandakumar
Abstract: Multi-agent systems (MAS) based on Large Language Models (LLMs) have the potential to solve tasks that are beyond the reach of any single LLM. However, this potential can only be realized when the collaboration mechanism between agents is optimized. Specifically, optimizing the communication structure between agents is critical for fruitful collaboration. Most existing approaches rely on fixed topologies, pretrained graph generators, optimization over edges, or employ external LLM judges, thereby adding to the complexity. In this work, we introduce a response-conditioned framework that adapts communication on-the-fly. Agents independently generate responses to the user query and assess peer contributions using an approximation of the Shapley value. A directed acyclic graph (DAG) is then constructed to regulate the propagation of the responses among agents, which ensures stable and efficient message transmission from high-contributing agents to others. This graph is dynamically updated based on the agent responses from the previous collaboration round. Since the proposed framework enables the self-organization of agents without additional supervision or training, we refer to it as SelfOrg. The SelfOrg framework goes beyond task- and query-level optimization and takes into account the stochastic nature of agent responses. Experiments with both strong and weak LLM backends demonstrate robust performance, with significant gains in the weak regime where prior methods collapse. We also theoretically show that multiple agents increase the chance of correctness and that the correct responses naturally dominate the information flow.
Authors: Giovanni Minelli, Giulio Turrisi, Victor Barasuol, Claudio Semini
Abstract: Learning robotic manipulation policies through supervised learning from demonstrations remains challenging when policies encounter execution variations not explicitly covered during training. While incorporating historical context through attention mechanisms can improve robustness, standard approaches process all past states in a sequence without explicitly modeling the temporal structure that demonstrations may include, such as failure and recovery patterns. We propose a Cross-State Transition Attention Transformer that employs a novel State Transition Attention (STA) mechanism to modulate standard attention weights based on learned state evolution patterns, enabling policies to better adapt their behavior based on execution history. Our approach combines this structured attention with temporal masking during training, where visual information is randomly removed from recent timesteps to encourage temporal reasoning from historical context. Evaluation in simulation shows that STA consistently outperforms standard attention approach and temporal modeling methods like TCN and LSTM networks, achieving more than 2x improvement over cross-attention on precision-critical tasks. The source code and data can be accessed at https://github.com/iit-DLSLab/croSTAta
Authors: Eduardo Fernandes Montesuma, Yassir Bendou, Mike Gartrell
Abstract: Wasserstein barycenters provide a principled approach for aggregating probability measures, while preserving the geometry of their ambient space. Existing discrete methods are not scalable as they assume access to the complete set of samples from the input measures. Meanwhile, neural network approaches do scale well, but rely on complex optimization problems and cannot easily incorporate label information. We address these limitations through gradient flows in the space of probability measures. Through time discretization, we achieve a scalable algorithm that i) relies on mini-batch optimal transport, ii) accepts modular regularization through task-aware functions, and iii) seamlessly integrates supervised information into the ground-cost. We empirically validate our approach on domain adaptation benchmarks that span computer vision, neuroscience, and chemical engineering. Our method establishes a new state-of-the-art barycenter solver, with labeled barycenters consistently outperforming unlabeled ones.
Authors: Jiale Fan, Andrei Cramariuc, Tifanny Portela, Marco Hutter
Abstract: The pretraining-finetuning paradigm has facilitated numerous transformative advancements in artificial intelligence research in recent years. However, in the domain of reinforcement learning (RL) for robot locomotion, individual skills are often learned from scratch despite the high likelihood that some generalizable knowledge is shared across all task-specific policies belonging to the same robot embodiment. This work aims to define a paradigm for pretraining neural network models that encapsulate such knowledge and can subsequently serve as a basis for warm-starting the RL process in classic actor-critic algorithms, such as Proximal Policy Optimization (PPO). We begin with a task-agnostic exploration-based data collection algorithm to gather diverse, dynamic transition data, which is then used to train a Proprioceptive Inverse Dynamics Model (PIDM) through supervised learning. The pretrained weights are then loaded into both the actor and critic networks to warm-start the policy optimization of actual tasks. We systematically validated our proposed method with 9 distinct robot locomotion RL environments comprising 3 different robot embodiments, showing significant benefits of this initialization strategy. Our proposed approach on average improves sample efficiency by 36.9% and task performance by 7.3% compared to random initialization. We further present key ablation studies and empirical analyses that shed light on the mechanisms behind the effectiveness of this method.
Authors: Roger Creus Castanyer, Faisal Mohamed, Pablo Samuel Castro, Cyrus Neary, Glen Berseth
Abstract: Reinforcement learning (RL) algorithms are highly sensitive to reward function specification, which remains a central challenge limiting their broad applicability. We present ARM-FM: Automated Reward Machines via Foundation Models, a framework for automated, compositional reward design in RL that leverages the high-level reasoning capabilities of foundation models (FMs). Reward machines (RMs) -- an automata-based formalism for reward specification -- are used as the mechanism for RL objective specification, and are automatically constructed via the use of FMs. The structured formalism of RMs yields effective task decompositions, while the use of FMs enables objective specifications in natural language. Concretely, we (i) use FMs to automatically generate RMs from natural language specifications; (ii) associate language embeddings with each RM automata-state to enable generalization across tasks; and (iii) provide empirical evidence of ARM-FM's effectiveness in a diverse suite of challenging environments, including evidence of zero-shot generalization.
Authors: Alex Alberts, Ilias Bilionis
Abstract: Quantifying the uncertainty in the output of a neural network is essential for deployment in scientific or engineering applications where decisions must be made under limited or noisy data. Bayesian neural networks (BNNs) provide a framework for this purpose by constructing a Bayesian posterior distribution over the network parameters. However, the prior, which is of key importance in any Bayesian setting, is rarely meaningful for BNNs. This is because the complexity of the input-to-output map of a BNN makes it difficult to understand how certain distributions enforce any interpretable constraint on the output space of the network. Gaussian processes (GPs), on the other hand, are often preferred in uncertainty quantification tasks due to their interpretability. The drawback is that GPs are limited to small datasets without advanced techniques, which often rely on the covariance kernel having a specific structure. To address these challenges, we introduce a new class of priors for BNNs, called Mercer priors, such that the resulting BNN has samples which approximate that of a specified GP. The method works by defining a prior directly over the network parameters from the Mercer representation of the covariance kernel, and does not rely on the network having a specific structure. In doing so, we can exploit the scalability of BNNs in a meaningful Bayesian way.
Authors: Atsuyuki Miyai, Mashiro Toyooka, Takashi Otonari, Zaiying Zhao, Kiyoharu Aizawa
Abstract: Understanding the current capabilities and risks of AI Scientist systems is essential for ensuring trustworthy and sustainable AI-driven scientific progress while preserving the integrity of the academic ecosystem. To this end, we develop Jr. AI Scientist, a state-of-the-art autonomous AI scientist system that mimics the core research workflow of a novice student researcher: Given the baseline paper from the human mentor, it analyzes its limitations, formulates novel hypotheses for improvement, validates them through rigorous experimentation, and writes a paper with the results. Unlike previous approaches that assume full automation or operate on small-scale code, Jr. AI Scientist follows a well-defined research workflow and leverages modern coding agents to handle complex, multi-file implementations, leading to scientifically valuable contributions. Through our experiments, the Jr. AI Scientist successfully generated new research papers that build upon real NeurIPS, IJCV, and ICLR works by proposing and implementing novel algorithms. For evaluation, we conducted automated assessments using AI Reviewers, author-led evaluations, and submissions to Agents4Science, a venue dedicated to AI-driven scientific contributions. The findings demonstrate that Jr. AI Scientist generates papers receiving higher review scores by DeepReviewer than existing fully automated systems. Nevertheless, we identify important limitations from both the author evaluation and the Agents4Science reviews, indicating the potential risks of directly applying current AI Scientist systems and key challenges for future research. Finally, we comprehensively report various risks identified during development. We believe this study clarifies the current role and limitations of AI Scientist systems, offering insights into the areas that still require human expertise and the risks that may emerge as these systems evolve.
Authors: Jerry Lin, Zeyuan Hu, Tom Beucler, Katherine Frields, Hannah Christensen, Walter Hannah, Helge Heuer, Peter Ukkonnen, Laura A. Mansfield, Tian Zheng, Liran Peng, Ritwik Gupta, Pierre Gentine, Yusef Al-Naher, Mingjiang Duan, Kyo Hattori, Weiliang Ji, Chunhan Li, Kippei Matsuda, Naoki Murakami, Shlomo Ron, Marec Serlin, Hongjian Song, Yuma Tanabe, Daisuke Yamamoto, Jianyao Zhou, Mike Pritchard
Abstract: Subgrid machine-learning (ML) parameterizations have the potential to introduce a new generation of climate models that incorporate the effects of higher-resolution physics without incurring the prohibitive computational cost associated with more explicit physics-based simulations. However, important issues, ranging from online instability to inconsistent online performance, have limited their operational use for long-term climate projections. To more rapidly drive progress in solving these issues, domain scientists and machine learning researchers opened up the offline aspect of this problem to the broader machine learning and data science community with the release of ClimSim, a NeurIPS Datasets and Benchmarks publication, and an associated Kaggle competition. This paper reports on the downstream results of the Kaggle competition by coupling emulators inspired by the winning teams' architectures to an interactive climate model (including full cloud microphysics, a regime historically prone to online instability) and systematically evaluating their online performance. Our results demonstrate that online stability in the low-resolution, real-geography setting is reproducible across multiple diverse architectures, which we consider a key milestone. All tested architectures exhibit strikingly similar offline and online biases, though their responses to architecture-agnostic design choices (e.g., expanding the list of input variables) can differ significantly. Multiple Kaggle-inspired architectures achieve state-of-the-art (SOTA) results on certain metrics such as zonal mean bias patterns and global RMSE, indicating that crowdsourcing the essence of the offline problem is one path to improving online performance in hybrid physics-AI climate simulation.
Authors: Abdelghafour Halimi, Ali Alibrahim, Didier Barradas-Bautista, Ronell Sicat, Abdulkader M. Afifi
Abstract: This study presents a comprehensive deep learning pipeline for the automated classification of foraminifera species using 2D micro-CT slices derived from 3D scans. We curated a scientifically rigorous dataset of 97 micro-CT scanned specimens spanning 27 species, from which we selected 12 representative species with sufficient specimen counts (at least four 3D models each) for robust classification. To ensure methodological integrity and prevent data leakage, we employed specimen-level data splitting, resulting in 109,617 high-quality 2D slices (44,103 for training, 14,046 for validation, and 51,468 for testing). We evaluated seven state-of-the-art 2D convolutional neural network (CNN) architectures using transfer learning. Our final ensemble model, ForamDeepSlice (FDS), combining ConvNeXt-Large and EfficientNetV2-Small, achieved a test accuracy of 95.64%, with a top-3 accuracy of 99.6% and an area under the ROC curve (AUC) of 0.998 across all species. To facilitate practical deployment, we developed an interactive advanced dashboard that supports real-time slice classification and 3D slice matching using advanced similarity metrics, including SSIM, NCC, and the Dice coefficient. This work establishes new benchmarks for AI-assisted micropaleontological identification and provides a fully reproducible framework for foraminifera classification research, bridging the gap between deep learning and applied geosciences.
Authors: Yan Yang, George Bebis, Mircea Nicolescu
Abstract: The absence of large-scale masked face datasets poses challenges for masked face detection and recognition. We propose a two-step generative data augmentation framework combining rule-based mask warping with unpaired image-to-image translation using GANs, producing masked face samples that go beyond rule-based geometric overlays. Trained on 3390 images, about 0.7% of the training data used by IAMGAN, the proposed approach yields consistent improvements over rule-based warping alone and achieves complementary results at a small fraction of IAMGAN data scale, showing that both steps contribute. Evaluation is conducted directly on the generated samples and is qualitative; quantitative metrics like FID and KID were not applied as any real reference distribution would unfairly favor the model with closer training data. We introduce a non-mask preservation loss to reduce non-mask distortions and stabilize training, and stochastic noise injection to enhance sample diversity. Note: This paper originated as a coursework submission completed under resource constraints. Following an inexplicable termination of scholarship, the author took on part-time employment to maintain research continuity, which introduced a mid-semester domain pivot from medical imaging to masked face tasks due to restrictions on company data. The work was completed alongside concurrent coursework with delayed compute access and without AI assistance of any kind. It was submitted to a small venue at the semester end under an obligatory publication requirement and accepted without revision requests. Subsequent invitations to submit to first-tier venues were not pursued due to continued funding absence. Downstream evaluation on recognition or detection performance was not completed by the submission deadline. These notes are added in response to subsequent comparisons and criticisms that did not account for these conditions.
Authors: Ananta R. Bhattarai, Helge Rhodin
Abstract: Monocular depth estimation remains challenging, as foundation models such as Depth Anything V2 (DA-V2) struggle with real-world images that are far from the training distribution. We introduce Re-Depth Anything, a test-time self-supervision framework that bridges this domain gap by fusing foundation models with the powerful priors of large-scale 2D diffusion models. Our method performs label-free refinement directly on the input image by re-lighting the predicted depth map and augmenting the input. This re-synthesis method replaces classical photometric reconstruction by leveraging shape from shading (SfS) cues in a new, generative context with Score Distillation Sampling (SDS). To prevent optimization collapse, our framework updates only intermediate embeddings and the decoder's weights, rather than optimizing the depth tensor directly or fine-tuning the full model. Across diverse benchmarks, Re-Depth Anything yields substantial gains in depth accuracy and realism over DA-V2, and applied on top of Depth Anything 3 (DA3) achieves state-of-the-art results, showcasing new avenues for self-supervision by geometric reasoning.
Authors: Yu Jiang, Xindi Tong, Ziyao Liu, Xiaoxi Zhang, Kwok-Yan Lam, Chee Wei Tan
Abstract: Federated unlearning has become an attractive approach to address privacy concerns in collaborative machine learning, for situations when sensitive data is remembered by AI models during the machine learning process. It enables the removal of specific data influences from trained models, aligning with the growing emphasis on the "right to be forgotten." While extensively studied in horizontal federated learning, unlearning in vertical federated learning (VFL) remains challenging due to the distributed feature architecture. VFL unlearning includes sample unlearning that removes specific data points' influence and label unlearning that removes entire classes. Since different parties hold complementary features of the same samples, unlearning tasks require cross-party coordination, creating computational overhead and complexities from feature interdependencies. To address such challenges, we propose FedORA (Federated Optimization for data Removal via primal-dual Algorithm), designed for sample and label unlearning in VFL. FedORA formulates the removal of certain samples or labels as a constrained optimization problem solved using a primal-dual framework. Our approach introduces a new unlearning loss function that promotes classification uncertainty rather than misclassification. An adaptive step size enhances stability, while an asymmetric batch design, considering the prior influence of the remaining data on the model, handles unlearning and retained data differently to efficiently reduce computational costs. We provide theoretical analysis proving that the model difference between FedORA and Train-from-scratch is bounded, establishing guarantees for unlearning effectiveness. Experiments on tabular and image datasets demonstrate that FedORA achieves unlearning effectiveness and utility preservation comparable to Retrain with reduced computation and communication overhead.
Authors: Anna Calissano, Etienne Lasalle
Abstract: Spatial graphs are particular graphs for which the nodes are localized in space (e.g., public transport network, molecules, branching biological structures). In this work, we consider the problem of spatial graph reduction, that aims to find a smaller spatial graph (i.e., with less nodes) with the same overall structure as the initial one. In this context, performing the graph reduction while preserving the main topological features of the initial graph is particularly relevant, due to the additional spatial information. Thus, we propose a topological spatial graph coarsening approach based on a new framework that finds a trade-off between the graph reduction and the preservation of the topological characteristics. The coarsening is realized by collapsing short edges. In order to capture the topological information required to calibrate the reduction level, we adapt the construction of classical topological descriptors made for point clouds (the so-called persistent diagrams) to spatial graphs. This construction relies on the introduction of a new filtration called triangle-aware graph filtration. Our coarsening approach is parameter-free and we prove that it is equivariant under rotations, translations and scaling of the initial spatial graph. We evaluate the performances of our method on synthetic and real spatial graphs, and show that it significantly reduces the graph sizes while preserving the relevant topological information.
Authors: Nam Phuong Tran, Andi Nika, Goran Radanovic, Long Tran-Thanh, Debmalya Mandal
Abstract: We investigate robustness to strong data corruption in offline sparse reinforcement learning (RL). In our setting, an adversary may arbitrarily perturb a fraction of the collected trajectories from a high-dimensional but sparse Markov decision process, and our goal is to estimate a near optimal policy. The main challenge is that, in the high-dimensional regime where the number of samples $N$ is smaller than the feature dimension $d$, exploiting sparsity is essential for obtaining non-vacuous guarantees but has not been systematically studied in offline RL. We analyse the problem under uniform coverage and sparse single-concentrability assumptions. While Least Square Value Iteration (LSVI), a standard approach for robust offline RL, performs well under uniform coverage, we show that integrating sparsity into LSVI is unnatural, and its analysis may break down due to overly pessimistic bonuses. To overcome this, we propose actor-critic methods with sparse robust estimator oracles, which avoid the use of pointwise pessimistic bonuses and provide the first non-vacuous guarantees for sparse offline RL under single-policy concentrability coverage. Moreover, we extend our results to the contaminated setting and show that our algorithm remains robust under strong contamination. Our results provide the first non-vacuous guarantees in high-dimensional sparse MDPs with single-policy concentrability coverage and corruption, showing that learning a near-optimal policy remains possible in regimes where traditional robust offline RL techniques may fail.
Authors: Benedikt Fluhr
Abstract: Group convolutional layers with respect to some group $G$ are modeled by convolutions or cross-correlations with a filter, and they provide the fundamental building block for group convolutional neural networks. For entirely unconstrained filters and $G$ a non-abelian group, any hidden layer of such a network requires as many nodes as vertices in a fine enough discretization of $G$. In order to reduce the necessary number of nodes, certain constraints on filters were proposed in the literature. We propose weaker constraints retaining this benefit while also resolving an incompatibility previous constraints have for group actions with non-compact stabilizers. Moreover, we generalize previous results to group actions that are not necessarily transitive, and we weaken the common assumption that $G$ is unimodular.
Authors: Svenja Jedhoff, Elizaveta Semenova, Aura Raulo, Anne Meyer, Paul-Christian B\"urkner
Abstract: Graphs arise across diverse domains, from biology and chemistry to social and information networks, as well as in transportation and logistics. Inference on graph-structured data requires methods that are permutation-invariant, scalable across varying sizes and sparsities, and capable of capturing complex long-range dependencies, making posterior estimation on graph parameters particularly challenging. Amortized Bayesian Inference (ABI) is a simulation-based framework that employs generative neural networks to enable fast, likelihood-free posterior inference. We adapt ABI to graph data to address these challenges to perform inference on node-, edge-, and graph-level parameters. Our approach couples permutation-invariant graph encoders with flexible neural posterior estimators in a two-module pipeline: a summary network maps attributed graphs to fixed-length representations, and an inference network approximates the posterior over parameters. In this setting, several neural architectures can serve as the summary network. In this work we evaluate multiple architectures and assess their performance on controlled synthetic settings and two real-world domains - biology and logistics - in terms of recovery and calibration.
Authors: Siguang Chen, Chunli Lv, Miao Xie
Abstract: Large language models (LLMs) have become powerful and widely used systems for language understanding and generation, while multi-armed bandit (MAB) algorithms provide a principled framework for adaptive decision-making under uncertainty. This survey explores the potential at the intersection of these two fields. As we know, it is the first survey to systematically review the bidirectional interaction between large language models and multi-armed bandits at the component level. We highlight the bidirectional benefits: MAB algorithms address critical LLM challenges, spanning from pre-training to retrieval-augmented generation (RAG) and personalization. Conversely, LLMs enhance MAB systems by redefining core components such as arm definition and environment modeling, thereby improving decision-making in sequential tasks. We analyze existing LLM-enhanced bandit systems and bandit-enhanced LLM systems, providing insights into their design, methodologies, and performance. Key challenges and representative findings are identified to help guide future research. An accompanying GitHub repository that indexes relevant literature is available at https://github.com/bucky1119/Awesome-LLM-Bandit-Interaction.
URLs: https://github.com/bucky1119/Awesome-LLM-Bandit-Interaction.
Authors: Zhengchi Ma, Anru R. Zhang
Abstract: Imbalanced classification often causes standard training procedures to prioritize the majority class and perform poorly on rare but important cases. A classic and widely used remedy is to augment the minority class with synthetic samples, but two basic questions remain under-resolved: when does synthetic augmentation actually help, and how many synthetic samples should be generated? We develop a unified statistical framework for synthetic augmentation in imbalanced learning, studying models trained on imbalanced data augmented with synthetic minority samples. Our theory shows that synthetic data is not always beneficial. In a "local symmetry" regime, imbalance is not the dominant source of error, so adding synthetic samples cannot improve learning rates and can even degrade performance by amplifying generator mismatch. When augmentation can help ("local asymmetry"), the optimal synthetic size depends on generator accuracy and on whether the generator's residual mismatch is directionally aligned with the intrinsic majority-minority shift. This structure can make the best synthetic size deviate from naive full balancing. Practically, we recommend Validation-Tuned Synthetic Size (VTSS): select the synthetic size by minimizing balanced validation loss over a range centered near the fully balanced baseline, while allowing meaningful departures. Extensive simulations and real data analysis further support our findings.
Authors: Yuanhong Wu, Wei Ye, Jingyan Xu, D. Frank Hsu
Abstract: In this work, we propose to apply a new model fusion and learning paradigm, known as Combinatorial Fusion Analysis (CFA), to the field of Bitcoin price prediction. Price prediction of financial product has always been a big topic in finance, as the successful prediction of the price can yield significant profit. Every machine learning model has its own strength and weakness, which hinders progress toward robustness. CFA has been used to enhance models by leveraging rank-score characteristic (RSC) function and cognitive diversity in the combination of a moderate set of diverse and relatively well-performed models. Our method utilizes both score and rank combinations as well as other weighted combination techniques. Key metrics such as RMSE and MAPE are used to evaluate our methodology performance. Our proposal presents a notable MAPE performance of 0.19\%. The proposed method greatly improves upon individual model performance, as well as outperforms other Bitcoin price prediction models.
Authors: V\'ictor Yeste, Paolo Rosso
Abstract: Human value detection from single sentences is a sparse, imbalanced multi-label task. We study whether Schwartz higher-order (HO) categories help this setting on ValueEval'24 / ValuesML (74K English sentences) under a compute-frugal budget. Rather than proposing a new architecture, we compare direct supervised transformers, hard HO$\rightarrow$values pipelines, Presence$\rightarrow$HO$\rightarrow$values cascades, compact instruction-tuned large language models (LLMs), QLoRA, and low-cost upgrades such as threshold tuning and small ensembles. HO categories are learnable: the easiest bipolar pair, Growth vs. Self-Protection, reaches Macro-$F_1=0.58$. The most reliable gains come from calibration and ensembling: threshold tuning improves Social Focus vs. Personal Focus from $0.41$ to $0.57$ ($+0.16$), transformer soft voting lifts Growth from $0.286$ to $0.303$, and a Transformer+LLM hybrid reaches $0.353$ on Self-Protection. In contrast, hard hierarchical gating does not consistently improve the end task. Compact LLMs also underperform supervised encoders as stand-alone systems, although they sometimes add useful diversity in hybrid ensembles. Under this benchmark, the HO structure is more useful as an inductive bias than as a rigid routing rule.
Authors: Muxin Fu, Xiangyuan Xue, Yafu Li, Zefeng He, Siyuan Huang, Xiaoye Qu, Yu Cheng, Yang Yang
Abstract: Large language model (LLM)-powered multi-agent systems (MAS) demonstrate remarkable collective intelligence, wherein multi-agent memory serves as a pivotal mechanism for continual adaptation. However, existing multi-agent memory designs remain constrained by two fundamental bottlenecks: (i) memory homogenization arising from the absence of role-aware customization, and (ii) information overload induced by excessively fine-grained memory entries. To address these limitations, we propose LatentMem, a learnable multi-agent memory framework designed to customize agent-specific memories in a token-efficient manner. Specifically, LatentMem comprises an experience bank that stores raw interaction trajectories in a lightweight form, and a memory composer that synthesizes compact latent memories conditioned on retrieved experience and agent-specific contexts. Further, we introduce Latent Memory Policy Optimization (LMPO), which propagates task-level optimization signals through latent memories to the composer, encouraging it to produce compact and high-utility representations. Extensive experiments across diverse benchmarks and mainstream MAS frameworks show that LatentMem achieves a performance gain of up to $19.36$% over vanilla settings and consistently outperforms existing memory architectures, without requiring any modifications to the underlying frameworks.
Authors: Ariel Fogel, Omer Hofman, Eilon Cohen, Roman Vainshtein
Abstract: Open-weight language models are increasingly used in production settings, raising new security challenges. One prominent threat in this context is backdoor attacks, in which adversaries embed hidden behaviors in language models that activate under specific conditions. Previous work has assumed that adversaries have access to training pipelines or deployment infrastructure. We propose a novel attack surface requiring neither, which utilizes the chat template. Chat templates are executable Jinja2 programs invoked at every inference call, occupying a privileged position between user input and model processing. We show that an adversary who distributes a model with a maliciously modified template can implant an inference-time backdoor without modifying model weights, poisoning training data, or controlling runtime infrastructure. We evaluated this attack vector by constructing template backdoors targeting two objectives: degrading factual accuracy and inducing emission of attacker-controlled URLs, and applied them across eighteen models spanning seven families and four inference engines. Under triggered conditions, factual accuracy drops from 90% to 15% on average while attacker-controlled URLs are emitted with success rates exceeding 80%; benign inputs show no measurable degradation. Backdoors generalize across inference runtimes and evade all automated security scans applied by the largest open-weight distribution platform. These results establish chat templates as a reliable and currently undefended attack surface in the LLM supply chain.
Authors: Scott Thornton
Abstract: Hybrid Retrieval-Augmented Generation (RAG) pipelines combine vector similarity search with knowledge graph expansion for multi-hop reasoning. We show that this composition introduces a distinct security failure mode: a vector-retrieved "seed" chunk can pivot via entity links into sensitive graph neighborhoods, causing cross-tenant data leakage that does not occur in vector-only retrieval. We formalize this risk as Retrieval Pivot Risk (RPR) and introduce companion metrics Leakage@k, Amplification Factor, and Pivot Depth (PD) to quantify leakage magnitude and traversal structure. We present seven Retrieval Pivot Attacks that exploit the vector-to-graph boundary and show that adversarial injection is not required: naturally shared entities create cross-tenant pivot paths organically. Across a synthetic multi-tenant enterprise corpus and the Enron email corpus, the undefended hybrid pipeline exhibits high pivot risk (RPR up to 0.95) with multiple unauthorized items returned per query. Leakage consistently appears at PD=2, which we attribute to the bipartite chunk-entity topology and formalize as a proposition. We then show that enforcing authorization at a single location, the graph expansion boundary, eliminates measured leakage (RPR near 0) across both corpora, all attack variants, and label forgery rates up to 10 percent, with minimal overhead. Our results indicate the root cause is boundary enforcement, not inherently complex defenses: two individually secure retrieval components can compose into an insecure system unless authorization is re-checked at the transition point.
Authors: Bo Wang, Yuxuan Zhang, Yueqin Hu, Hanchao Hou, Kaiping Peng, Shiguang Ni
Abstract: Psychological scale refinement traditionally relies on response-based methods such as factor analysis, item response theory, and network psychometrics to optimize item composition. Although rigorous, these approaches require large samples and may be constrained by data availability and cross-cultural comparability. Recent advances in natural language processing suggest that the semantic structure of questionnaire items may encode latent construct organization, offering a complementary response-free perspective. We introduce a topic-modeling framework that operationalizes semantic latent structure for scale simplification. Items are encoded using contextual sentence embeddings and grouped via density-based clustering to discover latent semantic factors without predefining their number. Class-based term weighting derives interpretable topic representations that approximate constructs and enable merging of semantically adjacent clusters. Representative items are selected using membership criteria within an integrated reduction pipeline. We benchmarked the framework across DASS, IPIP, and EPOCH, evaluating structural recovery, internal consistency, factor congruence, correlation preservation, and reduction efficiency. The proposed method recovered coherent factor-like groupings aligned with established constructs. Selected items reduced scale length by 60.5% on average while maintaining psychometric adequacy. Simplified scales showed high concordance with original factor structures and preserved inter-factor correlations, indicating that semantic latent organization provides a response-free approximation of measurement structure. Our framework formalizes semantic structure as an inspectable front-end for scale construction and reduction. To facilitate adoption, we provide a visualization-supported tool enabling one-click semantic analysis and structured simplification.
Authors: Naveen Vakada, Kartik Hegde, Arvind Krishna Sridhar, Yinyi Guo, Erik Visser
Abstract: Long-duration audio is increasingly common in industrial and consumer settings, yet reviewing multi-hour recordings is impractical, motivating systems that answer natural-language queries with precise temporal grounding and minimal hallucination. Existing audio-language models show promise, but long-audio question answering remains difficult due to context-length limits. We introduce LongAudio-RAG (LA-RAG), a hybrid framework that grounds Large Language Model outputs in retrieved, timestamped acoustic event detections rather than raw audio. Multi-hour streams are converted into structured event records stored in an SQL database, and at inference time the system resolves natural-language time references, classifies intent, retrieves only the relevant events, and generates answers using this constrained evidence. To evaluate performance, we construct a synthetic long-audio benchmark by concatenating recordings with preserved timestamps and generating template-based question-answer pairs for detection, counting, and summarization tasks. Finally, we demonstrate the practicality of our approach by deploying it in a hybrid edge-cloud environment, where the audio grounding model runs on-device on IoT-class hardware while the LLM is hosted on a GPU-backed server. This architecture enables low-latency event extraction at the edge and high-quality language reasoning in the cloud. Experiments show that structured, event-level retrieval significantly improves accuracy compared to vanilla Retrieval-Augmented Generation (RAG) or text-to-SQL approaches.
Authors: Rong Fu, Ziming Wang, Shuo Yin, Haiyun Wei, Kun Liu, Xianda Li, Zeli Su, Simon Fong
Abstract: Emotional expression underpins natural communication and effective human-computer interaction. We present Emotion Collider (EC-Net), a hyperbolic hypergraph framework for multimodal emotion and sentiment modeling. EC-Net represents modality hierarchies using Poincare-ball embeddings and performs fusion through a hypergraph mechanism that passes messages bidirectionally between nodes and hyperedges. To sharpen class separation, contrastive learning is formulated in hyperbolic space with decoupled radial and angular objectives. High-order semantic relations across time steps and modalities are preserved via adaptive hyperedge construction. Empirical results on standard multimodal emotion benchmarks show that EC-Net produces robust, semantically coherent representations and consistently improves accuracy, particularly when modalities are partially available or contaminated by noise. These findings indicate that explicit hierarchical geometry combined with hypergraph fusion is effective for resilient multimodal affect understanding.
Authors: Petrus H. Zwart
Abstract: Deployed conformal predictors are long-lived decision infrastructure reused over finite operational windows. In practice, stakeholders care not only about marginal coverage, but also about operational quantities: how often the system commits versus defers, and what error exposure it induces when it acts. These deployment-facing quantities are not determined by coverage alone: identical calibrated thresholds can yield markedly different operational profiles depending on score geometry. We develop tools for operational certification and planning beyond coverage for split conformal prediction. First, Small-Sample Beta Correction (SSBC) inverts the exact finite-sample rank/Beta law to map a user request $(\alpha^\star,\delta)$ to a concrete calibration grid point with PAC-style semantics, yielding explicit finite-window coverage guarantees for a reused deployed rule. Second, because no distribution-free pivot exists beyond coverage, we propose Calibrate-and-Audit: an independent audit set supports certified finite-window predictive envelopes (Binomial/Beta-Binomial) for key operational quantities -- commitment frequency, deferral, and decisive error exposure -- and related metrics via linear projection, without committing to a scalar objective. Third, we give a geometric characterization of the feasibility constraints and regime boundaries induced by a fixed conformal partition, clarifying why operational quantities are coupled and how calibration navigation trades them off. The result is an operational menu rthat traces attainable operational profiles (Pareto trade-offs) and attach finite-window uncertainty envelopes to each regime. We illustrate the approach on benchmark molecular toxicity and aqueous solubility datasets.
Authors: Minh Dinh, St\'ephane Deny
Abstract: Despite the successes of deep learning in computer vision, difficulties persist in recognizing objects that have undergone group-symmetric transformations rarely seen during training$\unicode{x2013}$for example objects seen in unusual poses, scales, positions, or combinations thereof. Equivariant neural networks are a solution to the problem of generalizing across symmetric transformations, but require knowledge of transformations a priori. An alternative family of architectures proposes to learn equivariant operators in a latent space, from examples of symmetric transformations. Here, using simple datasets of rotated and translated noisy MNIST, we illustrate how such architectures can successfully be harnessed for out-of-distribution classification, thus overcoming the limitations of both traditional and equivariant networks. While conceptually enticing, we discuss challenges ahead on the path of scaling these architectures to more complex datasets. Our code is available at https://github.com/BRAIN-Aalto/equivariant_operator.
Authors: Aymen Khouja, Imen Jendoubi, Oumayma Mahjoub, Oussama Mahfoudhi, Ruan De Kock, Siddarth Singh, Claude Formanek
Abstract: The optimization of urban energy systems is crucial for the advancement of sustainable and resilient smart cities, which are becoming increasingly complex with multiple decision-making units. To address scalability and coordination concerns, Multi-Agent Reinforcement Learning (MARL) is a promising solution. This paper addresses the imperative need for comprehensive and reliable benchmarking of MARL algorithms on energy management tasks. CityLearn is used as a case study environment because it realistically simulates urban energy systems, incorporates multiple storage systems, and utilizes renewable energy sources. By doing so, our work sets a new standard for evaluation, conducting a comparative study across multiple key performance indicators (KPIs). This approach illuminates the key strengths and weaknesses of various algorithms, moving beyond traditional KPI averaging which often masks critical insights. Our experiments utilize widely accepted baselines such as Proximal Policy Optimization (PPO) and Soft Actor Critic (SAC), and encompass diverse training schemes including Decentralized Training with Decentralized Execution (DTDE) and Centralized Training with Decentralized Execution (CTDE) approaches and different neural network architectures. Our work also proposes novel KPIs that tackle real world implementation challenges such as individual building contribution and battery storage lifetime. Our findings show that DTDE consistently outperforms CTDE in both average and worst-case performance. Additionally, temporal dependency learning improved control on memory dependent KPIs such as ramping and battery usage, contributing to more sustainable battery operation. Results also reveal robustness to agent or resource removal, highlighting both the resilience and decentralizability of the learned policies.
Authors: Daniel Tamayo, I\~naki Lacunza, Paula Rivera-Hidalgo, Severino Da Dalt, Javier Aula-Blasco, Aitor Gonzalez-Agirre, Marta Villegas
Abstract: We introduce MrBERT, a family of 150M-300M parameter encoders built on the ModernBERT architecture and pre-trained on 35 languages and code. Through targeted adaptation, this model family achieves state-of-the-art results on Catalan- and Spanish-specific tasks, while establishing robust performance across specialized biomedical and legal domains. To bridge the gap between research and production, we incorporate Matryoshka Representation Learning (MRL), enabling flexible vector sizing that significantly reduces inference and storage costs. Ultimately, the MrBERT family demonstrates that modern encoder architectures can be optimized for both localized linguistic excellence and efficient, high-stakes domain specialization. We open source the complete model family on Huggingface.
Authors: Evangelia Christakopoulou, Vivekkumar Patel, Hemanth Velaga, Sandip Gaikwad, Sean Suchter, Venkat Sundaranatha
Abstract: Large-scale commercial search systems optimize for relevance to drive successful sessions that help users find what they are looking for. To maximize relevance, we leverage two complementary objectives: behavioral relevance (results users tend to click or download) and textual relevance (a result's semantic fit to the query). A persistent challenge is the scarcity of expert-provided textual relevance labels relative to abundant behavioral relevance labels. We first address this by systematically evaluating LLM configurations, finding that a specialized, fine-tuned model significantly outperforms a much larger pre-trained one in providing highly relevant labels. Using this optimal model as a force multiplier, we generate millions of textual relevance labels to overcome the data scarcity. We show that augmenting our production ranker with these textual relevance labels leads to a significant outward shift of the Pareto frontier: offline NDCG improves for behavioral relevance while simultaneously increasing for textual relevance. These offline gains were validated by a worldwide A/B test on the App Store ranker, which demonstrated a statistically significant +0.24% increase in conversion rate, with the most substantial performance gains occurring in tail queries, where the new textual relevance labels provide a robust signal in the absence of reliable behavioral relevance labels.
Authors: Meisam Mohammady, Qin Yang, Nicholas Stout, Ayesha Samreen, Han Wang, Christopher J Quinn, Yuan Hong
Abstract: Differentially Private Stochastic Gradient Descent (DP-SGD) is a cornerstone technique for ensuring privacy in deep learning, widely used in both training from scratch and fine-tuning large-scale language models. While DP-SGD predominantly relies on the Gaussian mechanism, the Laplace mechanism remains underutilized due to its reliance on L1 norm clipping. This constraint severely limits its practicality in high-dimensional models because the L1 norm of an n-dimensional gradient can be up to sqrt(n) times larger than its L2 norm. As a result, the required noise scale grows significantly with model size, leading to poor utility or untrainable models. In this work, we introduce Lap2, a new solution that enables L2 clipping for Laplace DP-SGD while preserving strong privacy guarantees. We overcome the dimensionality-driven clipping barrier by computing coordinate-wise moment bounds and applying majorization theory to construct a tight, data-independent upper bound over the full model. By exploiting the Schur-convexity of the moment accountant function, we aggregate these bounds using a carefully designed majorization set that respects the L2 clipping constraint. This yields a multivariate privacy accountant that scales gracefully with model dimension and enables the use of thousands of moments. Empirical evaluations demonstrate that our approach significantly improves the performance of Laplace DP-SGD, achieving results comparable to or better than Gaussian DP-SGD under strong privacy constraints. For instance, fine-tuning RoBERTa-base (125M parameters) on SST-2 achieves 87.88% accuracy at epsilon=0.54, outperforming Gaussian (87.16%) and standard Laplace (48.97%) under the same budget.
Authors: Omar Alterkait, C\'esar Jes\'us-Valls, Ryo Matsumoto, Patrick de Perio, Kazuhiro Terao
Abstract: Large-scale homogeneous detectors with optical readouts are widely used in particle detection, with Cherenkov and scintillator neutrino detectors as prominent examples. Analyses in experimental physics rely on high-fidelity simulators to translate sensor-level information into physical quantities of interest. This task critically depends on accurate calibration, which aligns simulation behavior with real detector data, and on tracking, which infers particle properties from optical signals. We present the first end-to-end differentiable optical particle detector simulator, enabling simultaneous calibration and reconstruction through gradient-based optimization. Our approach unifies simulation, calibration, and tracking, which are traditionally treated as separate problems, within a single differentiable framework. We demonstrate that it achieves smooth and physically meaningful gradients across all key stages of light generation, propagation, and detection while maintaining computational efficiency. We show that gradient-based calibration and reconstruction greatly simplify existing analysis pipelines while matching or surpassing the performance of conventional non-differentiable methods in both accuracy and speed. Moreover, the framework's modularity allows straightforward adaptation to diverse detector geometries and target materials, providing a flexible foundation for experiment design and optimization. The results demonstrate the readiness of this technique for adoption in current and future optical detector experiments, establishing a new paradigm for simulation and reconstruction in particle physics.
Authors: Xiaoda Xu
Abstract: We study the expected star discrepancy under a newly designed class of non-equal volume partitions. The main contributions are twofold. First, we establish a strong partition principle for the star discrepancy, showing that our newly designed non-equal volume partitions yield stratified sampling point sets with lower expected star discrepancy than classical jittered sampling. Specifically, we prove that $\mathbb{E}(D^{*}_{N}(Z)) < \mathbb{E}(D^{*}_{N}(Y))$, where $Y$ and $Z$ represent jittered sampling and our non-equal volume partition sampling, respectively. Second, we derive explicit upper bounds for the expected star discrepancy under our non-equal volume partition models, which improve upon existing bounds for jittered sampling. Our results provide a theoretical foundation for using non-equal volume partitions in high-dimensional numerical integration.
Authors: Maxwell A. Xu, Harish Haresamudram, Catherine W. Liu, Patrick Langer, Jathurshan Pradeepkumar, Wanting Mao, Sunita J. Ferns, Aradhana Verma, Jimeng Sun, Paul Schmiedmayer, Xin Liu, Daniel McDuff, Emily B. Fox, James M. Rehg
Abstract: While multimodal large language models offer a promising solution to the "black box" nature of health AI by generating interpretable reasoning traces, verifying the validity of these traces remains a critical challenge. Existing evaluation methods are either unscalable, relying on manual clinician review, or superficial, utilizing proxy metrics (e.g. QA) that fail to capture the semantic correctness of clinical logic. In this work, we introduce a reproducible framework for evaluating reasoning in ECG signals. We propose decomposing reasoning into two distinct, components: (i) Perception, the accurate identification of patterns within the raw signal, and (ii) Deduction, the logical application of domain knowledge to those patterns. To evaluate Perception, we employ an agentic framework that generates code to empirically verify the temporal structures described in the reasoning trace. To evaluate Deduction, we measure the alignment of the model's logic against a structured database of established clinical criteria in a retrieval-based approach. This dual-verification method enables the scalable assessment of "true" reasoning capabilities.
Authors: Kalliopi Kleisarchaki
Abstract: The 2026 Formula 1 technical regulations introduce a fundamental change to energy strategy: under a 50/50 internal combustion engine / battery power split with unlimited regeneration and a driver-controlled Override Mode (abbreviated MOM throughout), the optimal energy deployment policy depends not only on a driver's own state but on the hidden state of rival cars. This creates a Partially Observable Stochastic Game that cannot be solved by single-agent optimisation methods. We present a tractable two-layer inference and decision framework. The first layer is a 30-state Hidden Markov Model (HMM) that infers a probability distribution over each rival's ERS charge level, Override Mode status, and tyre degradation state from five publicly observable telemetry signals. The second layer is a Deep Q-Network (DQN) policy that takes the HMM belief state as input and selects between energy deployment strategies. We formally characterise the counter-harvest trap -- a deceptive strategy in which a car deliberately suppresses observable deployment signals to induce a rival into a failed attack -- and show that detecting it requires belief-state inference rather than reactive threshold rules. On synthetic races generated from the model's own assumptions, the HMM achieves 92.3% ERS inference accuracy (random baseline: 33.3%) and detects counter-harvest trap conditions with 95.7% recall. Pre-registration -- empirical validation begins Australian Grand Prix, 8 March 2026.
Authors: Nikhil Raghav, Md Sahidullah
Abstract: This report presents the TCG CREST system description for Track 1 (Speaker Diarization) of the DISPLACE-M challenge, focusing on naturalistic medical conversations in noisy rural-healthcare scenarios. Our study evaluates the impact of various voice activity detection (VAD) methods and advanced clustering algorithms on overall speaker diarization (SD) performance. We compare and analyze two SD frameworks: a modular pipeline utilizing SpeechBrain with ECAPA-TDNN embeddings, and a state-of-the-art (SOTA) hybrid end-to-end neural diarization system, Diarizen, built on top of a pre-trained WavLM. With these frameworks, we explore diverse clustering techniques, including agglomerative hierarchical clustering (AHC), and multiple novel variants of spectral clustering, such as SC-adapt, SC-PNA, and SC-MK. Experimental results demonstrate that the Diarizen system provides an approximate $39\%$ relative improvement in the diarization error rate (DER) on the post-evaluation analysis of Phase~I compared to the SpeechBrain baseline. Our best-performing submitted system employing the Diarizen baseline with AHC employing a median filtering with a larger context window of $29$ achieved a DER of 10.37\% on the development and 9.21\% on the evaluation sets, respectively. Our team ranked fifth out of the 11 participating teams after the Phase~I evaluation.
Authors: Harikrishnan Unnikrishnan
Abstract: Background: Accurate glottal segmentation in high-speed videoendoscopy (HSV) is essential for extracting kinematic biomarkers of laryngeal function. However, existing deep learning models often produce spurious artifacts in non-glottal frames and fail to generalize across different clinical settings. Methods: We propose a detection-gated pipeline that integrates a localizer with a segmenter. A temporal consistency wrapper ensures robustness by suppressing false positives during glottal closure and occlusion. The segmenter was trained on a limited subset of the GIRAFE dataset (600 frames), while the localizer was trained on the BAGLS training set. The in-distribution localizer provides a tight region of interest (ROI), removing geometric anatomical variations and enabling cross-dataset generalization without fine-tuning. Results: The pipeline achieved state-of-the-art performance on the GIRAFE (DSC=0.81) and BAGLS (DSC=0.85) benchmarks and demonstrated superior generalizability. Notably, the framework maintained robust cross-dataset generalization (DSC=0.77). Downstream validation on a 65-subject clinical cohort confirmed that automated kinematic features - specifically the Open Quotient and Glottal Area Waveform (GAW) - remained consistent with clinical benchmarks. The coefficient of variation (CV) of the glottal area was a significant marker for distinguishing healthy from pathological vocal function (p=0.006). Conclusions: This architecture provides a computationally efficient solution (~35 frames/s) suitable for real-time clinical use. By overcoming cross-dataset variability, this framework facilitates the standardized, large-scale extraction of clinical biomarkers across diverse endoscopy platforms. Code, trained weights, and evaluation scripts are released at https://github.com/hari-krishnan/openglottal.
Authors: Youngjun Jun, Seil Kang, Woojung Han, Seong Jae Hwang
Abstract: Video Diffusion Transformers (DiTs) have been synthesizing high-quality video with high fidelity from given text descriptions involving motion. However, understanding how Video DiTs convert motion words into video remains insufficient. Furthermore, while prior studies on interpretable saliency maps primarily target objects, motion-related behavior in Video DiTs remains largely unexplored. In this paper, we investigate concrete motion features that specify when and which object moves for a given motion concept. First, to spatially localize, we introduce GramCol, which adaptively produces per-frame saliency maps for any text concept, including both motion and non-motion. Second, we propose a motion-feature selection algorithm to obtain an Interpretable Motion-Attentive Map (IMAP) that localizes motion spatially and temporally. Our method discovers concept saliency maps without the need for any gradient calculation or parameter update. Experimentally, our method shows outstanding localization capability on the motion localization task and zero-shot video semantic segmentation, providing interpretable and clearer saliency maps for both motion and non-motion concepts.
Authors: Marcel Torne, Karl Pertsch, Homer Walke, Kyle Vedder, Suraj Nair, Brian Ichter, Allen Z. Ren, Haohuan Wang, Jiaming Tang, Kyle Stachowicz, Karan Dhabalia, Michael Equi, Quan Vuong, Jost Tobias Springenberg, Sergey Levine, Chelsea Finn, Danny Driess
Abstract: Conventionally, memory in end-to-end robotic learning involves inputting a sequence of past observations into the learned policy. However, in complex multi-stage real-world tasks, the robot's memory must represent past events at multiple levels of granularity: from long-term memory that captures abstracted semantic concepts (e.g., a robot cooking dinner should remember which stages of the recipe are already done) to short-term memory that captures recent events and compensates for occlusions (e.g., a robot remembering the object it wants to pick up once its arm occludes it). In this work, our main insight is that an effective memory architecture for long-horizon robotic control should combine multiple modalities to capture these different levels of abstraction. We introduce Multi-Scale Embodied Memory (MEM), an approach for mixed-modal long-horizon memory in robot policies. MEM combines video-based short-horizon memory, compressed via a video encoder, with text-based long-horizon memory. Together, they enable robot policies to perform tasks that span up to fifteen minutes, like cleaning up a kitchen, or preparing a grilled cheese sandwich. Additionally, we find that memory enables MEM policies to intelligently adapt manipulation strategies in-context.
Authors: Haian Jin, Rundi Wu, Tianyuan Zhang, Ruiqi Gao, Jonathan T. Barron, Noah Snavely, Aleksander Holynski
Abstract: Feed-forward transformer models have driven rapid progress in 3D vision, but state-of-the-art methods such as VGGT and $\pi^3$ have a computational cost that scales quadratically with the number of input images, making them inefficient when applied to large image collections. Sequential-reconstruction approaches reduce this cost but sacrifice reconstruction quality. We introduce ZipMap, a stateful feed-forward model that achieves linear-time, bidirectional 3D reconstruction while matching or surpassing the accuracy of quadratic-time methods. ZipMap employs test-time training layers to zip an entire image collection into a compact hidden scene state in a single forward pass, enabling reconstruction of over 700 frames in under 10 seconds on a single H100 GPU, more than $20\times$ faster than state-of-the-art methods such as VGGT. Moreover, we demonstrate the benefits of having a stateful representation in real-time scene-state querying and its extension to sequential streaming reconstruction.
Authors: Ali Shamsaddinlou
Abstract: Recent advances in vision language models (VLMs) have shown strong potential for spatial reasoning and 3D scene layout generation from open-ended language instructions. However, generating layouts that are not only semantically coherent but also feasible for interaction by embodied agents remains challenging, particularly in physically constrained indoor environments. In this paper, RoboLayout is introduced as an extension of LayoutVLM that augments the original framework with agent-aware reasoning and improved optimization stability. RoboLayout integrates explicit reachability constraints into a differentiable layout optimization process, enabling the generation of layouts that are navigable and actionable by embodied agents. Importantly, the agent abstraction is not limited to a specific robot platform and can represent diverse entities with distinct physical capabilities, such as service robots, warehouse robots, humans of different age groups, or animals, allowing environment design to be tailored to the intended agent. In addition, a local refinement stage is proposed that selectively reoptimizes problematic object placements while keeping the remainder of the scene fixed, improving convergence efficiency without increasing global optimization iterations. Overall, RoboLayout preserves the strong semantic alignment and physical plausibility of LayoutVLM while enhancing applicability to agent-centric indoor scene generation, as demonstrated by experimental results across diverse scene configurations.
Authors: Vishal Thengane, Zhaochong An, Tianjin Huang, Son Lam Phung, Abdesselam Bouzerdoum, Lu Yin, Na Zhao, Xiatian Zhu
Abstract: Incremental Few-Shot (IFS) segmentation aims to learn new categories over time from only a few annotations. Although widely studied in 2D, it remains underexplored for 3D point clouds. Existing methods suffer from catastrophic forgetting or fail to learn discriminative prototypes under sparse supervision, and often overlook a key cue: novel categories frequently appear as unlabelled background in base-training scenes. We introduce SCOPE (Scene-COntextualised Prototype Enrichment), a plug-and-play background-guided prototype enrichment framework that integrates with any prototype-based 3D segmentation method. After base training, a class-agnostic segmentation model extracts high-confidence pseudo-instances from background regions to build a prototype pool. When novel classes arrive with few labelled samples, relevant background prototypes are retrieved and fused with few-shot prototypes to form enriched representations without retraining the backbone or adding parameters. Experiments on ScanNet and S3DIS show that SCOPE achieves SOTA performance, improving novel-class IoU by up to 6.98% and 3.61%, and mean IoU by 2.25% and 1.70%, respectively, while maintaining low forgetting. Code is available https://github.com/Surrey-UP-Lab/SCOPE.