Authors: Leyang Zhang
Abstract: The linear independence of neurons plays a significant role in theoretical analysis of neural networks. Specifically, given neurons $H_1, ..., H_n: \bR^N \times \bR^d \to \bR$, we are interested in the following question: when are $\{H_1(\theta_1, \cdot), ..., H_n(\theta_n, \cdot)\}$ are linearly independent as the parameters $\theta_1, ..., \theta_n$ of these functions vary over $\bR^N$. Previous works give a complete characterization of two-layer neurons without bias, for generic smooth activation functions. In this paper, we study the problem for neurons with arbitrary layers and widths, giving a simple but complete characterization for generic analytic activation functions.
Authors: Kaushal Paneri, Michael Munje, Kailash Singh Maurya, Adith Swaminathan, Yifan Shi
Abstract: Growing scale of recommender systems require extensive tuning to respond to market dynamics and system changes. We address the challenge of tuning a large-scale ads recommendation platform with multiple continuous parameters influencing key performance indicators (KPIs). Traditional methods like open-box Monte Carlo simulators, while accurate, are computationally expensive due to the high cost of evaluating numerous parameter settings. To mitigate this, we propose a hybrid approach Simulator-Guided Importance Sampling (SGIS) that combines open-box simulation with importance sampling (IS). SGIS leverages the strengths of both techniques: it performs a coarse enumeration over the parameter space to identify promising initial settings and then uses IS to iteratively refine these settings. This approach significantly reduces computational costs while maintaining high accuracy in KPI estimation. We demonstrate the effectiveness of SGIS through simulations as well as real-world experiments, showing that it achieves substantial improvements in KPIs with lower computational overhead compared to traditional methods.
Authors: A. Yark{\i}n Y{\i}ld{\i}z, Asli Kalayci
Abstract: Medical diagnosis is a crucial task in the medical field, in terms of providing accurate classification and respective treatments. Having near-precise decisions based on correct diagnosis can affect a patient's life itself, and may extremely result in a catastrophe if not classified correctly. Several traditional machine learning (ML), such as support vector machines (SVMs) and logistic regression, and state-of-the-art tabular deep learning (DL) methods, including TabNet and TabTransformer, have been proposed and used over tabular medical datasets. Additionally, due to the superior performances, lower computational costs, and easier optimization over different tasks, ensemble methods have been used in the field more recently. They offer a powerful alternative in terms of providing successful medical decision-making processes in several diagnosis tasks. In this study, we investigated the benefits of ensemble methods, especially the Gradient Boosting Decision Tree (GBDT) algorithms in medical classification tasks over tabular data, focusing on XGBoost, CatBoost, and LightGBM. The experiments demonstrate that GBDT methods outperform traditional ML and deep neural network architectures and have the highest average rank over several benchmark tabular medical diagnosis datasets. Furthermore, they require much less computational power compared to DL models, creating the optimal methodology in terms of high performance and lower complexity.
Authors: David Krame Kadurha
Abstract: The goal of this work is to serve as a foundation for deep studies of the topology of state, action, and policy spaces in reinforcement learning. By studying these spaces from a mathematical perspective, we expect to gain more insight into how to build better algorithms to solve decision problems. Therefore, we focus on presenting the connection between the Banach fixed point theorem and the convergence of reinforcement learning algorithms, and we illustrate how the insights gained from this can practically help in designing more efficient algorithms. Before doing so, however, we first introduce relevant concepts such as metric spaces, normed spaces and Banach spaces for better understanding, before expressing the entire reinforcement learning problem in terms of Markov decision processes. This allows us to properly introduce the Banach contraction principle in a language suitable for reinforcement learning, and to write the Bellman equations in terms of operators on Banach spaces to show why reinforcement learning algorithms converge. Finally, we show how the insights gained from the mathematical study of convergence are helpful in reasoning about the best ways to make reinforcement learning algorithms more efficient.
Authors: Juan Pablo Mu\~noz, Jinjie Yuan, Nilesh Jain
Abstract: Large pre-trained models (LPMs), such as large language models, have become ubiquitous and are employed in many applications. These models are often adapted to a desired domain or downstream task through a fine-tuning stage. This paper proposes SQFT, an end-to-end solution for low-precision sparse parameter-efficient fine-tuning of LPMs, allowing for effective model manipulation in resource-constrained environments. Additionally, an innovative strategy enables the merging of sparse weights with low-rank adapters without losing sparsity and accuracy, overcoming the limitations of previous approaches. SQFT also addresses the challenge of having quantized weights and adapters with different numerical precisions, enabling merging in the desired numerical format without sacrificing accuracy. Multiple adaptation scenarios, models, and comprehensive sparsity levels demonstrate the effectiveness of SQFT. Models and code are available at https://github.com/IntelLabs/Hardware-Aware-Automated-Machine-Learning.
URLs: https://github.com/IntelLabs/Hardware-Aware-Automated-Machine-Learning.
Authors: Dengsheng Chen, Jie Hu, Xiaoming Wei, Enhua Wu
Abstract: Joint-embedding predictive architectures (JEPAs) have shown substantial promise in self-supervised representation learning, yet their application in generative modeling remains underexplored. Conversely, diffusion models have demonstrated significant efficacy in modeling arbitrary probability distributions. In this paper, we introduce Denoising with a Joint-Embedding Predictive Architecture (D-JEPA), pioneering the integration of JEPA within generative modeling. By recognizing JEPA as a form of masked image modeling, we reinterpret it as a generalized next-token prediction strategy, facilitating data generation in an auto-regressive manner. Furthermore, we incorporate diffusion loss to model the per-token probability distribution, enabling data generation in a continuous space. We also adapt flow matching loss as an alternative to diffusion loss, thereby enhancing the flexibility of D-JEPA. Empirically, with increased GFLOPs, D-JEPA consistently achieves lower FID scores with fewer training epochs, indicating its good scalability. Our base, large, and huge models outperform all previous generative models across all scales on class-conditional ImageNet benchmarks. Beyond image generation, D-JEPA is well-suited for other continuous data modeling, including video and audio.
Authors: Naman Agarwal, Xinyi Chen, Evan Dogariu, Vlad Feinberg, Daniel Suo, Peter Bartlett, Elad Hazan
Abstract: We address the challenge of efficient auto-regressive generation in sequence prediction models by introducing FutureFill: a method for fast generation that applies to any sequence prediction algorithm based on convolutional operators. Our approach reduces the generation time requirement from linear to square root relative to the context length. Additionally, FutureFill requires a prefill cache sized only by the number of tokens generated, which is smaller than the cache requirements for standard convolutional and attention-based models. We validate our theoretical findings with experimental evidence demonstrating correctness and efficiency gains in a synthetic generation task.
Authors: B\'alint Csan\'ady, L\'or\'ant Nagy, D\'aniel Boros, Iv\'an Ivkovic, D\'avid Kov\'acs, Dalma T\'oth-Lakits, L\'aszl\'o M\'arkus, Andr\'as Luk\'acs
Abstract: We present a purely deep neural network-based approach for estimating long memory parameters of time series models that incorporate the phenomenon of long-range dependence. Parameters, such as the Hurst exponent, are critical in characterizing the long-range dependence, roughness, and self-similarity of stochastic processes. The accurate and fast estimation of these parameters holds significant importance across various scientific disciplines, including finance, physics, and engineering. We harnessed efficient process generators to provide high-quality synthetic training data, enabling the training of scale-invariant 1D Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) models. Our neural models outperform conventional statistical methods, even those augmented with neural networks. The precision, speed, consistency, and robustness of our estimators are demonstrated through experiments involving fractional Brownian motion (fBm), the Autoregressive Fractionally Integrated Moving Average (ARFIMA) process, and the fractional Ornstein-Uhlenbeck (fOU) process. We believe that our work will inspire further research in the field of stochastic process modeling and parameter estimation using deep learning techniques.
Authors: Huayu Deng, Xiangming Zhu, Yunbo Wang, Xiaokang Yang
Abstract: Graph neural networks have emerged as a powerful tool for large-scale mesh-based physics simulation. Existing approaches primarily employ hierarchical, multi-scale message passing to capture long-range dependencies within the graph. However, these graph hierarchies are typically fixed and manually designed, which do not adapt to the evolving dynamics present in complex physical systems. In this paper, we introduce a novel neural network named DHMP, which learns Dynamic Hierarchies for Message Passing networks through a differentiable node selection method. The key component is the anisotropic message passing mechanism, which operates at both intra-level and inter-level interactions. Unlike existing methods, it first supports directionally non-uniform aggregation of dynamic features between adjacent nodes within each graph hierarchy. Second, it determines node selection probabilities for the next hierarchy according to different physical contexts, thereby creating more flexible message shortcuts for learning remote node relations. Our experiments demonstrate the effectiveness of DHMP, achieving 22.7% improvement on average compared to recent fixed-hierarchy message passing networks across five classic physics simulation datasets.
Authors: Changdae Oh, Yixuan Li, Kyungwoo Song, Sangdoo Yun, Dongyoon Han
Abstract: Adapting a pre-trained foundation model on downstream tasks should ensure robustness against distribution shifts without the need to retrain the whole model. Although existing weight interpolation methods are simple yet effective, we argue their static nature limits downstream performance while achieving efficiency. In this work, we propose DaWin, a training-free dynamic weight interpolation method that leverages the entropy of individual models over each unlabeled test sample to assess model expertise, and compute per-sample interpolation coefficients dynamically. Unlike previous works that typically rely on additional training to learn such coefficients, our approach requires no training. Then, we propose a mixture modeling approach that greatly reduces inference overhead raised by dynamic interpolation. We validate DaWin on the large-scale visual recognition benchmarks, spanning 14 tasks across robust fine-tuning -- ImageNet and derived five distribution shift benchmarks -- and multi-task learning with eight classification tasks. Results demonstrate that DaWin achieves significant performance gain in considered settings, with minimal computational overhead. We further discuss DaWin's analytic behavior to explain its empirical success.
Authors: Jaemoo Choi, Yongxin Chen, Jaewoong Choi
Abstract: Optimal Transport (OT) theory investigates the cost-minimizing transport map that moves a source distribution to a target distribution. Recently, several approaches have emerged for learning the optimal transport map for a given cost function using neural networks. We refer to these approaches as the OT Map. OT Map provides a powerful tool for diverse machine learning tasks, such as generative modeling and unpaired image-to-image translation. However, existing methods that utilize max-min optimization often experience training instability and sensitivity to hyperparameters. In this paper, we propose a novel method to improve stability and achieve a better approximation of the OT Map by exploiting displacement interpolation, dubbed Displacement Interpolation Optimal Transport Model (DIOTM). We derive the dual formulation of displacement interpolation at specific time $t$ and prove how these dual problems are related across time. This result allows us to utilize the entire trajectory of displacement interpolation in learning the OT Map. Our method improves the training stability and achieves superior results in estimating optimal transport maps. We demonstrate that DIOTM outperforms existing OT-based models on image-to-image translation tasks.
Authors: Xishun Liao, Yifan Liu, Chenchen Kuai, Haoxuan Ma, Yueshuai He, Shangqing Cao, Chris Stanford, Jiaqi Ma
Abstract: Understanding human mobility patterns is crucial for urban planning, transportation management, and public health. This study tackles two primary challenges in the field: the reliance on trajectory data, which often fails to capture the semantic interdependencies of activities, and the inherent incompleteness of real-world trajectory data. We have developed a model that reconstructs and learns human mobility patterns by focusing on semantic activity chains. We introduce a semi-supervised iterative transfer learning algorithm to adapt models to diverse geographical contexts and address data scarcity. Our model is validated using comprehensive datasets from the United States, where it effectively reconstructs activity chains and generates high-quality synthetic mobility data, achieving a low Jensen-Shannon Divergence (JSD) value of 0.001, indicating a close similarity between synthetic and real data. Additionally, sparse GPS data from Egypt is used to evaluate the transfer learning algorithm, demonstrating successful adaptation of US mobility patterns to Egyptian contexts, achieving a 64\% of increase in similarity, i.e., a JSD reduction from 0.09 to 0.03. This mobility reconstruction model and the associated transfer learning algorithm show significant potential for global human mobility modeling studies, enabling policymakers and researchers to design more effective and culturally tailored transportation solutions.
Authors: Muhammad Asif Khan, Ridha Hamila, Hamid Menouar
Abstract: The success of modern deep learning is attributed to two key elements: huge amounts of training data and large model sizes. Where a vast amount of data allows the model to learn more features, the large model architecture boosts the learning capability of the model. However, both these factors result in prolonged training time. In some practical applications such as edge-based learning and federated learning, limited-time budgets necessitate more efficient training methods. This paper proposes an effective technique for training arbitrary deep learning models within fixed time constraints utilizing sample importance and dynamic ranking. The proposed method is extensively evaluated in both classification and regression tasks in computer vision. The results consistently show clear gains achieved by the proposed method in improving the learning performance of various state-of-the-art deep learning models in both regression and classification tasks.
Authors: Nan Huang, Haishuai Wang, Zihuai He, Marinka Zitnik, Xiang Zhang
Abstract: Medical time series (MedTS) classification is critical for a wide range of healthcare applications such as Alzheimer's Disease diagnosis. However, its real-world deployment is severely challenged by poor generalizability due to inter- and intra-dataset heterogeneity in MedTS, including variations in channel configurations, time series lengths, and diagnostic tasks. Here, we propose FORMED, a foundation classification model that leverages a pre-trained backbone and tackles these challenges through re-purposing. FORMED integrates the general representation learning enabled by the backbone foundation model and the medical domain knowledge gained on a curated cohort of MedTS datasets. FORMED can adapt seamlessly to unseen MedTS datasets, regardless of the number of channels, sample lengths, or medical tasks. Experimental results show that, without any task-specific adaptation, the repurposed FORMED achieves performance that is competitive with, and often superior to, 11 baseline models trained specifically for each dataset. Furthermore, FORMED can effectively adapt to entirely new, unseen datasets, with lightweight parameter updates, consistently outperforming baselines. Our results highlight FORMED as a versatile and scalable model for a wide range of MedTS classification tasks, positioning it as a strong foundation model for future research in MedTS analysis.
Authors: Ying Liu, Lihong Liu, Cai Xu, Xiangyu Song, Ziyu Guan, Wei Zhao
Abstract: Multi-view learning methods often focus on improving decision accuracy, while neglecting the decision uncertainty, limiting their suitability for safety-critical applications. To mitigate this, researchers propose trusted multi-view learning methods that estimate classification probabilities and uncertainty by learning the class distributions for each instance. However, these methods assume that the data from each view can effectively differentiate all categories, ignoring the semantic vagueness phenomenon in real-world multi-view data. Our findings demonstrate that this phenomenon significantly suppresses the learning of view-specific evidence in existing methods. We propose a Consistent and Complementary-aware trusted Multi-view Learning (CCML) method to solve this problem. We first construct view opinions using evidential deep neural networks, which consist of belief mass vectors and uncertainty estimates. Next, we dynamically decouple the consistent and complementary evidence. The consistent evidence is derived from the shared portions across all views, while the complementary evidence is obtained by averaging the differing portions across all views. We ensure that the opinion constructed from the consistent evidence strictly aligns with the ground-truth category. For the opinion constructed from the complementary evidence, we allow it for potential vagueness in the evidence. We compare CCML with state-of-the-art baselines on one synthetic and six real-world datasets. The results validate the effectiveness of the dynamic evidence decoupling strategy and show that CCML significantly outperforms baselines on accuracy and reliability. The code is released at https://github.com/Lihong-Liu/CCML.
Authors: Xavier Warin
Abstract: A new Kolmogorov-Arnold network (KAN) is proposed to approximate potentially irregular functions in high dimension. We show that it outperforms multilayer perceptrons in terms of accuracy and converges faster. We also compare it with ReLU-KAN, a recently proposed network: it is more time consuming than ReLU-KAN, but more accurate.
Authors: Yanchen Luo, Junfeng Fang, Sihang Li, Zhiyuan Liu, Jiancan Wu, An Zhang, Wenjie Du, Xiang Wang
Abstract: The de novo generation of molecules with targeted properties is crucial in biology, chemistry, and drug discovery. Current generative models are limited to using single property values as conditions, struggling with complex customizations described in detailed human language. To address this, we propose the text guidance instead, and introduce TextSMOG, a new Text-guided Small Molecule Generation Approach via 3D Diffusion Model which integrates language and diffusion models for text-guided small molecule generation. This method uses textual conditions to guide molecule generation, enhancing both stability and diversity. Experimental results show TextSMOG's proficiency in capturing and utilizing information from textual descriptions, making it a powerful tool for generating 3D molecular structures in response to complex textual customizations.
Authors: Ignacio Aguilera-Martos, Andr\'es Herrera-Poyatos, Juli\'an Luengo, Francisco Herrera
Abstract: Transformers have become the leading choice in natural language processing over other deep learning architectures. This trend has also permeated the field of time series analysis, especially for long-horizon forecasting, showcasing promising results both in performance and running time. In this paper, we introduce Local Attention Mechanism (LAM), an efficient attention mechanism tailored for time series analysis. This mechanism exploits the continuity properties of time series to reduce the number of attention scores computed. We present an algorithm for implementing LAM in tensor algebra that runs in time and memory O(nlogn), significantly improving upon the O(n^2) time and memory complexity of traditional attention mechanisms. We also note the lack of proper datasets to evaluate long-horizon forecast models. Thus, we propose a novel set of datasets to improve the evaluation of models addressing long-horizon forecasting challenges. Our experimental analysis demonstrates that the vanilla transformer architecture magnified with LAM surpasses state-of-the-art models, including the vanilla attention mechanism. These results confirm the effectiveness of our approach and highlight a range of future challenges in long-sequence time series forecasting.
Authors: Jiaxiang Dong, Haixu Wu, Yuxuan Wang, Li Zhang, Jianmin Wang, Mingsheng Long
Abstract: Time series forecasting is prevalent in extensive real-world applications, such as financial analysis and energy planning. Previous studies primarily focus on time series modality, endeavoring to capture the intricate variations and dependencies inherent in time series. Beyond numerical time series data, we notice that metadata (e.g.~dataset and variate descriptions) also carries valuable information essential for forecasting, which can be used to identify the application scenario and provide more interpretable knowledge than digit sequences. Inspired by this observation, we propose a Metadata-informed Time Series Transformer (MetaTST), which incorporates multiple levels of context-specific metadata into Transformer forecasting models to enable informative time series forecasting. To tackle the unstructured nature of metadata, MetaTST formalizes them into natural languages by pre-designed templates and leverages large language models (LLMs) to encode these texts into metadata tokens as a supplement to classic series tokens, resulting in an informative embedding. Further, a Transformer encoder is employed to communicate series and metadata tokens, which can extend series representations by metadata information for more accurate forecasting. This design also allows the model to adaptively learn context-specific patterns across various scenarios, which is particularly effective in handling large-scale, diverse-scenario forecasting tasks. Experimentally, MetaTST achieves state-of-the-art compared to advanced time series models and LLM-based methods on widely acknowledged short- and long-term forecasting benchmarks, covering both single-dataset individual and multi-dataset joint training settings.
Authors: Ruifeng Ren, Zhicong Li, Yong Liu
Abstract: Transformers have been the cornerstone of current Large Language Models (LLMs); however, its linear growth in overhead during inference with respect to sequence length poses challenges for modeling long sequences. In this context, Mamba has gradually attracted attention due to its constant-level size during inference and existing empirical results have shown that it can perform comparably to Transformers in sequence modeling while offering significant savings. However, one may ask that, can Mamba always enjoy the ``free lunch"? In this paper, we focus on analyzing the expressive ability of Mamba from a theoretical standpoint. First, inspired by the connection between Mamba and linear attention, we investigate potential shortcomings of the Mamba when performing the COPY operation. Our results indicate that Mamba with constant size may encounter bottlenecks when handling COPY, while it can achieve perfect performance when the size scales linearly with sequence length. Based on this observation, we analyze Mamba's ability to tackle DP problems when equipped with Chain of Thought (CoT). Our findings suggest that to solve arbitrary DP problems, the total cost of Mamba is comparable to standard and efficient Transformers. However, similar to efficient Transformers, when facing DP problems with favorable properties such as locality, Mamba can provide savings in overhead. Our results contribute to a deeper understanding of Mamba.
Authors: Grzegorz Stefa\'nski, Pawe{\l} Daniluk, Artur Szumaczuk, Jakub Tkaczuk
Abstract: Consumer electronics used to follow the miniaturization trend described by Moore's Law. Despite increased processing power in Microcontroller Units (MCUs), MCUs used in the smallest appliances are still not capable of running even moderately big, state-of-the-art artificial neural networks (ANNs) especially in time-sensitive scenarios. In this work, we present a novel method called Scattered Online Inference (SOI) that aims to reduce the computational complexity of ANNs. SOI leverages the continuity and seasonality of time-series data and model predictions, enabling extrapolation for processing speed improvements, particularly in deeper layers. By applying compression, SOI generates more general inner partial states of ANN, allowing skipping full model recalculation at each inference.
Authors: Ching-Yun Ko, Pin-Yu Chen, Payel Das, Youssef Mroueh, Soham Dan, Georgios Kollias, Subhajit Chaudhury, Tejaswini Pedapati, Luca Daniel
Abstract: Reducing the likelihood of generating harmful and toxic output is an essential task when aligning large language models (LLMs). Existing methods mainly rely on training an external reward model (i.e., another language model) or fine-tuning the LLM using self-generated data to influence the outcome. In this paper, we show that LLMs have the capability of self-detoxification without the use of an additional reward model or re-training. We propose \textit{Self-disciplined Autoregressive Sampling (SASA)}, a lightweight controlled decoding algorithm for toxicity reduction of LLMs. SASA leverages the contextual representations from an LLM to learn linear subspaces characterizing toxic v.s. non-toxic output in analytical forms. When auto-completing a response token-by-token, SASA dynamically tracks the margin of the current output to steer the generation away from the toxic subspace, by adjusting the autoregressive sampling strategy. Evaluated on LLMs of different scale and nature, namely Llama-3.1-Instruct (8B), Llama-2 (7B), and GPT2-L models with the RealToxicityPrompts, BOLD, and AttaQ benchmarks, SASA markedly enhances the quality of the generated sentences relative to the original models and attains comparable performance to state-of-the-art detoxification techniques, significantly reducing the toxicity level by only using the LLM's internal representations.
Authors: Meng Ding, Jinhui Xu, Kaiyi Ji
Abstract: Machine Unlearning has emerged as a significant area of research, focusing on 'removing' specific subsets of data from a trained model. Fine-tuning (FT) methods have become one of the fundamental approaches for approximating unlearning, as they effectively retain model performance. However, it is consistently observed that naive FT methods struggle to forget the targeted data. In this paper, we present the first theoretical analysis of FT methods for machine unlearning within a linear regression framework, providing a deeper exploration of this phenomenon. We investigate two scenarios with distinct features and overlapping features. Our findings reveal that FT models can achieve zero remaining loss yet fail to forget the forgetting data, unlike golden models (trained from scratch without the forgetting data). This analysis reveals that naive FT methods struggle with forgetting because the pretrained model retains information about the forgetting data, and the fine-tuning process has no impact on this retained information. To address this issue, we first propose a theoretical approach to mitigate the retention of forgetting data in the pretrained model. Our analysis shows that removing the forgetting data's influence allows FT models to match the performance of the golden model. Building on this insight, we introduce a discriminative regularization term to practically reduce the unlearning loss gap between the fine-tuned model and the golden model. Our experiments on both synthetic and real-world datasets validate these theoretical insights and demonstrate the effectiveness of the proposed regularization method.
Authors: Jiawei Liu, Thanh Nguyen, Mingyue Shang, Hantian Ding, Xiaopeng Li, Yu Yu, Varun Kumar, Zijian Wang
Abstract: Large Language Models (LLMs) have recently demonstrated remarkable coding capabilities. However, assessing code generation based on well-formed properties and aligning it with developer preferences remains challenging. In this paper, we explore two key questions under the new challenge of code preference learning: (i) How do we train models to predict meaningful preferences for code? and (ii) How do human and LLM preferences align with verifiable code properties and developer code tastes? To this end, we propose CodeFavor, a framework for training pairwise code preference models from synthetic evolution data, including code commits and code critiques. To evaluate code preferences, we introduce CodePrefBench, a benchmark comprising 1364 rigorously curated code preference tasks to cover three verifiable properties-correctness, efficiency, and security-along with human preference. Our evaluation shows that CodeFavor holistically improves the accuracy of model-based code preferences by up to 28.8%. Meanwhile, CodeFavor models can match the performance of models with 6-9x more parameters while being 34x more cost-effective. We also rigorously validate the design choices in CodeFavor via a comprehensive set of controlled experiments. Furthermore, we discover the prohibitive costs and limitations of human-based code preference: despite spending 23.4 person-minutes on each task, 15.1-40.3% of tasks remain unsolved. Compared to model-based preference, human preference tends to be more accurate under the objective of code correctness, while being sub-optimal for non-functional objectives.
Authors: Simon Sinong Zhan, Qingyuan Wu, Philip Wang, Yixuan Wang, Ruochen Jiao, Chao Huang, Qi Zhu
Abstract: In this paper, we aim to tackle the limitation of the Adversarial Inverse Reinforcement Learning (AIRL) method in stochastic environments where theoretical results cannot hold and performance is degraded. To address this issue, we propose a novel method which infuses the dynamics information into the reward shaping with the theoretical guarantee for the induced optimal policy in the stochastic environments. Incorporating our novel model-enhanced rewards, we present a novel Model-Enhanced AIRL framework, which integrates transition model estimation directly into reward shaping. Furthermore, we provide a comprehensive theoretical analysis of the reward error bound and performance difference bound for our method. The experimental results in MuJoCo benchmarks show that our method can achieve superior performance in stochastic environments and competitive performance in deterministic environments, with significant improvement in sample efficiency, compared to existing baselines.
Authors: Ziyi Liu, Idan Attias, Daniel M. Roy
Abstract: We study the fundamental problem of sequential probability assignment, also known as online learning with logarithmic loss, with respect to an arbitrary, possibly nonparametric hypothesis class. Our goal is to obtain a complexity measure for the hypothesis class that characterizes the minimax regret and to determine a general, minimax optimal algorithm. Notably, the sequential $\ell_{\infty}$ entropy, extensively studied in the literature (Rakhlin and Sridharan, 2015, Bilodeau et al., 2020, Wu et al., 2023), was shown to not characterize minimax risk in general. Inspired by the seminal work of Shtarkov (1987) and Rakhlin, Sridharan, and Tewari (2010), we introduce a novel complexity measure, the \emph{contextual Shtarkov sum}, corresponding to the Shtarkov sum after projection onto a multiary context tree, and show that the worst case log contextual Shtarkov sum equals the minimax regret. Using the contextual Shtarkov sum, we derive the minimax optimal strategy, dubbed \emph{contextual Normalized Maximum Likelihood} (cNML). Our results hold for sequential experts, beyond binary labels, which are settings rarely considered in prior work. To illustrate the utility of this characterization, we provide a short proof of a new regret upper bound in terms of sequential $\ell_{\infty}$ entropy, unifying and sharpening state-of-the-art bounds by Bilodeau et al. (2020) and Wu et al. (2023).
Authors: Teresa Salazar, Helder Ara\'ujo, Alberto Cano, Pedro Henriques Abreu
Abstract: Group fairness in machine learning is a critical area of research focused on achieving equitable outcomes across different groups defined by sensitive attributes such as race or gender. Federated learning, a decentralized approach to training machine learning models across multiple devices or organizations without sharing raw data, amplifies the need for fairness due to the heterogeneous data distributions across clients, which can exacerbate biases. The intersection of federated learning and group fairness has attracted significant interest, with 47 research works specifically dedicated to addressing this issue. However, no dedicated survey has focused comprehensively on group fairness in federated learning. In this work, we present an in-depth survey on this topic, addressing the critical challenges and reviewing related works in the field. We create a novel taxonomy of these approaches based on key criteria such as data partitioning, location, and applied strategies. Additionally, we explore broader concerns related to this problem and investigate how different approaches handle the complexities of various sensitive groups and their intersections. Finally, we review the datasets and applications commonly used in current research. We conclude by highlighting key areas for future research, emphasizing the need for more methods to address the complexities of achieving group fairness in federated systems.
Authors: Kaleb D. Ruscitti, Leland McInnes
Abstract: We propose an improvement to the Mapper algorithm that removes the assumption of a single resolution scale across semantic space, and improves the robustness of the results under change of parameters. This eases parameter selection, especially for datasets with highly variable local density in the Morse function $f$ used for Mapper. This is achieved by incorporating this density into the choice of cover for Mapper. Furthermore, we prove that for covers with some natural hypotheses, the graph output by Mapper still converges in bottleneck distance to the Reeb graph of the Rips complex of the data, but captures more topological features than when using the usual Mapper cover. Finally, we discuss implementation details, and include the results of computational experiments. We also provide an accompanying reference implementation.
Authors: Michael Ibrahim, Heraldo Rozas, Nagi Gebraeel, Weijun Xie
Abstract: The training of classification models for fault diagnosis tasks using geographically dispersed data is a crucial task for original parts manufacturers (OEMs) seeking to provide long-term service contracts (LTSCs) to their customers. Due to privacy and bandwidth constraints, such models must be trained in a federated fashion. Moreover, due to harsh industrial settings the data often suffers from feature and label uncertainty. Therefore, we study the problem of training a distributionally robust (DR) support vector machine (SVM) in a federated fashion over a network comprised of a central server and $G$ clients without sharing data. We consider the setting where the local data of each client $g$ is sampled from a unique true distribution $\mathbb{P}_g$, and the clients can only communicate with the central server. We propose a novel Mixture of Wasserstein Balls (MoWB) ambiguity set that relies on local Wasserstein balls centered at the empirical distribution of the data at each client. We study theoretical aspects of the proposed ambiguity set, deriving its out-of-sample performance guarantees and demonstrating that it naturally allows for the separability of the DR problem. Subsequently, we propose two distributed optimization algorithms for training the global FDR-SVM: i) a subgradient method-based algorithm, and ii) an alternating direction method of multipliers (ADMM)-based algorithm. We derive the optimization problems to be solved by each client and provide closed-form expressions for the computations performed by the central server during each iteration for both algorithms. Finally, we thoroughly examine the performance of the proposed algorithms in a series of numerical experiments utilizing both simulation data and popular real-world datasets.
Authors: Xinwei Zhang, Zhiqi Bu, Borja Balle, Mingyi Hong, Meisam Razaviyayn, Vahab Mirrokni
Abstract: Differential privacy (DP) offers a robust framework for safeguarding individual data privacy. To utilize DP in training modern machine learning models, differentially private optimizers have been widely used in recent years. A popular approach to privatize an optimizer is to clip the individual gradients and add sufficiently large noise to the clipped gradient. This approach led to the development of DP optimizers that have comparable performance with their non-private counterparts in fine-tuning tasks or in tasks with a small number of training parameters. However, a significant performance drop is observed when these optimizers are applied to large-scale training. This degradation stems from the substantial noise injection required to maintain DP, which disrupts the optimizer's dynamics. This paper introduces DiSK, a novel framework designed to significantly enhance the performance of DP optimizers. DiSK employs Kalman filtering, a technique drawn from control and signal processing, to effectively denoise privatized gradients and generate progressively refined gradient estimations. To ensure practicality for large-scale training, we simplify the Kalman filtering process, minimizing its memory and computational demands. We establish theoretical privacy-utility trade-off guarantees for DiSK, and demonstrate provable improvements over standard DP optimizers like DPSGD in terms of iteration complexity upper-bound. Extensive experiments across diverse tasks, including vision tasks such as CIFAR-100 and ImageNet-1k and language fine-tuning tasks such as GLUE, E2E, and DART, validate the effectiveness of DiSK. The results showcase its ability to significantly improve the performance of DP optimizers, surpassing state-of-the-art results under the same privacy constraints on several benchmarks.
Authors: Fabian Akkerman, Nils Knofius, Matthieu van der Heijden, Martijn Mes
Abstract: This paper investigates dual sourcing problems with supply mode dependent failure rates, particularly relevant in managing spare parts for downtime-critical assets. To enhance resilience, businesses increasingly adopt dual sourcing strategies using both conventional and additive manufacturing techniques. This paper explores how these strategies can optimise sourcing by addressing variations in part properties and failure rates. A significant challenge is the distinct failure characteristics of parts produced by these methods, which influence future demand. To tackle this, we propose a new iterative heuristic and several reinforcement learning techniques combined with an endogenous parameterised learning (EPL) approach. This EPL approach - compatible with any learning method - allows a single policy to handle various input parameters for multiple items. In a stylised setting, our best policy achieves an average optimality gap of 0.4%. In a case study within the energy sector, our policies outperform the baseline in 91.1% of instances, yielding average cost savings up to 22.6%.
Authors: Yang Li, Junier Oliva
Abstract: Many real-world situations allow for the acquisition of additional relevant information when making decisions with limited or uncertain data. However, traditional RL approaches either require all features to be acquired beforehand (e.g. in a MDP) or regard part of them as missing data that cannot be acquired (e.g. in a POMDP). In this work, we consider RL models that may actively acquire features from the environment to improve the decision quality and certainty, while automatically balancing the cost of feature acquisition process and the reward of task decision process. We propose the Active-Acquisition POMDP and identify two types of the acquisition process for different application domains. In order to assist the agent in the actively-acquired partially-observed environment and alleviate the exploration-exploitation dilemma, we develop a model-based approach, where a deep generative model is utilized to capture the dependencies of the features and impute the unobserved features. The imputations essentially represent the beliefs of the agent. Equipped with the dynamics model, we develop hierarchical RL algorithms to resolve both types of the AA-POMDPs. Empirical results demonstrate that our approach achieves considerably better performance than existing POMDP-RL solutions.
Authors: Yiming Zhang, Athul Paul Jacob, Vivian Lai, Daniel Fried, Daphne Ippolito
Abstract: Chess has long been a testbed for AI's quest to match human intelligence, and in recent years, chess AI systems have surpassed the strongest humans at the game. However, these systems are not human-aligned; they are unable to match the skill levels of all human partners or model human-like behaviors beyond piece movement. In this paper, we introduce Allie, a chess-playing AI designed to bridge the gap between artificial and human intelligence in this classic game. Allie is trained on log sequences of real chess games to model the behaviors of human chess players across the skill spectrum, including non-move behaviors such as pondering times and resignations In offline evaluations, we find that Allie exhibits humanlike behavior: it outperforms the existing state-of-the-art in human chess move prediction and "ponders" at critical positions. The model learns to reliably assign reward at each game state, which can be used at inference as a reward function in a novel time-adaptive Monte-Carlo tree search (MCTS) procedure, where the amount of search depends on how long humans would think in the same positions. Adaptive search enables remarkable skill calibration; in a large-scale online evaluation against players with ratings from 1000 to 2600 Elo, our adaptive search method leads to a skill gap of only 49 Elo on average, substantially outperforming search-free and standard MCTS baselines. Against grandmaster-level (2500 Elo) opponents, Allie with adaptive search exhibits the strength of a fellow grandmaster, all while learning exclusively from humans.
Authors: Ying-Chun Lin, Jennifer Neville
Abstract: Graphs model complex relationships between entities, with nodes and edges capturing intricate connections. Node representation learning involves transforming nodes into low-dimensional embeddings. These embeddings are typically used as features for downstream tasks. Therefore, their quality has a significant impact on task performance. Existing approaches for node representation learning span (semi-)supervised, unsupervised, and self-supervised paradigms. In graph domains, (semi-)supervised learning often only optimizes models based on class labels, neglecting other abundant graph signals, which limits generalization. While self-supervised or unsupervised learning produces representations that better capture underlying graph signals, the usefulness of these captured signals for downstream target tasks can vary. To bridge this gap, we introduce Target-Aware Contrastive Learning (Target-aware CL) which aims to enhance target task performance by maximizing the mutual information between the target task and node representations with a self-supervised learning process. This is achieved through a sampling function, XGBoost Sampler (XGSampler), to sample proper positive examples for the proposed Target-Aware Contrastive Loss (XTCL). By minimizing XTCL, Target-aware CL increases the mutual information between the target task and node representations, such that model generalization is improved. Additionally, XGSampler enhances the interpretability of each signal by showing the weights for sampling the proper positive examples. We show experimentally that XTCL significantly improves the performance on two target tasks: node classification and link prediction tasks, compared to state-of-the-art models.
Authors: Yang Li, Junier Oliva
Abstract: Human agents routinely reason on instances with incomplete and muddied data (and weigh the cost of obtaining further features). In contrast, much of ML is devoted to the unrealistic, sterile environment where all features are observed and further information on an instance is obviated. Here we extend past static ML and develop an active feature acquisition (AFA) framework that interacts with the environment to obtain new information on-the-fly and can: 1) make inferences on an instance in the face of incomplete features, 2) determine a plan for feature acquisitions to obtain additional information on the instance at hand. We build our AFA framework on a backbone of understanding the information and conditional dependencies that are present in the data. First, we show how to build generative models that can capture dependencies over arbitrary subsets of features and employ these models for acquisitions in a greedy scheme. After, we show that it is possible to guide the training of RL agents for AFA via side-information and auxiliary rewards stemming from our generative models. We also examine two important factors for deploying AFA models in real-world scenarios, namely interpretability and robustness. Extensive experiments demonstrate the state-of-the-art performance of our AFA framework.
Authors: Branislav Kveton, Boris Oreshkin, Youngsuk Park, Aniket Deshmukh, Rui Song
Abstract: Posterior sampling in contextual bandits with a Gaussian prior can be implemented exactly or approximately using the Laplace approximation. The Gaussian prior is computationally efficient but it cannot describe complex distributions. In this work, we propose approximate posterior sampling algorithms for contextual bandits with a diffusion model prior. The key idea is to sample from a chain of approximate conditional posteriors, one for each stage of the reverse process, which are estimated in a closed form using the Laplace approximation. Our approximations are motivated by posterior sampling with a Gaussian prior, and inherit its simplicity and efficiency. They are asymptotically consistent and perform well empirically on a variety of contextual bandit problems.
Authors: Edoardo Urettini, Daniele Atzeni, Reshawn J. Ramjattan, Antonio Carta
Abstract: Despite their popularity, deep neural networks (DNNs) applied to time series forecasting often fail to beat simpler statistical models. One of the main causes of this suboptimal performance is the data non-stationarity present in many processes. In particular, changes in the mean and variance of the input data can disrupt the predictive capability of a DNN. In this paper, we first show how DNN forecasting models fail in simple non-stationary settings. We then introduce GAS-Norm, a novel methodology for adaptive time series normalization and forecasting based on the combination of a Generalized Autoregressive Score (GAS) model and a Deep Neural Network. The GAS approach encompasses a score-driven family of models that estimate the mean and variance at each new observation, providing updated statistics to normalize the input data of the deep model. The output of the DNN is eventually denormalized using the statistics forecasted by the GAS model, resulting in a hybrid approach that leverages the strengths of both statistical modeling and deep learning. The adaptive normalization improves the performance of the model in non-stationary settings. The proposed approach is model-agnostic and can be applied to any DNN forecasting model. To empirically validate our proposal, we first compare GAS-Norm with other state-of-the-art normalization methods. We then combine it with state-of-the-art DNN forecasting models and test them on real-world datasets from the Monash open-access forecasting repository. Results show that deep forecasting models improve their performance in 21 out of 25 settings when combined with GAS-Norm compared to other normalization methods.
Authors: Tianyi Wei, Shu Yang, Davoud Ataee Tarzanagh, Jingxuan Bao, Jia Xu, Patryk Orzechowski, Joost B. Wagenaar, Qi Long, Li Shen
Abstract: Alzheimer's disease (AD) is a complex neurodegenerative disorder that affects millions of people worldwide. Due to the heterogeneous nature of AD, its diagnosis and treatment pose critical challenges. Consequently, there is a growing research interest in identifying homogeneous AD subtypes that can assist in addressing these challenges in recent years. In this study, we aim to identify subtypes of AD that represent distinctive clinical features and underlying pathology by utilizing unsupervised clustering with graph diffusion and similarity learning. We adopted SIMLR, a multi-kernel similarity learning framework, and graph diffusion to perform clustering on a group of 829 patients with AD and mild cognitive impairment (MCI, a prodromal stage of AD) based on their cortical thickness measurements extracted from magnetic resonance imaging (MRI) scans. Although the clustering approach we utilized has not been explored for the task of AD subtyping before, it demonstrated significantly better performance than several commonly used clustering methods. Specifically, we showed the power of graph diffusion in reducing the effects of noise in the subtype detection. Our results revealed five subtypes that differed remarkably in their biomarkers, cognitive status, and some other clinical features. To evaluate the resultant subtypes further, a genetic association study was carried out and successfully identified potential genetic underpinnings of different AD subtypes. Our source code is available at: https://github.com/PennShenLab/AD-SIMLR.
Authors: T. Konstantin Rusch, Daniela Rus
Abstract: We propose Linear Oscillatory State-Space models (LinOSS) for efficiently learning on long sequences. Inspired by cortical dynamics of biological neural networks, we base our proposed LinOSS model on a system of forced harmonic oscillators. A stable discretization, integrated over time using fast associative parallel scans, yields the proposed state-space model. We prove that LinOSS produces stable dynamics only requiring nonnegative diagonal state matrix. This is in stark contrast to many previous state-space models relying heavily on restrictive parameterizations. Moreover, we rigorously show that LinOSS is universal, i.e., it can approximate any continuous and causal operator mapping between time-varying functions, to desired accuracy. In addition, we show that an implicit-explicit discretization of LinOSS perfectly conserves the symmetry of time reversibility of the underlying dynamics. Together, these properties enable efficient modeling of long-range interactions, while ensuring stable and accurate long-horizon forecasting. Finally, our empirical results, spanning a wide range of time-series tasks from mid-range to very long-range classification and regression, as well as long-horizon forecasting, demonstrate that our proposed LinOSS model consistently outperforms state-of-the-art sequence models. Notably, LinOSS outperforms Mamba by nearly 2x and LRU by 2.5x on a sequence modeling task with sequences of length 50k.
Authors: Jean-S\'ebastien Giroux, Simon-Philippe Breton, Julie Carreau
Abstract: As climate change intensifies, the shift to cleaner energy sources becomes increasingly urgent. With wind energy production set to accelerate, reliable wind probabilistic forecasts are essential to ensure its efficient use. However, since numerical weather prediction models are computationally expensive, probabilistic forecasts are produced at resolutions too coarse to capture all mesoscale wind behaviors. Statistical downscaling, typically applied to enchance the resolution of climate model simulations, presents a viable solution with lower computational costs by learning a mapping from low-resolution (LR) variables to high-resolution (HR) meteorological variables. Leveraging deep learning, we evaluate a downscaling model based on a state-of-the-art U-Net architecture, applied to an ensemble member from a coarse-scale probabilistic forecast of wind velocity. The architecture is modified to incorporate (1) a learned grid alignment strategy to resolve LR-HR grid mismatches and (2) a processing module for multi-level atmospheric predictors. To extend the downscaling model's applicability from fixed spatial domains to the entire Canadian region, we assess a transfer learning approach. Our results show that the learned grid alignment strategy performs as well as conventional pre-processing interpolation steps and that LR wind speed at multiple levels is sufficient as a predictor, enabling a more compact architecture. Additionally, they suggest that extending to new spatial domains using transfer learning is promising, and that downscaled wind velocities demonstrate potential in improving the detection of wind power ramps, a critical phenomenon for wind energy.
Authors: Wenquan Dong, Songyan Zhu, Jian Xu, Casey M. Ryan, Man Chen, Jingya Zeng, Hao Yu, Congfeng Cao, Jiancheng Shi
Abstract: Gross Primary Productivity (GPP), the amount of carbon plants fixed by photosynthesis, is pivotal for understanding the global carbon cycle and ecosystem functioning. Process-based models built on the knowledge of ecological processes are susceptible to biases stemming from their assumptions and approximations. These limitations potentially result in considerable uncertainties in global GPP estimation, which may pose significant challenges to our Net Zero goals. This study presents UFLUX v2.0, a process-informed model that integrates state-of-art ecological knowledge and advanced machine learning techniques to reduce uncertainties in GPP estimation by learning the biases between process-based models and eddy covariance (EC) measurements. In our findings, UFLUX v2.0 demonstrated a substantial improvement in model accuracy, achieving an R^2 of 0.79 with a reduced RMSE of 1.60 g C m^-2 d^-1, compared to the process-based model's R^2 of 0.51 and RMSE of 3.09 g C m^-2 d^-1. Our global GPP distribution analysis indicates that while UFLUX v2.0 and the process-based model achieved similar global total GPP (137.47 Pg C and 132.23 Pg C, respectively), they exhibited large differences in spatial distribution, particularly in latitudinal gradients. These differences are very likely due to systematic biases in the process-based model and differing sensitivities to climate and environmental conditions. This study offers improved adaptability for GPP modelling across diverse ecosystems, and further enhances our understanding of global carbon cycles and its responses to environmental changes.
Authors: Elie Attias, Cengiz Pehlevan, Dina Obeid
Abstract: Convolutional Neural Networks (CNNs) excel in many visual tasks, but they tend to be sensitive to slight input perturbations that are imperceptible to the human eye, often resulting in task failures. Recent studies indicate that training CNNs with regularizers that promote brain-like representations, using neural recordings, can improve model robustness. However, the requirement to use neural data severely restricts the utility of these methods. Is it possible to develop regularizers that mimic the computational function of neural regularizers without the need for neural recordings, thereby expanding the usability and effectiveness of these techniques? In this work, we inspect a neural regularizer introduced in Li et al. (2019) to extract its underlying strength. The regularizer uses neural representational similarities, which we find also correlate with pixel similarities. Motivated by this finding, we introduce a new regularizer that retains the essence of the original but is computed using image pixel similarities, eliminating the need for neural recordings. We show that our regularization method 1) significantly increases model robustness to a range of black box attacks on various datasets and 2) is computationally inexpensive and relies only on original datasets. Our work explores how biologically motivated loss functions can be used to drive the performance of artificial neural networks.
Authors: Amir Eskandari, Aman Anand, Drishti Sharma, Farhana Zulkernine
Abstract: In various applications, the multivariate time series often suffers from missing data. This issue can significantly disrupt systems that rely on the data. Spatial and temporal dependencies can be leveraged to impute the missing samples. Existing imputation methods often ignore dynamic changes in spatial dependencies. We propose a Spatial Dynamic Aware Graph Recurrent Imputation Network (SDA-GRIN) which is capable of capturing dynamic changes in spatial dependencies.SDA-GRIN leverages a multi-head attention mechanism to adapt graph structures with time. SDA-GRIN models multivariate time series as a sequence of temporal graphs and uses a recurrent message-passing architecture for imputation. We evaluate SDA-GRIN on four real-world datasets: SDA-GRIN improves MSE by 9.51% for the AQI and 9.40% for AQI-36. On the PEMS-BAY dataset, it achieves a 1.94% improvement in MSE. Detailed ablation study demonstrates the effect of window sizes and missing data on the performance of the method. Project page:https://ameskandari.github.io/sda-grin/
Authors: Gang Li, Wendi Yu, Yao Yao, Wei Tong, Yingbin Liang, Qihang Lin, Tianbao Yang
Abstract: In the real world, a learning-enabled system usually undergoes multiple cycles of model development to enhance the system's ability to handle difficult or emerging tasks. This continual model development process raises a significant issue that the model development for acquiring new or improving existing capabilities may inadvertently lose capabilities of the old model, also known as catastrophic forgetting. Existing continual learning studies focus on mitigating catastrophic forgetting by trading off performance on previous tasks and new tasks to ensure good average performance. However, they are inadequate for many applications especially in safety-critical domains, as failure to strictly preserve the performance of the old model not only introduces safety risks and uncertainties but also imposes substantial expenses in the re-improving and re-validation of existing properties. To address this issue, we introduce model developmental safety as a guarantee of a learning system such that in the model development process the new model should strictly preserve the existing protected capabilities of the old model while improving its performance on target tasks. To ensure the model developmental safety, we present a safety-centric framework by formulating the model developmental safety as data-dependent constraints. Under this framework, we study how to develop a pretrained vision-language model (aka the CLIP model) for acquiring new capabilities or improving existing capabilities of image classification. We propose an efficient constrained optimization algorithm with theoretical guarantee and use its insights to finetune a CLIP model with task-dependent heads for promoting the model developmental safety. Our experiments on improving vision perception capabilities on autonomous driving and scene recognition datasets demonstrate the efficacy of the proposed approach.
Authors: Aurick Qiao, Zhewei Yao, Samyam Rajbhandari, Yuxiong He
Abstract: LLM inference for popular enterprise use cases, such as summarization, RAG, and code-generation, typically observes orders of magnitude longer prompt lengths than generation lengths. This characteristic leads to high cost of prefill and increased response latency. In this paper, we present SwiftKV, a novel model transformation and distillation procedure specifically designed to reduce the time and cost of processing prompt tokens while preserving high quality of generated tokens. SwiftKV combines three key mechanisms: i) SingleInputKV, which prefills later layers' KV cache using a much earlier layer's output, allowing prompt tokens to skip much of the model computation, ii) AcrossKV, which merges the KV caches of neighboring layers to reduce the memory footprint and support larger batch size for higher throughput, and iii) a knowledge-preserving distillation procedure that can adapt existing LLMs for SwiftKV with minimal accuracy impact and low compute and data requirement. For Llama-3.1-8B and 70B, SwiftKV reduces the compute requirement of prefill by 50% and the memory requirement of the KV cache by 62.5% while incurring minimum quality degradation across a wide range of tasks. In the end-to-end inference serving using an optimized vLLM implementation, SwiftKV realizes up to 2x higher aggregate throughput and 60% lower time per output token. It can achieve a staggering 560 TFlops/GPU of normalized inference throughput, which translates to 16K tokens/s for Llama-3.1-70B in 16-bit precision on 4x H100 GPUs.
Authors: Hengyi Wang, Shiwei Tan, Zhiqing Hong, Desheng Zhang, Hao Wang
Abstract: Foundation Language Models (FLMs) such as BERT and its variants have achieved remarkable success in natural language processing. To date, the interpretability of FLMs has primarily relied on the attention weights in their self-attention layers. However, these attention weights only provide word-level interpretations, failing to capture higher-level structures, and are therefore lacking in readability and intuitiveness. To address this challenge, we first provide a formal definition of conceptual interpretation and then propose a variational Bayesian framework, dubbed VAriational Language Concept (VALC), to go beyond word-level interpretations and provide concept-level interpretations. Our theoretical analysis shows that our VALC finds the optimal language concepts to interpret FLM predictions. Empirical results on several real-world datasets show that our method can successfully provide conceptual interpretation for FLMs.
Authors: Sijin Chen, Omar Hagrass, Jason M. Klusowski
Abstract: Decoding strategies play a pivotal role in text generation for modern language models, yet a puzzling gap divides theory and practice. Surprisingly, strategies that should intuitively be optimal, such as Maximum a Posteriori (MAP), often perform poorly in practice. Meanwhile, popular heuristic approaches like Top-$k$ and Nucleus sampling, which employ truncation and normalization of the conditional next-token probabilities, have achieved great empirical success but lack theoretical justifications. In this paper, we propose Decoding Game, a comprehensive theoretical framework which reimagines text generation as a two-player zero-sum game between Strategist, who seeks to produce text credible in the true distribution, and Nature, who distorts the true distribution adversarially. After discussing the decomposibility of multi-step generation, we derive the optimal strategy in closed form for one-step Decoding Game. It is shown that the adversarial Nature imposes an implicit regularization on likelihood maximization, and truncation-normalization methods are first-order approximations to the optimal strategy under this regularization. Additionally, by generalizing the objective and parameters of Decoding Game, near-optimal strategies encompass diverse methods such as greedy search, temperature scaling, and hybrids thereof. Numerical experiments are conducted to complement our theoretical analysis.
Authors: Ann Huang, Satpreet H. Singh, Kanaka Rajan
Abstract: Task-trained recurrent neural networks (RNNs) are versatile models of dynamical processes widely used in machine learning and neuroscience. While RNNs are easily trained to perform a wide range of tasks, the nature and extent of the degeneracy in the resultant solutions (i.e., the variability across trained RNNs) remain poorly understood. Here, we provide a unified framework for analyzing degeneracy across three levels: behavior, neural dynamics, and weight space. We analyzed RNNs trained on diverse tasks across machine learning and neuroscience domains, including N-bit flip-flop, sine wave generation, delayed discrimination, and path integration. Our key finding is that the variability across RNN solutions, quantified on the basis of neural dynamics and trained weights, depends primarily on network capacity and task characteristics such as complexity. We introduce information-theoretic measures to quantify task complexity and demonstrate that increasing task complexity consistently reduces degeneracy in neural dynamics and generalization behavior while increasing degeneracy in weight space. These relationships hold across diverse tasks and can be used to control the degeneracy of the solution space of task-trained RNNs. Furthermore, we provide several strategies to control solution degeneracy, enabling task-trained RNNs to learn more consistent or diverse solutions as needed. We envision that these insights will lead to more reliable machine learning models and could inspire strategies to better understand and control degeneracy observed in neuroscience experiments.
Authors: Jianxin Zhang, Josh Viktorov, Doosan Jung, Emily Pitler
Abstract: Neural Stochastic Differential Equations (Neural SDEs) have emerged as powerful mesh-free generative models for continuous stochastic processes, with critical applications in fields such as finance, physics, and biology. Previous state-of-the-art methods have relied on adversarial training, such as GANs, or on minimizing distance measures between processes using signature kernels. However, GANs suffer from issues like instability, mode collapse, and the need for specialized training techniques, while signature kernel-based methods require solving linear PDEs and backpropagating gradients through the solver, whose computational complexity scales quadratically with the discretization steps. In this paper, we identify a novel class of strictly proper scoring rules for comparing continuous Markov processes. This theoretical finding naturally leads to a novel approach called Finite Dimensional Matching (FDM) for training Neural SDEs. Our method leverages the Markov property of SDEs to provide a computationally efficient training objective. This scoring rule allows us to bypass the computational overhead associated with signature kernels and reduces the training complexity from $O(D^2)$ to $O(D)$ per epoch, where $D$ represents the number of discretization steps of the process. We demonstrate that FDM achieves superior performance, consistently outperforming existing methods in terms of both computational efficiency and generative quality.
Authors: Ugochukwu O. Ugwu, Michael Kirby
Abstract: This paper presents approaches to compute sparse solutions of Generalized Singular Value Problem (GSVP). The GSVP is regularized by $\ell_1$-norm and $\ell_q$-penalty for $0
Authors: Haozhe Huang, Leo Kaixuan Cheng, Kaiwen Chen, Al\'an Aspuru-Guzik
Abstract: In machine learning datasets with symmetries, the paradigm for backward compatibility with symmetry-breaking has been to relax equivariant architectural constraints, engineering extra weights to differentiate symmetries of interest. However, this process becomes increasingly over-engineered as models are geared towards specific symmetries/asymmetries hardwired of a particular set of equivariant basis functions. In this work, we introduce symmetry-cloning, a method for inducing equivariance in machine learning models. We show that general machine learning architectures (i.e., MLPs) can learn symmetries directly as a supervised learning task from group equivariant architectures and retain/break the learned symmetry for downstream tasks. This simple formulation enables machine learning models with group-agnostic architectures to capture the inductive bias of group-equivariant architectures.
Authors: Woojin Cho, Kookjin Lee, Noseong Park, Donsub Rim, Gerrit Welper
Abstract: We introduce Sparse Physics Informed Backpropagation (SPInProp), a new class of methods for accelerating backpropagation for a specialized neural network architecture called Low Rank Neural Representation (LRNR). The approach exploits the low rank structure within LRNR and constructs a reduced neural network approximation that is much smaller in size. We call the smaller network FastLRNR. We show that backpropagation of FastLRNR can be substituted for that of LRNR, enabling a significant reduction in complexity. We apply SPInProp to a physics informed neural networks framework and demonstrate how the solution of parametrized partial differential equations is accelerated.
Authors: Menglin Yang, Aosong Feng, Bo Xiong, Jihong Liu, Irwin King, Rex Ying
Abstract: Large language models (LLMs) have demonstrated remarkable performance on various tasks. However, it remains an open question whether the default Euclidean space is the most suitable choice for embedding tokens in LLMs. In this study, we first investigate the non-Euclidean characteristics of LLMs. Our findings reveal that token frequency follows a power-law distribution, with high-frequency tokens clustering near the origin and low-frequency tokens positioned farther away. Additionally, token embeddings exhibit a high degree of hyperbolicity, indicating a latent tree-like structure in the embedding space. Building on the observation, we propose to efficiently fine-tune LLMs in hyperbolic space to better exploit the underlying complex structures. However, we found that this fine-tuning in hyperbolic space cannot be achieved with naive application of exponential and logarithmic maps, when the embedding and weight matrices both reside in Euclidean space. To address this technique issue, we introduce a new method called hyperbolic low-rank efficient fine-tuning, HypLoRA, that performs low-rank adaptation directly on the hyperbolic manifold, avoiding the cancellation effect caused by the exponential and logarithmic maps, thus preserving the hyperbolic modeling capabilities. Through extensive experiments, we demonstrate that HypLoRA significantly enhances the performance of LLMs on reasoning tasks, particularly for complex reasoning problems. In particular, HypLoRA improves the performance in the complex AQuA dataset by up to 13.0%, showcasing its effectiveness in handling complex reasoning challenges
Authors: Xiaodong Lu, Leilei Sun, Tongyu Zhu, Weifeng Lv
Abstract: Temporal link prediction, aiming at predicting future interactions among entities based on historical interactions, is crucial for a series of real-world applications. Although previous methods have demonstrated the importance of relative encodings for effective temporal link prediction, computational efficiency remains a major concern in constructing these encodings. Moreover, existing relative encodings are usually constructed based on structural connectivity, where temporal information is seldom considered. To address the aforementioned issues, we first analyze existing relative encodings and unify them as a function of temporal walk matrices. This unification establishes a connection between relative encodings and temporal walk matrices, providing a more principled way for analyzing and designing relative encodings. Based on this analysis, we propose a new temporal graph neural network called TPNet, which introduces a temporal walk matrix that incorporates the time decay effect to simultaneously consider both temporal and structural information. Moreover, TPNet designs a random feature propagation mechanism with theoretical guarantees to implicitly maintain the temporal walk matrices, which improves the computation and storage efficiency. Experimental results on 13 benchmark datasets verify the effectiveness and efficiency of TPNet, where TPNet outperforms other baselines on most datasets and achieves a maximum speedup of $33.3 \times$ compared to the SOTA baseline. Our code can be found at \url{https://github.com/lxd99/TPNet}.
Authors: Yixuan Wang, Zhenwu Chen, Kangshuai Zhang, Yunduan Cui, Lei Peng
Abstract: With the sharp increase in the number of vehicles, the issue of parking difficulties has emerged as an urgent challenge that many cities need to address promptly. In the task of predicting large-scale urban parking data, existing research often lacks effective deep learning models and strategies. To tackle this challenge, this paper proposes an innovative framework for predicting large-scale urban parking graphs leveraging real-time service capabilities, aimed at improving the accuracy and efficiency of parking predictions. Specifically, we introduce a graph attention mechanism that assesses the real-time service capabilities of parking lots to construct a dynamic parking graph that accurately reflects real preferences in parking behavior. To effectively handle large-scale parking data, this study combines graph coarsening techniques with temporal convolutional autoencoders to achieve unified dimension reduction of the complex urban parking graph structure and features. Subsequently, we use a spatio-temporal graph convolutional model to make predictions based on the coarsened graph, and a pre-trained autoencoder-decoder module restores the predicted results to their original data dimensions, completing the task. Our methodology has been rigorously tested on a real dataset from parking lots in Shenzhen. The experimental results indicate that compared to traditional parking prediction models, our framework achieves improvements of 46.8\% and 30.5\% in accuracy and efficiency, respectively. Remarkably, with the expansion of the graph's scale, our framework's advantages become even more apparent, showcasing its substantial potential for solving complex urban parking dilemmas in practical scenarios.
Authors: Wen Ye, Yizhou Zhang, Wei Yang, Lumingyuan Tang, Defu Cao, Jie Cai, Yan Liu
Abstract: In recent decades, there has been substantial advances in time series models and benchmarks across various individual tasks, such as time series forecasting, classification, and anomaly detection. Meanwhile, compositional reasoning in time series prevalent in real-world applications (e.g., decision-making and compositional question answering) is in great demand. Unlike simple tasks that primarily focus on predictive accuracy, compositional reasoning emphasizes the synthesis of diverse information from both time series data and various domain knowledge, making it distinct and extremely more challenging. In this paper, we introduce Compositional Time Series Reasoning, a new task of handling intricate multistep reasoning tasks from time series data. Specifically, this new task focuses on various question instances requiring structural and compositional reasoning abilities on time series data, such as decision-making and compositional question answering. As an initial attempt to tackle this novel task, we developed TS-Reasoner, a program-aided approach that utilizes large language model (LLM) to decompose a complex task into steps of programs that leverage existing time series models and numerical subroutines. Unlike existing reasoning work which only calls off-the-shelf modules, TS-Reasoner allows for the creation of custom modules and provides greater flexibility to incorporate domain knowledge as well as user-specified constraints. We demonstrate the effectiveness of our method through a comprehensive set of experiments. These promising results indicate potential opportunities in the new task of time series reasoning and highlight the need for further research.
Authors: Xinjian Zhao, Wei Pang, Xiangru Jian, Yaoyao Xu, Chaolong Ying, Tianshu Yu
Abstract: Graph self-supervised learning (GSSL) has emerged as a compelling framework for extracting informative representations from graph-structured data without extensive reliance on labeled inputs. In this study, we introduce Graph Interplay (GIP), an innovative and versatile approach that significantly enhances the performance equipped with various existing GSSL methods. To this end, GIP advocates direct graph-level communications by introducing random inter-graph edges within standard batches. Against GIP's simplicity, we further theoretically show that \textsc{GIP} essentially performs a principled manifold separation via combining inter-graph message passing and GSSL, bringing about more structured embedding manifolds and thus benefits a series of downstream tasks. Our empirical study demonstrates that GIP surpasses the performance of prevailing GSSL methods across multiple benchmarks by significant margins, highlighting its potential as a breakthrough approach. Besides, GIP can be readily integrated into a series of GSSL methods and consistently offers additional performance gain. This advancement not only amplifies the capability of GSSL but also potentially sets the stage for a novel graph learning paradigm in a broader sense.
Authors: Fatemeh Pesaran Zadeh, Juyeon Kim, Jin-Hwa Kim, Gunhee Kim
Abstract: Large language models (LLMs) have demonstrated strong capabilities across various language tasks, notably through instruction-tuning methods. However, LLMs face challenges in visualizing complex, real-world data through charts and plots. Firstly, existing datasets rarely cover a full range of chart types, such as 3D, volumetric, and gridded charts. Secondly, supervised fine-tuning methods do not fully leverage the intricate relationships within rich datasets, including text, code, and figures. To address these challenges, we propose a hierarchical pipeline and a new dataset for chart generation. Our dataset, Text2Chart31, includes 31 unique plot types referring to the Matplotlib library, with 11.1K tuples of descriptions, code, data tables, and plots. Moreover, we introduce a reinforcement learning-based instruction tuning technique for chart generation tasks without requiring human feedback. Our experiments show that this approach significantly enhances the model performance, enabling smaller models to outperform larger open-source models and be comparable to state-of-the-art proprietary models in data visualization tasks. We make the code and dataset available at https://github.com/fatemehpesaran310/Text2Chart31.
Authors: Ruiyuan Huang, Zengfeng Huang
Abstract: Motivated by applications in online bidding and sleeping bandits, we examine the problem of contextual bandits with cross learning, where the learner observes the loss associated with the action across all possible contexts, not just the current round's context. Our focus is on a setting where losses are chosen adversarially, and contexts are sampled i.i.d. from a specific distribution. This problem was first studied by Balseiro et al. (2019), who proposed an algorithm that achieves near-optimal regret under the assumption that the context distribution is known in advance. However, this assumption is often unrealistic. To address this issue, Schneider and Zimmert (2023) recently proposed a new algorithm that achieves nearly optimal expected regret. It is well-known that expected regret can be significantly weaker than high-probability bounds. In this paper, we present a novel, in-depth analysis of their algorithm and demonstrate that it actually achieves near-optimal regret with high probability. There are steps in the original analysis by Schneider and Zimmert (2023) that lead only to an expected bound by nature. In our analysis, we introduce several new insights. Specifically, we make extensive use of the weak dependency structure between different epochs, which was overlooked in previous analyses. Additionally, standard martingale inequalities are not directly applicable, so we refine martingale inequalities to complete our analysis.
Authors: Allahdadi Fatemeh, Mahdian Toroghi Rahil, Zareian Hassan
Abstract: Query-by-example spoken term detection (QbE-STD) is typically constrained by transcribed data scarcity and language specificity. This paper introduces a novel, language-agnostic QbE-STD model leveraging image processing techniques and transformer architecture. By employing a pre-trained XLSR-53 network for feature extraction and a Hough transform for detection, our model effectively searches for user-defined spoken terms within any audio file. Experimental results across four languages demonstrate significant performance gains (19-54%) over a CNN-based baseline. While processing time is improved compared to DTW, accuracy remains inferior. Notably, our model offers the advantage of accurately counting query term repetitions within the target audio.
Authors: Tianchi Yu, Jingwei Qiu, Jiang Yang, Ivan Oseledets
Abstract: In this paper, we propose to use Sinc interpolation in the context of Kolmogorov-Arnold Networks, neural networks with learnable activation functions, which recently gained attention as alternatives to multilayer perceptron. Many different function representations have already been tried, but we show that Sinc interpolation proposes a viable alternative, since it is known in numerical analysis to represent well both smooth functions and functions with singularities. This is important not only for function approximation but also for the solutions of partial differential equations with physics-informed neural networks. Through a series of experiments, we show that SincKANs provide better results in almost all of the examples we have considered.
Authors: Anas Barakat, Souradip Chakraborty, Peihong Yu, Pratap Tokekar, Amrit Singh Bedi
Abstract: Reinforcement learning with general utilities has recently gained attention thanks to its ability to unify several problems, including imitation learning, pure exploration, and safe RL. However, prior work for solving this general problem in a unified way has mainly focused on the tabular setting. This is restrictive when considering larger state-action spaces because of the need to estimate occupancy measures during policy optimization. In this work, we address this issue and propose to approximate occupancy measures within a function approximation class using maximum likelihood estimation (MLE). We propose a simple policy gradient algorithm (PG-OMA) where an actor updates the policy parameters to maximize the general utility objective whereas a critic approximates the occupancy measure using MLE. We provide a sample complexity analysis of PG-OMA showing that our occupancy measure estimation error only scales with the dimension of our function approximation class rather than the size of the state action space. Under suitable assumptions, we establish first order stationarity and global optimality performance bounds for the proposed PG-OMA algorithm for nonconcave and concave general utilities respectively. We complement our methodological and theoretical findings with promising empirical results showing the scalability potential of our approach compared to existing tabular count-based approaches.
Authors: Swadesh Swain, Shree Singhi
Abstract: Integrated Gradients (IG) is a widely used algorithm for attributing the outputs of a deep neural network to its input features. Due to the absence of closed-form integrals for deep learning models, inaccurate Riemann Sum approximations are used to calculate IG. This often introduces undesirable errors in the form of high levels of noise, leading to false insights in the model's decision-making process. We introduce a framework, RiemannOpt, that minimizes these errors by optimizing the sample point selection for the Riemann Sum. Our algorithm is highly versatile and applicable to IG as well as its derivatives like Blur IG and Guided IG. RiemannOpt achieves up to 20% improvement in Insertion Scores. Additionally, it enables its users to curtail computational costs by up to four folds, thereby making it highly functional for constrained environments.
Authors: Charles Jones, Fabio de Sousa Ribeiro, M\'elanie Roschewitz, Daniel C. Castro, Ben Glocker
Abstract: We investigate the prominent class of fair representation learning methods for bias mitigation. Using causal reasoning to define and formalise different sources of dataset bias, we reveal important implicit assumptions inherent to these methods. We prove fundamental limitations on fair representation learning when evaluation data is drawn from the same distribution as training data and run experiments across a range of medical modalities to examine the performance of fair representation learning under distribution shifts. Our results explain apparent contradictions in the existing literature and reveal how rarely considered causal and statistical aspects of the underlying data affect the validity of fair representation learning. We raise doubts about current evaluation practices and the applicability of fair representation learning methods in performance-sensitive settings. We argue that fine-grained analysis of dataset biases should play a key role in the field moving forward.
Authors: Jun Li, Aaron Aguirre, Junior Moura, Che Liu, Lanhai Zhong, Chenxi Sun, Gari Clifford, Brandon Westover, Shenda Hong
Abstract: Artificial Intelligence (AI) has shown great promise in electrocardiogram (ECG) analysis and cardiovascular disease detection. However, developing a general AI-ECG model has been challenging due to inter-individual variability and the diversity of ECG diagnoses, limiting existing models to specific diagnostic tasks and datasets. Moreover, current AI-ECG models struggle to achieve comparable performance between single-lead and 12-lead ECGs, limiting the application of AI-ECG to portable and wearable ECG devices. To address these limitations, we introduce an ECG Foundation Model (ECGFounder), a general-purpose model that leverages real-world ECG annotations from cardiology experts to broaden the diagnostic capabilities of ECG analysis. ECGFounder is trained on over 10 million ECGs with 150 label categories from the Harvard-Emory ECG Database, enabling comprehensive cardiovascular disease diagnosis through ECG analysis. The model is designed to be both effective out-of-the-box and fine-tunable for downstream tasks, maximizing usability. More importantly, we extend its application to single-lead ECGs, enabling complex condition diagnoses and supporting various downstream tasks in mobile and remote monitoring scenarios. Experimental results demonstrate that ECGFounder achieves expert-level performance on internal validation sets for both 12-lead and single-lead ECGs, while also exhibiting strong classification performance and generalization across various diagnoses on external validation sets. When fine-tuned, ECGFounder outperforms baseline models in demographics detection, clinical event detection, and cross-modality cardiac rhythm diagnosis. The trained model and data will be publicly released upon publication through the bdsp.io. Our code is available at https://github.com/bdsp-core/ECGFounder.
Authors: Vikram S Chundawat, Pushkar Niroula, Prasanna Dhungana, Stefan Schoepf, Murari Mandal, Alexandra Brintrup
Abstract: Federated learning (FL) has enabled collaborative model training across decentralized data sources or clients. While adding new participants to a shared model does not pose great technical hurdles, the removal of a participant and their related information contained in the shared model remains a challenge. To address this problem, federated unlearning has emerged as a critical research direction, seeking to remove information from globally trained models without harming the model performance on the remaining data. Most modern federated unlearning methods use costly approaches such as the use of remaining clients data to retrain the global model or methods that would require heavy computation on client or server side. We introduce Contribution Dampening (ConDa), a framework that performs efficient unlearning by tracking down the parameters which affect the global model for each client and performs synaptic dampening on the parameters of the global model that have privacy infringing contributions from the forgetting client. Our technique does not require clients data or any kind of retraining and it does not put any computational overhead on either the client or server side. We perform experiments on multiple datasets and demonstrate that ConDa is effective to forget a client's data. In experiments conducted on the MNIST, CIFAR10, and CIFAR100 datasets, ConDa proves to be the fastest federated unlearning method, outperforming the nearest state of the art approach by at least 100x. Our emphasis is on the non-IID Federated Learning setting, which presents the greatest challenge for unlearning. Additionally, we validate ConDa's robustness through backdoor and membership inference attacks. We envision this work as a crucial component for FL in adhering to legal and ethical requirements.
Authors: Robin Frehner, Kurt Stockinger
Abstract: Anomaly detection is an important problem with applications in various domains such as fraud detection, pattern recognition or medical diagnosis. Several algorithms have been introduced using classical computing approaches. However, using quantum computing for solving anomaly detection problems in time series data is a widely unexplored research field. This paper explores the application of quantum autoencoders to time series anomaly detection. We investigate two primary techniques for classifying anomalies: (1) Analyzing the reconstruction error generated by the quantum autoencoder and (2) latent representation analysis. Our simulated experimental results, conducted across various ansaetze, demonstrate that quantum autoencoders consistently outperform classical deep learning-based autoencoders across multiple datasets. Specifically, quantum autoencoders achieve superior anomaly detection performance while utilizing 60-230 times fewer parameters and requiring five times fewer training iterations. In addition, we implement our quantum encoder on real quantum hardware. Our experimental results demonstrate that quantum autoencoders achieve anomaly detection performance on par with their simulated counterparts.
Authors: Abbas Abdolmaleki, Bilal Piot, Bobak Shahriari, Jost Tobias Springenberg, Tim Hertweck, Rishabh Joshi, Junhyuk Oh, Michael Bloesch, Thomas Lampe, Nicolas Heess, Jonas Buchli, Martin Riedmiller
Abstract: Existing preference optimization methods are mainly designed for directly learning from human feedback with the assumption that paired examples (preferred vs. dis-preferred) are available. In contrast, we propose a method that can leverage unpaired preferred or dis-preferred examples, and works even when only one type of feedback (positive or negative) is available. This flexibility allows us to apply it in scenarios with varying forms of feedback and models, including training generative language models based on human feedback as well as training policies for sequential decision-making problems, where learned (value) functions are available. Our approach builds upon the probabilistic framework introduced in (Dayan and Hinton, 1997), which proposes to use expectation-maximization (EM) to directly optimize the probability of preferred outcomes (as opposed to classic expected reward maximization). To obtain a practical algorithm, we identify and address a key limitation in current EM-based methods: when applied to preference optimization, they solely maximize the likelihood of preferred examples, while neglecting dis-preferred samples. We show how one can extend EM algorithms to explicitly incorporate dis-preferred outcomes, leading to a novel, theoretically grounded, preference optimization algorithm that offers an intuitive and versatile way to learn from both positive and negative feedback.
Authors: Sebastian Basterrech
Abstract: Concept drift typically refers to the analysis of changes in data distribution. A drift in the input data can have negative consequences on a learning predictor and the system's stability. The majority of concept drift methods emphasize the analysis of statistical changes in non-stationary data over time. In this context, we consider another perspective, where the concept drift also integrates substantial changes in the topological characteristics of the data stream. In this article, we introduce a novel framework for monitoring changes in multi-dimensional data streams. We explore a generalization of the standard concept drift focusing on the changes in the topological characteristics of the data. Our developed approach is based on persistent entropy and topology-preserving projections in a continual learning scenario. The framework operates in both unsupervised and supervised environments. To demonstrate the utility of the proposed framework, we analyze the model across three scenarios using data streams generated with MNIST samples. The obtained results reveal the potential of applying topological data analysis for shift detection and encourage further research in this area.
Authors: Xinlei Lin, Dunhui Xiao
Abstract: Numerical solving parameterised partial differential equations (P-PDEs) is highly practical yet computationally expensive, driving the development of reduced-order models (ROMs). Recently, methods that combine latent space identification techniques with deep learning algorithms (e.g., autoencoders) have shown great potential in describing the dynamical system in the lower dimensional latent space, for example, LaSDI, gLaSDI and GPLaSDI. In this paper, a new parametric latent identification of nonlinear dynamics neural networks, P-TLDINets, is introduced, which relies on a novel neural network structure based on Taylor series expansion and ResNets to learn the ODEs that govern the reduced space dynamics. During the training process, Taylor series-based Latent Dynamic Neural Networks (TLDNets) and identified equations are trained simultaneously to generate a smoother latent space. In order to facilitate the parameterised study, a $k$-nearest neighbours (KNN) method based on an inverse distance weighting (IDW) interpolation scheme is introduced to predict the identified ODE coefficients using local information. Compared to other latent dynamics identification methods based on autoencoders, P-TLDINets remain the interpretability of the model. Additionally, it circumvents the building of explicit autoencoders, avoids dependency on specific grids, and features a more lightweight structure, which is easy to train with high generalisation capability and accuracy. Also, it is capable of using different scales of meshes. P-TLDINets improve training speeds nearly hundred times compared to GPLaSDI and gLaSDI, maintaining an $L_2$ error below $2\%$ compared to high-fidelity models.
Authors: Tuan Truong, Quyen Tran, Quan Pham-Ngoc, Nhat Ho, Dinh Phung, Trung Le
Abstract: We introduce Flat Hilbert Bayesian Inference (FHBI), an algorithm designed to enhance generalization in Bayesian inference. Our approach involves an iterative two-step procedure with an adversarial functional perturbation step and a functional descent step within the reproducing kernel Hilbert spaces. This methodology is supported by a theoretical analysis that extends previous findings on generalization ability from finite-dimensional Euclidean spaces to infinite-dimensional functional spaces. To evaluate the effectiveness of FHBI, we conduct comprehensive comparisons against seven baseline methods on the VTAB-1K benchmark, which encompasses 19 diverse datasets across various domains with diverse semantics. Empirical results demonstrate that FHBI consistently outperforms the baselines by notable margins, highlighting its practical efficacy.
Authors: Md. Tarek Hasan, Mohammad Nazmush Shamael, H. M. Mutasim Billah, Arifa Akter, Md Al Emran Hossain, Sumayra Islam, Salekul Islam, Swakkhar Shatabda
Abstract: Peer review is the quality assessment of a manuscript by one or more peer experts. Papers are submitted by the authors to scientific venues, and these papers must be reviewed by peers or other authors. The meta-reviewers then gather the peer reviews, assess them, and create a meta-review and decision for each manuscript. As the number of papers submitted to these venues has grown in recent years, it becomes increasingly challenging for meta-reviewers to collect these peer evaluations on time while still maintaining the quality that is the primary goal of meta-review creation. In this paper, we address two peer review aggregation challenges a meta-reviewer faces: paper acceptance decision-making and meta-review generation. Firstly, we propose to automate the process of acceptance decision prediction by applying traditional machine learning algorithms. We use pre-trained word embedding techniques BERT to process the reviews written in natural language text. For the meta-review generation, we propose a transfer learning model based on the T5 model. Experimental results show that BERT is more effective than the other word embedding techniques, and the recommendation score is an important feature for the acceptance decision prediction. In addition, we figure out that fine-tuned T5 outperforms other inference models. Our proposed system takes peer reviews and other relevant features as input to produce a meta-review and make a judgment on whether or not the paper should be accepted. In addition, experimental results show that the acceptance decision prediction system of our task outperforms the existing models, and the meta-review generation task shows significantly improved scores compared to the existing models. For the statistical test, we utilize the Wilcoxon signed-rank test to assess whether there is a statistically significant improvement between paired observations.
Authors: Theo (Moe), Putterman, Derek Lim, Yoav Gelberg, Stefanie Jegelka, Haggai Maron
Abstract: Low-rank adaptations (LoRAs) have revolutionized the finetuning of large foundation models, enabling efficient adaptation even with limited computational resources. The resulting proliferation of LoRAs presents exciting opportunities for applying machine learning techniques that take these low-rank weights themselves as inputs. In this paper, we investigate the potential of Learning on LoRAs (LoL), a paradigm where LoRA weights serve as input to machine learning models. For instance, an LoL model that takes in LoRA weights as inputs could predict the performance of the finetuned model on downstream tasks, detect potentially harmful finetunes, or even generate novel model edits without traditional training methods. We first identify the inherent parameter symmetries of low rank decompositions of weights, which differ significantly from the parameter symmetries of standard neural networks. To efficiently process LoRA weights, we develop several symmetry-aware invariant or equivariant LoL models, using tools such as canonicalization, invariant featurization, and equivariant layers. We finetune thousands of text-to-image diffusion models and language models to collect datasets of LoRAs. In numerical experiments on these datasets, we show that our LoL architectures are capable of processing low rank weight decompositions to predict CLIP score, finetuning data attributes, finetuning data membership, and accuracy on downstream tasks.
Authors: Viet-Hoang Tran, Thieu N. Vo, An Nguyen The, Tho Tran Huu, Minh-Khoi Nguyen-Nhat, Thanh Tran, Duy-Tung Pham, Tan Minh Nguyen
Abstract: This paper systematically explores neural functional networks (NFN) for transformer architectures. NFN are specialized neural networks that treat the weights, gradients, or sparsity patterns of a deep neural network (DNN) as input data and have proven valuable for tasks such as learnable optimizers, implicit data representations, and weight editing. While NFN have been extensively developed for MLP and CNN, no prior work has addressed their design for transformers, despite the importance of transformers in modern deep learning. This paper aims to address this gap by providing a systematic study of NFN for transformers. We first determine the maximal symmetric group of the weights in a multi-head attention module as well as a necessary and sufficient condition under which two sets of hyperparameters of the multi-head attention module define the same function. We then define the weight space of transformer architectures and its associated group action, which leads to the design principles for NFN in transformers. Based on these, we introduce Transformer-NFN, an NFN that is equivariant under this group action. Additionally, we release a dataset of more than 125,000 Transformers model checkpoints trained on two datasets with two different tasks, providing a benchmark for evaluating Transformer-NFN and encouraging further research on transformer training and performance.
Authors: Thieu N. Vo, Viet-Hoang Tran, Tho Tran Huu, An Nguyen The, Thanh Tran, Minh-Khoi Nguyen-Nhat, Duy-Tung Pham, Tan Minh Nguyen
Abstract: Neural Functional Networks (NFNs) have gained increasing interest due to their wide range of applications, including extracting information from implicit representations of data, editing network weights, and evaluating policies. A key design principle of NFNs is their adherence to the permutation and scaling symmetries inherent in the connectionist structure of the input neural networks. Recent NFNs have been proposed with permutation and scaling equivariance based on either graph-based message-passing mechanisms or parameter-sharing mechanisms. However, graph-based equivariant NFNs suffer from high memory consumption and long running times. On the other hand, parameter-sharing-based NFNs built upon equivariant linear layers exhibit lower memory consumption and faster running time, yet their expressivity is limited due to the large size of the symmetric group of the input neural networks. The challenge of designing a permutation and scaling equivariant NFN that maintains low memory consumption and running time while preserving expressivity remains unresolved. In this paper, we propose a novel solution with the development of MAGEP-NFN (Monomial mAtrix Group Equivariant Polynomial NFN). Our approach follows the parameter-sharing mechanism but differs from previous works by constructing a nonlinear equivariant layer represented as a polynomial in the input weights. This polynomial formulation enables us to incorporate additional relationships between weights from different input hidden layers, enhancing the model's expressivity while keeping memory consumption and running time low, thereby addressing the aforementioned challenge. We provide empirical evidence demonstrating that MAGEP-NFN achieves competitive performance and efficiency compared to existing baselines.
Authors: Gang Liu, Michael Sun, Wojciech Matusik, Meng Jiang, Jie Chen
Abstract: While large language models (LLMs) have integrated images, adapting them to graphs remains challenging, limiting their applications in materials and drug design. This difficulty stems from the need for coherent autoregressive generation across texts and graphs. To address this, we introduce Llamole, the first multimodal LLM capable of interleaved text and graph generation, enabling molecular inverse design with retrosynthetic planning. Llamole integrates a base LLM with the Graph Diffusion Transformer and Graph Neural Networks for multi-conditional molecular generation and reaction inference within texts, while the LLM, with enhanced molecular understanding, flexibly controls activation among the different graph modules. Additionally, Llamole integrates A* search with LLM-based cost functions for efficient retrosynthetic planning. We create benchmarking datasets and conduct extensive experiments to evaluate Llamole against in-context learning and supervised fine-tuning. Llamole significantly outperforms 14 adapted LLMs across 12 metrics for controllable molecular design and retrosynthetic planning.
Authors: Dmitry Yarotsky, Maksim Velikanov
Abstract: An important open problem is the theoretically feasible acceleration of mini-batch SGD-type algorithms on quadratic problems with power-law spectrum. In the non-stochastic setting, the optimal exponent $\xi$ in the loss convergence $L_t\sim C_Lt^{-\xi}$ is double that in plain GD and is achievable using Heavy Ball (HB) with a suitable schedule; this no longer works in the presence of mini-batch noise. We address this challenge by considering first-order methods with an arbitrary fixed number $M$ of auxiliary velocity vectors (*memory-$M$ algorithms*). We first prove an equivalence between two forms of such algorithms and describe them in terms of suitable characteristic polynomials. Then we develop a general expansion of the loss in terms of signal and noise propagators. Using it, we show that losses of stationary stable memory-$M$ algorithms always retain the exponent $\xi$ of plain GD, but can have different constants $C_L$ depending on their effective learning rate that generalizes that of HB. We prove that in memory-1 algorithms we can make $C_L$ arbitrarily small while maintaining stability. As a consequence, we propose a memory-1 algorithm with a time-dependent schedule that we show heuristically and experimentally to improve the exponent $\xi$ of plain SGD.
Authors: Zi Wang, Divyam Anshumaan, Ashish Hooda, Yudong Chen, Somesh Jha
Abstract: Optimization methods are widely employed in deep learning to identify and mitigate undesired model responses. While gradient-based techniques have proven effective for image models, their application to language models is hindered by the discrete nature of the input space. This study introduces a novel optimization approach, termed the \emph{functional homotopy} method, which leverages the functional duality between model training and input generation. By constructing a series of easy-to-hard optimization problems, we iteratively solve these problems using principles derived from established homotopy methods. We apply this approach to jailbreak attack synthesis for large language models (LLMs), achieving a $20\%-30\%$ improvement in success rate over existing methods in circumventing established safe open-source models such as Llama-2 and Llama-3.
Authors: Andrea Napoli, Paul White
Abstract: Domain shifts are ubiquitous in machine learning, and can substantially degrade a model's performance when deployed to real-world data. To address this, distribution alignment methods aim to learn feature representations which are invariant across domains, by minimising the discrepancy between the distributions. However, the discrepancy estimates can be extremely noisy when training via stochastic gradient descent (SGD), and shifts in the relative proportions of different subgroups can lead to domain misalignments; these can both stifle the benefits of the method. This paper proposes to improve these estimates by inducing diversity in each sampled minibatch. This simultaneously balances the data and reduces the variance of the gradients, thereby enhancing the model's generalisation ability. We describe two options for diversity-based data samplers, based on the k-determinantal point process (k-DPP) and the k-means++ algorithm, which can function as drop-in replacements for a standard random sampler. On a real-world domain shift task of bioacoustic event detection, we show that both options 1) yield minibatches which are more representative of the full dataset; 2) reduce the distance estimation error between distributions, for a given sample size; and 3) improve out-of-distribution accuracy for two distribution alignment algorithms, as well as standard ERM.
Authors: Maria Luz Gamiz, Fernando Navas-Gomez, Rafael Nozal-Ca\~nadas, Rocio Raya-Miranda
Abstract: Studying the reliability of complex systems using machine learning techniques involves facing a series of technical and practical challenges, ranging from the intrinsic nature of the system and data to the difficulties in modeling and effectively deploying models in real-world scenarios. This study compares the effectiveness of classical statistical techniques and machine learning methods for improving complex system analysis in reliability assessments. We aim to demonstrate that classical statistical algorithms often yield more precise and interpretable results than black-box machine learning approaches in many practical applications. The evaluation is conducted using both real-world data and simulated scenarios. We report the results obtained from statistical modeling algorithms, as well as from machine learning methods including neural networks, K-nearest neighbors, and random forests.
Authors: Gilchan Park, Paul Baity, Byung-Jun Yoon, Adolfy Hoisie
Abstract: Quantum computing is rapidly evolving in both physics and computer science, offering the potential to solve complex problems and accelerate computational processes. The development of quantum chips necessitates understanding the correlations among diverse experimental conditions. Semantic networks built on scientific literature, representing meaningful relationships between concepts, have been used across various domains to identify knowledge gaps and novel concept combinations. Neural network-based approaches have shown promise in link prediction within these networks. This study proposes initializing node features using LLMs to enhance node representations for link prediction tasks in graph neural networks. LLMs can provide rich descriptions, reducing the need for manual feature creation and lowering costs. Our method, evaluated using various link prediction models on a quantum computing semantic network, demonstrated efficacy compared to traditional node embedding techniques.
Authors: Yiming Qin, Manuel Madeira, Dorina Thanou, Pascal Frossard
Abstract: Graph generation is fundamental in diverse scientific applications, due to its ability to reveal the underlying distribution of complex data, and eventually generate new, realistic data points. Despite the success of diffusion models in this domain, those face limitations in sampling efficiency and flexibility, stemming from the tight coupling between the training and sampling stages. To address this, we propose DeFoG, a novel framework using discrete flow matching for graph generation. DeFoG employs a flow-based approach that features an efficient linear interpolation noising process and a flexible denoising process based on a continuous-time Markov chain formulation. We leverage an expressive graph transformer and ensure desirable node permutation properties to respect graph symmetry. Crucially, our framework enables a disentangled design of the training and sampling stages, enabling more effective and efficient optimization of model performance. We navigate this design space by introducing several algorithmic improvements that boost the model performance, consistently surpassing existing diffusion models. We also theoretically demonstrate that, for general discrete data, discrete flow models can faithfully replicate the ground truth distribution - a result that naturally extends to graph data and reinforces DeFoG's foundations. Extensive experiments show that DeFoG achieves state-of-the-art results on synthetic and molecular datasets, improving both training and sampling efficiency over diffusion models, and excels in conditional generation on a digital pathology dataset.
Authors: Josh Alman, Hantao Yu
Abstract: The Transformer architecture is widely deployed in many popular and impactful Large Language Models. At its core is the attention mechanism for calculating correlations between pairs of tokens. Performing an attention computation takes quadratic time in the input size, and had become the time bottleneck for transformer operations. In order to circumvent this, researchers have used a variety of approaches, including designing heuristic algorithms for performing attention computations faster, and proposing alternatives to the attention mechanism which can be computed more quickly. For instance, state space models such as Mamba were designed to replace attention with an almost linear time alternative. In this paper, we prove that any such approach cannot perform important tasks that Transformer is able to perform (assuming a popular conjecture from fine-grained complexity theory). We focus on document similarity tasks, where one is given as input many documents and would like to find a pair which is (approximately) the most similar. We prove that Transformer is able to perform this task, and we prove that this task cannot be performed in truly subquadratic time by any algorithm. Thus, any model which can be evaluated in subquadratic time - whether because of subquadratic-time heuristics for attention, faster attention replacements like Mamba, or any other reason - cannot perform this task. In other words, in order to perform tasks that (implicitly or explicitly) involve document similarity, one may as well use Transformer and cannot avoid its quadratic running time.
Authors: Abdul Hameed Azeemi, Ihsan Ayyub Qazi, Agha Ali Raza
Abstract: Active learning (AL) optimizes data labeling efficiency by selecting the most informative instances for annotation. A key component in this procedure is an acquisition function that guides the selection process and identifies the suitable instances for labeling from the unlabeled pool. However, these acquisition methods suffer from high computational costs with large unlabeled data pools, posing a roadblock to their applicability on large datasets. To address this challenge and bridge this gap, we introduce a novel plug-and-play unlabeled data pruning strategy, ActivePrune, which leverages language models to prune the unlabeled pool. ActivePrune implements a two-stage pruning process: an initial fast evaluation using perplexity scores from an n-gram language model, followed by a high-quality selection using metrics for data quality computed through a quantized LLM. Additionally, to enhance the diversity in the unlabeled pool, we propose a novel perplexity reweighting method that systematically brings forward underrepresented instances for selection in subsequent labeling iterations. Experiments on translation, sentiment analysis, topic classification, and summarization tasks on four diverse datasets and four active learning strategies demonstrate that ActivePrune outperforms existing data pruning methods. Finally, we compare the selection quality $\leftrightarrow$ efficiency tradeoff of the data pruning methods and demonstrate that ActivePrune is computationally more efficient than other LLM score-based pruning methods, and provides up to 74% reduction in the end-to-end time required for active learning.
Authors: Emi Zeger, Mert Pilanci
Abstract: We show that training deep neural networks (DNNs) with absolute value activation and arbitrary input dimension can be formulated as equivalent convex Lasso problems with novel features expressed using geometric algebra. This formulation reveals geometric structures encoding symmetry in neural networks. Using the equivalent Lasso form of DNNs, we formally prove a fundamental distinction between deep and shallow networks: deep networks inherently favor symmetric structures in their fitted functions, with greater depth enabling multilevel symmetries, i.e., symmetries within symmetries. Moreover, Lasso features represent distances to hyperplanes that are reflected across training points. These reflection hyperplanes are spanned by training data and are orthogonal to optimal weight vectors. Numerical experiments support theory and demonstrate theoretically predicted features when training networks using embeddings generated by Large Language Models.
Authors: Mengfang Sun, Wenying Sun, Ying Sun, Shaobo Liu, Mohan Jiang, Zhen Xu
Abstract: This paper presents a novel approach to credit risk prediction by employing Graph Convolutional Neural Networks (GCNNs) to assess the creditworthiness of borrowers. Leveraging the power of big data and artificial intelligence, the proposed method addresses the challenges faced by traditional credit risk assessment models, particularly in handling imbalanced datasets and extracting meaningful features from complex relationships. The paper begins by transforming raw borrower data into graph-structured data, where borrowers and their relationships are represented as nodes and edges, respectively. A classic subgraph convolutional model is then applied to extract local features, followed by the introduction of a hybrid GCNN model that integrates both local and global convolutional operators to capture a comprehensive representation of node features. The hybrid model incorporates an attention mechanism to adaptively select features, mitigating issues of over-smoothing and insufficient feature consideration. The study demonstrates the potential of GCNNs in improving the accuracy of credit risk prediction, offering a robust solution for financial institutions seeking to enhance their lending decision-making processes.
Authors: Donald Loveland, Danai Koutra
Abstract: Graph Neural Networks (GNNs) often struggle to generalize when graphs exhibit both homophily (same-class connections) and heterophily (different-class connections). Specifically, GNNs tend to underperform for nodes with local homophily levels that differ significantly from the global homophily level. This issue poses a risk in user-centric applications where underrepresented homophily levels are present. Concurrently, fairness within GNNs has received substantial attention due to the potential amplification of biases via message passing. However, the connection between local homophily and fairness in GNNs remains underexplored. In this work, we move beyond global homophily and explore how local homophily levels can lead to unfair predictions. We begin by formalizing the challenge of fair predictions for underrepresented homophily levels as an out-of-distribution (OOD) problem. We then conduct a theoretical analysis that demonstrates how local homophily levels can alter predictions for differing sensitive attributes. We additionally introduce three new GNN fairness benchmarks, as well as a novel semi-synthetic graph generator, to empirically study the OOD problem. Across extensive analysis we find that two factors can promote unfairness: (a) OOD distance, and (b) heterophilous nodes situated in homophilous graphs. In cases where these two conditions are met, fairness drops by up to 24% on real world datasets, and 30% in semi-synthetic datasets. Together, our theoretical insights, empirical analysis, and algorithmic contributions unveil a previously overlooked source of unfairness rooted in the graph's homophily information.
Authors: Oskar {\AA}str\"om, Carina Geldhauser, Markus Grillitsch, Ola Hall, Alexandros Sopasakis
Abstract: We propose a methodology to enhance local CO2 monitoring by integrating satellite data from the Orbiting Carbon Observatories (OCO-2 and OCO-3) with ground level observations from the Integrated Carbon Observation System (ICOS) and weather data from the ECMWF Reanalysis v5 (ERA5). Unlike traditional methods that downsample national data, our approach uses multimodal data fusion for high-resolution CO2 estimations. We employ weighted K-nearest neighbor (KNN) interpolation with machine learning models to predict ground level CO2 from satellite measurements, achieving a Root Mean Squared Error of 3.92 ppm. Our results show the effectiveness of integrating diverse data sources in capturing local emission patterns, highlighting the value of high-resolution atmospheric transport models. The developed model improves the granularity of CO2 monitoring, providing precise insights for targeted carbon mitigation strategies, and represents a novel application of neural networks and KNN in environmental monitoring, adaptable to various regions and temporal scales.
Authors: Stanis{\l}aw Ka\'zmierczak, Jacek Ma\'ndziuk
Abstract: Random forests utilize bootstrap sampling to create an individual training set for each component tree. This involves sampling with replacement, with the number of instances equal to the size of the original training set ($N$). Research literature indicates that drawing fewer than $N$ observations can also yield satisfactory results. The ratio of the number of observations in each bootstrap sample to the total number of training instances is called the bootstrap rate (BR). Sampling more than $N$ observations (BR $>$ 1) has been explored in the literature only to a limited extent and has generally proven ineffective. In this paper, we re-examine this approach using 36 diverse datasets and consider BR values ranging from 1.2 to 5.0. Contrary to previous findings, we show that such parameterization can result in statistically significant improvements in classification accuracy compared to standard settings (BR $\leq$ 1). Furthermore, we investigate what the optimal BR depends on and conclude that it is more a property of the dataset than a dependence on the random forest hyperparameters. Finally, we develop a binary classifier to predict whether the optimal BR is $\leq$ 1 or $>$ 1 for a given dataset, achieving between 81.88\% and 88.81\% accuracy, depending on the experiment configuration.
Authors: Caitlin Ho, Andrea Arnold
Abstract: Incorporating a priori physics knowledge into machine learning leads to more robust and interpretable algorithms. In this work, we combine deep learning techniques and classic numerical methods for differential equations to solve two challenging problems in dynamical systems theory: dynamics discovery and parameter estimation. Results demonstrate the effectiveness of the proposed approaches on a suite of test problems exhibiting oscillatory and chaotic dynamics. When comparing the performance of various numerical schemes, such as the Runge-Kutta and linear multistep families of methods, we observe promising results in predicting the system dynamics and estimating physical parameters, given appropriate choices of spatial and temporal discretization schemes and numerical method orders.
Authors: Quyen Tran, Minh Le, Tuan Truong, Dinh Phung, Linh Ngo, Thien Nguyen, Nhat Ho, Trung Le
Abstract: Drawing inspiration from human learning behaviors, this work proposes a novel approach to mitigate catastrophic forgetting in Prompt-based Continual Learning models by exploiting the relationships between continuously emerging class data. We find that applying human habits of organizing and connecting information can serve as an efficient strategy when training deep learning models. Specifically, by building a hierarchical tree structure based on the expanding set of labels, we gain fresh insights into the data, identifying groups of similar classes could easily cause confusion. Additionally, we delve deeper into the hidden connections between classes by exploring the original pretrained model's behavior through an optimal transport-based approach. From these insights, we propose a novel regularization loss function that encourages models to focus more on challenging knowledge areas, thereby enhancing overall performance. Experimentally, our method demonstrated significant superiority over the most robust state-of-the-art models on various benchmarks.
Authors: Alex Cloud, Jacob Goldman-Wetzler, Ev\v{z}en Wybitul, Joseph Miller, Alexander Matt Turner
Abstract: Neural networks are trained primarily based on their inputs and outputs, without regard for their internal mechanisms. These neglected mechanisms determine properties that are critical for safety, like (i) transparency; (ii) the absence of sensitive information or harmful capabilities; and (iii) reliable generalization of goals beyond the training distribution. To address this shortcoming, we introduce gradient routing, a training method that isolates capabilities to specific subregions of a neural network. Gradient routing applies data-dependent, weighted masks to gradients during backpropagation. These masks are supplied by the user in order to configure which parameters are updated by which data points. We show that gradient routing can be used to (1) learn representations which are partitioned in an interpretable way; (2) enable robust unlearning via ablation of a pre-specified network subregion; and (3) achieve scalable oversight of a reinforcement learner by localizing modules responsible for different behaviors. Throughout, we find that gradient routing localizes capabilities even when applied to a limited, ad-hoc subset of the data. We conclude that the approach holds promise for challenging, real-world applications where quality data are scarce.
Authors: Yahong Yang
Abstract: In this paper, we investigate the application of operator learning, specifically DeepONet, to solve partial differential equations (PDEs). Unlike function learning methods that require training separate neural networks for each PDE, operator learning generalizes across different PDEs without retraining. We focus on the performance of DeepONet in Sobolev training, addressing two key questions: the approximation ability of deep branch and trunk networks, and the generalization error in Sobolev norms. Our findings highlight that deep branch networks offer significant performance benefits, while trunk networks are best kept simple. Moreover, standard sampling methods without adding derivative information in the encoding part are sufficient for minimizing generalization error in Sobolev training, based on generalization analysis. This paper fills a theoretical gap by providing error estimations for a wide range of physics-informed machine learning models and applications.
Authors: Bingxuan Li, Pengyi Shi, Amy Ward
Abstract: Predictive modeling often faces challenges due to limited data availability and quality, especially in domains where collected features are weakly correlated with outcomes and where additional feature collection is constrained by ethical or practical difficulties. Traditional machine learning (ML) models struggle to incorporate unobserved yet critical factors. In this work, we introduce an effective approach to formulate latent feature mining as text-to-text propositional logical reasoning. We propose FLAME (Faithful Latent Feature Mining for Predictive Model Enhancement), a framework that leverages large language models (LLMs) to augment observed features with latent features and enhance the predictive power of ML models in downstream tasks. Our framework is generalizable across various domains with necessary domain-specific adaptation, as it is designed to incorporate contextual information unique to each area, ensuring effective transfer to different areas facing similar data availability challenges. We validate our framework with two case studies: (1) the criminal justice system, a domain characterized by limited and ethically challenging data collection; (2) the healthcare domain, where patient privacy concerns and the complexity of medical data limit comprehensive feature collection. Our results show that inferred latent features align well with ground truth labels and significantly enhance the downstream classifier.
Authors: Ziqian Zhong, Jacob Andreas
Abstract: Trained transformer models have been found to implement interpretable procedures for tasks like arithmetic and associative recall, but little is understood about how the circuits that implement these procedures originate during training. To what extent do they depend on the supervisory signal provided to models, and to what extent are they attributable to behavior already present in models at the beginning of training? To investigate these questions, we investigate what functions can be learned by randomly initialized transformers in which only the embedding layers are optimized, so that the only input--output mappings learnable from data are those already implemented (up to a choice of encoding scheme) by the randomly initialized model. We find that these random transformers can perform a wide range of meaningful algorithmic tasks, including modular arithmetic, in-weights and in-context associative recall, decimal addition, parenthesis balancing, and even some aspects of natural language text generation. Our results indicate that some algorithmic capabilities are present in transformers (and accessible via appropriately structured inputs) even before these models are trained. Code is available at https://github.com/fjzzq2002/random_transformers.
Authors: Shakila Mahjabin Tonni, Pedro Faustini, Mark Dras
Abstract: Adversarial examples pose a significant challenge to deep neural networks (DNNs) across both image and text domains, with the intent to degrade model performance through meticulously altered inputs. Adversarial texts, however, are distinct from adversarial images due to their requirement for semantic similarity and the discrete nature of the textual contents. This study delves into the concept of human suspiciousness, a quality distinct from the traditional focus on imperceptibility found in image-based adversarial examples. Unlike images, where adversarial changes are meant to be indistinguishable to the human eye, textual adversarial content must often remain undetected or non-suspicious to human readers, even when the text's purpose is to deceive NLP systems or bypass filters. In this research, we expand the study of human suspiciousness by analyzing how individuals perceive adversarial texts. We gather and publish a novel dataset of Likert-scale human evaluations on the suspiciousness of adversarial sentences, crafted by four widely used adversarial attack methods and assess their correlation with the human ability to detect machine-generated alterations. Additionally, we develop a regression-based model to quantify suspiciousness and establish a baseline for future research in reducing the suspiciousness in adversarial text generation. We also demonstrate how the regressor-generated suspicious scores can be incorporated into adversarial generation methods to produce texts that are less likely to be perceived as computer-generated. We make our human suspiciousness annotated data and our code available.
Authors: Xinyi Xu, Shuaiqi Wang, Chuan-Sheng Foo, Bryan Kian Hsiang Low, Giulia Fanti
Abstract: Data valuation is a class of techniques for quantitatively assessing the value of data for applications like pricing in data marketplaces. Existing data valuation methods define a value for a discrete dataset. However, in many use cases, users are interested in not only the value of the dataset, but that of the distribution from which the dataset was sampled. For example, consider a buyer trying to evaluate whether to purchase data from different vendors. The buyer may observe (and compare) only a small preview sample from each vendor, to decide which vendor's data distribution is most useful to the buyer and purchase. The core question is how should we compare the values of data distributions from their samples? Under a Huber characterization of the data heterogeneity across vendors, we propose a maximum mean discrepancy (MMD)-based valuation method which enables theoretically principled and actionable policies for comparing data distributions from samples. We empirically demonstrate that our method is sample-efficient and effective in identifying valuable data distributions against several existing baselines, on multiple real-world datasets (e.g., network intrusion detection, credit card fraud detection) and downstream applications (classification, regression).
Authors: Peiyuan Liu, Beiliang Wu, Yifan Hu, Naiqi Li, Tao Dai, Jigang Bao, Shu-tao Xia
Abstract: Non-stationarity poses significant challenges for multivariate time series forecasting due to the inherent short-term fluctuations and long-term trends that can lead to spurious regressions or obscure essential long-term relationships. Most existing methods either eliminate or retain non-stationarity without adequately addressing its distinct impacts on short-term and long-term modeling. Eliminating non-stationarity is essential for avoiding spurious regressions and capturing local dependencies in short-term modeling, while preserving it is crucial for revealing long-term cointegration across variates. In this paper, we propose TimeBridge, a novel framework designed to bridge the gap between non-stationarity and dependency modeling in long-term time series forecasting. By segmenting input series into smaller patches, TimeBridge applies Integrated Attention to mitigate short-term non-stationarity and capture stable dependencies within each variate, while Cointegrated Attention preserves non-stationarity to model long-term cointegration across variates. Extensive experiments show that TimeBridge consistently achieves state-of-the-art performance in both short-term and long-term forecasting. Additionally, TimeBridge demonstrates exceptional performance in financial forecasting on the CSI 500 and S&P 500 indices, further validating its robustness and effectiveness. Code is available at \url{https://github.com/Hank0626/TimeBridge}.
Authors: Chen Yu
Abstract: Accurate calibration of gravity adaptation zones is of great significance in fields such as underwater navigation, geophysical exploration, and marine engineering. With the increasing application of gravity field data in these areas, traditional calibration methods based on single features are becoming inadequate for capturing the complex characteristics of gravity fields and addressing the intricate interrelationships among multidimensional data. This paper proposes an attention-enhanced algorithm for gravity adaptation zone calibration. By introducing an attention mechanism, the algorithm adaptively fuses multidimensional gravity field features and dynamically assigns feature weights, effectively solving the problems of multicollinearity and redundancy inherent in traditional feature selection methods, significantly improving calibration accuracy and robustness.In addition, a large-scale gravity field dataset with over 10,000 sampling points was constructed, and Kriging interpolation was used to enhance the spatial resolution of the data, providing a reliable data foundation for model training and evaluation. We conducted both qualitative and quantitative experiments on several classical machine learning models (such as SVM, GBDT, and RF), and the results demonstrate that the proposed algorithm significantly improves performance across these models, outperforming other traditional feature selection methods. The method proposed in this paper provides a new solution for gravity adaptation zone calibration, showing strong generalization ability and potential for application in complex environments. The code is available at \href{this link} {https://github.com/hulnifox/RF-ATTN}.
Authors: Ruinan Jin, Xiao Li, Yaoliang Yu, Baoxiang Wang
Abstract: Adaptive Moment Estimation (Adam) is a cornerstone optimization algorithm in deep learning, widely recognized for its flexibility with adaptive learning rates and efficiency in handling large-scale data. However, despite its practical success, the theoretical understanding of Adam's convergence has been constrained by stringent assumptions, such as almost surely bounded stochastic gradients or uniformly bounded gradients, which are more restrictive than those typically required for analyzing stochastic gradient descent (SGD). In this paper, we introduce a novel and comprehensive framework for analyzing the convergence properties of Adam. This framework offers a versatile approach to establishing Adam's convergence. Specifically, we prove that Adam achieves asymptotic (last iterate sense) convergence in both the almost sure sense and the \(L_1\) sense under the relaxed assumptions typically used for SGD, namely \(L\)-smoothness and the ABC inequality. Meanwhile, under the same assumptions, we show that Adam attains non-asymptotic sample complexity bounds similar to those of SGD.
Authors: Hyeonah Kim, Minsu Kim, Taeyoung Yun, Sanghyeok Choi, Emmanuel Bengio, Alex Hern\'andez-Garc\'ia, Jinkyoo Park
Abstract: Designing biological sequences with desired properties is a significant challenge due to the combinatorially vast search space and the high cost of evaluating each candidate sequence. To address these challenges, reinforcement learning (RL) methods, such as GFlowNets, utilize proxy models for rapid reward evaluation and annotated data for policy training. Although these approaches have shown promise in generating diverse and novel sequences, the limited training data relative to the vast search space often leads to the misspecification of proxy for out-of-distribution inputs. We introduce $\delta$-Conservative Search, a novel off-policy search method for training GFlowNets designed to improve robustness against proxy misspecification. The key idea is to incorporate conservativeness, controlled by parameter $\delta$, to constrain the search to reliable regions. Specifically, we inject noise into high-score offline sequences by randomly masking tokens with a Bernoulli distribution of parameter $\delta$ and then denoise masked tokens using the GFlowNet policy. Additionally, $\delta$ is adaptively adjusted based on the uncertainty of the proxy model for each data point. This enables the reflection of proxy uncertainty to determine the level of conservativeness. Experimental results demonstrate that our method consistently outperforms existing machine learning methods in discovering high-score sequences across diverse tasks-including DNA, RNA, protein, and peptide design-especially in large-scale scenarios.
Authors: Renye Yan, Yaozhong Gan, You Wu, Junliang Xing, Ling Liangn, Yeshang Zhu, Yimao Cai
Abstract: In sparse reward scenarios of reinforcement learning (RL), the memory mechanism provides promising shortcuts to policy optimization by reflecting on past experiences like humans. However, current memory-based RL methods simply store and reuse high-value policies, lacking a deeper refining and filtering of diverse past experiences and hence limiting the capability of memory. In this paper, we propose AdaMemento, an adaptive memory-enhanced RL framework. Instead of just memorizing positive past experiences, we design a memory-reflection module that exploits both positive and negative experiences by learning to predict known local optimal policies based on real-time states. To effectively gather informative trajectories for the memory, we further introduce a fine-grained intrinsic motivation paradigm, where nuances in similar states can be precisely distinguished to guide exploration. The exploitation of past experiences and exploration of new policies are then adaptively coordinated by ensemble learning to approach the global optimum. Furthermore, we theoretically prove the superiority of our new intrinsic motivation and ensemble mechanism. From 59 quantitative and visualization experiments, we confirm that AdaMemento can distinguish subtle states for better exploration and effectively exploiting past experiences in memory, achieving significant improvement over previous methods.
Authors: Berker Demirel, Lingjing Kong, Kun Zhang, Theofanis Karaletsos, Celestine Mendler-D\"unner, Francesco Locatello
Abstract: With the widespread deployment of deep learning models, they influence their environment in various ways. The induced distribution shifts can lead to unexpected performance degradation in deployed models. Existing methods to anticipate performativity typically incorporate information about the deployed model into the feature vector when predicting future outcomes. While enjoying appealing theoretical properties, modifying the input dimension of the prediction task is often not practical. To address this, we propose a novel technique to adjust pretrained backbones for performativity in a modular way, achieving better sample efficiency and enabling the reuse of existing deep learning assets. Focusing on performative label shift, the key idea is to train a shallow adapter module to perform a Bayes-optimal label shift correction to the backbone's logits given a sufficient statistic of the model to be deployed. As such, our framework decouples the construction of input-specific feature embeddings from the mechanism governing performativity. Motivated by dynamic benchmarking as a use-case, we evaluate our approach under adversarial sampling, for vision and language tasks. We show how it leads to smaller loss along the retraining trajectory and enables us to effectively select among candidate models to anticipate performance degradations. More broadly, our work provides a first baseline for addressing performativity in deep learning.
Authors: Sebastian Pineda Arango, Maciej Janowski, Lennart Purucker, Arber Zela, Frank Hutter, Josif Grabocka
Abstract: Ensemble methods are known for enhancing the accuracy and robustness of machine learning models by combining multiple base learners. However, standard approaches like greedy or random ensembles often fall short, as they assume a constant weight across samples for the ensemble members. This can limit expressiveness and hinder performance when aggregating the ensemble predictions. In this study, we explore employing neural networks as ensemble methods, emphasizing the significance of dynamic ensembling to leverage diverse model predictions adaptively. Motivated by the risk of learning low-diversity ensembles, we propose regularizing the model by randomly dropping base model predictions during the training. We demonstrate this approach lower bounds the diversity within the ensemble, reducing overfitting and improving generalization capabilities. Our experiments showcase that the dynamic neural ensemblers yield competitive results compared to strong baselines in computer vision, natural language processing, and tabular data.
Authors: Berker Demirel, Marco Fumero, Francesco Locatello
Abstract: Deep learning systems deployed in real-world applications often encounter data that is different from their in-distribution (ID). A reliable system should ideally abstain from making decisions in this out-of-distribution (OOD) setting. Existing state-of-the-art methods primarily focus on feature distances, such as k-th nearest neighbors and distances to decision boundaries, either overlooking or ineffectively using in-distribution statistics. In this work, we propose a novel angle-based metric for OOD detection that is computed relative to the in-distribution structure. We demonstrate that the angles between feature representations and decision boundaries, viewed from the mean of in-distribution features, serve as an effective discriminative factor between ID and OOD data. Our method achieves state-of-the-art performance on CIFAR-10 and ImageNet benchmarks, reducing FPR95 by 0.88% and 7.74% respectively. Our score function is compatible with existing feature space regularization techniques, enhancing performance. Additionally, its scale-invariance property enables creating an ensemble of models for OOD detection via simple score summation.
Authors: Shoaib Ahmed Siddiqui, Yanzhi Chen, Juyeon Heo, Menglin Xia, Adrian Weller
Abstract: Recent works have successfully applied Large Language Models (LLMs) to function modeling tasks. However, the reasons behind this success remain unclear. In this work, we propose a new evaluation framework to comprehensively assess LLMs' function modeling abilities. By adopting a Bayesian perspective of function modeling, we discover that LLMs are relatively weak in understanding patterns in raw data, but excel at utilizing prior knowledge about the domain to develop a strong understanding of the underlying function. Our findings offer new insights about the strengths and limitations of LLMs in the context of function modeling.
Authors: Friso de Kruiff, Erik Bekkers, Ozan \"Oktem, Carola-Bibiane Sch\"onlieb, Willem Diepeveen
Abstract: We propose Pullback Flow Matching (PFM), a novel framework for generative modeling on data manifolds. Unlike existing methods that assume or learn restrictive closed-form manifold mappings for training Riemannian Flow Matching (RFM) models, PFM leverages pullback geometry and isometric learning to preserve the underlying manifold's geometry while enabling efficient generation and precise interpolation in latent space. This approach not only facilitates closed-form mappings on the data manifold but also allows for designable latent spaces, using assumed metrics on both data and latent manifolds. By enhancing isometric learning through Neural ODEs and proposing a scalable training objective, we achieve a latent space more suitable for interpolation, leading to improved manifold learning and generative performance. We demonstrate PFM's effectiveness through applications in synthetic data, protein dynamics and protein sequence data, generating novel proteins with specific properties. This method shows strong potential for drug discovery and materials science, where generating novel samples with specific properties is of great interest.
Authors: Yutaka Shimizu, Masayoshi Tomizuka
Abstract: Model-based reinforcement learning has shown promise for improving sample efficiency and decision-making in complex environments. However, existing methods face challenges in training stability, robustness to noise, and computational efficiency. In this paper, we propose Bisimulation Metric for Model Predictive Control (BS-MPC), a novel approach that incorporates bisimulation metric loss in its objective function to directly optimize the encoder. This time-step-wise direct optimization enables the learned encoder to extract intrinsic information from the original state space while discarding irrelevant details and preventing the gradients and errors from diverging. BS-MPC improves training stability, robustness against input noise, and computational efficiency by reducing training time. We evaluate BS-MPC on both continuous control and image-based tasks from the DeepMind Control Suite, demonstrating superior performance and robustness compared to state-of-the-art baseline methods.
Authors: Junwei Deng, Ting-Wei Li, Shiyuan Zhang, Shixuan Liu, Yijun Pan, Hao Huang, Xinhe Wang, Pingbang Hu, Xingjian Zhang, Jiaqi W. Ma
Abstract: Data attribution methods aim to quantify the influence of individual training samples on the prediction of artificial intelligence (AI) models. As training data plays an increasingly crucial role in the modern development of large-scale AI models, data attribution has found broad applications in improving AI performance and safety. However, despite a surge of new data attribution methods being developed recently, there lacks a comprehensive library that facilitates the development, benchmarking, and deployment of different data attribution methods. In this work, we introduce $\texttt{dattri}$, an open-source data attribution library that addresses the above needs. Specifically, $\texttt{dattri}$ highlights three novel design features. Firstly, $\texttt{dattri}$ proposes a unified and easy-to-use API, allowing users to integrate different data attribution methods into their PyTorch-based machine learning pipeline with a few lines of code changed. Secondly, $\texttt{dattri}$ modularizes low-level utility functions that are commonly used in data attribution methods, such as Hessian-vector product, inverse-Hessian-vector product or random projection, making it easier for researchers to develop new data attribution methods. Thirdly, $\texttt{dattri}$ provides a comprehensive benchmark framework with pre-trained models and ground truth annotations for a variety of benchmark settings, including generative AI settings. We have implemented a variety of state-of-the-art efficient data attribution methods that can be applied to large-scale neural network models, and will continuously update the library in the future. Using the developed $\texttt{dattri}$ library, we are able to perform a comprehensive and fair benchmark analysis across a wide range of data attribution methods. The source code of $\texttt{dattri}$ is available at https://github.com/TRAIS-Lab/dattri.
Authors: Andreas Mueller, Julien Siems, Harsha Nori, David Salinas, Arber Zela, Rich Caruana, Frank Hutter
Abstract: Generalized Additive Models (GAMs) are widely recognized for their ability to create fully interpretable machine learning models for tabular data. Traditionally, training GAMs involves iterative learning algorithms, such as splines, boosted trees, or neural networks, which refine the additive components through repeated error reduction. In this paper, we introduce GAMformer, the first method to leverage in-context learning to estimate shape functions of a GAM in a single forward pass, representing a significant departure from the conventional iterative approaches to GAM fitting. Building on previous research applying in-context learning to tabular data, we exclusively use complex, synthetic data to train GAMformer, yet find it extrapolates well to real-world data. Our experiments show that GAMformer performs on par with other leading GAMs across various classification benchmarks while generating highly interpretable shape functions.
Authors: Stefano Calzavara, Lorenzo Cazzaro, Donald Gera, Salvatore Orlando
Abstract: Protecting the intellectual property of machine learning models is a hot topic and many watermarking schemes for deep neural networks have been proposed in the literature. Unfortunately, prior work largely neglected the investigation of watermarking techniques for other types of models, including decision tree ensembles, which are a state-of-the-art model for classification tasks on non-perceptual data. In this paper, we present the first watermarking scheme designed for decision tree ensembles, focusing in particular on random forest models. We discuss watermark creation and verification, presenting a thorough security analysis with respect to possible attacks. We finally perform an experimental evaluation of the proposed scheme, showing excellent results in terms of accuracy and security against the most relevant threats.
Authors: Aakriti Agrawal, Mucong Ding, Zora Che, Chenghao Deng, Anirudh Satheesh, John Langford, Furong Huang
Abstract: How can we harness the collective capabilities of multiple Large Language Models (LLMs) to create an even more powerful model? This question forms the foundation of our research, where we propose an innovative approach to weak-to-strong (w2s) generalization-a critical problem in AI alignment. Our work introduces an easy-to-hard (e2h) framework for studying the feasibility of w2s generalization, where weak models trained on simpler tasks collaboratively supervise stronger models on more complex tasks. This setup mirrors real-world challenges, where direct human supervision is limited. To achieve this, we develop a novel AdaBoost-inspired ensemble method, demonstrating that an ensemble of weak supervisors can enhance the performance of stronger LLMs across classification and generative tasks on difficult QA datasets. In several cases, our ensemble approach matches the performance of models trained on ground-truth data, establishing a new benchmark for w2s generalization. We observe an improvement of up to 14% over existing baselines and average improvements of 5% and 4% for binary classification and generative tasks, respectively. This research points to a promising direction for enhancing AI through collective supervision, especially in scenarios where labeled data is sparse or insufficient.
Authors: Zhichao Hou, MohamadAli Torkamani, Hamid Krim, Xiaorui Liu
Abstract: This work tackles an intriguing and fundamental open challenge in representation learning: Given a well-trained deep learning model, can it be reprogrammed to enhance its robustness against adversarial or noisy input perturbations without altering its parameters? To explore this, we revisit the core feature transformation mechanism in representation learning and propose a novel non-linear robust pattern matching technique as a robust alternative. Furthermore, we introduce three model reprogramming paradigms to offer flexible control of robustness under different efficiency requirements. Comprehensive experiments and ablation studies across diverse learning models ranging from basic linear model and MLPs to shallow and modern deep ConvNets demonstrate the effectiveness of our approaches. This work not only opens a promising and orthogonal direction for improving adversarial defenses in deep learning beyond existing methods but also provides new insights into designing more resilient AI systems with robust statistics.
Authors: Qiqiang Lin, Muning Wen, Qiuying Peng, Guanyu Nie, Junwei Liao, Jun Wang, Xiaoyun Mo, Jiamu Zhou, Cheng Cheng, Yin Zhao, Jun Wang, Weinan Zhang
Abstract: Large language models have demonstrated impressive value in performing as autonomous agents when equipped with external tools and API calls. Nonetheless, effectively harnessing their potential for executing complex tasks crucially relies on enhancements in their function calling capabilities. This paper identifies a critical gap in existing function calling models, where performance varies significantly across benchmarks, often due to being misled by specific naming conventions. To address such an issue, we introduce Hammer, a novel family of foundation models specifically engineered for on-device function calling. Hammer employs an augmented dataset that enhances models' sensitivity to irrelevant functions and incorporates function masking techniques to minimize misleading. Our empirical evaluations reveal that Hammer not only outperforms larger models but also demonstrates robust generalization across diverse benchmarks, achieving sota results. Our open source contributions include a specialized dataset for irrelevance detection, a tuning framework for enhanced generalization, and the Hammer models, establishing a new standard for function calling performance.
Authors: Zhaolin Gao, Wenhao Zhan, Jonathan D. Chang, Gokul Swamy, Kiant\'e Brantley, Jason D. Lee, Wen Sun
Abstract: Large Language Models (LLMs) have achieved remarkable success at tasks like summarization that involve a single turn of interaction. However, they can still struggle with multi-turn tasks like dialogue that require long-term planning. Previous works on multi-turn dialogue extend single-turn reinforcement learning from human feedback (RLHF) methods to the multi-turn setting by treating all prior dialogue turns as a long context. Such approaches suffer from covariate shift: the conversations in the training set have previous turns generated by some reference policy, which means that low training error may not necessarily correspond to good performance when the learner is actually in the conversation loop. In response, we introduce REgressing the RELative FUture (REFUEL), an efficient policy optimization approach designed to address multi-turn RLHF in LLMs. REFUEL employs a single model to estimate $Q$-values and trains on self-generated data, addressing the covariate shift issue. REFUEL frames the multi-turn RLHF problem as a sequence of regression tasks on iteratively collected datasets, enabling ease of implementation. Theoretically, we prove that REFUEL can match the performance of any policy covered by the training set. Empirically, we evaluate our algorithm by using Llama-3.1-70B-it to simulate a user in conversation with our model. REFUEL consistently outperforms state-of-the-art methods such as DPO and REBEL across various settings. Furthermore, despite having only 8 billion parameters, Llama-3-8B-it fine-tuned with REFUEL outperforms Llama-3.1-70B-it on long multi-turn dialogues. Implementation of REFUEL can be found at https://github.com/ZhaolinGao/REFUEL/, and models trained by REFUEL can be found at https://huggingface.co/Cornell-AGI.
URLs: https://github.com/ZhaolinGao/REFUEL/,, https://huggingface.co/Cornell-AGI.
Authors: David X. Wu, Anant Sahai
Abstract: The classic teacher-student model in machine learning posits that a strong teacher supervises a weak student to improve the student's capabilities. We instead consider the inverted situation, where a weak teacher supervises a strong student with imperfect pseudolabels. This paradigm was recently brought forth by Burns et al.'23 and termed \emph{weak-to-strong generalization}. We theoretically investigate weak-to-strong generalization for binary and multilabel classification in a stylized overparameterized spiked covariance model with Gaussian covariates where the weak teacher's pseudolabels are asymptotically like random guessing. Under these assumptions, we provably identify two asymptotic phases of the strong student's generalization after weak supervision: (1) successful generalization and (2) random guessing. Our techniques should eventually extend to weak-to-strong multiclass classification. Towards doing so, we prove a tight lower tail inequality for the maximum of correlated Gaussians, which may be of independent interest. Understanding the multilabel setting reinforces the value of using logits for weak supervision when they are available.
Authors: Jason Kurz, Sean Oughton, Shitao Liu
Abstract: Operator networks are designed to approximate nonlinear operators, which provide mappings between infinite-dimensional spaces such as function spaces. These networks are playing an increasingly important role in machine learning, with their most notable contributions in the field of scientific computing. Their significance stems from their ability to handle the type of data often encountered in scientific applications. For instance, in climate modeling or fluid dynamics, input data typically consists of discretized continuous fields (like temperature distributions or velocity fields). We introduce the radial basis operator network (RBON), which represents a significant advancement as the first operator network capable of learning an operator in both the time domain and frequency domain when adjusted to accept complex-valued inputs. Despite the small, single hidden-layer structure, the RBON boasts small $L^2$ relative test error for both in- and out-of-distribution data (OOD) of less than $1\times 10^{-7}$ in some benchmark cases. Moreover, the RBON maintains small error on OOD data from entirely different function classes from the training data.
Authors: Alexander Atanasov, Alexandru Meterez, James B. Simon, Cengiz Pehlevan
Abstract: We consider neural networks (NNs) where the final layer is down-scaled by a fixed hyperparameter $\gamma$. Recent work has identified $\gamma$ as controlling the strength of feature learning. As $\gamma$ increases, network evolution changes from ``lazy'' kernel dynamics to ``rich'' feature-learning dynamics, with a host of associated benefits including improved performance on common tasks. In this work, we conduct a thorough empirical investigation of the effect of scaling $\gamma$ across a variety of models and datasets in the online training setting. We first examine the interaction of $\gamma$ with the learning rate $\eta$, identifying several scaling regimes in the $\gamma$-$\eta$ plane which we explain theoretically using a simple model. We find that the optimal learning rate $\eta^*$ scales non-trivially with $\gamma$. In particular, $\eta^* \propto \gamma^2$ when $\gamma \ll 1$ and $\eta^* \propto \gamma^{2/L}$ when $\gamma \gg 1$ for a feed-forward network of depth $L$. Using this optimal learning rate scaling, we proceed with an empirical study of the under-explored ``ultra-rich'' $\gamma \gg 1$ regime. We find that networks in this regime display characteristic loss curves, starting with a long plateau followed by a drop-off, sometimes followed by one or more additional staircase steps. We find networks of different large $\gamma$ values optimize along similar trajectories up to a reparameterization of time. We further find that optimal online performance is often found at large $\gamma$ and could be missed if this hyperparameter is not tuned. Our findings indicate that analytical study of the large-$\gamma$ limit may yield useful insights into the dynamics of representation learning in performant models.
Authors: Shane E. Loeffler, Zan Ahmad, Syed Yusuf Ali, Carolyna Yamamoto, Dan M. Popescu, Alana Yee, Yash Lal, Natalia Trayanova, Mauro Maggioni
Abstract: Predicting time-dependent dynamics of complex systems governed by non-linear partial differential equations (PDEs) with varying parameters and domains is a challenging task motivated by applications across various fields. We introduce a novel family of neural operators based on our Graph Fourier Neural Kernels, designed to learn solution generators for nonlinear PDEs in which the highest-order term is diffusive, across multiple domains and parameters. G-FuNK combines components that are parameter- and domain-adapted with others that are not. The domain-adapted components are constructed using a weighted graph on the discretized domain, where the graph Laplacian approximates the highest-order diffusive term, ensuring boundary condition compliance and capturing the parameter and domain-specific behavior. Meanwhile, the learned components transfer across domains and parameters via Fourier Neural Operators. This approach naturally embeds geometric and directional information, improving generalization to new test domains without need for retraining the network. To handle temporal dynamics, our method incorporates an integrated ODE solver to predict the evolution of the system. Experiments show G-FuNK's capability to accurately approximate heat, reaction diffusion, and cardiac electrophysiology equations across various geometries and anisotropic diffusivity fields. G-FuNK achieves low relative errors on unseen domains and fiber fields, significantly accelerating predictions compared to traditional finite-element solvers.
Authors: Ethan Wilson, Kai Yue, Chau-Wai Wong, Huaiyu Dai
Abstract: Federated learning (FL) is a privacy-preserving machine learning framework that enables multiple nodes to train models on their local data and periodically average weight updates to benefit from other nodes' training. Each node's goal is to collaborate with other nodes to improve the model's performance while keeping its training data private. However, this framework does not guarantee data privacy. Prior work has shown that the gradient-sharing steps in FL can be vulnerable to data reconstruction attacks from an honest-but-curious central server. In this work, we show that an honest-but-curious node/client can also launch attacks to reconstruct peers' image data in a centralized system, presenting a severe privacy risk. We demonstrate that a single client can silently reconstruct other clients' private images using diluted information available within consecutive updates. We leverage state-of-the-art diffusion models to enhance the perceptual quality and recognizability of the reconstructed images, further demonstrating the risk of information leakage at a semantic level. This highlights the need for more robust privacy-preserving mechanisms that protect against silent client-side attacks during federated training.
Authors: Yongyi Su, Yushu Li, Nanqing Liu, Kui Jia, Xulei Yang, Chuan-Sheng Foo, Xun Xu
Abstract: Test-time adaptation (TTA) updates the model weights during the inference stage using testing data to enhance generalization. However, this practice exposes TTA to adversarial risks. Existing studies have shown that when TTA is updated with crafted adversarial test samples, also known as test-time poisoned data, the performance on benign samples can deteriorate. Nonetheless, the perceived adversarial risk may be overstated if the poisoned data is generated under overly strong assumptions. In this work, we first review realistic assumptions for test-time data poisoning, including white-box versus grey-box attacks, access to benign data, attack budget, and more. We then propose an effective and realistic attack method that better produces poisoned samples without access to benign samples, and derive an effective in-distribution attack objective. We also design two TTA-aware attack objectives. Our benchmarks of existing attack methods reveal that the TTA methods are more robust than previously believed. In addition, we analyze effective defense strategies to help develop adversarially robust TTA methods.
Authors: Dylan Xu, Juan-Pablo Rivera
Abstract: Recent advances in deep learning have brought attention to the possibility of creating advanced, general AI systems that outperform humans across many tasks. However, if these systems pursue unintended goals, there could be catastrophic consequences. A key prerequisite for AI systems pursuing unintended goals is whether they will behave in a coherent and goal-directed manner in the first place, optimizing for some unknown goal; there exists significant research trying to evaluate systems for said behaviors. However, the most rigorous definitions of goal-directedness we currently have are difficult to compute in real-world settings. Drawing upon this previous literature, we explore policy goal-directedness within reinforcement learning (RL) environments. In our findings, we propose a different family of definitions of the goal-directedness of a policy that analyze whether it is well-modeled as near-optimal for many (sparse) reward functions. We operationalize this preliminary definition of goal-directedness and test it in toy Markov decision process (MDP) environments. Furthermore, we explore how goal-directedness could be measured in frontier large-language models (LLMs). Our contribution is a definition of goal-directedness that is simpler and more easily computable in order to approach the question of whether AI systems could pursue dangerous goals. We recommend further exploration of measuring coherence and goal-directedness, based on our findings.
Authors: Qingyu Yin, Xuzheng He, Luoao Deng, Chak Tou Leong, Fan Wang, Yanzhao Yan, Xiaoyu Shen, Qiang Zhang
Abstract: Fine-tuning and in-context learning (ICL) are two prevalent methods in imbuing large language models with task-specific knowledge. It is commonly believed that fine-tuning can surpass ICL given sufficient training samples as it allows the model to adjust its internal parameters based on the data. However, this paper presents a counterintuitive finding: For tasks with implicit patterns, ICL captures these patterns significantly better than fine-tuning. We developed several datasets featuring implicit patterns, such as sequences determining answers through parity or identifying reducible terms in calculations. We then evaluated the models' understanding of these patterns under both fine-tuning and ICL across models ranging from 0.5B to 7B parameters. The results indicate that models employing ICL can quickly grasp deep patterns and significantly improve accuracy. In contrast, fine-tuning, despite utilizing thousands of times more training samples than ICL, achieved only limited improvements. We also proposed circuit shift theory from a mechanistic interpretability's view to explain why ICL wins.
Authors: Hoang-Viet Tran, Thieu N. Vo, Tho Tran Huu, Tan Minh Nguyen
Abstract: Designing neural network architectures that can handle data symmetry is crucial. This is especially important for geometric graphs whose properties are equivariance under Euclidean transformations. Current equivariant graph neural networks (EGNNs), particularly those using message passing, have a limitation in expressive power. Recent high-order graph neural networks can overcome this limitation, yet they lack equivariance properties, representing a notable drawback in certain applications in chemistry and physical sciences. In this paper, we introduce the Clifford Group Equivariant Graph Neural Networks (CG-EGNNs), a novel EGNN that enhances high-order message passing by integrating high-order local structures in the context of Clifford algebras. As a key benefit of using Clifford algebras, CG-EGNN can learn functions that capture equivariance from positional features. By adopting the high-order message passing mechanism, CG-EGNN gains richer information from neighbors, thus improving model performance. Furthermore, we establish the universality property of the $k$-hop message passing framework, showcasing greater expressive power of CG-EGNNs with additional $k$-hop message passing mechanism. We empirically validate that CG-EGNNs outperform previous methods on various benchmarks including n-body, CMU motion capture, and MD17, highlighting their effectiveness in geometric deep learning.
Authors: Minjung Kim, Yusuke Hioka, Michael Witbrock
Abstract: Neural time-series analysis has traditionally focused on modeling data in the time domain, often with some approaches incorporating equivalent Fourier domain representations as auxiliary spectral features. In this work, we shift the main focus to frequency representations, modeling time-series data fully and directly in the Fourier domain. We introduce Neural Fourier Modelling (NFM), a compact yet powerful solution for time-series analysis. NFM is grounded in two key properties of the Fourier transform (FT): (i) the ability to model finite-length time series as functions in the Fourier domain, treating them as continuous-time elements in function space, and (ii) the capacity for data manipulation (such as resampling and timespan extension) within the Fourier domain. We reinterpret Fourier-domain data manipulation as frequency extrapolation and interpolation, incorporating this as a core learning mechanism in NFM, applicable across various tasks. To support flexible frequency extension with spectral priors and effective modulation of frequency representations, we propose two learning modules: Learnable Frequency Tokens (LFT) and Implicit Neural Fourier Filters (INFF). These modules enable compact and expressive modeling in the Fourier domain. Extensive experiments demonstrate that NFM achieves state-of-the-art performance on a wide range of tasks (forecasting, anomaly detection, and classification), including challenging time-series scenarios with previously unseen sampling rates at test time. Moreover, NFM is highly compact, requiring fewer than 40K parameters in each task, with time-series lengths ranging from 100 to 16K.
Authors: Mehul Damani, Idan Shenfeld, Andi Peng, Andreea Bobu, Jacob Andreas
Abstract: Computationally intensive decoding procedures--including search, reranking, and self-critique--can improve the quality of language model (LM) outputs in problems spanning code generation, numerical reasoning, and dialog. Existing work typically applies the same decoding procedure for every input to an LM. But not all inputs require the same amount of computation to process. Can we allocate decoding computation adaptively, using more resources to answer questions whose answers will be harder to compute? We present an approach that predicts the distribution of rewards given an input and computation budget, then allocates additional computation to inputs for which it is predicted to be most useful. We apply this approach in two decoding procedures: first, an adaptive best-of-k procedure that dynamically selects the number of samples to generate as input to a reranker; second, a routing procedure that dynamically responds to a query using a decoding procedure that is expensive but accurate, or one that is cheaper but less capable. Across a suite of programming, mathematics, and dialog tasks, we show that accurate computation-allocation procedures can be learned, and reduce computation by up to 50% at no cost to response quality, or improve quality by up to 10% at a fixed computational budget.
Authors: Ankur Mali, Tommaso Salvatori, Alexander Ororbia
Abstract: Energy-based learning algorithms, such as predictive coding (PC), have garnered significant attention in the machine learning community due to their theoretical properties, such as local operations and biologically plausible mechanisms for error correction. In this work, we rigorously analyze the stability, robustness, and convergence of PC through the lens of dynamical systems theory. We show that, first, PC is Lyapunov stable under mild assumptions on its loss and residual energy functions, which implies intrinsic robustness to small random perturbations due to its well-defined energy-minimizing dynamics. Second, we formally establish that the PC updates approximate quasi-Newton methods by incorporating higher-order curvature information, which makes them more stable and able to converge with fewer iterations compared to models trained via backpropagation (BP). Furthermore, using this dynamical framework, we provide new theoretical bounds on the similarity between PC and other algorithms, i.e., BP and target propagation (TP), by precisely characterizing the role of higher-order derivatives. These bounds, derived through detailed analysis of the Hessian structures, show that PC is significantly closer to quasi-Newton updates than TP, providing a deeper understanding of the stability and efficiency of PC compared to conventional learning methods.
Authors: Hyungjin Chung, Dohun Lee, Jong Chul Ye
Abstract: Autoregressive models (ARMs) and diffusion models (DMs) represent two leading paradigms in generative modeling, each excelling in distinct areas: ARMs in global context modeling and long-sequence generation, and DMs in generating high-quality local contexts, especially for continuous data such as images and short videos. However, ARMs often suffer from exponential error accumulation over long sequences, leading to physically implausible results, while DMs are limited by their local context generation capabilities. In this work, we introduce Autoregressive Coherent multimodal generation with Diffusion Correction (ACDC), a zero-shot approach that combines the strengths of both ARMs and DMs at the inference stage without the need for additional fine-tuning. ACDC leverages ARMs for global context generation and memory-conditioned DMs for local correction, ensuring high-quality outputs by correcting artifacts in generated multimodal tokens. In particular, we propose a memory module based on large language models (LLMs) that dynamically adjusts the conditioning texts for the DMs, preserving crucial global context information. Our experiments on multimodal tasks, including coherent multi-frame story generation and autoregressive video generation, demonstrate that ACDC effectively mitigates the accumulation of errors and significantly enhances the quality of generated outputs, achieving superior performance while remaining agnostic to specific ARM and DM architectures. Project page: https://acdc2025.github.io/
Authors: Xuanrui Zeng
Abstract: One recent research demonstrated successful application of the label alignment property for unsupervised domain adaptation in a linear regression settings. Instead of regularizing representation learning to be domain invariant, the research proposed to regularize the linear regression model to align with the top singular vectors of the data matrix from the target domain. In this work we expand upon this idea and generalize it to the case of deep learning, where we derive an alternative formulation of the original adaptation algorithm exploiting label alignment suitable for deep neural network. We also perform experiments to demonstrate that our approach achieves comparable performance to mainstream unsupervised domain adaptation methods while having stabler convergence. All experiments and implementations in our work can be found at the following codebase: \url{https://github.com/xuanrui-work/DeepLabelAlignment}.
Authors: Guangzhi Xiong, Sanchit Sinha, Aidong Zhang
Abstract: Generalized additive models (GAMs) have long been a powerful white-box tool for the intelligible analysis of tabular data, revealing the influence of each feature on the model predictions. Despite the success of neural networks (NNs) in various domains, their application as NN-based GAMs in tabular data analysis remains suboptimal compared to tree-based ones, and the opacity of encoders in NN-GAMs also prevents users from understanding how networks learn the functions. In this work, we propose a new deep tabular learning method, termed Prototypical Neural Additive Model (ProtoNAM), which introduces prototypes into neural networks in the framework of GAMs. With the introduced prototype-based feature activation, ProtoNAM can flexibly model the irregular mapping from tabular features to the outputs while maintaining the explainability of the final prediction. We also propose a gradient-boosting inspired hierarchical shape function modeling method, facilitating the discovery of complex feature patterns and bringing transparency into the learning process of each network layer. Our empirical evaluations demonstrate that ProtoNAM outperforms all existing NN-based GAMs, while providing additional insights into the shape function learned for each feature. The source code of ProtoNAM is available at \url{https://github.com/Teddy-XiongGZ/ProtoNAM}.
Authors: Deqing Fu, Tong Xiao, Rui Wang, Wang Zhu, Pengchuan Zhang, Guan Pang, Robin Jia, Lawrence Chen
Abstract: Although reward models have been successful in improving multimodal large language models, the reward models themselves remain brutal and contain minimal information. Notably, existing reward models only mimic human annotations by assigning only one binary feedback to any text, no matter how long the text is. In the realm of multimodal language models, where models are required to process both images and texts, a naive reward model may learn implicit biases toward texts and become less grounded in images. In this paper, we propose a $\textbf{T}$oken-$\textbf{L}$evel $\textbf{D}$etective $\textbf{R}$eward Model ($\textbf{TLDR}$) to provide fine-grained annotations to each text token. We first introduce a perturbation-based method to generate synthetic hard negatives and their token-level labels to train TLDR models. Then we show the rich usefulness of TLDR models both in assisting off-the-shelf models to self-correct their generations, and in serving as a hallucination evaluation tool. Finally, we show that TLDR models can significantly speed up human annotation by 3 times to acquire a broader range of high-quality vision language data.
Authors: Hongjun Wang, Jiyuan Chen, Tong Pan, Zheng Dong, Lingyu Zhang, Renhe Jiang, Xuan Song
Abstract: Spatiotemporal neural networks have shown great promise in urban scenarios by effectively capturing temporal and spatial correlations. However, urban environments are constantly evolving, and current model evaluations are often limited to traffic scenarios and use data mainly collected only a few weeks after training period to evaluate model performance. The generalization ability of these models remains largely unexplored. To address this, we propose a Spatiotemporal Out-of-Distribution (ST-OOD) benchmark, which comprises six urban scenario: bike-sharing, 311 services, pedestrian counts, traffic speed, traffic flow, ride-hailing demand, and bike-sharing, each with in-distribution (same year) and out-of-distribution (next years) settings. We extensively evaluate state-of-the-art spatiotemporal models and find that their performance degrades significantly in out-of-distribution settings, with most models performing even worse than a simple Multi-Layer Perceptron (MLP). Our findings suggest that current leading methods tend to over-rely on parameters to overfit training data, which may lead to good performance on in-distribution data but often results in poor generalization. We also investigated whether dropout could mitigate the negative effects of overfitting. Our results showed that a slight dropout rate could significantly improve generalization performance on most datasets, with minimal impact on in-distribution performance. However, balancing in-distribution and out-of-distribution performance remains a challenging problem. We hope that the proposed benchmark will encourage further research on this critical issue.
Authors: Aye Phyu Phyu Aung, Xinrun Wang, Ruiyu Wang, Hau Chan, Bo An, Xiaoli Li, J. Senthilnath
Abstract: In this paper, we propose a new approach to train deep learning models using game theory concepts including Generative Adversarial Networks (GANs) and Adversarial Training (AT) where we deploy a double-oracle framework using best response oracles. GAN is essentially a two-player zero-sum game between the generator and the discriminator. The same concept can be applied to AT with attacker and classifier as players. Training these models is challenging as a pure Nash equilibrium may not exist and even finding the mixed Nash equilibrium is difficult as training algorithms for both GAN and AT have a large-scale strategy space. Extending our preliminary model DO-GAN, we propose the methods to apply the double oracle framework concept to Adversarial Neural Architecture Search (NAS for GAN) and Adversarial Training (NAS for AT) algorithms. We first generalize the players' strategies as the trained models of generator and discriminator from the best response oracles. We then compute the meta-strategies using a linear program. For scalability of the framework where multiple network models of best responses are stored in the memory, we prune the weakly-dominated players' strategies to keep the oracles from becoming intractable. Finally, we conduct experiments on MNIST, CIFAR-10 and TinyImageNet for DONAS-GAN. We also evaluate the robustness under FGSM and PGD attacks on CIFAR-10, SVHN and TinyImageNet for DONAS-AT. We show that all our variants have significant improvements in both subjective qualitative evaluation and quantitative metrics, compared with their respective base architectures.
Authors: A. Quadir, M. Sajid, M. Tanveer
Abstract: On Efficient and Scalable Computation of the Nonparametric Maximum Likelihood Estimator in Mixture ModelsTwin support vector machine (TSVM) is an emerging machine learning model with versatile applicability in classification and regression endeavors. Nevertheless, TSVM confronts noteworthy challenges: $(i)$ the imperative demand for matrix inversions presents formidable obstacles to its efficiency and applicability on large-scale datasets; $(ii)$ the omission of the structural risk minimization (SRM) principle in its primal formulation heightens the vulnerability to overfitting risks; and $(iii)$ the TSVM exhibits a high susceptibility to noise and outliers, and also demonstrates instability when subjected to resampling. In view of the aforementioned challenges, we propose the granular ball twin support vector machine (GBTSVM). GBTSVM takes granular balls, rather than individual data points, as inputs to construct a classifier. These granular balls, characterized by their coarser granularity, exhibit robustness to resampling and reduced susceptibility to the impact of noise and outliers. We further propose a novel large-scale granular ball twin support vector machine (LS-GBTSVM). LS-GBTSVM's optimization formulation ensures two critical facets: $(i)$ it eliminates the need for matrix inversions, streamlining the LS-GBTSVM's computational efficiency, and $(ii)$ it incorporates the SRM principle through the incorporation of regularization terms, effectively addressing the issue of overfitting. The proposed LS-GBTSVM exemplifies efficiency, scalability for large datasets, and robustness against noise and outliers. We conduct a comprehensive evaluation of the GBTSVM and LS-GBTSVM models on benchmark datasets from UCI, KEEL, and NDC datasets. Our experimental findings and statistical analyses affirm the superior generalization prowess of the proposed GBTSVM and LS-GBTSVM models.
Authors: Taesun Yeom, Sangyoon Lee, Jaeho Lee
Abstract: Neural fields are an emerging paradigm that represent data as continuous functions parameterized by neural networks. Despite many advantages, neural fields often have a high training cost, which prevents a broader adoption. In this paper, we focus on a popular family of neural fields, called sinusoidal neural fields (SNFs), and study how it should be initialized to maximize the training speed. We find that the standard initialization scheme for SNFs -- designed based on the signal propagation principle -- is suboptimal. In particular, we show that by simply multiplying each weight (except for the last layer) by a constant, we can accelerate SNF training by 10$\times$. This method, coined $\textit{weight scaling}$, consistently provides a significant speedup over various data domains, allowing the SNFs to train faster than more recently proposed architectures. To understand why the weight scaling works well, we conduct extensive theoretical and empirical analyses which reveal that the weight scaling not only resolves the spectral bias quite effectively but also enjoys a well-conditioned optimization trajectory.
Authors: Yong Liu, Guo Qin, Xiangdong Huang, Jianmin Wang, Mingsheng Long
Abstract: We present Timer-XL, a generative Transformer for unified time series forecasting. To uniformly predict 1D and 2D time series, we generalize next token prediction, predominantly adopted for causal generation of 1D sequences, to multivariate next token prediction. The proposed paradigm uniformly formulates various forecasting scenarios as a long-context generation problem. We opt for the generative Transformer, which can capture global-range and causal dependencies while providing contextual flexibility, to implement unified forecasting on univariate series characterized by non-stationarity, multivariate time series with complicated dynamics and correlations, and covariate-informed contexts that include both endogenous and exogenous variables. Technically, we propose a universal TimeAttention to facilitate generative Transformers on time series, which can effectively capture fine-grained intra- and inter-series dependencies of flattened time series tokens (patches) and is further strengthened by position embeddings in both temporal and variable dimensions. Timer-XL achieves state-of-the-art performance across challenging forecasting benchmarks through a unified approach. As a large time series model, it demonstrates notable model transferability by large-scale pre-training, as well as contextual flexibility in token lengths, positioning it as a one-for-all forecaster.
Authors: Haokun Chen, Hang Li, Yao Zhang, Gengyuan Zhang, Jinhe Bi, Philip Torr, Jindong Gu, Denis Krompass, Volker Tresp
Abstract: One-Shot Federated Learning (OSFL), a special decentralized machine learning paradigm, has recently gained significant attention. OSFL requires only a single round of client data or model upload, which reduces communication costs and mitigates privacy threats compared to traditional FL. Despite these promising prospects, existing methods face challenges due to client data heterogeneity and limited data quantity when applied to real-world OSFL systems. Recently, Latent Diffusion Models (LDM) have shown remarkable advancements in synthesizing high-quality images through pretraining on large-scale datasets, thereby presenting a potential solution to overcome these issues. However, directly applying pretrained LDM to heterogeneous OSFL results in significant distribution shifts in synthetic data, leading to performance degradation in classification models trained on such data. This issue is particularly pronounced in rare domains, such as medical imaging, which are underrepresented in LDM's pretraining data. To address this challenge, we propose Federated Bi-Level Personalization (FedBiP), which personalizes the pretrained LDM at both instance-level and concept-level. Hereby, FedBiP synthesizes images following the client's local data distribution without compromising the privacy regulations. FedBiP is also the first approach to simultaneously address feature space heterogeneity and client data scarcity in OSFL. Our method is validated through extensive experiments on three OSFL benchmarks with feature space heterogeneity, as well as on challenging medical and satellite image datasets with label heterogeneity. The results demonstrate the effectiveness of FedBiP, which substantially outperforms other OSFL methods.
Authors: Manuel Brenner, Elias Weber, Georgia Koppe, Daniel Durstewitz
Abstract: In science, we are often interested in obtaining a generative model of the underlying system dynamics from observed time series. While powerful methods for dynamical systems reconstruction (DSR) exist when data come from a single domain, how to best integrate data from multiple dynamical regimes and leverage it for generalization is still an open question. This becomes particularly important when individual time series are short, and group-level information may help to fill in for gaps in single-domain data. At the same time, averaging is not an option in DSR, as it will wipe out crucial dynamical properties (e.g., limit cycles in one domain vs. chaos in another). Hence, a framework is needed that enables to efficiently harvest group-level (multi-domain) information while retaining all single-domain dynamical characteristics. Here we provide such a hierarchical approach and showcase it on popular DSR benchmarks, as well as on neuroscientific and medical time series. In addition to faithful reconstruction of all individual dynamical regimes, our unsupervised methodology discovers common low-dimensional feature spaces in which datasets with similar dynamics cluster. The features spanning these spaces were further dynamically highly interpretable, surprisingly in often linear relation to control parameters that govern the dynamics of the underlying system. Finally, we illustrate transfer learning and generalization to new parameter regimes.
Authors: MoonJeong Park, Dongwoo Kim
Abstract: Oversmoothing has been claimed as a primary bottleneck for multi-layered graph neural networks (GNNs). Multiple analyses have examined how and why oversmoothing occurs. However, none of the prior work addressed how optimization is performed under the oversmoothing regime. In this work, we show the presence of $\textit{gradient oversmoothing}$ preventing optimization during training. We further analyze that GNNs with residual connections, a well-known solution to help gradient flow in deep architecture, introduce $\textit{gradient expansion}$, a phenomenon of the gradient explosion in diverse directions. Therefore, adding residual connections cannot be a solution for making a GNN deep. Our analysis reveals that constraining the Lipschitz bound of each layer can neutralize the gradient expansion. To this end, we provide a simple yet effective normalization method to prevent the gradient expansion. An empirical study shows that the residual GNNs with hundreds of layers can be efficiently trained with the proposed normalization without compromising performance. Additional studies show that the empirical observations corroborate our theoretical analysis.
Authors: Elvis Dohmatob, Yunzhen Feng, Julia Kempe
Abstract: Within the scaling laws paradigm, which underpins the training of large neural networks like ChatGPT and Llama, we consider a supervised regression setting and establish the existance of a strong form of the model collapse phenomenon, a critical performance degradation due to synthetic data in the training corpus. Our results show that even the smallest fraction of synthetic data (e.g., as little as 1\% of the total training dataset) can still lead to model collapse: larger and larger training sets do not enhance performance. We further investigate whether increasing model size, an approach aligned with current trends in training large language models, exacerbates or mitigates model collapse. In a simplified regime where neural networks are approximated via random projections of tunable size, we both theoretically and empirically show that larger models can amplify model collapse. Interestingly, our theory also indicates that, beyond the interpolation threshold (which can be extremely high for very large datasets), larger models may mitigate the collapse, although they do not entirely prevent it. Our theoretical findings are empirically verified through experiments on language models and feed-forward neural networks for images.
Authors: Ao Hu, Dongkai Wang, Yong Dai, Shiyi Qi, Liangjian Wen, Jun Wang, Zhi Chen, Xun Zhou, Zenglin Xu, Jiang Duan
Abstract: Time series forecasting is extensively applied across diverse domains. Transformer-based models demonstrate significant potential in modeling cross-time and cross-variable interaction. However, we notice that the cross-variable correlation of multivariate time series demonstrates multifaceted (positive and negative correlations) and dynamic progression over time, which is not well captured by existing Transformer-based models. To address this issue, we propose a TimeCNN model to refine cross-variable interactions to enhance time series forecasting. Its key innovation is timepoint-independent, where each time point has an independent convolution kernel, allowing each time point to have its independent model to capture relationships among variables. This approach effectively handles both positive and negative correlations and adapts to the evolving nature of variable relationships over time. Extensive experiments conducted on 12 real-world datasets demonstrate that TimeCNN consistently outperforms state-of-the-art models. Notably, our model achieves significant reductions in computational requirements (approximately 60.46%) and parameter count (about 57.50%), while delivering inference speeds 3 to 4 times faster than the benchmark iTransformer model
Authors: Yu Chen, Juntong Lin, Zhichao Shu
Abstract: We have developed a high-performance Chinese Chess AI that operates without reliance on search algorithms. This AI has demonstrated the capability to compete at a level commensurate with the top 0.1\% of human players. By eliminating the search process typically associated with such systems, this AI achieves a Queries Per Second (QPS) rate that exceeds those of systems based on the Monte Carlo Tree Search (MCTS) algorithm by over a thousandfold and surpasses those based on the AlphaBeta pruning algorithm by more than a hundredfold. The AI training system consists of two parts: supervised learning and reinforcement learning. Supervised learning provides an initial human-like Chinese chess AI, while reinforcement learning, based on supervised learning, elevates the strength of the entire AI to a new level. Based on this training system, we carried out enough ablation experiments and discovered that 1. The same parameter amount of Transformer architecture has a higher performance than CNN on Chinese chess; 2. Possible moves of both sides as features can greatly improve the training process; 3. Selective opponent pool, compared to pure self-play training, results in a faster improvement curve and a higher strength limit. 4. Value Estimation with Cutoff(VECT) improves the original PPO algorithm training process and we will give the explanation.
Authors: Bingrui Li, Wei Huang, Andi Han, Zhanpeng Zhou, Taiji Suzuki, Jun Zhu, Jianfei Chen
Abstract: The Adam optimizer is widely used for transformer optimization in practice, which makes understanding the underlying optimization mechanisms an important problem. However, due to the Adam's complexity, theoretical analysis of how it optimizes transformers remains a challenging task. Fortunately, Sign Gradient Descent (SignGD) serves as an effective surrogate for Adam. Despite its simplicity, theoretical understanding of how SignGD optimizes transformers still lags behind. In this work, we study how SignGD optimizes a two-layer transformer -- consisting of a softmax attention layer with trainable query-key parameterization followed by a linear layer -- on a linearly separable noisy dataset. We identify four stages in the training dynamics, each exhibiting intriguing behaviors. Based on the training dynamics, we prove the fast convergence but poor generalization of the learned transformer on the noisy dataset. We also show that Adam behaves similarly to SignGD in terms of both optimization and generalization in this setting. Additionally, we find that the poor generalization of SignGD is not solely due to data noise, suggesting that both SignGD and Adam requires high-quality data for real-world tasks. Finally, experiments on synthetic and real-world datasets empirically support our theoretical results.
Authors: Lars Henry Berge Olsen, Martin Jullum
Abstract: Shapley values are a popular model-agnostic explanation framework for explaining predictions made by complex machine learning models. The framework provides feature contribution scores that sum to the predicted response and represent each feature's importance. The computation of exact Shapley values is computationally expensive due to estimating an exponential amount of non-trivial conditional expectations. The KernelSHAP framework enables us to approximate the Shapley values using a sampled subset of weighted conditional expectations. We propose three main novel contributions: a stabilizing technique to reduce the variance of the weights in the current state-of-the-art strategy, a novel weighing scheme that corrects the Shapley kernel weights based on sampled subsets, and a straightforward strategy that includes the important subsets and integrates them with the corrected Shapley kernel weights. We compare these new approximation strategies against existing ones by evaluating their Shapley value accuracy as a function of the number of subsets. The results demonstrate that our sampling strategies significantly enhance the accuracy of the approximated Shapley value explanations, making them more reliable in practical applications. This work provides valuable insights and practical recommendations for researchers and practitioners seeking to implement Shapley value-based explainability of their models.
Authors: Arthur Jacot, Peter S\'uken\'ik, Zihan Wang, Marco Mondelli
Abstract: Deep neural networks (DNNs) at convergence consistently represent the training data in the last layer via a highly symmetric geometric structure referred to as neural collapse. This empirical evidence has spurred a line of theoretical research aimed at proving the emergence of neural collapse, mostly focusing on the unconstrained features model. Here, the features of the penultimate layer are free variables, which makes the model data-agnostic and, hence, puts into question its ability to capture DNN training. Our work addresses the issue, moving away from unconstrained features and studying DNNs that end with at least two linear layers. We first prove generic guarantees on neural collapse that assume (i) low training error and balancedness of the linear layers (for within-class variability collapse), and (ii) bounded conditioning of the features before the linear part (for orthogonality of class-means, as well as their alignment with weight matrices). We then show that such assumptions hold for gradient descent training with weight decay: (i) for networks with a wide first layer, we prove low training error and balancedness, and (ii) for solutions that are either nearly optimal or stable under large learning rates, we additionally prove the bounded conditioning. Taken together, our results are the first to show neural collapse in the end-to-end training of DNNs.
Authors: {\L}ukasz Staniszewski, Katarzyna Zaleska, Kamil Deja
Abstract: Recent personalization methods for diffusion models, such as Dreambooth, allow fine-tuning pre-trained models to generate new concepts. However, applying these techniques across multiple tasks in order to include, e.g., several new objects or styles, leads to mutual interference between their adapters. While recent studies attempt to mitigate this issue by combining trained adapters across tasks after fine-tuning, we adopt a more rigorous regime and investigate the personalization of large diffusion models under a continual learning scenario, where such interference leads to catastrophic forgetting of previous knowledge. To that end, we evaluate the na\"ive continual fine-tuning of customized models and compare this approach with three methods for consecutive adapters' training: sequentially merging new adapters, merging orthogonally initialized adapters, and updating only relevant parameters according to the task. In our experiments, we show that the proposed approaches mitigate forgetting when compared to the na\"ive approach.
Authors: Xiao Yang, Kai Zhou, Yuni Lai, Gaolei Li
Abstract: With the trend of large graph learning models, business owners tend to employ a model provided by a third party to deliver business services to users. However, these models might be backdoored, and malicious users can submit trigger-embedded inputs to manipulate the model predictions. Current graph backdoor defenses have several limitations: 1) depending on model-related details, 2) requiring additional model fine-tuning, and 3) relying upon extra explainability tools, all of which are infeasible under stringent privacy policies. To address those limitations, we propose GraphProt, which allows resource-constrained business owners to rely on third parties to avoid backdoor attacks on GNN-based graph classifiers. Our GraphProt is model-agnostic and only relies on the input graph. The key insight is to leverage subgraph information for prediction, thereby mitigating backdoor effects induced by triggers. GraphProt comprises two components: clustering-based trigger elimination and robust subgraph ensemble. Specifically, we first propose feature-topology clustering that aims to remove most of the anomalous subgraphs (triggers). Moreover, we design subgraph sampling strategies based on feature-topology clustering to build a robust classifier via majority vote. Experimental results across three backdoor attacks and six benchmark datasets demonstrate that GraphProt significantly reduces the backdoor attack success rate while preserving the model accuracy on regular graph classification tasks.
Authors: Tankred Saanum, Luca M. Schulze Buschoff, Peter Dayan, Eric Schulz
Abstract: Compositional representations are thought to enable humans to generalize across combinatorially vast state spaces. Models with learnable object slots, which encode information about objects in separate latent codes, have shown promise for this type of generalization but rely on strong architectural priors. Models with distributed representations, on the other hand, use overlapping, potentially entangled neural codes, and their ability to support compositional generalization remains underexplored. In this paper we examine whether distributed models can develop linearly separable representations of objects, like slotted models, through unsupervised training on videos of object interactions. We show that, surprisingly, models with distributed representations often match or outperform models with object slots in downstream prediction tasks. Furthermore, we find that linearly separable object representations can emerge without object-centric priors, with auxiliary objectives like next-state prediction playing a key role. Finally, we observe that distributed models' object representations are never fully disentangled, even if they are linearly separable: Multiple objects can be encoded through partially overlapping neural populations while still being highly separable with a linear classifier. We hypothesize that maintaining partially shared codes enables distributed models to better compress object dynamics, potentially enhancing generalization.
Authors: Irene Cannistraci, Emanuele Rodol\`a, Bastian Rieck
Abstract: Deep neural networks often learn similar internal representations, both across different models and within their own layers. While inter-network similarities have enabled techniques such as model stitching and merging, intra-network similarities present new opportunities for designing more efficient architectures. In this paper, we investigate the emergence of these internal similarities across different layers in diverse neural architectures, showing that similarity patterns emerge independently of the datataset used. We introduce a simple metric, Block Redundancy, to detect redundant blocks, providing a foundation for future architectural optimization methods. Building on this, we propose Redundant Blocks Approximation (RBA), a general framework that identifies and approximates one or more redundant computational blocks using simpler transformations. We show that the transformation $\mathcal{T}$ between two representations can be efficiently computed in closed-form, and it is enough to replace the redundant blocks from the network. RBA reduces model parameters and time complexity while maintaining good performance. We validate our method on classification tasks in the vision domain using a variety of pretrained foundational models and datasets.
Authors: Emanuele Sansone, Tim Lebailly, Tinne Tuytelaars
Abstract: We identify sufficient conditions to avoid known failure modes, including representation, dimensional, cluster and intracluster collapses, occurring in non-contrastive self-supervised learning. Based on these findings, we propose a principled design for the projector and loss function. We theoretically demonstrate that this design introduces an inductive bias that promotes learning representations that are both decorrelated and clustered without explicit enforcing these properties and leading to improved generalization. To the best of our knowledge, this is the first solution that achieves robust training with respect to these failure modes while guaranteeing enhanced generalization performance in downstream tasks. We validate our theoretical findings on image datasets including SVHN, CIFAR10, CIFAR100 and ImageNet-100, and show that our solution, dubbed FALCON, outperforms existing feature decorrelation and cluster-based self-supervised learning methods in terms of generalization to clustering and linear classification tasks.
Authors: Jasmine Bayrooti, Carl Henrik Ek, Amanda Prorok
Abstract: Learning complex robot behavior through interactions with the environment necessitates principled exploration. Effective strategies should prioritize exploring regions of the state-action space that maximize rewards, with optimistic exploration emerging as a promising direction aligned with this idea and enabling sample-efficient reinforcement learning. However, existing methods overlook a crucial aspect: the need for optimism to be informed by a belief connecting the reward and state. To address this, we propose a practical, theoretically grounded approach to optimistic exploration based on Thompson sampling. Our model structure is the first that allows for reasoning about joint uncertainty over transitions and rewards. We apply our method on a set of MuJoCo and VMAS continuous control tasks. Our experiments demonstrate that optimistic exploration significantly accelerates learning in environments with sparse rewards, action penalties, and difficult-to-explore regions. Furthermore, we provide insights into when optimism is beneficial and emphasize the critical role of model uncertainty in guiding exploration.
Authors: Hugo Thimonier, Jos\'e Lucas De Melo Costa, Fabrice Popineau, Arpad Rimmel, Bich-Li\^en Doan
Abstract: Self-supervision is often used for pre-training to foster performance on a downstream task by constructing meaningful representations of samples. Self-supervised learning (SSL) generally involves generating different views of the same sample and thus requires data augmentations that are challenging to construct for tabular data. This constitutes one of the main challenges of self-supervision for structured data. In the present work, we propose a novel augmentation-free SSL method for tabular data. Our approach, T-JEPA, relies on a Joint Embedding Predictive Architecture (JEPA) and is akin to mask reconstruction in the latent space. It involves predicting the latent representation of one subset of features from the latent representation of a different subset within the same sample, thereby learning rich representations without augmentations. We use our method as a pre-training technique and train several deep classifiers on the obtained representation. Our experimental results demonstrate a substantial improvement in both classification and regression tasks, outperforming models trained directly on samples in their original data space. Moreover, T-JEPA enables some methods to consistently outperform or match the performance of traditional methods likes Gradient Boosted Decision Trees. To understand why, we extensively characterize the obtained representations and show that T-JEPA effectively identifies relevant features for downstream tasks without access to the labels. Additionally, we introduce regularization tokens, a novel regularization method critical for training of JEPA-based models on structured data.
Authors: Pol G. Recasens, \'Ad\'am Horv\'ath, Alberto Gutierrez-Torre, Jordi Torres, Josep Ll. Berral, Bal\'azs Pej\'o
Abstract: Federated learning is increasingly popular as it enables multiple parties with limited datasets and resources to train a high-performing machine learning model collaboratively. However, similarly to other collaborative systems, federated learning is vulnerable to free-riders -- participants who do not contribute to the training but still benefit from the shared model. Free-riders not only compromise the integrity of the learning process but also slow down the convergence of the global model, resulting in increased costs for the honest participants. To address this challenge, we propose FRIDA: free-rider detection using privacy attacks, a framework that leverages inference attacks to detect free-riders. Unlike traditional methods that only capture the implicit effects of free-riding, FRIDA directly infers details of the underlying training datasets, revealing characteristics that indicate free-rider behaviour. Through extensive experiments, we demonstrate that membership and property inference attacks are effective for this purpose. Our evaluation shows that FRIDA outperforms state-of-the-art methods, especially in non-IID settings.
Authors: Alex Iacob, Lorenzo Sani, Meghdad Kurmanji, William F. Shen, Xinchi Qiu, Dongqi Cai, Yan Gao, Nicholas D. Lane
Abstract: Language Model pre-training benefits from a broader data mixture to enhance performance across domains and languages. However, training on such heterogeneous text corpora is complex, requiring extensive and cost-intensive efforts. Since these data sources vary in lexical, syntactic, and semantic aspects, they cause negative interference or the "curse of multilinguality". We propose a novel pre-training framework to alleviate this curse. Our method, DEPT, decouples the embedding layers from the transformer body while simultaneously training the latter in multiple contexts. DEPT enables the model to train without being bound to a shared global vocabulary. DEPT: (1) can train robustly and effectively under significant data heterogeneity, (2) reduces the parameter count of the token embeddings by up to 80% and the communication costs by 675x for billion-scale models (3) enhances model generalization and plasticity in adapting to new languages and domains, and (4) allows training with custom optimized vocabulary per data source. We prove DEPT's potential by performing the first vocabulary-agnostic federated multilingual pre-training of a 1.3 billion-parameter model across high and low-resource languages, reducing its parameter count by 409 million.
Authors: Marco Bagatella, Jonas H\"ubotter, Georg Martius, Andreas Krause
Abstract: Pre-trained generalist policies are rapidly gaining relevance in robot learning due to their promise of fast adaptation to novel, in-domain tasks. This adaptation often relies on collecting new demonstrations for a specific task of interest and applying imitation learning algorithms, such as behavioral cloning. However, as soon as several tasks need to be learned, we must decide which tasks should be demonstrated and how often? We study this multi-task problem and explore an interactive framework in which the agent adaptively selects the tasks to be demonstrated. We propose AMF (Active Multi-task Fine-tuning), an algorithm to maximize multi-task policy performance under a limited demonstration budget by collecting demonstrations yielding the largest information gain on the expert policy. We derive performance guarantees for AMF under regularity assumptions and demonstrate its empirical effectiveness to efficiently fine-tune neural policies in complex and high-dimensional environments.
Authors: Adam Kania, Marko Mihajlovic, Sergey Prokudin, Jacek Tabor, Przemys{\l}aw Spurek
Abstract: Implicit Neural Representations (INRs) have recently gained attention as a powerful approach for continuously representing signals such as images, videos, and 3D shapes using multilayer perceptrons (MLPs). However, MLPs are known to exhibit a low-frequency bias, limiting their ability to capture high-frequency details accurately. This limitation is typically addressed by incorporating high-frequency input embeddings or specialized activation layers. In this work, we demonstrate that these embeddings and activations are often configured with hyperparameters that perform well on average but are suboptimal for specific input signals under consideration, necessitating a costly grid search to identify optimal settings. Our key observation is that the initial frequency spectrum of an untrained model's output correlates strongly with the model's eventual performance on a given target signal. Leveraging this insight, we propose frequency shifting (or FreSh), a method that selects embedding hyperparameters to align the frequency spectrum of the model's initial output with that of the target signal. We show that this simple initialization technique improves performance across various neural representation methods and tasks, achieving results comparable to extensive hyperparameter sweeps but with only marginal computational overhead compared to training a single model with default hyperparameters.
Authors: Han Qi (Harvard University), Haocheng Yin (Harvard University, ETH Z\"urich), Heng Yang (ETH Z\"urich)
Abstract: We initiate a study of the geometry of the visual representation space -- the information channel from the vision encoder to the action decoder -- in an image-based control pipeline learned from behavior cloning. Inspired by the phenomenon of neural collapse (NC) in image classification, we investigate whether a similar law of clustering emerges in the visual representation space. Since image-based control is a regression task without explicitly defined classes, the central piece of the puzzle lies in determining according to what implicit classes the visual features cluster, if such a law exists. Focusing on image-based planar pushing, we posit the most important role of the visual representation in a control task is to convey a goal to the action decoder. We then classify training samples of expert demonstrations into eight "control-oriented" classes based on (a) the relative pose between the object and the target in the input or (b) the relative pose of the object induced by expert actions in the output, where one class corresponds to one relative pose orthant (REPO). Across four different instantiations of architecture, we report the prevalent emergence of control-oriented clustering in the visual representation space according to the eight REPOs. Beyond empirical observation, we show such a law of clustering can be leveraged as an algorithmic tool to improve test-time performance when training a policy with limited expert demonstrations. Particularly, we pretrain the vision encoder using NC as a regularization to encourage control-oriented clustering of the visual features. Surprisingly, such an NC-pretrained vision encoder, when finetuned end-to-end with the action decoder, boosts the test-time performance by 10% to 35% in the low-data regime. Real-world vision-based planar pushing experiments confirmed the surprising advantage of control-oriented visual representation pretraining.
Authors: Andrew Lamperski, Siddharth Salapaka
Abstract: Neural networks are widely used to approximate unknown functions in control. A common neural network architecture uses a single hidden layer (i.e. a shallow network), in which the input parameters are fixed in advance and only the output parameters are trained. The typical formal analysis asserts that if output parameters exist to approximate the unknown function with sufficient accuracy, then desired control performance can be achieved. A long-standing theoretical gap was that no conditions existed to guarantee that, for the fixed input parameters, required accuracy could be obtained by training the output parameters. Our recent work has partially closed this gap by demonstrating that if input parameters are chosen randomly, then for any sufficiently smooth function, with high-probability there are output parameters resulting in $O((1/m)^{1/2})$ approximation errors, where $m$ is the number of neurons. However, some applications, notably continuous-time value function approximation, require that the network approximates the both the unknown function and its gradient with sufficient accuracy. In this paper, we show that randomly generated input parameters and trained output parameters result in gradient errors of $O((\log(m)/m)^{1/2})$, and additionally, improve the constants from our prior work. We show how to apply the result to policy evaluation problems.
Authors: Lijie Yang, Zhihao Zhang, Zhuofu Chen, Zikun Li, Zhihao Jia
Abstract: Large language models (LLMs) have driven significant advancements across diverse NLP tasks, with long-context models gaining prominence for handling extended inputs. However, the expanding key-value (KV) cache size required by Transformer architectures intensifies the memory constraints, particularly during the decoding phase, creating a significant bottleneck. Existing sparse attention mechanisms designed to address this bottleneck have two limitations: (1) they often fail to reliably identify the most relevant tokens for attention, and (2) they overlook the spatial coherence of token selection across consecutive Transformer layers, which can lead to performance degradation and substantial overhead in token selection. This paper introduces TidalDecode, a simple yet effective algorithm and system for fast and accurate LLM decoding through position persistent sparse attention. TidalDecode leverages the spatial coherence of tokens selected by existing sparse attention methods and introduces a few token selection layers that perform full attention to identify the tokens with the highest attention scores, while all other layers perform sparse attention with the pre-selected tokens. This design enables TidalDecode to substantially reduce the overhead of token selection for sparse attention without sacrificing the quality of the generated results. Evaluation on a diverse set of LLMs and tasks shows that TidalDecode closely matches the generative performance of full attention methods while reducing the LLM decoding latency by up to 2.1x.
Authors: David Heurtel-Depeiges, Anian Ruoss, Joel Veness, Tim Genewein
Abstract: Foundation models have recently been shown to be strong data compressors. However, when accounting for their excessive parameter count, their compression ratios are actually inferior to standard compression algorithms. Moreover, naively reducing the number of parameters may not necessarily help as it leads to worse predictions and thus weaker compression. In this paper, we conduct a large-scale empirical study to investigate whether there is a sweet spot where competitive compression ratios with pre-trained vanilla transformers are possible. To this end, we train families of models on 165GB of raw byte sequences of either text, image, or audio data (and all possible combinations of the three) and then compress 1GB of out-of-distribution (OOD) data from each modality. We find that relatively small models (i.e., millions of parameters) can outperform standard general-purpose compression algorithms (gzip, LZMA2) and even domain-specific compressors (PNG, JPEG 2000, FLAC) - even when factoring in parameter count. We achieve, e.g., the lowest compression ratio of 0.49 on OOD audio data (vs. 0.54 for FLAC). To study the impact of model- and dataset scale, we conduct extensive ablations and hyperparameter sweeps, and we investigate the effect of unimodal versus multimodal training. We find that even small models can be trained to perform well on multiple modalities, but, in contrast to previously reported results with large-scale foundation models, transfer to unseen modalities is generally weak.
Authors: Xinyu Zhou, Simin Fan, Martin Jaggi
Abstract: Influence functions provide a principled method to assess the contribution of individual training samples to a specific target. Yet, their high computational costs limit their applications on large-scale models and datasets. Existing methods proposed for influence function approximation have significantly reduced the computational overheads. However, they mostly suffer from inaccurate estimation due to the lack of strong convergence guarantees from the algorithm. The family of hyperpower methods are well-known for their rigorous convergence guarantees on matrix inverse approximation, while the matrix multiplication operation can involve intractable memory and computation costs on large-scale models. We propose HyperINF, an efficient and accurate influence function approximation method which leverages the hyperpower method, specifically Schulz's iterative algorithm. To deal with the computation-intensive matrix multiplication, we incorporate the generalized fisher information (GFIM) as a low-rank approximation of the Hessian matrix, which reduces the memory and computation overheads to constant costs independent of ranks on LoRA-tuned models. We first demonstrate the superior accuracy and stability of \method compared to other baselines through a synthetic convergence simulation for matrix inversion. We further validate the efficacy of \method through extensive real-world data attribution tasks, including mislabeled data detection and data selection for LLM and VLM fine-tuning. On LoRA-tuned models, HyperINF achieves superior downstream performance with minimal memory and computational overhead, while other baselines suffer from significant degradation. Our codebase is available at https://github.com/Blackzxy/HyperINF.
Authors: Konstantin Sch\"urholt
Abstract: This thesis addresses the challenge of understanding Neural Networks through the lens of their most fundamental component: the weights, which encapsulate the learned information and determine the model behavior. At the core of this thesis is a fundamental question: Can we learn general, task-agnostic representations from populations of Neural Network models? The key contribution of this thesis to answer that question are hyper-representations, a self-supervised method to learn representations of NN weights. Work in this thesis finds that trained NN models indeed occupy meaningful structures in the weight space, that can be learned and used. Through extensive experiments, this thesis demonstrates that hyper-representations uncover model properties, such as their performance, state of training, or hyperparameters. Moreover, the identification of regions with specific properties in hyper-representation space allows to sample and generate model weights with targeted properties. This thesis demonstrates applications for fine-tuning, and transfer learning to great success. Lastly, it presents methods that allow hyper-representations to generalize beyond model sizes, architectures, and tasks. The practical implications of that are profound, as it opens the door to foundation models of Neural Networks, which aggregate and instantiate their knowledge across models and architectures. Ultimately, this thesis contributes to the deeper understanding of Neural Networks by investigating structures in their weights which leads to more interpretable, efficient, and adaptable models. By laying the groundwork for representation learning of NN weights, this research demonstrates the potential to change the way Neural Networks are developed, analyzed, and used.
Authors: Ayano Hiranaka, Shang-Fu Chen, Chieh-Hsin Lai, Dongjun Kim, Naoki Murata, Takashi Shibuya, Wei-Hsiang Liao, Shao-Hua Sun, Yuki Mitsufuji
Abstract: Controllable generation through Stable Diffusion (SD) fine-tuning aims to improve fidelity, safety, and alignment with human guidance. Existing reinforcement learning from human feedback methods usually rely on predefined heuristic reward functions or pretrained reward models built on large-scale datasets, limiting their applicability to scenarios where collecting such data is costly or difficult. To effectively and efficiently utilize human feedback, we develop a framework, HERO, which leverages online human feedback collected on the fly during model learning. Specifically, HERO features two key mechanisms: (1) Feedback-Aligned Representation Learning, an online training method that captures human feedback and provides informative learning signals for fine-tuning, and (2) Feedback-Guided Image Generation, which involves generating images from SD's refined initialization samples, enabling faster convergence towards the evaluator's intent. We demonstrate that HERO is 4x more efficient in online feedback for body part anomaly correction compared to the best existing method. Additionally, experiments show that HERO can effectively handle tasks like reasoning, counting, personalization, and reducing NSFW content with only 0.5K online feedback.
Authors: Fan Chen, Dylan J. Foster, Yanjun Han, Jian Qian, Alexander Rakhlin, Yunbei Xu
Abstract: In this paper, we develop a unified framework for lower bound methods in statistical estimation and interactive decision making. Classical lower bound techniques -- such as Fano's inequality, Le Cam's method, and Assouad's lemma -- have been central to the study of minimax risk in statistical estimation, yet they are insufficient for the analysis of methods that collect data in an interactive manner. The recent minimax lower bounds for interactive decision making via the Decision-Estimation Coefficient (DEC) appear to be genuinely different from the classical methods. We propose a unified view of these distinct methodologies through a general algorithmic lower bound method. We further introduce a novel complexity measure, decision dimension, which facilitates the derivation of new lower bounds for interactive decision making. In particular, decision dimension provides a characterization of bandit learnability for any structured bandit model class. Further, we characterize the sample complexity of learning convex model class up to a polynomial gap with the decision dimension, addressing the remaining gap between upper and lower bounds in Foster et al. (2021, 2023).
Authors: Ali Ebrahimpour-Boroojeny, Hari Sundaram, Varun Chandrasekaran
Abstract: Transferability of adversarial examples is a well-known property that endangers all classification models, even those that are only accessible through black-box queries. Prior work has shown that an ensemble of models is more resilient to transferability: the probability that an adversarial example is effective against most models of the ensemble is low. Thus, most ongoing research focuses on improving ensemble diversity. Another line of prior work has shown that Lipschitz continuity of the models can make models more robust since it limits how a model's output changes with small input perturbations. In this paper, we study the effect of Lipschitz continuity on transferability rates. We show that although a lower Lipschitz constant increases the robustness of a single model, it is not as beneficial in training robust ensembles as it increases the transferability rate of adversarial examples across models in the ensemble. Therefore, we introduce LOTOS, a new training paradigm for ensembles, which counteracts this adverse effect. It does so by promoting orthogonality among the top-$k$ sub-spaces of the transformations of the corresponding affine layers of any pair of models in the ensemble. We theoretically show that $k$ does not need to be large for convolutional layers, which makes the computational overhead negligible. Through various experiments, we show LOTOS increases the robust accuracy of ensembles of ResNet-18 models by $6$ percentage points (p.p) against black-box attacks on CIFAR-10. It is also capable of combining with the robustness of prior state-of-the-art methods for training robust ensembles to enhance their robust accuracy by $10.7$ p.p.
Authors: Yifan Yang, Hao Ban, Minhui Huang, Shiqian Ma, Kaiyi Ji
Abstract: Bilevel optimization has recently attracted considerable attention due to its abundant applications in machine learning problems. However, existing methods rely on prior knowledge of problem parameters to determine stepsizes, resulting in significant effort in tuning stepsizes when these parameters are unknown. In this paper, we propose two novel tuning-free algorithms, D-TFBO and S-TFBO. D-TFBO employs a double-loop structure with stepsizes adaptively adjusted by the "inverse of cumulative gradient norms" strategy. S-TFBO features a simpler fully single-loop structure that updates three variables simultaneously with a theory-motivated joint design of adaptive stepsizes for all variables. We provide a comprehensive convergence analysis for both algorithms and show that D-TFBO and S-TFBO respectively require $O(\frac{1}{\epsilon})$ and $O(\frac{1}{\epsilon}\log^4(\frac{1}{\epsilon}))$ iterations to find an $\epsilon$-accurate stationary point, (nearly) matching their well-tuned counterparts using the information of problem parameters. Experiments on various problems show that our methods achieve performance comparable to existing well-tuned approaches, while being more robust to the selection of initial stepsizes. To the best of our knowledge, our methods are the first to completely eliminate the need for stepsize tuning, while achieving theoretical guarantees.
Authors: Mengjian Hua, Matthieu Lauri\`ere, Eric Vanden-Eijnden
Abstract: We propose a simulation-free algorithm for the solution of generic problems in stochastic optimal control (SOC). Unlike existing methods, our approach does not require the solution of an adjoint problem, but rather leverages Girsanov theorem to directly calculate the gradient of the SOC objective on-policy. This allows us to speed up the optimization of control policies parameterized by neural networks since it completely avoids the expensive back-propagation step through stochastic differential equations (SDEs) used in the Neural SDE framework. In particular, it enables us to solve SOC problems in high dimension and on long time horizons. We demonstrate the efficiency of our approach in various domains of applications, including standard stochastic optimal control problems, sampling from unnormalized distributions via construction of a Schr\"odinger-F\"ollmer process, and fine-tuning of pre-trained diffusion models. In all cases our method is shown to outperform the existing methods in both the computing time and memory efficiency.
Authors: Kaiyue Wen, Zhiyuan Li, Jason Wang, David Hall, Percy Liang, Tengyu Ma
Abstract: Training language models currently requires pre-determining a fixed compute budget because the typical cosine learning rate schedule depends on the total number of steps. In contrast, the Warmup-Stable-Decay (WSD) schedule uses a constant learning rate to produce a main branch of iterates that can in principle continue indefinitely without a pre-specified compute budget. Then, given any compute budget, one can branch out from the main branch at a proper at any time with a rapidly decaying learning rate to produce a strong model. Empirically, WSD generates a non-traditional loss curve: the loss remains elevated during the stable phase but sharply declines during the decay phase. Towards explaining this phenomenon, we conjecture that pretraining loss exhibits a river valley landscape, which resembles a deep valley with a river at its bottom. Under this assumption, we show that during the stable phase, the iterate undergoes large oscillations due to the high learning rate, yet it progresses swiftly along the river. During the decay phase, the rapidly dropping learning rate minimizes the iterate's oscillations, moving it closer to the river and revealing true optimization progress. Therefore, the sustained high learning rate phase and fast decaying phase are responsible for progress in the river and the mountain directions respectively, and are both critical. Our analysis predicts phenomenons consistent with empirical observations and shows that this landscape can emerge from pretraining on a simple bi-gram dataset. Inspired by the theory, we introduce WSD-S, a variant of WSD that reuses previous checkpoints' decay phases and keeps only one main branch, where we resume from a decayed checkpoint. WSD-S empirically outperforms WSD and Cyclic-Cosine in obtaining multiple language model checkpoints across various compute budgets in a single run for parameters scaling from 0.1B to 1.2B.
Authors: Toni J. B. Liu, Nicolas Boull\'e, Rapha\"el Sarfati, Christopher J. Earls
Abstract: Large language models (LLMs) demonstrate remarkable emergent abilities to perform in-context learning across various tasks, including time series forecasting. This work investigates LLMs' ability to estimate probability density functions (PDFs) from data observed in-context; such density estimation (DE) is a fundamental task underlying many probabilistic modeling problems. We leverage the Intensive Principal Component Analysis (InPCA) to visualize and analyze the in-context learning dynamics of LLaMA-2 models. Our main finding is that these LLMs all follow similar learning trajectories in a low-dimensional InPCA space, which are distinct from those of traditional density estimation methods like histograms and Gaussian kernel density estimation (KDE). We interpret the LLaMA in-context DE process as a KDE with an adaptive kernel width and shape. This custom kernel model captures a significant portion of LLaMA's behavior despite having only two parameters. We further speculate on why LLaMA's kernel width and shape differs from classical algorithms, providing insights into the mechanism of in-context probabilistic reasoning in LLMs.
Authors: Riccardo Fogliato, Pratik Patil, Nil-Jana Akpinar, Mathew Monfort
Abstract: How can we precisely estimate a large language model's (LLM) accuracy on questions belonging to a specific topic within a larger question-answering dataset? The standard direct estimator, which averages the model's accuracy on the questions in each subgroup, may exhibit high variance for subgroups (topics) with small sample sizes. Synthetic regression modeling, which leverages the model's accuracy on questions about other topics, may yield biased estimates that are too unreliable for large subgroups. We prescribe a simple yet effective solution: an empirical Bayes (EB) estimator that balances direct and regression estimates for each subgroup separately, improving the precision of subgroup-level estimates of model performance. Our experiments on multiple datasets show that this approach consistently provides more precise estimates of the LLM performance compared to the direct and regression approaches, achieving substantial reductions in the mean squared error. Confidence intervals for EB estimates also have near-nominal coverage and are narrower compared to those for the direct estimator. Additional experiments on tabular and vision data validate the benefits of this EB approach.
Authors: Ehsan Futuhi, Shayan Karimi, Chao Gao, Martin M\"uller
Abstract: We consider deep deterministic policy gradient (DDPG) in the context of reinforcement learning with sparse rewards. To enhance exploration, we introduce a search procedure, \emph{${\epsilon}{t}$-greedy}, which generates exploratory options for exploring less-visited states. We prove that search using $\epsilon t$-greedy has polynomial sample complexity under mild MDP assumptions. To more efficiently use the information provided by rewarded transitions, we develop a new dual experience replay buffer framework, \emph{GDRB}, and implement \emph{longest n-step returns}. The resulting algorithm, \emph{ETGL-DDPG}, integrates all three techniques: \bm{$\epsilon t$}-greedy, \textbf{G}DRB, and \textbf{L}ongest $n$-step, into DDPG. We evaluate ETGL-DDPG on standard benchmarks and demonstrate that it outperforms DDPG, as well as other state-of-the-art methods, across all tested sparse-reward continuous environments. Ablation studies further highlight how each strategy individually enhances the performance of DDPG in this setting.
Authors: Iman Mirzadeh, Keivan Alizadeh, Hooman Shahrokhi, Oncel Tuzel, Samy Bengio, Mehrdad Farajtabar
Abstract: Recent advancements in Large Language Models (LLMs) have sparked interest in their formal reasoning capabilities, particularly in mathematics. The GSM8K benchmark is widely used to assess the mathematical reasoning of models on grade-school-level questions. While the performance of LLMs on GSM8K has significantly improved in recent years, it remains unclear whether their mathematical reasoning capabilities have genuinely advanced, raising questions about the reliability of the reported metrics. To address these concerns, we conduct a large-scale study on several SOTA open and closed models. To overcome the limitations of existing evaluations, we introduce GSM-Symbolic, an improved benchmark created from symbolic templates that allow for the generation of a diverse set of questions. GSM-Symbolic enables more controllable evaluations, providing key insights and more reliable metrics for measuring the reasoning capabilities of models.Our findings reveal that LLMs exhibit noticeable variance when responding to different instantiations of the same question. Specifically, the performance of all models declines when only the numerical values in the question are altered in the GSM-Symbolic benchmark. Furthermore, we investigate the fragility of mathematical reasoning in these models and show that their performance significantly deteriorates as the number of clauses in a question increases. We hypothesize that this decline is because current LLMs cannot perform genuine logical reasoning; they replicate reasoning steps from their training data. Adding a single clause that seems relevant to the question causes significant performance drops (up to 65%) across all state-of-the-art models, even though the clause doesn't contribute to the reasoning chain needed for the final answer. Overall, our work offers a more nuanced understanding of LLMs' capabilities and limitations in mathematical reasoning.
Authors: Onur Efe, Arkadas Ozakin
Abstract: We develop a new, unsupervised symmetry learning method that starts with raw data, and gives the minimal (discrete) generator of an underlying Lie group of symmetries, together with a symmetry equivariant representation of the data. The method is able to learn the pixel translation operator from a dataset with only an approximate translation symmetry, and can learn quite different types of symmetries which are not apparent to the naked eye, equally well. The method is based on the formulation of an information-theoretic loss function that measures both the degree to which the dataset is symmetric under a given candidate symmetry, and also, the degree of locality of the samples in the dataset with respect to this symmetry. We demonstrate that this coupling between symmetry and locality, together with a special optimization technique developed for entropy estimation, results in a highly stable system that gives reproducible results. The symmetry actions we consider are group representations, however, we believe the approach has the potential to be generalized to more general, nonlinear actions of non-commutative Lie groups.
Authors: Taha Bouhsine, Imad El Aaroussi, Atik Faysal, Wang Huaxia
Abstract: We introduce a novel anchor-free contrastive learning (AFCL) method leveraging our proposed Similarity-Orthogonality (SimO) loss. Our approach minimizes a semi-metric discriminative loss function that simultaneously optimizes two key objectives: reducing the distance and orthogonality between embeddings of similar inputs while maximizing these metrics for dissimilar inputs, facilitating more fine-grained contrastive learning. The AFCL method, powered by SimO loss, creates a fiber bundle topological structure in the embedding space, forming class-specific, internally cohesive yet orthogonal neighborhoods. We validate the efficacy of our method on the CIFAR-10 dataset, providing visualizations that demonstrate the impact of SimO loss on the embedding space. Our results illustrate the formation of distinct, orthogonal class neighborhoods, showcasing the method's ability to create well-structured embeddings that balance class separation with intra-class variability. This work opens new avenues for understanding and leveraging the geometric properties of learned representations in various machine learning tasks.
Authors: Mengzhao Chen, Yi Liu, Jiahao Wang, Yi Bin, Wenqi Shao, Ping Luo
Abstract: Quantization is essential for deploying Large Language Models (LLMs) by enhancing memory efficiency and inference speed. Existing methods for activation quantization mainly address channel-wise outliers, often neglecting token-wise outliers, leading to reliance on costly per-token dynamic quantization. To address this, we introduce PrefixQuant, a novel technique that isolates outlier tokens offline without re-training. Specifically, PrefixQuant identifies high-frequency outlier tokens and prefixes them in the KV cache, preventing the generation of outlier tokens during inference and simplifying quantization. To our knowledge, PrefixQuant is the first to enable efficient per-tensor static quantization to outperform expensive per-token dynamic quantization. For instance, in W4A4KV4 (4- bit weight, 4-bit activation, and 4-bit KV cache) Llama-3-8B, PrefixQuant with per-tensor static quantization achieves a 7.43 WikiText2 perplexity and 71.08% average accuracy on 5 common-sense reasoning tasks, outperforming previous per-token dynamic quantization methods like QuaRot with 0.98 perplexity improvement and +5.98 points accuracy. Additionally, the inference speed of W4A4 quantized models using PrefixQuant is 1.60x to 2.81x faster than FP16 models and exceeds QuaRot models by 1.2x to 1.3x. Our code is available at \url{https://github.com/ChenMnZ/PrefixQuant}.
Authors: Boyu Han, Qianqian Xu, Zhiyong Yang, Shilong Bao, Peisong Wen, Yangbangyan Jiang, Qingming Huang
Abstract: The Area Under the ROC Curve (AUC) is a well-known metric for evaluating instance-level long-tail learning problems. In the past two decades, many AUC optimization methods have been proposed to improve model performance under long-tail distributions. In this paper, we explore AUC optimization methods in the context of pixel-level long-tail semantic segmentation, a much more complicated scenario. This task introduces two major challenges for AUC optimization techniques. On one hand, AUC optimization in a pixel-level task involves complex coupling across loss terms, with structured inner-image and pairwise inter-image dependencies, complicating theoretical analysis. On the other hand, we find that mini-batch estimation of AUC loss in this case requires a larger batch size, resulting in an unaffordable space complexity. To address these issues, we develop a pixel-level AUC loss function and conduct a dependency-graph-based theoretical analysis of the algorithm's generalization ability. Additionally, we design a Tail-Classes Memory Bank (T-Memory Bank) to manage the significant memory demand. Finally, comprehensive experiments across various benchmarks confirm the effectiveness of our proposed AUCSeg method. The code is available at https://github.com/boyuh/AUCSeg.
Authors: Philippe Krajsic, Viola Suess, Zehong Cao, Ryszard Kowalczyk, Bogdan Franczyk
Abstract: Efficient satellite communications play an enormously important role in all of our daily lives. This includes the transmission of data for communication purposes, the operation of IoT applications or the provision of data for ground stations. More and more, AI-based methods are finding their way into these areas. This paper gives an overview of current research in the field of intelligent optimization of satellite communication. For this purpose, a text-mining based literature review was conducted and the identified papers were thematically clustered and analyzed. The identified clusters cover the main topics of routing, resource allocation and, load balancing. Through such a clustering of the literature in overarching topics, a structured analysis of the research papers was enabled, allowing the identification of latest technologies and approaches as well as research needs for intelligent optimization of satellite communication.
Authors: Cynthia C. S. Liem, Do\u{g}a Ta\c{s}c{\i}lar, Andrew M. Demetriou
Abstract: As audio machine learning outcomes are deployed in societally impactful applications, it is important to have a sense of the quality and origins of the data used. Noticing that being explicit about this sense is not trivially rewarded in academic publishing in applied machine learning domains, and neither is included in typical applied machine learning curricula, we present a study into dataset usage connected to the top-5 cited papers at the International Conference on Acoustics, Speech, and Signal Processing (ICASSP). In this, we conduct thorough depth-first analyses towards origins of used datasets, often leading to searches that had to go beyond what was reported in official papers, and ending into unclear or entangled origins. Especially in the current pull towards larger, and possibly generative AI models, awareness of the need for accountability on data provenance is increasing. With this, we call on the community to not only focus on engineering larger models, but create more room and reward for explicitizing the foundations on which such models should be built.
Authors: Itamar Cohen, Gil Einziger
Abstract: Efficient number representation is essential for federated learning, natural language processing, and network measurement solutions. Due to timing, area, and power constraints, such applications use narrow bit-width (e.g., 8-bit) number systems. The widely used floating-point systems exhibit a trade-off between the counting range and accuracy. This paper introduces Floating-Floating-Point (F2P) - a floating point number that varies the partition between mantissa and exponent. Such flexibility leads to a large counting range combined with improved accuracy over a selected sub-range. Our evaluation demonstrates that moving to F2P from the state-of-the-art improves network measurement accuracy and federated learning.
Authors: Gyanendra Shrestha, Soumya Tejaswi Vadlamani
Abstract: People now frequently meet and develop relationships through online dating. Yet, due to their limited accessibility, utilizing dating services can be difficult and irritating for people with visual impairments. The significance of the research issue can be attributed to the fact that dating websites are becoming more and more common and have a significant impact on how people establish romantic connections. It can be challenging for people with visual impairments to use dating services and develop lasting relationships because many of them are not created with their requirements in mind. We can encourage people with visual impairments to participate more completely in online dating and possibly enhance the success of their romantic relationships by making dating websites more accessible. There is some existing implementation that can automatically recognize the facial expression, age, gender, presence of child(ren) and other common objects from a profile photo in a dating platform. The goal of this project is incorporate additional features (presence of any common pets, indoor vs. outdoor image) to further enhance the capability of existing system and come up with test viable solutions to accessibility issues that people with visual impairments face when using dating websites.
Authors: Laura Gutierrez-Martin (Department of Electronics, Universidad Carlos III de Madrid, Spain), Celia Lopez Ongil (Department of Electronics, Universidad Carlos III de Madrid, Spain, Gender Studies Institute, Universidad Carlos III de Madrid, Spain), Jose M. Lanza-Gutierrez (Department of Computer Science, Universidad de Alcala, Spain), Jose A. Miranda Calero (Embedded Systems Laboratory, Ecole Polytechnique Federale de Lausanne, Switzerland)
Abstract: Emotion recognition through artificial intelligence and smart sensing of physical and physiological signals (Affective Computing) is achieving very interesting results in terms of accuracy, inference times, and user-independent models. In this sense, there are applications related to the safety and well-being of people (sexual aggressions, gender-based violence, children and elderly abuse, mental health, etc.) that require even more improvements. Emotion detection should be done with fast, discrete, and non-luxurious systems working in real-time and real life (wearable devices, wireless communications, battery-powered). Furthermore, emotional reactions to violence are not equal in all people. Then, large general models cannot be applied to a multiuser system for people protection, and customized and simple AI models would be welcomed by health and social workers and law enforcement agents. These customized models will be applicable to clusters of subjects sharing similarities in their emotional reactions to external stimuli. This customization requires several steps: creating clusters of subjects with similar behaviors, creating AI models for every cluster, continually updating these models with new data, and enrolling new subjects in clusters when required. A methodology for clustering data compiled (physical and physiological data, together with emotional labels) is presented in this work, as well as the method for including new subjects once the AI model is generated. Experimental results demonstrate an improvement of 4% in accuracy and 3% in f1-score w.r.t. the general model, along with a 14% reduction in variability.
Authors: Ali Mehrabian, Ehsan Hoseinzade, Mahdi Mazloum, Xiaohong Chen
Abstract: Stock markets play an important role in the global economy, where accurate stock price predictions can lead to significant financial returns. While existing transformer-based models have outperformed long short-term memory networks and convolutional neural networks in financial time series prediction, their high computational complexity and memory requirements limit their practicality for real-time trading and long-sequence data processing. To address these challenges, we propose SAMBA, an innovative framework for stock return prediction that builds on the Mamba architecture and integrates graph neural networks. SAMBA achieves near-linear computational complexity by utilizing a bidirectional Mamba block to capture long-term dependencies in historical price data and employing adaptive graph convolution to model dependencies between daily stock features. Our experimental results demonstrate that SAMBA significantly outperforms state-of-the-art baseline models in prediction accuracy, maintaining low computational complexity. The code and datasets are available at github.com/Ali-Meh619/SAMBA.
Authors: Mohit Raghavendra, Vaskar Nath, Sean Hendryx
Abstract: The Superficial Alignment Hypothesis posits that almost all of a language model's abilities and knowledge are learned during pre-training, while post-training is about giving a model the right style and format. We re-examine these claims by empirically studying the scaling behavior of post-training with increasing finetuning examples and evaluating them using objective task-specific standardized benchmarks. Through experiments with the Llama-3, Mistral, and Llama-2 model families of multiple sizes, we observe that, similar to the pre-training scaling laws, post-training task performance scales as a power law against the number of finetuning examples. This power law relationship holds across a broad array of capabilities, including mathematical reasoning, coding, instruction following, and multihop-reasoning. In addition, for tasks like math and multihop reasoning, we observe that a handful of examples merely align the model stylistically but do not saturate performance on the benchmarks. Model performance is instead correlated with its reasoning ability and it improves significantly with more examples, illustrating the need for holistic evaluation programs leveraging objective benchmarks in addition to measurement of alignment to human preferences. We also observe that language models are not necessarily limited to using knowledge learned during pre-training. With appropriate post-training, a model's ability to integrate new knowledge greatly improves on downstream tasks like multihop question-answering. Taken together, these results shed new light on the Superficial Alignment Hypothesis, suggesting that it is, at best, an over-simplification.
Authors: Zhixiao Xiong, Fangyu Zong, Huigen Ye, Hua Xu
Abstract: Machine Learning (ML) optimization frameworks have gained attention for their ability to accelerate the optimization of large-scale Quadratically Constrained Quadratic Programs (QCQPs) by learning shared problem structures. However, existing ML frameworks often rely heavily on strong problem assumptions and large-scale solvers. This paper introduces NeuralQP, a general hypergraph-based framework for large-scale QCQPs. NeuralQP features two main components: Hypergraph-based Neural Prediction, which generates embeddings and predicted solutions for QCQPs without problem assumptions, and Parallel Neighborhood Optimization, which employs a McCormick relaxation-based repair strategy to identify and correct illegal variables, iteratively improving the solution with a small-scale solver. We further prove that our framework UniEGNN with our hypergraph representation is equivalent to the Interior-Point Method (IPM) for quadratic programming. Experiments on two benchmark problems and large-scale real-world instances from QPLIB demonstrate that NeuralQP outperforms state-of-the-art solvers (e.g., Gurobi and SCIP) in both solution quality and time efficiency, further validating the efficiency of ML optimization frameworks for QCQPs.
Authors: Yifei Ming, Senthil Purushwalkam, Shrey Pandit, Zixuan Ke, Xuan-Phi Nguyen, Caiming Xiong, Shafiq Joty
Abstract: Ensuring faithfulness to context in large language models (LLMs) and retrieval-augmented generation (RAG) systems is crucial for reliable deployment in real-world applications, as incorrect or unsupported information can erode user trust. Despite advancements on standard benchmarks, faithfulness hallucination-where models generate responses misaligned with the provided context-remains a significant challenge. In this work, we introduce FaithEval, a novel and comprehensive benchmark tailored to evaluate the faithfulness of LLMs in contextual scenarios across three diverse tasks: unanswerable, inconsistent, and counterfactual contexts. These tasks simulate real-world challenges where retrieval mechanisms may surface incomplete, contradictory, or fabricated information. FaithEval comprises 4.9K high-quality problems in total, validated through a rigorous four-stage context construction and validation framework, employing both LLM-based auto-evaluation and human validation. Our extensive study across a wide range of open-source and proprietary models reveals that even state-of-the-art models often struggle to remain faithful to the given context, and that larger models do not necessarily exhibit improved faithfulness.Project is available at: \url{https://github.com/SalesforceAIResearch/FaithEval}.
Authors: Barak Gahtan, Robert J. Sahala, Alex M. Bronstein, Reuven Cohen
Abstract: QUIC, a new and increasingly used transport protocol, addresses and resolves the limitations of TCP by offering improved security, performance, and features such as stream multiplexing and connection migration. These features, however, also present challenges for network operators who need to monitor and analyze web traffic. In this paper, we introduce VisQUIC, a labeled dataset comprising over 100,000 QUIC traces from more than 44,000 websites (URLs), collected over a four-month period. These traces provide the foundation for generating more than seven million images, with configurable parameters of window length, pixel resolution, normalization, and labels. These images enable an observer looking at the interactions between a client and a server to analyze and gain insights about QUIC encrypted connections. To illustrate the dataset's potential, we offer a use-case example of an observer estimating the number of HTTP/3 responses/requests pairs in a given QUIC, which can reveal server behavior, client--server interactions, and the load imposed by an observed connection. We formulate the problem as a discrete regression problem, train a machine learning (ML) model for it, and then evaluate it using the proposed dataset on an example use case.
Authors: Sebastien Origer, Dario Izzo, Giacomo Acciarini, Francesco Biscani, Rita Mastroianni, Max Bannach, Harry Holt
Abstract: We perform uncertainty propagation on an event manifold for Guidance & Control Networks (G&CNETs), aiming to enhance the certification tools for neural networks in this field. This work utilizes three previously solved optimal control problems with varying levels of dynamics nonlinearity and event manifold complexity. The G&CNETs are trained to represent the optimal control policies of a time-optimal interplanetary transfer, a mass-optimal landing on an asteroid and energy-optimal drone racing, respectively. For each of these problems, we describe analytically the terminal conditions on an event manifold with respect to initial state uncertainties. Crucially, this expansion does not depend on time but solely on the initial conditions of the system, thereby making it possible to study the robustness of the G&CNET at any specific stage of a mission defined by the event manifold. Once this analytical expression is found, we provide confidence bounds by applying the Cauchy-Hadamard theorem and perform uncertainty propagation using moment generating functions. While Monte Carlo-based (MC) methods can yield the results we present, this work is driven by the recognition that MC simulations alone may be insufficient for future certification of neural networks in guidance and control applications.
Authors: Mehdi Ali, Michael Fromm, Klaudia Thellmann, Jan Ebert, Alexander Arno Weber, Richard Rutmann, Charvi Jain, Max L\"ubbering, Daniel Steinigen, Johannes Leveling, Katrin Klug, Jasper Schulze Buschhoff, Lena Jurkschat, Hammam Abdelwahab, Benny J\"org Stein, Karl-Heinz Sylla, Pavel Denisov, Nicolo Brandizzi, Qasid Saleem, Bhowmick Anirban, Chelsea John, Pedro Ortiz Suarez, Malte Ostendorff, Alex Jude, Lalith Manjunath, Samuel Weinbach, Carolin Penke, Shima Asaadi, Fabio Barth, Rafet Sifa, Fabian K\"uch, Ren\'e J\"akel, Georg Rehm, Stefan Kesselheim, Joachim K\"ohler, Nicolas Flores-Herr
Abstract: We present preliminary results of the project OpenGPT-X. At present, the project has developed two multilingual LLMs designed to embrace Europe's linguistic diversity by supporting all 24 official languages of the European Union. Trained on a dataset comprising around 60% non-English data and utilizing a custom multilingual tokenizer, our models address the limitations of existing LLMs that predominantly focus on English or a few high-resource languages. We detail the models' development principles, data processing techniques, tokenizer optimization, and training methodologies. The models demonstrate competitive performance across multilingual benchmarks, as evidenced by its performance on European versions of ARC, HellaSwag, MMLU, and TruthfulQA.
Authors: Nooruddin Noonari, Daniel Corujo, Rui L. Aguiar, Francisco J. Ferrao
Abstract: The rapid growth in mobile broadband usage and increasing subscribers have made it crucial to ensure reliable network performance. As mobile networks grow more complex, especially during peak hours, manual collection of Key Performance Indicators (KPIs) is time-consuming due to the vast data involved. Detecting network failures and identifying unusual behavior during busy periods is vital to assess network health. Researchers have applied Deep Learning (DL) and Machine Learning (ML) techniques to understand network behavior by predicting throughput, analyzing call records, and detecting outages. However, these methods often require significant computational power, large labeled datasets, and are typically specialized, making retraining for new scenarios costly and time-intensive. This study introduces a novel approach Multi-Scale Convolutional LSTM with Transfer Learning (TL) to detect anomalies in cellular networks. The model is initially trained from scratch using a publicly available dataset to learn typical network behavior. Transfer Learning is then employed to fine-tune the model by applying learned weights to different datasets. We compare the performance of the model trained from scratch with that of the fine-tuned model using TL. To address class imbalance and gain deeper insights, Exploratory Data Analysis (EDA) and the Synthetic Minority Over-sampling Technique (SMOTE) are applied. Results demonstrate that the model trained from scratch achieves 99% accuracy after 100 epochs, while the fine-tuned model reaches 95% accuracy on a different dataset after just 20 epochs.
Authors: David Grangier, Simin Fan, Skyler Seto, Pierre Ablin
Abstract: Specialist language models (LMs) focus on a specific task or domain on which they often outperform generalist LMs of the same size. However, the specialist data needed to pretrain these models is only available in limited amount for most tasks. In this work, we build specialist models from large generalist training sets instead. We adjust the training distribution of the generalist data with guidance from the limited domain-specific data. We explore several approaches, with clustered importance sampling standing out. This method clusters the generalist dataset and samples from these clusters based on their frequencies in the smaller specialist dataset. It is scalable, suitable for pretraining and continued pretraining, it works well in multi-task settings. Our findings demonstrate improvements across different domains in terms of language modeling perplexity and accuracy on multiple-choice question tasks. We also present ablation studies that examine the impact of dataset sizes, clustering configurations, and model sizes.
Authors: Evgeny Saveliev, Tim Schubert, Thomas Pouplin, Vasilis Kosmoliaptsis, Mihaela van der Schaar
Abstract: Despite its significant promise and continuous technical advances, real-world applications of artificial intelligence (AI) remain limited. We attribute this to the "domain expert-AI-conundrum": while domain experts, such as clinician scientists, should be able to build predictive models such as risk scores, they face substantial barriers in accessing state-of-the-art (SOTA) tools. While automated machine learning (AutoML) has been proposed as a partner in clinical predictive modeling, many additional requirements need to be fulfilled to make machine learning accessible for clinician scientists. To address this gap, we introduce CliMB, a no-code AI-enabled partner designed to empower clinician scientists to create predictive models using natural language. CliMB guides clinician scientists through the entire medical data science pipeline, thus empowering them to create predictive models from real-world data in just one conversation. CliMB also creates structured reports and interpretable visuals. In evaluations involving clinician scientists and systematic comparisons against a baseline GPT-4, CliMB consistently demonstrated superior performance in key areas such as planning, error prevention, code execution, and model performance. Moreover, in blinded assessments involving 45 clinicians from diverse specialties and career stages, more than 80% preferred CliMB over GPT-4. Overall, by providing a no-code interface with clear guidance and access to SOTA methods in the fields of data-centric AI, AutoML, and interpretable ML, CliMB empowers clinician scientists to build robust predictive models.
Authors: Fatemeh Lotfi, Fatemeh Afghah
Abstract: As wireless networks grow to support more complex applications, the Open Radio Access Network (O-RAN) architecture, with its smart RAN Intelligent Controller (RIC) modules, becomes a crucial solution for real-time network data collection, analysis, and dynamic management of network resources including radio resource blocks and downlink power allocation. Utilizing artificial intelligence (AI) and machine learning (ML), O-RAN addresses the variable demands of modern networks with unprecedented efficiency and adaptability. Despite progress in using ML-based strategies for network optimization, challenges remain, particularly in the dynamic allocation of resources in unpredictable environments. This paper proposes a novel Meta Deep Reinforcement Learning (Meta-DRL) strategy, inspired by Model-Agnostic Meta-Learning (MAML), to advance resource block and downlink power allocation in O-RAN. Our approach leverages O-RAN's disaggregated architecture with virtual distributed units (DUs) and meta-DRL strategies, enabling adaptive and localized decision-making that significantly enhances network efficiency. By integrating meta-learning, our system quickly adapts to new network conditions, optimizing resource allocation in real-time. This results in a 19.8% improvement in network management performance over traditional methods, advancing the capabilities of next-generation wireless networks.
Authors: Ziyi Ye, Xiangsheng Li, Qiuchi Li, Qingyao Ai, Yujia Zhou, Wei Shen, Dong Yan, Yiqun Liu
Abstract: Learning from preference feedback is a common practice for aligning large language models~(LLMs) with human value. Conventionally, preference data is learned and encoded into a scalar reward model that connects a value head with an LLM to produce a scalar score as preference or reward. However, scalar models lack interpretability and are known to be susceptible to biases in datasets. This paper investigates leveraging the generation capability of LLMs to address both limitations in one shot. Specifically, we prompt the pre-trained LLM to generate positive and negative judgments, both supported with rationales in natural language form. The self-generated contrastive judgment pairs are used to train the generative judge with Direct Preference Optimization (DPO). This proposal of training the generative Judge using self-generated Contrastive judgments (Con-J) ensures natural interpretability due to the generated rationales together with the judgments, as well as high robustness against bias without the need for an additional reward head. Experimental results show that the performance of Con-J is comparable to the scalar reward model trained on the same collection of preference data, and demonstrate its superior interpretability and robustness in encoding human preferences.
Authors: Yiming Ju, Ziyi Ni, Xingrun Xing, Zhixiong Zeng, hanyu Zhao, Siqi Fan, Zheng Zhang
Abstract: Supervised fine-tuning (SFT) is crucial for adapting Large Language Models (LLMs) to specific tasks. In this work, we demonstrate that the order of training data can lead to significant training imbalances, potentially resulting in performance degradation. Consequently, we propose to mitigate this imbalance by merging SFT models fine-tuned with different data orders, thereby enhancing the overall effectiveness of SFT. Additionally, we introduce a novel technique, "parameter-selection merging," which outperforms traditional weighted-average methods on five datasets. Further, through analysis and ablation studies, we validate the effectiveness of our method and identify the sources of performance improvements.
Authors: Ahmed Hussein, Alaa Elsetohy, Sama Hadhoud, Tameem Bakr, Yasser Rohaim, Badr AlKhamissi
Abstract: Designing expressive typography that visually conveys a word's meaning while maintaining readability is a complex task, known as semantic typography. It involves selecting an idea, choosing an appropriate font, and balancing creativity with legibility. We introduce an end-to-end system that automates this process. First, a Large Language Model (LLM) generates imagery ideas for the word, useful for abstract concepts like freedom. Then, the FontCLIP pre-trained model automatically selects a suitable font based on its semantic understanding of font attributes. The system identifies optimal regions of the word for morphing and iteratively transforms them using a pre-trained diffusion model. A key feature is our OCR-based loss function, which enhances readability and enables simultaneous stylization of multiple characters. We compare our method with other baselines, demonstrating great readability enhancement and versatility across multiple languages and writing scripts.
Authors: K. Lian (Columbia University), L. S. Liebovitch (Columbia University), M. Wild (Columbia University), H. West (Columbia University), P. T. Coleman (Columbia University), F. Chen (Toyota Research Institute), E. Kimani (Toyota Research Institute), K. Sieck (Toyota Research Institute)
Abstract: This paper presents a machine learning approach to classify countries as peaceful or non-peaceful using linguistic patterns extracted from global media articles. We employ vector embeddings and cosine similarity to develop a supervised classification model that effectively identifies peaceful countries. Additionally, we explore the impact of dataset size on model performance, investigating how shrinking the dataset influences classification accuracy. Our results highlight the challenges and opportunities associated with using large-scale text data for peace studies.
Authors: Judah Goldfeder, John Sipple
Abstract: Commercial office buildings contribute 17 percent of Carbon Emissions in the US, according to the US Energy Information Administration (EIA), and improving their efficiency will reduce their environmental burden and operating cost. A major contributor of energy consumption in these buildings are the Heating, Ventilation, and Air Conditioning (HVAC) devices. HVAC devices form a complex and interconnected thermodynamic system with the building and outside weather conditions, and current setpoint control policies are not fully optimized for minimizing energy use and carbon emission. Given a suitable training environment, a Reinforcement Learning (RL) agent is able to improve upon these policies, but training such a model, especially in a way that scales to thousands of buildings, presents many practical challenges. Most existing work on applying RL to this important task either makes use of proprietary data, or focuses on expensive and proprietary simulations that may not be grounded in the real world. We present the Smart Buildings Control Suite, the first open source interactive HVAC control dataset extracted from live sensor measurements of devices in real office buildings. The dataset consists of two components: six years of real-world historical data from three buildings, for offline RL, and a lightweight interactive simulator for each of these buildings, calibrated using the historical data, for online and model-based RL. For ease of use, our RL environments are all compatible with the OpenAI gym environment standard. We also demonstrate a novel method of calibrating the simulator, as well as baseline results on training an RL agent on the simulator, predicting real-world data, and training an RL agent directly from data. We believe this benchmark will accelerate progress and collaboration on building optimization and environmental sustainability research.
Authors: Luis Fredes (IMB), Bernard Bercu (IMB), Em\'eric Gbaguidi (IMB)
Abstract: Stochastic optimization naturally appear in many application areas, including machine learning. Our goal is to go further in the analysis of the Stochastic Average Gradient Accelerated (SAGA) algorithm. To achieve this, we introduce a new $\lambda$-SAGA algorithm which interpolates between the Stochastic Gradient Descent ($\lambda=0$) and the SAGA algorithm ($\lambda=1$). Firstly, we investigate the almost sure convergence of this new algorithm with decreasing step which allows us to avoid the restrictive strong convexity and Lipschitz gradient hypotheses associated to the objective function. Secondly, we establish a central limit theorem for the $\lambda$-SAGA algorithm. Finally, we provide the non-asymptotic $\mathbb{L}^p$ rates of convergence.
Authors: Yuntong Hu, Zhuofeng Li, Zheng Zhang, Chen Ling, Raasikh Kanjiani, Boxin Zhao, Liang Zhao
Abstract: In this work, we present HiReview, a novel framework for hierarchical taxonomy-driven automatic literature review generation. With the exponential growth of academic documents, manual literature reviews have become increasingly labor-intensive and time-consuming, while traditional summarization models struggle to generate comprehensive document reviews effectively. Large language models (LLMs), with their powerful text processing capabilities, offer a potential solution; however, research on incorporating LLMs for automatic document generation remains limited. To address key challenges in large-scale automatic literature review generation (LRG), we propose a two-stage taxonomy-then-generation approach that combines graph-based hierarchical clustering with retrieval-augmented LLMs. First, we retrieve the most relevant sub-community within the citation network, then generate a hierarchical taxonomy tree by clustering papers based on both textual content and citation relationships. In the second stage, an LLM generates coherent and contextually accurate summaries for clusters or topics at each hierarchical level, ensuring comprehensive coverage and logical organization of the literature. Extensive experiments demonstrate that HiReview significantly outperforms state-of-the-art methods, achieving superior hierarchical organization, content relevance, and factual accuracy in automatic literature review generation tasks.
Authors: T. Prasad (Columbia University), L. S. Liebovitch (Columbia University), M. Wild (Columbia University), H. West (Columbia University), P. T. Coleman (Columbia University)
Abstract: We used data from LexisNexis to determine the words in news media that best classifies countries as higher or lower peace. We found that higher peace news is characterized by themes of finance, daily actitivities, and health and that lower peace news is characterized by themes of politics, government, and legal issues. This work provides a starting point to measure levels of peace and identify the social processes that underly those words.
Authors: Jingcun Wang, Yu-Guang Chen, Ing-Chao Lin, Bing Li, Grace Li Zhang
Abstract: Large Language Models (LLMs) have achieved remarkable breakthroughs. However, the huge number of parameters in LLMs require significant amount of memory storage in inference, which prevents their practical deployment in many applications. To reduce memory storage of LLMs, singular value decomposition (SVD) provides a promising solution to approximate weight matrices for compressing LLMs. In this paper, we take a step further to explore parameter sharing across different layers with SVD to achieve more effective compression for LLMs. Specifically, weight matrices in different layers are decomposed and represented as a linear combination of a set of shared basis vectors and unique coefficients. The types of weight matrices and the layer selection for basis sharing are examined when compressing LLMs to maintain the performance. Comprehensive experiments demonstrate that Basis Sharing outperforms state-of-the-art SVD-based compression approaches and parameter sharing techniques, especially under large compression ratios. Code is available at: https://github.com/TUDa-HWAI/Basis_Sharing
Authors: Alihan H\"uy\"uk, Xinnuo Xu, Jacqueline Maasch, Aditya V. Nori, Javier Gonz\'alez
Abstract: Despite the increasing effectiveness of language models, their reasoning capabilities remain underdeveloped. In particular, causal reasoning through counterfactual question answering is lacking. This work aims to bridge this gap. We first derive novel metrics that balance accuracy in factual and counterfactual questions, capturing a more complete view of the reasoning abilities of language models than traditional factual-only based metrics. Second, we propose several fine-tuning approaches that aim to elicit better reasoning mechanisms, in the sense of the proposed metrics. Finally, we evaluate the performance of the fine-tuned language models in a variety of realistic scenarios. In particular, we investigate to what extent our fine-tuning approaches systemically achieve better generalization with respect to the base models in several problems that require, among others, inductive and deductive reasoning capabilities.
Authors: Yohan Mathew, Ollie Matthews, Robert McCarthy, Joan Velja, Christian Schroeder de Witt, Dylan Cope, Nandi Schoots
Abstract: The rapid proliferation of frontier model agents promises significant societal advances but also raises concerns about systemic risks arising from unsafe interactions. Collusion to the disadvantage of others has been identified as a central form of undesirable agent cooperation. The use of information hiding (steganography) in agent communications could render collusion practically undetectable. This underscores the need for evaluation frameworks to monitor and mitigate steganographic collusion capabilities. We address a crucial gap in the literature by demonstrating, for the first time, that robust steganographic collusion in LLMs can arise indirectly from optimization pressure. To investigate this problem we design two approaches -- a gradient-based reinforcement learning (GBRL) method and an in-context reinforcement learning (ICRL) method -- for reliably eliciting sophisticated LLM-generated linguistic text steganography. Importantly, we find that emergent steganographic collusion can be robust to both passive steganalytic oversight of model outputs and active mitigation through communication paraphrasing. We contribute a novel model evaluation framework and discuss limitations and future work. Our findings imply that effective risk mitigation from steganographic collusion post-deployment requires innovation in passive and active oversight techniques.
Authors: Ruiyuan Zhang, Yuyao Chen, Yuchi Huo, Jiaxiang Liu, Dianbing Xi, Jie Liu, Chao Wu
Abstract: Multi-task-learning(MTL) is a multi-target optimization task. Neural networks try to realize each target using a shared interpretative space within MTL. However, as the scale of datasets expands and the complexity of tasks increases, knowledge sharing becomes increasingly challenging. In this paper, we first re-examine previous cross-attention MTL methods from the perspective of noise. We theoretically analyze this issue and identify it as a flaw in the cross-attention mechanism. To address this issue, we propose an information bottleneck knowledge extraction module (KEM). This module aims to reduce inter-task interference by constraining the flow of information, thereby reducing computational complexity. Furthermore, we have employed neural collapse to stabilize the knowledge-selection process. That is, before input to KEM, we projected the features into ETF space. This mapping makes our method more robust. We implemented and conducted comparative experiments with this method on multiple datasets. The results demonstrate that our approach significantly outperforms existing methods in multi-task learning.
Authors: Thang Nguyen, Peter Chin, Yu-Wing Tai
Abstract: In this paper, we introduce Reward-RAG, a novel approach designed to enhance the Retrieval-Augmented Generation (RAG) model through Reward-Driven Supervision. Unlike previous RAG methodologies, which focus on training language models (LMs) to utilize external knowledge retrieved from external sources, our method adapts retrieval information to specific domains by employing CriticGPT to train a dedicated reward model. This reward model generates synthesized datasets for fine-tuning the RAG encoder, aligning its outputs more closely with human preferences. The versatility of our approach allows it to be effectively applied across various domains through domain-specific fine-tuning. We evaluate Reward-RAG on publicly available benchmarks from multiple domains, comparing it to state-of-the-art methods. Our experimental results demonstrate significant improvements in performance, highlighting the effectiveness of Reward-RAG in improving the relevance and quality of generated responses. These findings underscore the potential of integrating reward models with RAG to achieve superior outcomes in natural language generation tasks.
Authors: Keyu Chen, Ziqian Bi, Tianyang Wang, Yizhu Wen, Pohsun Feng, Qian Niu, Junyu Liu, Benji Peng, Sen Zhang, Ming Li, Xuanhe Pan, Jiawei Xu, Jinlang Wang, Caitlyn Heqi Yin, Ming Liu
Abstract: This book, Design Patterns in Machine Learning and Deep Learning: Advancing Big Data Analytics Management, presents a comprehensive study of essential design patterns tailored for large-scale machine learning and deep learning applications. The book explores the application of classical software engineering patterns, Creational, Structural, Behavioral, and Concurrency Patterns, to optimize the development, maintenance, and scalability of big data analytics systems. Through practical examples and detailed Python implementations, it bridges the gap between traditional object-oriented design patterns and the unique demands of modern data analytics environments. Key design patterns such as Singleton, Factory, Observer, and Strategy are analyzed for their impact on model management, deployment strategies, and team collaboration, providing invaluable insights into the engineering of efficient, reusable, and flexible systems. This volume is an essential resource for developers, researchers, and engineers aiming to enhance their technical expertise in both machine learning and software design.
Authors: Giuseppe Alessio D'Inverno, Saeid Moradizadeh, Sajad Salavatidezfouli, Pasquale Claudio Africa, Gianluigi Rozza
Abstract: The complexity of the cardiovascular system needs to be accurately reproduced in order to promptly acknowledge health conditions; to this aim, advanced multifidelity and multiphysics numerical models are crucial. On one side, Full Order Models (FOMs) deliver accurate hemodynamic assessments, but their high computational demands hinder their real-time clinical application. In contrast, ROMs provide more efficient yet accurate solutions, essential for personalized healthcare and timely clinical decision-making. In this work, we explore the application of computational fluid dynamics (CFD) in cardiovascular medicine by integrating FOMs with ROMs for predicting the risk of aortic aneurysm growth and rupture. Wall Shear Stress (WSS) and the Oscillatory Shear Index (OSI), sampled at different growth stages of the abdominal aortic aneurysm, are predicted by means of Graph Neural Networks (GNNs). GNNs exploit the natural graph structure of the mesh obtained by the Finite Volume (FV) discretization, taking into account the spatial local information, regardless of the dimension of the input graph. Our experimental validation framework yields promising results, confirming our method as a valid alternative that overcomes the curse of dimensionality.
Authors: Matthieu Zimmer, Milan Gritta, Gerasimos Lampouras, Haitham Bou Ammar, Jun Wang
Abstract: The growth in the number of parameters of Large Language Models (LLMs) has led to a significant surge in computational requirements, making them challenging and costly to deploy. Speculative decoding (SD) leverages smaller models to efficiently propose future tokens, which are then verified by the LLM in parallel. Small models that utilise activations from the LLM currently achieve the fastest decoding speeds. However, we identify several limitations of SD models including the lack of on-policyness during training and partial observability. To address these shortcomings, we propose a more grounded architecture for small models by introducing a Mixture of Attentions for SD. Our novel architecture can be applied in two scenarios: a conventional single device deployment and a novel client-server deployment where the small model is hosted on a consumer device and the LLM on a server. In a single-device scenario, we demonstrate state-of-the-art speedups improving EAGLE-2 by 9.5% and its acceptance length by 25%. In a client-server setting, our experiments demonstrate: 1) state-of-the-art latencies with minimal calls to the server for different network conditions, and 2) in the event of a complete disconnection, our approach can maintain higher accuracy compared to other SD methods and demonstrates advantages over API calls to LLMs, which would otherwise be unable to continue the generation process.
Authors: Fajrian Yunus, Talel Abdessalem
Abstract: A lot of effort in recent years have been expended to explain machine learning systems. However, some machine learning methods are inherently explainable, and thus are not completely black box. This enables the developers to make sense of the output without a developing a complex and expensive explainability technique. Besides that, explainability should be tailored to suit the context of the problem. In a recommendation system which relies on collaborative filtering, the recommendation is based on the behaviors of similar users, therefore the explanation should tell which other users are similar to the current user. Similarly, if the recommendation system is based on sequence prediction, the explanation should also tell which input timesteps are the most influential. We demonstrate this philosophy/paradigm in STAN (Spatio-Temporal Attention Network for Next Location Recommendation), a next Point of Interest recommendation system based on collaborative filtering and sequence prediction. We also show that the explanation helps to "debug" the output.
Authors: Kuan-Chen Wang, Kai-Chun Liu, Ping-Cheng Yeh, Sheng-Yu Peng, Yu Tsao
Abstract: Surface electromyography (sEMG) is a widely employed bio-signal that captures human muscle activity via electrodes placed on the skin. Several studies have proposed methods to remove sEMG contaminants, as non-invasive measurements render sEMG susceptible to various contaminants. However, these approaches often rely on heuristic-based optimization and are sensitive to the contaminant type. A more potent, robust, and generalized sEMG denoising approach should be developed for various healthcare and human-computer interaction applications. This paper proposes a novel neural network (NN)-based sEMG denoising method called TrustEMG-Net. It leverages the potent nonlinear mapping capability and data-driven nature of NNs. TrustEMG-Net adopts a denoising autoencoder structure by combining U-Net with a Transformer encoder using a representation-masking approach. The proposed approach is evaluated using the Ninapro sEMG database with five common contamination types and signal-to-noise ratio (SNR) conditions. Compared with existing sEMG denoising methods, TrustEMG-Net achieves exceptional performance across the five evaluation metrics, exhibiting a minimum improvement of 20%. Its superiority is consistent under various conditions, including SNRs ranging from -14 to 2 dB and five contaminant types. An ablation study further proves that the design of TrustEMG-Net contributes to its optimality, providing high-quality sEMG and serving as an effective, robust, and generalized denoising solution for sEMG applications.
Authors: Yufei Tian, Zeyu Pan, Nanyun Peng
Abstract: The increasing capability of large language models (LLMs) to generate fluent long-form texts is presenting new challenges in distinguishing machine-generated outputs from human-written ones, which is crucial for ensuring authenticity and trustworthiness of expressions. Existing zero-shot detectors primarily focus on token-level distributions, which are vulnerable to real-world domain shifts, including different prompting and decoding strategies, and adversarial attacks. We propose a more robust method that incorporates abstract elements, such as event transitions, as key deciding factors to detect machine versus human texts by training a latent-space model on sequences of events or topics derived from human-written texts. In three different domains, machine-generated texts, which are originally inseparable from human texts on the token level, can be better distinguished with our latent-space model, leading to a 31% improvement over strong baselines such as DetectGPT. Our analysis further reveals that, unlike humans, modern LLMs like GPT-4 generate event triggers and their transitions differently, an inherent disparity that helps our method to robustly detect machine-generated texts.
Authors: Murong Yue, Wenlin Yao, Haitao Mi, Dian Yu, Ziyu Yao, Dong Yu
Abstract: Enhancing the capability of large language models (LLMs) in reasoning has gained significant attention in recent years. Previous studies have demonstrated the effectiveness of various prompting strategies in aiding LLMs in reasoning (called "reasoning actions"), such as step-by-step thinking, reflecting before answering, solving with programs, and their combinations. However, these approaches often applied static, predefined reasoning actions uniformly to all questions, without considering the specific characteristics of each question or the capability of the task-solving LLM. In this paper, we propose DOTS, an approach enabling LLMs to reason dynamically via optimal reasoning trajectory search, tailored to the specific characteristics of each question and the inherent capability of the task-solving LLM. Our approach involves three key steps: i) defining atomic reasoning action modules that can be composed into various reasoning action trajectories; ii) searching for the optimal action trajectory for each training question through iterative exploration and evaluation for the specific task-solving LLM; and iii) using the collected optimal trajectories to train an LLM to plan for the reasoning trajectories of unseen questions. In particular, we propose two learning paradigms, i.e., fine-tuning an external LLM as a planner to guide the task-solving LLM, or directly fine-tuning the task-solving LLM with an internalized capability for reasoning actions planning. Our experiments across eight reasoning tasks show that our method consistently outperforms static reasoning techniques and the vanilla instruction tuning approach. Further analysis reveals that our method enables LLMs to adjust their computation based on problem complexity, allocating deeper thinking and reasoning to harder problems.
Authors: Ricardo Di Pasquale, Soledad Represa
Abstract: In an era dominated by data, the management and utilization of domain-specific language have emerged as critical challenges in various application domains, particularly those with industry-specific requirements. Our work is driven by the need to effectively manage and process large volumes of short text documents inherent in specific application domains. By leveraging domain-specific knowledge and expertise, our approach aims to shape factual data within these domains, thereby facilitating enhanced utilization and understanding by end-users. Central to our methodology is the integration of domain-specific language models with graph-oriented databases, facilitating seamless processing, analysis, and utilization of textual data within targeted domains. Our work underscores the transformative potential of the partnership of domain-specific language models and graph-oriented databases. This cooperation aims to assist researchers and engineers in metric usage, mitigation of latency issues, boosting explainability, enhancing debug and improving overall model performance. Moving forward, we envision our work as a guide AI engineers, providing valuable insights for the implementation of domain-specific language models in conjunction with graph-oriented databases, and additionally provide valuable experience in full-life cycle maintenance of this kind of products.
Authors: Manish Jha, Jialin Qian, Michael Weber, Baozhong Yang
Abstract: We use generative AI to extract managerial expectations about their economic outlook from over 120,000 corporate conference call transcripts. The overall measure, AI Economy Score, robustly predicts future economic indicators such as GDP growth, production, and employment, both in the short term and to 10 quarters. This predictive power is incremental to that of existing measures, including survey forecasts. Moreover, industry and firm-level measures provide valuable information about sector-specific and individual firm activities. Our findings suggest that managerial expectations carry unique insights about economic activities, with implications for both macroeconomic and microeconomic decision-making.
Authors: John Phan, Hung-Fu Chang
Abstract: This paper investigates the application of machine learning models, Long Short-Term Memory (LSTM), one-dimensional Convolutional Neural Networks (1D CNN), and Logistic Regression (LR), for predicting stock trends based on fundamental analysis. Unlike most existing studies that predominantly utilize technical or sentiment analysis, we emphasize the use of a company's financial statements and intrinsic value for trend forecasting. Using a dataset of 269 data points from publicly traded companies across various sectors from 2019 to 2023, we employ key financial ratios and the Discounted Cash Flow (DCF) model to formulate two prediction tasks: Annual Stock Price Difference (ASPD) and Difference between Current Stock Price and Intrinsic Value (DCSPIV). These tasks assess the likelihood of annual profit and current profitability, respectively. Our results demonstrate that LR models outperform CNN and LSTM models, achieving an average test accuracy of 74.66% for ASPD and 72.85% for DCSPIV. This study contributes to the limited literature on integrating fundamental analysis into machine learning for stock prediction, offering valuable insights for both academic research and practical investment strategies. By leveraging fundamental data, our approach highlights the potential for long-term stock trend prediction, supporting portfolio managers in their decision-making processes.
Authors: Zihao Liang, Tianyu Zhou, Zehui Lu, Shaoshuai Mou
Abstract: This paper proposes an Online Control-Informed Learning (OCIL) framework, which synthesizes the well-established control theories to solve a broad class of learning and control tasks in real time. This novel integration effectively handles practical issues in machine learning such as noisy measurement data, online learning, and data efficiency. By considering any robot as a tunable optimal control system, we propose an online parameter estimator based on extended Kalman filter (EKF) to incrementally tune the system in real time, enabling it to complete designated learning or control tasks. The proposed method also improves robustness in learning by effectively managing noise in the data. Theoretical analysis is provided to demonstrate the convergence and regret of OCIL. Three learning modes of OCIL, i.e. Online Imitation Learning, Online System Identification, and Policy Tuning On-the-fly, are investigated via experiments, which validate their effectiveness.
Authors: Yongxian Wu, Qiang Zhu, Ray Luo
Abstract: In computational biochemistry and biophysics, understanding the role of electrostatic interactions is crucial for elucidating the structure, dynamics, and function of biomolecules. The Poisson-Boltzmann (PB) equation is a foundational tool for modeling these interactions by describing the electrostatic potential in and around charged molecules. However, solving the PB equation presents significant computational challenges due to the complexity of biomolecular surfaces and the need to account for mobile ions. While traditional numerical methods for solving the PB equation are accurate, they are computationally expensive and scale poorly with increasing system size. To address these challenges, we introduce PBNeF, a novel machine learning approach inspired by recent advancements in neural network-based partial differential equation solvers. Our method formulates the input and boundary electrostatic conditions of the PB equation into a learnable voxel representation, enabling the use of a neural field transformer to predict the PB solution and, subsequently, the reaction field potential energy. Extensive experiments demonstrate that PBNeF achieves over a 100-fold speedup compared to traditional PB solvers, while maintaining accuracy comparable to the Generalized Born (GB) model.
Authors: Amirhosein Ghasemabadi, Muhammad Kamran Janjua, Mohammad Salameh, Di Niu
Abstract: One key challenge to video restoration is to model the transition dynamics of video frames governed by motion. In this work, we propose TURTLE to learn the truncated causal history model for efficient and high-performing video restoration. Unlike traditional methods that process a range of contextual frames in parallel, TURTLE enhances efficiency by storing and summarizing a truncated history of the input frame latent representation into an evolving historical state. This is achieved through a sophisticated similarity-based retrieval mechanism that implicitly accounts for inter-frame motion and alignment. The causal design in TURTLE enables recurrence in inference through state-memorized historical features while allowing parallel training by sampling truncated video clips. We report new state-of-the-art results on a multitude of video restoration benchmark tasks, including video desnowing, nighttime video deraining, video raindrops and rain streak removal, video super-resolution, real-world and synthetic video deblurring, and blind video denoising while reducing the computational cost compared to existing best contextual methods on all these tasks.
Authors: Selim Furkan Tekin, Fatih Ilhan, Tiansheng Huang, Sihao Hu, Ling Liu
Abstract: Combining large language models during training or at inference time has shown substantial performance gain over component LLMs. This paper presents LLM-TOPLA, a diversity-optimized LLM ensemble method with three unique properties: (i) We introduce the focal diversity metric to capture the diversity-performance correlation among component LLMs of an ensemble. (ii) We develop a diversity-optimized ensemble pruning algorithm to select the top-k sub-ensembles from a pool of $N$ base LLMs. Our pruning method recommends top-performing LLM subensembles of size $S$, often much smaller than $N$. (iii) We generate new output for each prompt query by utilizing a learn-to-ensemble approach, which learns to detect and resolve the output inconsistency among all component LLMs of an ensemble. Extensive evaluation on four different benchmarks shows good performance gain over the best LLM ensemble methods: (i) In constrained solution set problems, LLM-TOPLA outperforms the best-performing ensemble (Mixtral) by 2.2\% in accuracy on MMLU and the best-performing LLM ensemble (MoreAgent) on GSM8k by 2.1\%. (ii) In generative tasks, LLM-TOPLA outperforms the top-2 performers (Llama70b/Mixtral) on SearchQA by $3.9\mathrm{x}$ in F1, and on XSum by more than $38$ in ROUGE-1. Our code and dataset, which contains outputs of 8 modern LLMs on 4 benchmarks is available at https://github.com/git-disl/llm-topla
Authors: Milena Gazdieva, Jaemoo Choi, Alexander Kolesov, Jaewoong Choi, Petr Mokrov, Alexander Korotin
Abstract: A common challenge in aggregating data from multiple sources can be formalized as an \textit{Optimal Transport} (OT) barycenter problem, which seeks to compute the average of probability distributions with respect to OT discrepancies. However, the presence of outliers and noise in the data measures can significantly hinder the performance of traditional statistical methods for estimating OT barycenters. To address this issue, we propose a novel, scalable approach for estimating the \textit{robust} continuous barycenter, leveraging the dual formulation of the \textit{(semi-)unbalanced} OT problem. To the best of our knowledge, this paper is the first attempt to develop an algorithm for robust barycenters under the continuous distribution setup. Our method is framed as a $\min$-$\max$ optimization problem and is adaptable to \textit{general} cost function. We rigorously establish the theoretical underpinnings of the proposed method and demonstrate its robustness to outliers and class imbalance through a number of illustrative experiments.
Authors: Hongjun Wang, Jiyuan Chen, Zhengwei Yin, Xuan Song, Yinqiang Zheng
Abstract: Cloth-Changing Person Re-Identification (CC-ReID) involves recognizing individuals in images regardless of clothing status. In this paper, we empirically and experimentally demonstrate that completely eliminating or fully retaining clothing features is detrimental to the task. Existing work, either relying on clothing labels, silhouettes, or other auxiliary data, fundamentally aim to balance the learning of clothing and identity features. However, we practically find that achieving this balance is challenging and nuanced. In this study, we introduce a novel module called Diverse Norm, which expands personal features into orthogonal spaces and employs channel attention to separate clothing and identity features. A sample re-weighting optimization strategy is also introduced to guarantee the opposite optimization direction. Diverse Norm presents a simple yet effective approach that does not require additional data. Furthermore, Diverse Norm can be seamlessly integrated ResNet50 and significantly outperforms the state-of-the-art methods.
Authors: Sathvik Joel, Jie JW Wu, Fatemeh H. Fard
Abstract: Large Language Models (LLMs) have shown impressive capabilities in code generation for popular programming languages. However, their performance on Low-Resource Programming Languages (LRPLs) and Domain-Specific Languages (DSLs) remains a significant challenge, affecting millions of developers-3.5 million users in Rust alone-who cannot fully utilize LLM capabilities. LRPLs and DSLs encounter unique obstacles, including data scarcity and, for DSLs, specialized syntax that is poorly represented in general-purpose datasets. Addressing these challenges is crucial, as LRPLs and DSLs enhance development efficiency in specialized domains, such as finance and science. While several surveys discuss LLMs in software engineering, none focus specifically on the challenges and opportunities associated with LRPLs and DSLs. Our survey fills this gap by systematically reviewing the current state, methodologies, and challenges in leveraging LLMs for code generation in these languages. We filtered 111 papers from over 27,000 published studies between 2020 and 2024 to evaluate the capabilities and limitations of LLMs in LRPLs and DSLs. We report the LLMs used, benchmarks, and metrics for evaluation, strategies for enhancing performance, and methods for dataset collection and curation. We identified four main evaluation techniques and several metrics for assessing code generation in LRPLs and DSLs. Our analysis categorizes improvement methods into six groups and summarizes novel architectures proposed by researchers. Despite various techniques and metrics, a standard approach and benchmark dataset for evaluating code generation in LRPLs and DSLs are lacking. This survey serves as a resource for researchers and practitioners at the intersection of LLMs, software engineering, and specialized programming languages, laying the groundwork for future advancements in code generation for LRPLs and DSLs.
Authors: Shuang Liang, Guido Mont\'ufar
Abstract: We examine the implicit bias of mirror flow in univariate least squares error regression with wide and shallow neural networks. For a broad class of potential functions, we show that mirror flow exhibits lazy training and has the same implicit bias as ordinary gradient flow when the network width tends to infinity. For ReLU networks, we characterize this bias through a variational problem in function space. Our analysis includes prior results for ordinary gradient flow as a special case and lifts limitations which required either an intractable adjustment of the training data or networks with skip connections. We further introduce scaled potentials and show that for these, mirror flow still exhibits lazy training but is not in the kernel regime. For networks with absolute value activations, we show that mirror flow with scaled potentials induces a rich class of biases, which generally cannot be captured by an RKHS norm. A takeaway is that whereas the parameter initialization determines how strongly the curvature of the learned function is penalized at different locations of the input space, the scaled potential determines how the different magnitudes of the curvature are penalized.
Authors: Haoqun Cao, Zizhuo Meng, Tianjun Ke, Feng Zhou
Abstract: Score matching estimators have gained widespread attention in recent years partly because they are free from calculating the integral of normalizing constant, thereby addressing the computational challenges in maximum likelihood estimation (MLE). Some existing works have proposed score matching estimators for point processes. However, this work demonstrates that the incompleteness of the estimators proposed in those works renders them applicable only to specific problems, and they fail for more general point processes. To address this issue, this work introduces the weighted score matching estimator to point processes. Theoretically, we prove the consistency of our estimator and establish its rate of convergence. Experimental results indicate that our estimator accurately estimates model parameters on synthetic data and yields results consistent with MLE on real data. In contrast, existing score matching estimators fail to perform effectively. Codes are publicly available at \url{https://github.com/KenCao2007/WSM_TPP}.
Authors: Monik Raj Behera, Suchetana Chakraborty
Abstract: Conventional federated learning frameworks suffer from several challenges including performance bottlenecks at the central aggregation server, data bias, poor model convergence, and exposure to model poisoning attacks, and limited trust in the centralized infrastructure. In the current paper, a novel game theory-based approach called pFedGame is proposed for decentralized federated learning, best suitable for temporally dynamic networks. The proposed algorithm works without any centralized server for aggregation and incorporates the problem of vanishing gradients and poor convergence over temporally dynamic topology among federated learning participants. The solution comprises two sequential steps in every federated learning round, for every participant. First, it selects suitable peers for collaboration in federated learning. Secondly, it executes a two-player constant sum cooperative game to reach convergence by applying an optimal federated learning aggregation strategy. Experiments performed to assess the performance of pFedGame in comparison to existing methods in decentralized federated learning have shown promising results with accuracy higher than 70% for heterogeneous data.
Authors: Yiran Wang, Masao Utiyama
Abstract: Unsupervised parsing, also known as grammar induction, aims to infer syntactic structure from raw text. Recently, binary representation has exhibited remarkable information-preserving capabilities at both lexicon and syntax levels. In this paper, we explore the possibility of leveraging this capability to deduce parsing trees from raw text, relying solely on the implicitly induced grammars within models. To achieve this, we upgrade the bit-level CKY from zero-order to first-order to encode the lexicon and syntax in a unified binary representation space, switch training from supervised to unsupervised under the contrastive hashing framework, and introduce a novel loss function to impose stronger yet balanced alignment signals. Our model shows competitive performance on various datasets, therefore, we claim that our method is effective and efficient enough to acquire high-quality parsing trees from pre-trained language models at a low cost.
Authors: Pankhi Kashyap, Pavni Tandon, Sunny Gupta, Abhishek Tiwari, Ritwik Kulkarni, Kshitij Sharad Jadhav
Abstract: Long-tailed problems in healthcare emerge from data imbalance due to variability in the prevalence and representation of different medical conditions, warranting the requirement of precise and dependable classification methods. Traditional loss functions such as cross-entropy and binary cross-entropy are often inadequate due to their inability to address the imbalances between the classes with high representation and the classes with low representation found in medical image datasets. We introduce a novel polynomial loss function based on Pade approximation, designed specifically to overcome the challenges associated with long-tailed classification. This approach incorporates asymmetric sampling techniques to better classify under-represented classes. We conducted extensive evaluations on three publicly available medical datasets and a proprietary medical dataset. Our implementation of the proposed loss function is open-sourced in the public repository:https://github.com/ipankhi/ALPA.
Authors: Maryam Viqar, Erdem Sahin, Violeta Madjarova, Elena Stoykova, Keehoon Hong
Abstract: In this work, we propose to leverage a deep-learning (DL) based reconstruction framework for high quality Swept-Source Optical Coherence Tomography (SS-OCT) images, by incorporating wavelength ({\lambda}) space interferometric fringes. Generally, the SS-OCT captured fringe is linear in wavelength space and if Inverse Discrete Fourier Transform (IDFT) is applied to extract depth-resolved spectral information, the resultant images are blurred due to the broadened Point Spread Function (PSF). Thus, the recorded wavelength space fringe is to be scaled to uniform grid in wavenumber (k) space using k-linearization and calibration involving interpolations which may result in loss of information along with increased system complexity. Another challenge in OCT is the speckle noise, inherent in the low coherence interferometry-based systems. Hence, we propose a systematic design methodology WAVE-UNET to reconstruct the high-quality OCT images directly from the {\lambda}-space to reduce the complexity. The novel design paradigm surpasses the linearization procedures and uses DL to enhance the realism and quality of raw {\lambda}-space scans. This framework uses modified UNET having attention gating and residual connections, with IDFT processed {\lambda}-space fringes as the input. The method consistently outperforms the traditional OCT system by generating good-quality B-scans with highly reduced time-complexity.
Authors: Stavros Kassinos, Alessio Alexiadis
Abstract: Transformer Neural Networks are driving an explosion of activity and discovery in the field of Large Language Models (LLMs). In contrast, there have been only a few attempts to apply Transformers in engineering physics. Aiming to offer an easy entry point to physics-centric Transformers, we introduce a physics-informed Transformer model for solving the heat conduction problem in a 2D plate with Dirichlet boundary conditions. The model is implemented in the machine learning framework MLX and leverages the unified memory of Apple M-series processors. The use of MLX means that the models can be trained and perform predictions efficiently on personal machines with only modest memory requirements. To train, validate and test the Transformer model we solve the 2D heat conduction problem using central finite differences. Each finite difference solution in these sets is initialized with four random Dirichlet boundary conditions, a uniform but random internal temperature distribution and a randomly selected thermal diffusivity. Validation is performed in-line during training to monitor against over-fitting. The excellent performance of the trained model is demonstrated by predicting the evolution of the temperature field to steady state for the unseen test set of conditions.
Authors: Ammar Daskin
Abstract: In this paper, we show that an equivalent implementation of KAN can be done on quantum computers by simply combining quantum signal processing circuits in layers. This provides a powerful and robust path for the applications of KAN on quantum computers.
Authors: Suryavardan Suresh, Anku Rani, Parth Patwa, Aishwarya Reganti, Vinija Jain, Aman Chadha, Amitava Das, Amit Sheth, Asif Ekbal
Abstract: Researchers have found that fake news spreads much times faster than real news. This is a major problem, especially in today's world where social media is the key source of news for many among the younger population. Fact verification, thus, becomes an important task and many media sites contribute to the cause. Manual fact verification is a tedious task, given the volume of fake news online. The Factify5WQA shared task aims to increase research towards automated fake news detection by providing a dataset with an aspect-based question answering based fact verification method. Each claim and its supporting document is associated with 5W questions that help compare the two information sources. The objective performance measure in the task is done by comparing answers using BLEU score to measure the accuracy of the answers, followed by an accuracy measure of the classification. The task had submissions using custom training setup and pre-trained language-models among others. The best performing team posted an accuracy of 69.56%, which is a near 35% improvement over the baseline.
Authors: Tom\'as Feith, Akhil Arora, Martin Gerlach, Debjit Paul, Robert West
Abstract: Links are a fundamental part of information networks, turning isolated pieces of knowledge into a network of information that is much richer than the sum of its parts. However, adding a new link to the network is not trivial: it requires not only the identification of a suitable pair of source and target entities but also the understanding of the content of the source to locate a suitable position for the link in the text. The latter problem has not been addressed effectively, particularly in the absence of text spans in the source that could serve as anchors to insert a link to the target entity. To bridge this gap, we introduce and operationalize the task of entity insertion in information networks. Focusing on the case of Wikipedia, we empirically show that this problem is, both, relevant and challenging for editors. We compile a benchmark dataset in 105 languages and develop a framework for entity insertion called LocEI (Localized Entity Insertion) and its multilingual variant XLocEI. We show that XLocEI outperforms all baseline models (including state-of-the-art prompt-based ranking with LLMs such as GPT-4) and that it can be applied in a zero-shot manner on languages not seen during training with minimal performance drop. These findings are important for applying entity insertion models in practice, e.g., to support editors in adding links across the more than 300 language versions of Wikipedia.
Authors: Julia Briden, Yilun Du, Enrico M. Zucchelli, Richard Linares
Abstract: This work introduces TrajDiffuser, a compositional diffusion-based flexible and concurrent trajectory generator for 6 degrees of freedom powered descent guidance. TrajDiffuser is a statistical model that learns the multi-modal distributions of a dataset of simulated optimal trajectories, each subject to only one or few constraints that may vary for different trajectories. During inference, the trajectory is generated simultaneously over time, providing stable long-horizon planning, and constraints can be composed together, increasing the model's generalizability and decreasing the training data required. The generated trajectory is then used to initialize an optimizer, increasing its robustness and speed.
Authors: Yoonsoo Nam, Chris Mingard, Seok Hyeong Lee, Soufiane Hayou, Ard Louis
Abstract: Deep neural networks (DNNs) exhibit a remarkable ability to automatically learn data representations, finding appropriate features without human input. Here we present a method for analysing feature learning by decomposing DNNs into 1) a forward feature-map $\Phi$ that maps the input dataspace to the post-activations of the penultimate layer, and 2) a final linear layer that classifies the data. We diagonalize $\Phi$ with respect to the gradient descent operator and track feature learning by measuring how the eigenfunctions and eigenvalues of $\Phi$ change during training. Across many popular architectures and classification datasets, we find that DNNs converge, after just a few epochs, to a minimal feature (MF) regime dominated by a number of eigenfunctions equal to the number of classes. This behaviour resembles the neural collapse phenomenon studied at longer training times. For other DNN-data combinations, such as a fully connected network on CIFAR10, we find an extended feature (EF) regime where significantly more features are used. Optimal generalisation performance upon hyperparameter tuning typically coincides with the MF regime, but we also find examples of poor performance within the MF regime. Finally, we recast the phenomenon of neural collapse into a kernel picture which can be extended to broader tasks such as regression.
Authors: Artavazd Maranjyan, Omar Shaikh Omar, Peter Richt\'arik
Abstract: We study the problem of minimizing the expectation of smooth nonconvex functions with the help of several parallel workers whose role is to compute stochastic gradients. In particular, we focus on the challenging situation where the workers' compute times are arbitrarily heterogeneous and random. In the simpler regime characterized by arbitrarily heterogeneous but deterministic compute times, Tyurin and Richt\'arik (NeurIPS 2023) recently designed the first theoretically optimal asynchronous SGD method, called Rennala SGD, in terms of a novel complexity notion called time complexity. The starting point of our work is the observation that Rennala SGD can have arbitrarily bad performance in the presence of random compute times -- a setting it was not designed to handle. To advance our understanding of stochastic optimization in this challenging regime, we propose a new asynchronous SGD method, for which we coin the name MindFlayer SGD. Our theory and empirical results demonstrate the superiority of MindFlayer SGD over existing baselines, including Rennala SGD, in cases when the noise is heavy tailed.
Authors: Daniel Otero, Rafael Mateus, Randall Balestriero
Abstract: Accurate anomaly detection is critical in vision-based infrastructure inspection, where it helps prevent costly failures and enhances safety. Self-Supervised Learning (SSL) offers a promising approach by learning robust representations from unlabeled data. However, its application in anomaly detection remains underexplored. This paper addresses this gap by providing a comprehensive evaluation of SSL methods for real-world anomaly detection, focusing on sewer infrastructure. Using the Sewer-ML dataset, we evaluate lightweight models such as ViT-Tiny and ResNet-18 across SSL frameworks, including BYOL, Barlow Twins, SimCLR, DINO, and MAE, under varying class imbalance levels. Through 250 experiments, we rigorously assess the performance of these SSL methods to ensure a robust and comprehensive evaluation. Our findings highlight the superiority of joint-embedding methods like SimCLR and Barlow Twins over reconstruction-based approaches such as MAE, which struggle to maintain performance under class imbalance. Furthermore, we find that the SSL model choice is more critical than the backbone architecture. Additionally, we emphasize the need for better label-free assessments of SSL representations, as current methods like RankMe fail to adequately evaluate representation quality, making cross-validation without labels infeasible. Despite the remaining performance gap between SSL and supervised models, these findings highlight the potential of SSL to enhance anomaly detection, paving the way for further research in this underexplored area of SSL applications.
Authors: Yannick Marcel Holtkamp, Emiliano Godinez-Ramirez, Ulrich Kleinekath\"ofer
Abstract: Although recent advances in simulating open quantum systems have lead to significant progress, the applicability of numerically exact methods is still restricted to rather small systems. Hence, more approximate methods remain relevant due to their computational efficiency, enabling simulations of larger systems over extended timescales. In this study, we present advances for one such method, namely the Numerical Integration of Schr\"odinger Equation (NISE). Firstly, we introduce a modified ensemble-averaging procedure that improves the long-time behavior of the thermalized variant of the NISE scheme, termed Thermalized NISE. Secondly, we demonstrate how to use the NISE in conjunction with (highly) structured spectral densities by utilizing a noise generating algorithm for arbitrary structured noise. This algorithm also serves as a tool for establishing best practices in determining spectral densities from excited state calculations along molecular dynamics or quantum mechanics/molecular mechanics trajectories. Finally, we assess the ability of the NISE approach to calculate absorption spectra and demonstrate the utility of the proposed modifications by determining population dynamics.
Authors: Ankit Bhardwaj, Ananth Balashankar, Shiva Iyer, Nita Soans, Anant Sudarshan, Rohini Pande, Lakshminarayanan Subramanian
Abstract: Effective air pollution management in urban areas relies on both monitoring and mitigation strategies, yet high costs often limit sensor networks to a few key pollution hotspots. In this paper, we show that New Delhi's public sensor network is insufficient for identifying all pollution hotspots. To address this, we augmented the city's network with 28 low-cost sensors, monitoring PM 2.5 concentrations over 30 months (May 2018 to November 2020). Our analysis uncovered 189 additional hotspots, supplementing the 660 already detected by the government network. We observed that Space-Time Kriging with limited but accurate sensor data provides a more robust and generalizable approach for identifying these hotspots, as compared to deep learning models that require large amounts of fine-grained multi-modal data (emissions inventory, meteorology, etc.) which was not reliably, frequently and accurately available in the New Delhi context. Using Space-Time Kriging, we achieved 98% precision and 95.4% recall in detecting hotspots with 50% sensor failure. Furthermore, this method proved effective in predicting hotspots in areas without sensors, achieving 95.3% precision and 88.5% recall in the case of 50% missing sensors. Our findings revealed that a significant portion of New Delhi's population, around 23 million people, was exposed to pollution hotspots for at least half of the study period. We also identified areas beyond the reach of the public sensor network that should be prioritized for pollution control. These results highlight the need for more comprehensive monitoring networks and suggest Space-Time Kriging as a viable solution for cities facing similar resource constraints.
Authors: Peiqi Wang, Barbara D. Lam, Yingcheng Liu, Ameneh Asgari-Targhi, Rameswar Panda, William M. Wells, Tina Kapur, Polina Golland
Abstract: We present a novel approach to calibrating linguistic expressions of certainty, e.g., "Maybe" and "Likely". Unlike prior work that assigns a single score to each certainty phrase, we model uncertainty as distributions over the simplex to capture their semantics more accurately. To accommodate this new representation of certainty, we generalize existing measures of miscalibration and introduce a novel post-hoc calibration method. Leveraging these tools, we analyze the calibration of both humans (e.g., radiologists) and computational models (e.g., language models) and provide interpretable suggestions to improve their calibration.
Authors: Rached Bouchoucha, Ahmed Haj Yahmed, Darshan Patil, Janarthanan Rajendran, Amin Nikanjam, Sarath Chandar, Foutse Khomh
Abstract: Deep reinforcement learning (DRL) has shown success in diverse domains such as robotics, computer games, and recommendation systems. However, like any other software system, DRL-based software systems are susceptible to faults that pose unique challenges for debugging and diagnosing. These faults often result in unexpected behavior without explicit failures and error messages, making debugging difficult and time-consuming. Therefore, automating the monitoring and diagnosis of DRL systems is crucial to alleviate the burden on developers. In this paper, we propose RLExplorer, the first fault diagnosis approach for DRL-based software systems. RLExplorer automatically monitors training traces and runs diagnosis routines based on properties of the DRL learning dynamics to detect the occurrence of DRL-specific faults. It then logs the results of these diagnoses as warnings that cover theoretical concepts, recommended practices, and potential solutions to the identified faults. We conducted two sets of evaluations to assess RLExplorer. Our first evaluation of faulty DRL samples from Stack Overflow revealed that our approach can effectively diagnose real faults in 83% of the cases. Our second evaluation of RLExplorer with 15 DRL experts/developers showed that (1) RLExplorer could identify 3.6 times more defects than manual debugging and (2) RLExplorer is easily integrated into DRL applications.
Authors: Dohun Lee, Bryan S Kim, Geon Yeong Park, Jong Chul Ye
Abstract: Text-to-image (T2I) diffusion models have revolutionized visual content creation, but extending these capabilities to text-to-video (T2V) generation remains a challenge, particularly in preserving temporal consistency. Existing methods that aim to improve consistency often cause trade-offs such as reduced imaging quality and impractical computational time. To address these issues we introduce VideoGuide, a novel framework that enhances the temporal consistency of pretrained T2V models without the need for additional training or fine-tuning. Instead, VideoGuide leverages any pretrained video diffusion model (VDM) or itself as a guide during the early stages of inference, improving temporal quality by interpolating the guiding model's denoised samples into the sampling model's denoising process. The proposed method brings about significant improvement in temporal consistency and image fidelity, providing a cost-effective and practical solution that synergizes the strengths of various video diffusion models. Furthermore, we demonstrate prior distillation, revealing that base models can achieve enhanced text coherence by utilizing the superior data prior of the guiding model through the proposed method. Project Page: http://videoguide2025.github.io/
Authors: Hadi Hosseini, Sanjukta Roy, Duohan Zhang
Abstract: Two-sided matching markets describe a large class of problems wherein participants from one side of the market must be matched to those from the other side according to their preferences. In many real-world applications (e.g. content matching or online labor markets), the knowledge about preferences may not be readily available and must be learned, i.e., one side of the market (aka agents) may not know their preferences over the other side (aka arms). Recent research on online settings has focused primarily on welfare optimization aspects (i.e. minimizing the overall regret) while paying little attention to the game-theoretic properties such as the stability of the final matching. In this paper, we exploit the structure of stable solutions to devise algorithms that improve the likelihood of finding stable solutions. We initiate the study of the sample complexity of finding a stable matching, and provide theoretical bounds on the number of samples needed to reach a stable matching with high probability. Finally, our empirical results demonstrate intriguing tradeoffs between stability and optimality of the proposed algorithms, further complementing our theoretical findings.
Authors: Javier Marin
Abstract: This paper introduces an innovative approach to analyzing and improving multi-hop reasoning in AI systems by drawing inspiration from Hamiltonian mechanics. We propose a novel framework that maps reasoning chains in embedding spaces to Hamiltonian systems, allowing us to leverage powerful analytical tools from classical physics. Our method defines a Hamiltonian function that balances the progression of reasoning (kinetic energy) against the relevance to the question at hand (potential energy). Using this framework, we analyze a large dataset of reasoning chains from a multi-hop question-answering task, revealing intriguing patterns that distinguish valid from invalid reasoning. We show that valid reasoning chains have lower Hamiltonian energy and move in ways that make the best trade-off between getting more information and answering the right question. Furthermore, we demonstrate the application of this framework to steer the creation of more efficient reasoning algorithms within AI systems. Our results not only provide new insights into the nature of valid reasoning but also open up exciting possibilities for physics-inspired approaches to understanding and improving artificial intelligence.
Authors: Zhengting Chen, Lei Cheng, Lianghui Ding, Quanshi Zhang
Abstract: This paper presents a method to explain the internal representation structure of a neural network for image generation. Specifically, our method disentangles primitive feature components from the intermediate-layer feature of the neural network, which ensures that each feature component is exclusively used to generate a specific set of image regions. In this way, the generation of the entire image can be considered as the superposition of different pre-encoded primitive regional patterns, each being generated by a feature component. We find that the feature component can be represented as an OR relationship between the demands for generating different image regions, which is encoded by the neural network. Therefore, we extend the Harsanyi interaction to represent such an OR interaction to disentangle the feature component. Experiments show a clear correspondence between each feature component and the generation of specific image regions.
Authors: Shivank Garg, Manyana Tiwari
Abstract: This study investigates the generation of unsafe or harmful content in state-of-the-art generative models, focusing on methods for restricting such generations. We introduce a novel training-free approach using attention reweighing to remove unsafe concepts without additional training during inference. We compare our method against existing ablation methods, evaluating the performance on both, direct and adversarial jailbreak prompts, using qualitative and quantitative metrics. We hypothesize potential reasons for the observed results and discuss the limitations and broader implications of content restriction.
Authors: Melanie Rieff, Fabian Holzberger, Oksana Lapina, Geir Ringstad, Lars Magnus Valnes, Bogna Warsza, Kent-Andre Mardal, Per Kristian Eide, Barbara Wohlmuth
Abstract: Previous work shows evidence that cerebrospinal fluid (CSF) plays a crucial role in brain waste clearance processes, and that altered flow patterns are associated with various diseases of the central nervous system. In this study, we investigate the potential of deep learning to predict the distribution in human brain of a gadolinium-based CSF contrast agent (tracer) administered intrathecal. For this, T1-weighted magnetic resonance imaging (MRI) scans taken at multiple time points before and after intrathecal injection were utilized. We propose a U-net-based supervised learning model to predict pixel-wise signal increases at their peak after 24 hours. Its performance is evaluated based on different tracer distribution stages provided during training, including predictions from baseline scans taken before injection. Our findings indicate that using imaging data from just the first two hours post-injection for training yields tracer flow predictions comparable to those trained with additional later-stage scans. The model was further validated by comparing ventricular reflux gradings provided by neuroradiologists, and inter-rater grading among medical experts and the model showed excellent agreement. Our results demonstrate the potential of deep learning-based methods for CSF flow prediction, suggesting that fewer MRI scans could be sufficient for clinical analysis, which might significantly improve clinical efficiency, patient well-being, and lower healthcare costs.
Authors: Georgii Novikov, Alexander Gneushev, Alexey Kadeishvili, Ivan Oseledets
Abstract: Nearest-neighbor search in large vector databases is crucial for various machine learning applications. This paper introduces a novel method using tensor-train (TT) low-rank tensor decomposition to efficiently represent point clouds and enable fast approximate nearest-neighbor searches. We propose a probabilistic interpretation and utilize density estimation losses like Sliced Wasserstein to train TT decompositions, resulting in robust point cloud compression. We reveal an inherent hierarchical structure within TT point clouds, facilitating efficient approximate nearest-neighbor searches. In our paper, we provide detailed insights into the methodology and conduct comprehensive comparisons with existing methods. We demonstrate its effectiveness in various scenarios, including out-of-distribution (OOD) detection problems and approximate nearest-neighbor (ANN) search tasks.
Authors: Jinhao Li, Jiaming Xu, Shan Huang, Yonghua Chen, Wen Li, Jun Liu, Yaoxiu Lian, Jiayi Pan, Li Ding, Hao Zhou, Guohao Dai
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities across various fields, from natural language understanding to text generation. Compared to non-generative LLMs like BERT and DeBERTa, generative LLMs like GPT series and Llama series are currently the main focus due to their superior algorithmic performance. The advancements in generative LLMs are closely intertwined with the development of hardware capabilities. Various hardware platforms exhibit distinct hardware characteristics, which can help improve LLM inference performance. Therefore, this paper comprehensively surveys efficient generative LLM inference on different hardware platforms. First, we provide an overview of the algorithm architecture of mainstream generative LLMs and delve into the inference process. Then, we summarize different optimization methods for different platforms such as CPU, GPU, FPGA, ASIC, and PIM/NDP, and provide inference results for generative LLMs. Furthermore, we perform a qualitative and quantitative comparison of inference performance with batch sizes 1 and 8 on different hardware platforms by considering hardware power consumption, absolute inference speed (tokens/s), and energy efficiency (tokens/J). We compare the performance of the same optimization methods across different hardware platforms, the performance across different hardware platforms, and the performance of different methods on the same hardware platform. This provides a systematic and comprehensive summary of existing inference acceleration work by integrating software optimization methods and hardware platforms, which can point to the future trends and potential developments of generative LLMs and hardware technology for edge-side scenarios.
Authors: Hakaze Cho, Mariko Kato, Yoshihiro Sakai, Naoya Inoue
Abstract: In-context Learning (ICL) is an emerging few-shot learning paradigm on Language Models (LMs) with inner mechanisms un-explored. There are already existing works describing the inner processing of ICL, while they struggle to capture all the inference phenomena in large language models. Therefore, this paper proposes a comprehensive circuit to model the inference dynamics and try to explain the observed phenomena of ICL. In detail, we divide ICL inference into 3 major operations: (1) Summarize: LMs encode every input text (demonstrations and queries) into linear representation in the hidden states with sufficient information to solve ICL tasks. (2) Semantics Merge: LMs merge the encoded representations of demonstrations with their corresponding label tokens to produce joint representations of labels and demonstrations. (3) Feature Retrieval and Copy: LMs search the joint representations similar to the query representation on a task subspace, and copy the searched representations into the query. Then, language model heads capture these copied label representations to a certain extent and decode them into predicted labels. The proposed inference circuit successfully captured many phenomena observed during the ICL process, making it a comprehensive and practical explanation of the ICL inference process. Moreover, ablation analysis by disabling the proposed steps seriously damages the ICL performance, suggesting the proposed inference circuit is a dominating mechanism. Additionally, we confirm and list some bypass mechanisms that solve ICL tasks in parallel with the proposed circuit.
Authors: Ismail Alkhouri, Shijun Liang, Cheng-Han Huang, Jimmy Dai, Qing Qu, Saiprasad Ravishankar, Rongrong Wang
Abstract: Diffusion models (DMs) are a class of generative models that allow sampling from a distribution learned over a training set. When applied to solving inverse imaging problems (IPs), the reverse sampling steps of DMs are typically modified to approximately sample from a measurement-conditioned distribution in the image space. However, these modifications may be unsuitable for certain settings (such as in the presence of measurement noise) and non-linear tasks, as they often struggle to correct errors from earlier sampling steps and generally require a large number of optimization and/or sampling steps. To address these challenges, we state three conditions for achieving measurement-consistent diffusion trajectories. Building on these conditions, we propose a new optimization-based sampling method that not only enforces the standard data manifold measurement consistency and forward diffusion consistency, as seen in previous studies, but also incorporates backward diffusion consistency that maintains a diffusion trajectory by optimizing over the input of the pre-trained model at every sampling step. By enforcing these conditions, either implicitly or explicitly, our sampler requires significantly fewer reverse steps. Therefore, we refer to our accelerated method as Step-wise Triple-Consistent Sampling (SITCOM). Compared to existing state-of-the-art baseline methods, under different levels of measurement noise, our extensive experiments across five linear and three non-linear image restoration tasks demonstrate that SITCOM achieves competitive or superior results in terms of standard image similarity metrics while requiring a significantly reduced run-time across all considered tasks.
Authors: Alon Beck, Noam Levi, Yohai Bar-Sinai
Abstract: We study the generalization properties of binary logistic classification in a simplified setting, for which a "memorizing" and "generalizing" solution can always be strictly defined, and elucidate empirically and analytically the mechanism underlying Grokking in its dynamics. We analyze the asymptotic long-time dynamics of logistic classification on a random feature model with a constant label and show that it exhibits Grokking, in the sense of delayed generalization and non-monotonic test loss. We find that Grokking is amplified when classification is applied to training sets which are on the verge of linear separability. Even though a perfect generalizing solution always exists, we prove the implicit bias of the logisitc loss will cause the model to overfit if the training data is linearly separable from the origin. For training sets that are not separable from the origin, the model will always generalize perfectly asymptotically, but overfitting may occur at early stages of training. Importantly, in the vicinity of the transition, that is, for training sets that are almost separable from the origin, the model may overfit for arbitrarily long times before generalizing. We gain more insights by examining a tractable one-dimensional toy model that quantitatively captures the key features of the full model. Finally, we highlight intriguing common properties of our findings with recent literature, suggesting that grokking generally occurs in proximity to the interpolation threshold, reminiscent of critical phenomena often observed in physical systems.
Authors: Beatrice Casey, Joanna C. S. Santos, Mehdi Mirakhorli
Abstract: The development of machine learning (ML) techniques has led to ample opportunities for developers to develop and deploy their own models. Hugging Face serves as an open source platform where developers can share and download other models in an effort to make ML development more collaborative. In order for models to be shared, they first need to be serialized. Certain Python serialization methods are considered unsafe, as they are vulnerable to object injection. This paper investigates the pervasiveness of these unsafe serialization methods across Hugging Face, and demonstrates through an exploitation approach, that models using unsafe serialization methods can be exploited and shared, creating an unsafe environment for ML developers. We investigate to what extent Hugging Face is able to flag repositories and files using unsafe serialization methods, and develop a technique to detect malicious models. Our results show that Hugging Face is home to a wide range of potentially vulnerable models.
Authors: Zhaorui Tan, Xi Yang, Qiufeng Wang, Anh Nguyen, Kaizhu Huang
Abstract: Vision models excel in image classification but struggle to generalize to unseen data, such as classifying images from unseen domains or discovering novel categories. In this paper, we explore the relationship between logical reasoning and deep learning generalization in visual classification. A logical regularization termed L-Reg is derived which bridges a logical analysis framework to image classification. Our work reveals that L-Reg reduces the complexity of the model in terms of the feature distribution and classifier weights. Specifically, we unveil the interpretability brought by L-Reg, as it enables the model to extract the salient features, such as faces to persons, for classification. Theoretical analysis and experiments demonstrate that L-Reg enhances generalization across various scenarios, including multi-domain generalization and generalized category discovery. In complex real-world scenarios where images span unknown classes and unseen domains, L-Reg consistently improves generalization, highlighting its practical efficacy.
Authors: Vy Nguyen, Chau Pham
Abstract: The increasing frequency of suicidal thoughts highlights the importance of early detection and intervention. Social media platforms, where users often share personal experiences and seek help, could be utilized to identify individuals at risk. However, the large volume of daily posts makes manual review impractical. This paper explores the use of Large Language Models (LLMs) to automatically detect suicidal content in text-based social media posts. We propose a novel method for generating pseudo-labels for unlabeled data by prompting LLMs, along with traditional classification fine-tuning techniques to enhance label accuracy. To create a strong suicide detection model, we develop an ensemble approach involving prompting with Qwen2-72B-Instruct, and using fine-tuned models such as Llama3-8B, Llama3.1-8B, and Gemma2-9B. We evaluate our approach on the dataset of the Suicide Ideation Detection on Social Media Challenge, a track of the IEEE Big Data 2024 Big Data Cup. Additionally, we conduct a comprehensive analysis to assess the impact of different models and fine-tuning strategies on detection performance. Experimental results show that the ensemble model significantly improves the detection accuracy, by 5% points compared with the individual models. It achieves a weight F1 score of 0.770 on the public test set, and 0.731 on the private test set, providing a promising solution for identifying suicidal content in social media. Our analysis shows that the choice of LLMs affects the prompting performance, with larger models providing better accuracy. Our code and checkpoints are publicly available at https://github.com/khanhvynguyen/Suicide_Detection_LLMs.
URLs: https://github.com/khanhvynguyen/Suicide_Detection_LLMs.
Authors: Han Yang, Kun Su, Yutong Zhang, Jiaben Chen, Kaizhi Qian, Gaowen Liu, Chuang Gan
Abstract: We introduce UniMuMo, a unified multimodal model capable of taking arbitrary text, music, and motion data as input conditions to generate outputs across all three modalities. To address the lack of time-synchronized data, we align unpaired music and motion data based on rhythmic patterns to leverage existing large-scale music-only and motion-only datasets. By converting music, motion, and text into token-based representation, our model bridges these modalities through a unified encoder-decoder transformer architecture. To support multiple generation tasks within a single framework, we introduce several architectural improvements. We propose encoding motion with a music codebook, mapping motion into the same feature space as music. We introduce a music-motion parallel generation scheme that unifies all music and motion generation tasks into a single transformer decoder architecture with a single training task of music-motion joint generation. Moreover, the model is designed by fine-tuning existing pre-trained single-modality models, significantly reducing computational demands. Extensive experiments demonstrate that UniMuMo achieves competitive results on all unidirectional generation benchmarks across music, motion, and text modalities. Quantitative results are available in the \href{https://hanyangclarence.github.io/unimumo_demo/}{project page}.
Authors: Wencong Cheng, Jiangjiang Xia, Chang Qu, Zhigang Wang, Xinyi Zeng, Fang Huang, Tianye Li
Abstract: To promote the practical application of AI Global Weather Forecasting Models (AIGWFM), we have developed an adaptable application platform named 'YanTian'. This platform enhances existing open-source AIGWFM with a suite of capability-enhancing modules and is constructed by a "loosely coupled" plug-in architecture. The goal of 'YanTian' is to address the limitations of current open-source AIGWFM in operational application, including improving local forecast accuracy, providing spatial high-resolution forecasts, increasing density of forecast intervals, and generating diverse products with the provision of AIGC capabilities. 'YianTian' also provides a simple, visualized user interface, allowing meteorologists easily access both basic and extended capabilities of the platform by simply configuring the platform UI. Users do not need to possess the complex artificial intelligence knowledge and the coding techniques. Additionally, 'YianTian' can be deployed on a PC with GPUs. We hope 'YianTian' can facilitate the operational widespread adoption of AIGWFMs.
Authors: Seonghwan Seo, Minsu Kim, Tony Shen, Martin Ester, Jinkyoo Park, Sungsoo Ahn, Woo Youn Kim
Abstract: Generative models in drug discovery have recently gained attention as efficient alternatives to brute-force virtual screening. However, most existing models do not account for synthesizability, limiting their practical use in real-world scenarios. In this paper, we propose RxnFlow, which sequentially assembles molecules using predefined molecular building blocks and chemical reaction templates to constrain the synthetic chemical pathway. We then train on this sequential generating process with the objective of generative flow networks (GFlowNets) to generate both highly rewarded and diverse molecules. To mitigate the large action space of synthetic pathways in GFlowNets, we implement a novel action space subsampling method. This enables RxnFlow to learn generative flows over extensive action spaces comprising combinations of 1.2 million building blocks and 71 reaction templates without significant computational overhead. Additionally, RxnFlow can employ modified or expanded action spaces for generation without retraining, allowing for the introduction of additional objectives or the incorporation of newly discovered building blocks. We experimentally demonstrate that RxnFlow outperforms existing reaction-based and fragment-based models in pocket-specific optimization across various target pockets. Furthermore, RxnFlow achieves state-of-the-art performance on CrossDocked2020 for pocket-conditional generation, with an average Vina score of -8.85kcal/mol and 34.8% synthesizability.
Authors: Amanda Aird, Elena \v{S}tefancov\'a, Cassidy All, Amy Voida, Martin Homola, Nicholas Mattei, Robin Burke
Abstract: Algorithmic fairness in recommender systems requires close attention to the needs of a diverse set of stakeholders that may have competing interests. Previous work in this area has often been limited by fixed, single-objective definitions of fairness, built into algorithms or optimization criteria that are applied to a single fairness dimension or, at most, applied identically across dimensions. These narrow conceptualizations limit the ability to adapt fairness-aware solutions to the wide range of stakeholder needs and fairness definitions that arise in practice. Our work approaches recommendation fairness from the standpoint of computational social choice, using a multi-agent framework. In this paper, we explore the properties of different social choice mechanisms and demonstrate the successful integration of multiple, heterogeneous fairness definitions across multiple data sets.
Authors: Sabrina Guidotti, Gregor Donabauer, Simone Somazzi, Udo Kruschwitz, Davide Taibi, Dimitri Ognibene
Abstract: The widespread use of social media has highlighted potential negative impacts on society and individuals, largely driven by recommendation algorithms that shape user behavior and social dynamics. Understanding these algorithms is essential but challenging due to the complex, distributed nature of social media networks as well as limited access to real-world data. This study proposes to use academic social networks as a proxy for investigating recommendation systems in social media. By employing Graph Neural Networks (GNNs), we develop a model that separates the prediction of academic infosphere from behavior prediction, allowing us to simulate recommender-generated infospheres and assess the model's performance in predicting future co-authorships. Our approach aims to improve our understanding of recommendation systems' roles and social networks modeling. To support the reproducibility of our work we publicly make available our implementations: https://github.com/DimNeuroLab/academic_network_project
URLs: https://github.com/DimNeuroLab/academic_network_project
Authors: Ehsan Ebrahimzadeh, Nikhil Monga, Hang Gao, Alex Cozzi, Abraham Bagherjeiran
Abstract: We develop a decision making framework to cast the problem of learning a ranking policy for search or recommendation engines in a two-sided e-commerce marketplace as an expected reward optimization problem using observational data. As a value allocation mechanism, the ranking policy allocates retrieved items to the designated slots so as to maximize the user utility from the slotted items, at any given stage of the shopping journey. The objective of this allocation can in turn be defined with respect to the underlying probabilistic user browsing model as the expected number of interaction events on presented items matching the user intent, given the ranking context. Through recognizing the effect of ranking as an intervention action to inform users' interactions with slotted items and the corresponding economic value of the interaction events for the marketplace, we formulate the expected reward of the marketplace as the collective value from all presented ranking actions. The key element in this formulation is a notion of context value distribution, which signifies not only the attribution of value to ranking interventions within a session but also the distribution of marketplace reward across user sessions. We build empirical estimates for the expected reward of the marketplace from observational data that account for the heterogeneity of economic value across session contexts as well as the distribution shifts in learning from observational user activity data. The ranking policy can then be trained by optimizing the empirical expected reward estimates via standard Bayesian inference techniques. We report empirical results for a product search ranking task in a major e-commerce platform demonstrating the fundamental trade-offs governed by ranking polices trained on empirical reward estimates with respect to extreme choices of the context value distribution.
Authors: Mehwish Ghafoor, Arif Mahmood, Muhammad Bilal
Abstract: In the field of 3D Human Pose Estimation from monocular videos, the presence of diverse occlusion types presents a formidable challenge. Prior research has made progress by harnessing spatial and temporal cues to infer 3D poses from 2D joint observations. This paper introduces a Dual Transformer Fusion (DTF) algorithm, a novel approach to obtain a holistic 3D pose estimation, even in the presence of severe occlusions. Confronting the issue of occlusion-induced missing joint data, we propose a temporal interpolation-based occlusion guidance mechanism. To enable precise 3D Human Pose Estimation, our approach leverages the innovative DTF architecture, which first generates a pair of intermediate views. Each intermediate-view undergoes spatial refinement through a self-refinement schema. Subsequently, these intermediate-views are fused to yield the final 3D human pose estimation. The entire system is end-to-end trainable. Through extensive experiments conducted on the Human3.6M and MPI-INF-3DHP datasets, our method's performance is rigorously evaluated. Notably, our approach outperforms existing state-of-the-art methods on both datasets, yielding substantial improvements. The code is available here: https://github.com/MehwishG/DTF.
Authors: Tianjian Li, Haoran Xu, Weiting Tan, Dongwei Jiang, Kenton Murray, Daniel Khashabi
Abstract: Data availability across domains often follows a long-tail distribution: a few domains have abundant data, while most face data scarcity. This imbalance poses challenges in training language models uniformly across all domains. In our study, we focus on multilingual settings, where data sizes vary significantly between high- and low-resource languages. Common strategies to address this include upsampling low-resource languages (Temperature Sampling) or upweighting their loss (Scalarization). Although often considered equivalent, this assumption has not been proven, which motivates our study. Through both theoretical and empirical analysis, we identify the conditions under which these approaches are equivalent and when they diverge. Specifically, we demonstrate that these two methods are equivalent under full gradient descent, but this equivalence breaks down with stochastic gradient descent. Empirically, we observe that Temperature Sampling converges more quickly but is prone to overfitting. We argue that this faster convergence is likely due to the lower variance in gradient estimations, as shown theoretically. Based on these insights, we propose Cooldown, a strategy that reduces sampling temperature during training, accelerating convergence without overfitting to low-resource languages. Our method is competitive with existing data re-weighting and offers computational efficiency.
Authors: Mathias Jackermeier, Alessandro Abate
Abstract: Linear temporal logic (LTL) has recently been adopted as a powerful formalism for specifying complex, temporally extended tasks in reinforcement learning (RL). However, learning policies that efficiently satisfy arbitrary specifications not observed during training remains a challenging problem. Existing approaches suffer from several shortcomings: they are often only applicable to finite-horizon fragments of LTL, are restricted to suboptimal solutions, and do not adequately handle safety constraints. In this work, we propose a novel learning approach to address these concerns. Our method leverages the structure of B\"uchi automata, which explicitly represent the semantics of LTL specifications, to learn policies conditioned on sequences of truth assignments that lead to satisfying the desired formulae. Experiments in a variety of discrete and continuous domains demonstrate that our approach is able to zero-shot satisfy a wide range of finite- and infinite-horizon specifications, and outperforms existing methods in terms of both satisfaction probability and efficiency.
Authors: Aniruddh Sriram, Fangyuan Xu, Eunsol Choi, Greg Durrett
Abstract: Recent work on fact-checking addresses a realistic setting where models incorporate evidence retrieved from the web to decide the veracity of claims. A bottleneck in this pipeline is in retrieving relevant evidence: traditional methods may surface documents directly related to a claim, but fact-checking complex claims requires more inferences. For instance, a document about how a vaccine was developed is relevant to addressing claims about what it might contain, even if it does not address them directly. We present Contrastive Fact-Checking Reranker (CFR), an improved retriever for this setting. By leveraging the AVeriTeC dataset, which annotates subquestions for claims with human written answers from evidence documents, we fine-tune Contriever with a contrastive objective based on multiple training signals, including distillation from GPT-4, evaluating subquestion answers, and gold labels in the dataset. We evaluate our model on both retrieval and end-to-end veracity judgments about claims. On the AVeriTeC dataset, we find a 6\% improvement in veracity classification accuracy. We also show our gains can be transferred to FEVER, ClaimDecomp, HotpotQA, and a synthetic dataset requiring retrievers to make inferences.
Authors: Chaithanya Bandi, Hari Bandi, Abir Harrasse
Abstract: This paper explores optimal architectures for evaluating the outputs of large language models (LLMs) using LLMs themselves. We propose a novel framework that interprets LLMs as advocates within an ensemble of interacting agents, allowing them to defend their answers and reach conclusions through a judge and jury system. This approach offers a more dynamic and comprehensive evaluation process compared to traditional human-based assessments or automated metrics. We discuss the motivation behind this framework, its key components, and comparative advantages. We also present a probabilistic model to evaluate the error reduction achieved by iterative advocate systems. Finally, we outline experiments to validate the effectiveness of multi-advocate architectures and discuss future research directions.
Authors: Christopher R. Wentland, Francesco Rizzi, Joshua Barnett, Irina Tezaur
Abstract: This paper presents and evaluates a framework for the coupling of subdomain-local projection-based reduced order models (PROMs) using the Schwarz alternating method following a domain decomposition (DD) of the spatial domain on which a given problem of interest is posed. In this approach, the solution on the full domain is obtained via an iterative process in which a sequence of subdomain-local problems are solved, with information propagating between subdomains through transmission boundary conditions (BCs). We explore several new directions involving the Schwarz alternating method aimed at maximizing the method's efficiency and flexibility, and demonstrate it on three challenging two-dimensional nonlinear hyperbolic problems: the shallow water equations, Burgers' equation, and the compressible Euler equations. We demonstrate that, for a cell-centered finite volume discretization and a non-overlapping DD, it is possible to obtain a stable and accurate coupled model utilizing Dirichlet-Dirichlet (rather than Robin-Robin or alternating Dirichlet-Neumann) transmission BCs on the subdomain boundaries. We additionally explore the impact of boundary sampling when utilizing the Schwarz alternating method to couple subdomain-local hyper-reduced PROMs. Our numerical results suggest that the proposed methodology has the potential to improve PROM accuracy by enabling the spatial localization of these models via domain decomposition, and achieve up to two orders of magnitude speedup over equivalent coupled full order model solutions and moderate speedups over analogous monolithic solutions.
Authors: Yanxi Hou, Xiaolan Xia, Guangyuan Gao
Abstract: Modeling insurance claim amounts and classifying claims into different risk levels are critical yet challenging tasks. Traditional predictive models for insurance claims often overlook the valuable information embedded in claim descriptions. This paper introduces a novel approach by developing a joint mixture model that integrates both claim descriptions and claim amounts. Our method establishes a probabilistic link between textual descriptions and loss amounts, enhancing the accuracy of claims clustering and prediction. In our proposed model, the latent topic/component indicator serves as a proxy for both the thematic content of the claim description and the component of loss distributions. Specifically, conditioned on the topic/component indicator, the claim description follows a multinomial distribution, while the claim amount follows a component loss distribution. We propose two methods for model calibration: an EM algorithm for maximum a posteriori estimates, and an MH-within-Gibbs sampler algorithm for the posterior distribution. The empirical study demonstrates that the proposed methods work effectively, providing interpretable claims clustering and prediction.
Authors: Minchan Kim, Myeonghun Jeong, Joun Yeop Lee, Nam Soo Kim
Abstract: We present SegINR, a novel approach to neural Text-to-Speech (TTS) that addresses sequence alignment without relying on an auxiliary duration predictor and complex autoregressive (AR) or non-autoregressive (NAR) frame-level sequence modeling. SegINR simplifies the process by converting text sequences directly into frame-level features. It leverages an optimal text encoder to extract embeddings, transforming each into a segment of frame-level features using a conditional implicit neural representation (INR). This method, named segment-wise INR (SegINR), models temporal dynamics within each segment and autonomously defines segment boundaries, reducing computational costs. We integrate SegINR into a two-stage TTS framework, using it for semantic token prediction. Our experiments in zero-shot adaptive TTS scenarios demonstrate that SegINR outperforms conventional methods in speech quality with computational efficiency.
Authors: Xiaomin Li, Mingye Gao, Zhiwei Zhang, Chang Yue, Hong Hu
Abstract: The quality of training data significantly impacts the performance of large language models (LLMs). There are increasing studies using LLMs to rate and select data based on several human-crafted metrics (rules). However, these conventional rule-based approaches often depend too heavily on human heuristics, lack effective metrics for assessing rules, and exhibit limited adaptability to new tasks. In our study, we introduce an innovative rule-based framework that utilizes the orthogonality of score vectors associated with rules as a novel metric for rule evaluations. Our approach includes an automated pipeline that first uses LLMs to generate a diverse set of rules, encompassing various rating dimensions to evaluate data quality. Then it rates a batch of data based on these rules and uses the determinantal point process (DPP) from random matrix theory to select the most orthogonal score vectors, thereby identifying a set of independent rules. These rules are subsequently used to evaluate all data, selecting samples with the highest average scores for downstream tasks such as LLM training. We verify the effectiveness of our method through two experimental setups: 1) comparisons with ground truth ratings and 2) benchmarking LLMs trained with the chosen data. Our comprehensive experiments cover a range of scenarios, including general pre-training and domain-specific fine-tuning in areas such as IMDB, Medical, Math, and Code. The outcomes demonstrate that our DPP-based rule rating method consistently outperforms other approaches, including rule-free rating, uniform sampling, importance resampling, and QuRating, in terms of both rating precision and model performance.
Authors: Dylan Zhang, Justin Wang, Francois Charton
Abstract: Understanding and accurately following instructions is critical for large language models (LLMs) to be effective across diverse tasks. In this work, we rigorously examine the key factors that enable models to generalize to unseen instructions, providing insights to guide the collection of data for instruction-tuning. Through controlled experiments, inspired by the Turing-complete Markov algorithm, we demonstrate that such generalization $\textbf{only emerges}$ when training data is diversified enough across semantic domains. Our findings also reveal that merely diversifying within limited domains fails to ensure robust generalization. In contrast, cross-domain data diversification, even under constrained data budgets, significantly enhances a model's adaptability. We further extend our analysis to real-world scenarios, including fine-tuning of $\textit{$\textbf{specialist}$}$ and $\textit{$\textbf{generalist}$}$ models. In both cases, we demonstrate that 1) better performance can be achieved by increasing the diversity of an established dataset while keeping the data size constant, and 2) when scaling up the data, diversifying the semantics of instructions is more effective than simply increasing the quantity of similar data. Our research provides important insights for dataset collation, particularly when optimizing model performance by expanding training data for both specialist and generalist scenarios. We show that careful consideration of data diversification is key: training specialist models with data extending beyond their core domain leads to significant performance improvements, while generalist models benefit from diverse data mixtures that enhance their overall instruction-following capabilities across a wide range of applications. Our results highlight the critical role of strategic diversification and offer clear guidelines for improving data quality.
Authors: Si-An Chen, Lesly Miculicich, Julian Martin Eisenschlos, Zifeng Wang, Zilong Wang, Yanfei Chen, Yasuhisa Fujii, Hsuan-Tien Lin, Chen-Yu Lee, Tomas Pfister
Abstract: Recent advancements in language models (LMs) have notably enhanced their ability to reason with tabular data, primarily through program-aided mechanisms that manipulate and analyze tables. However, these methods often require the entire table as input, leading to scalability challenges due to the positional bias or context length constraints. In response to these challenges, we introduce TableRAG, a Retrieval-Augmented Generation (RAG) framework specifically designed for LM-based table understanding. TableRAG leverages query expansion combined with schema and cell retrieval to pinpoint crucial information before providing it to the LMs. This enables more efficient data encoding and precise retrieval, significantly reducing prompt lengths and mitigating information loss. We have developed two new million-token benchmarks from the Arcade and BIRD-SQL datasets to thoroughly evaluate TableRAG's effectiveness at scale. Our results demonstrate that TableRAG's retrieval design achieves the highest retrieval quality, leading to the new state-of-the-art performance on large-scale table understanding.
Authors: Long Wu (University of Alberta), Xunyuan Yin (University of Alberta), Lei Pan (University of Alberta), Jinfeng Liu (University of Alberta)
Abstract: Integrated energy systems (IESs) are complex systems consisting of diverse operating units spanning multiple domains. To address its operational challenges, we propose a physics-informed hybrid time-series neural network (NN) surrogate to predict the dynamic performance of IESs across multiple time scales. This neural network-based modeling approach develops time-series multi-layer perceptrons (MLPs) for the operating units and integrates them with prior process knowledge about system structure and fundamental dynamics. This integration forms three hybrid NNs (long-term, slow, and fast MLPs) that predict the entire system dynamics across multiple time scales. Leveraging these MLPs, we design an NN-based scheduler and an NN-based economic model predictive control (NEMPC) framework to meet global operational requirements: rapid electrical power responsiveness to operators requests, adequate cooling supply to customers, and increased system profitability, while addressing the dynamic time-scale multiplicity present in IESs. The proposed day-ahead scheduler is formulated using the ReLU network-based MLP, which effectively represents IES performance under a broad range of conditions from a long-term perspective. The scheduler is then exactly recast into a mixed-integer linear programming problem for efficient evaluation. The real-time NEMPC, based on slow and fast MLPs, comprises two sequential distributed control agents: a slow NEMPC for the cooling-dominant subsystem with slower transient responses and a fast NEMPC for the power-dominant subsystem with faster responses. Extensive simulations demonstrate that the developed scheduler and NEMPC schemes outperform their respective benchmark scheduler and controller by about 25% and 40%. Together, they enhance overall system performance by over 70% compared to benchmark approaches.
Authors: Riyaz Ahuja, Jeremy Avigad, Prasad Tetali, Sean Welleck
Abstract: Large language models (LLMs) have been used to generate formal proofs of mathematical theorems in proofs assistants such as Lean. However, we often want to optimize a formal proof with respect to various criteria, depending on its downstream use. For example, we may want a proof to adhere to a certain style, or to be readable, concise, or modularly structured. Having suitably optimized proofs is also important for learning tasks, especially since human-written proofs may not optimal for that purpose. To this end, we study a new problem of automated proof optimization: rewriting a proof so that it is correct and optimizes for an arbitrary criterion, such as length or readability. As a first method for automated proof optimization, we present ImProver, a large-language-model agent that rewrites proofs to optimize arbitrary user-defined metrics in Lean. We find that naively applying LLMs to proof optimization falls short, and we incorporate various improvements into ImProver, such as the use of symbolic Lean context in a novel Chain-of-States technique, as well as error-correction and retrieval. We test ImProver on rewriting real-world undergraduate, competition, and research-level mathematics theorems, finding that ImProver is capable of rewriting proofs so that they are substantially shorter, more modular, and more readable.
Authors: Wooseong Yang, Chen Wang, Zihe Song, Weizhi Zhang, Philip S. Yu
Abstract: Session-based recommendation (SBR) aims to capture dynamic user preferences by analyzing item sequences within individual sessions. However, most existing approaches focus mainly on intra-session item relationships, neglecting the connections between items across different sessions (inter-session relationships), which limits their ability to fully capture complex item interactions. While some methods incorporate inter-session information, they often suffer from high computational costs, leading to longer training times and reduced efficiency. To address these challenges, we propose the CLIP-SBR (Cluster-aware Item Prompt learning for Session-Based Recommendation) framework. CLIP-SBR is composed of two modules: 1) an item relationship mining module that builds a global graph to effectively model both intra- and inter-session relationships, and 2) an item cluster-aware prompt learning module that uses soft prompts to integrate these relationships into SBR models efficiently. We evaluate CLIP-SBR across eight SBR models and three benchmark datasets, consistently demonstrating improved recommendation performance and establishing CLIP-SBR as a robust solution for session-based recommendation tasks.
Authors: Yuchen Wu, Yuxin Chen, Yuting Wei
Abstract: Diffusion models play a pivotal role in contemporary generative modeling, claiming state-of-the-art performance across various domains. Despite their superior sample quality, mainstream diffusion-based stochastic samplers like DDPM often require a large number of score function evaluations, incurring considerably higher computational cost compared to single-step generators like generative adversarial networks. While several acceleration methods have been proposed in practice, the theoretical foundations for accelerating diffusion models remain underexplored. In this paper, we propose and analyze a training-free acceleration algorithm for SDE-style diffusion samplers, based on the stochastic Runge-Kutta method. The proposed sampler provably attains $\varepsilon^2$ error -- measured in KL divergence -- using $\widetilde O(d^{3/2} / \varepsilon)$ score function evaluations (for sufficiently small $\varepsilon$), strengthening the state-of-the-art guarantees $\widetilde O(d^{3} / \varepsilon)$ in terms of dimensional dependency. Numerical experiments validate the efficiency of the proposed method.
Authors: Cong Shen, Yipeng Zhang, Fei Han, Kelin Xia
Abstract: Accurate and efficient prediction of polymer properties is of key importance for polymer design. Traditional experimental tools and density function theory (DFT)-based simulations for polymer property evaluation, are both expensive and time-consuming. Recently, a gigantic amount of graph-based molecular models have emerged and demonstrated huge potential in molecular data analysis. Even with the great progresses, these models tend to ignore the high-order and mutliscale information within the data. In this paper, we develop molecular topological deep learning (Mol-TDL) for polymer property analysis. Our Mol-TDL incorporates both high-order interactions and multiscale properties into topological deep learning architecture. The key idea is to represent polymer molecules as a series of simplicial complices at different scales and build up simplical neural networks accordingly. The aggregated information from different scales provides a more accurate prediction of polymer molecular properties.
Authors: Sarah H. Cen, Rohan Alur
Abstract: Artificial intelligence (AI) is increasingly intervening in our lives, raising widespread concern about its unintended and undeclared side effects. These developments have brought attention to the problem of AI auditing: the systematic evaluation and analysis of an AI system, its development, and its behavior relative to a set of predetermined criteria. Auditing can take many forms, including pre-deployment risk assessments, ongoing monitoring, and compliance testing. It plays a critical role in providing assurances to various AI stakeholders, from developers to end users. Audits may, for instance, be used to verify that an algorithm complies with the law, is consistent with industry standards, and meets the developer's claimed specifications. However, there are many operational challenges to AI auditing that complicate its implementation. In this work, we examine a key operational issue in AI auditing: what type of access to an AI system is needed to perform a meaningful audit? Addressing this question has direct policy relevance, as it can inform AI audit guidelines and requirements. We begin by discussing the factors that auditors balance when determining the appropriate type of access, and unpack the benefits and drawbacks of four types of access. We conclude that, at minimum, black-box access -- providing query access to a model without exposing its internal implementation -- should be granted to auditors, as it balances concerns related to trade secrets, data privacy, audit standardization, and audit efficiency. We then suggest a framework for determining how much further access (in addition to black-box access) to grant auditors. We show that auditing can be cast as a natural hypothesis test, draw parallels hypothesis testing and legal procedure, and argue that this framing provides clear and interpretable guidance on audit implementation.
Authors: Alessandro Montanari, Ashok Thangarajan, Khaldoon Al-Naimi, Andrea Ferlini, Yang Liu, Ananta Narayanan Balaji, Fahim Kawsar
Abstract: Sensory earables have evolved from basic audio enhancement devices into sophisticated platforms for clinical-grade health monitoring and wellbeing management. This paper introduces OmniBuds, an advanced sensory earable platform integrating multiple biosensors and onboard computation powered by a machine learning accelerator, all within a real-time operating system (RTOS). The platform's dual-ear symmetric design, equipped with precisely positioned kinetic, acoustic, optical, and thermal sensors, enables highly accurate and real-time physiological assessments. Unlike conventional earables that rely on external data processing, OmniBuds leverage real-time onboard computation to significantly enhance system efficiency, reduce latency, and safeguard privacy by processing data locally. This capability includes executing complex machine learning models directly on the device. We provide a comprehensive analysis of OmniBuds' design, hardware and software architecture demonstrating its capacity for multi-functional applications, accurate and robust tracking of physiological parameters, and advanced human-computer interaction.
Authors: Kazumoto Nakamura, Yuji Nozawa, Yu-Chieh Lin, Kengo Nakata, Youyang Ng
Abstract: The goal of this paper is to improve the performance of pretrained Vision Transformer (ViT) models, particularly DINOv2, in image clustering task without requiring re-training or fine-tuning. As model size increases, high-norm artifacts anomaly appears in the patches of multi-head attention. We observe that this anomaly leads to reduced accuracy in zero-shot image clustering. These artifacts are characterized by disproportionately large values in the attention map compared to other patch tokens. To address these artifacts, we propose an approach called Inference-Time Attention Engineering (ITAE), which manipulates attention function during inference. Specifically, we identify the artifacts by investigating one of the Query-Key-Value (QKV) patches in the multi-head attention and attenuate their corresponding attention values inside the pretrained models. ITAE shows improved clustering accuracy on multiple datasets by exhibiting more expressive features in latent space. Our findings highlight the potential of ITAE as a practical solution for reducing artifacts in pretrained ViT models and improving model performance in clustering tasks without the need for re-training or fine-tuning.
Authors: Matteo Risso, Alessia Goffi, Beatrice Alessandra Motetti, Alessio Burrello, Jean Baptiste Bove, Enrico Macii, Massimo Poncino, Daniele Jahier Pagliari, Giuseppe Maffeis
Abstract: Very High Resolution (VHR) geospatial image analysis is crucial for humanitarian assistance in both natural and anthropogenic crises, as it allows to rapidly identify the most critical areas that need support. Nonetheless, manually inspecting large areas is time-consuming and requires domain expertise. Thanks to their accuracy, generalization capabilities, and highly parallelizable workload, Deep Neural Networks (DNNs) provide an excellent way to automate this task. Nevertheless, there is a scarcity of VHR data pertaining to conflict situations, and consequently, of studies on the effectiveness of DNNs in those scenarios. Motivated by this, our work extensively studies the applicability of a collection of state-of-the-art Convolutional Neural Networks (CNNs) originally developed for natural disasters damage assessment in a war scenario. To this end, we build an annotated dataset with pre- and post-conflict images of the Ukrainian city of Mariupol. We then explore the transferability of the CNN models in both zero-shot and learning scenarios, demonstrating their potential and limitations. To the best of our knowledge, this is the first study to use sub-meter resolution imagery to assess building damage in combat zones.
Authors: Anna Varbella, Damien Briens, Blazhe Gjorgiev, Giuseppe Alessio D'Inverno, Giovanni Sansavini
Abstract: The energy transition is driving the integration of large shares of intermittent power sources in the electric power grid. Therefore, addressing the AC optimal power flow (AC-OPF) effectively becomes increasingly essential. The AC-OPF, which is a fundamental optimization problem in power systems, must be solved more frequently to ensure the safe and cost-effective operation of power systems. Due to its non-linear nature, AC-OPF is often solved in its linearized form, despite inherent inaccuracies. Non-linear solvers, such as the interior point method, are typically employed to solve the full OPF problem. However, these iterative methods may not converge for large systems and do not guarantee global optimality. This work explores a physics-informed graph neural network, PINCO, to solve the AC-OPF. We demonstrate that this method provides accurate solutions in a fraction of the computational time when compared to the established non-linear programming solvers. Remarkably, PINCO generalizes effectively across a diverse set of loading conditions in the power system. We show that our method can solve the AC-OPF without violating inequality constraints. Furthermore, it can function both as a solver and as a hybrid universal function approximator. Moreover, the approach can be easily adapted to different power systems with minimal adjustments to the hyperparameters, including systems with multiple generators at each bus. Overall, this work demonstrates an advancement in the field of power system optimization to tackle the challenges of the energy transition. The code and data utilized in this paper are available at https://anonymous.4open.science/r/opf_pinn_iclr-B83E/.
URLs: https://anonymous.4open.science/r/opf_pinn_iclr-B83E/.
Authors: Lucia Gordon, Nico Lang, Catherine Ressijac, Andrew Davies
Abstract: Multimodal aerial data are used to monitor natural systems, and machine learning can significantly accelerate the classification of landscape features within such imagery to benefit ecology and conservation. It remains under-explored, however, how these multiple modalities ought to be fused in a deep learning model. As a step towards filling this gap, we study three strategies (Early fusion, Late fusion, and Mixture of Experts) for fusing thermal, RGB, and LiDAR imagery using a dataset of spatially-aligned orthomosaics in these three modalities. In particular, we aim to map three ecologically-relevant biophysical landscape features in African savanna ecosystems: rhino middens, termite mounds, and water. The three fusion strategies differ in whether the modalities are fused early or late, and if late, whether the model learns fixed weights per modality for each class or generates weights for each class adaptively, based on the input. Overall, the three methods have similar macro-averaged performance with Late fusion achieving an AUC of 0.698, but their per-class performance varies strongly, with Early fusion achieving the best recall for middens and water and Mixture of Experts achieving the best recall for mounds.
Authors: Paul Jansonnie, Bingbing Wu, Julien Perez, Jan Peters
Abstract: Learning skills that interact with objects is of major importance for robotic manipulation. These skills can indeed serve as an efficient prior for solving various manipulation tasks. We propose a novel Skill Learning approach that discovers composable behaviors by solving a large and diverse number of autonomously generated tasks. Our method learns skills allowing the robot to consistently and robustly interact with objects in its environment. The discovered behaviors are embedded in primitives which can be composed with Hierarchical Reinforcement Learning to solve unseen manipulation tasks. In particular, we leverage Asymmetric Self-Play to discover behaviors and Multiplicative Compositional Policies to embed them. We compare our method to Skill Learning baselines and find that our skills are more interactive. Furthermore, the learned skills can be used to solve a set of unseen manipulation tasks, in simulation as well as on a real robotic platform.
Authors: Marie-Charlotte Brandenburg, Moritz Grillo, Christoph Hertrich
Abstract: In this paper we contribute to the frequently studied question of how to decompose a continuous piecewise linear (CPWL) function into a difference of two convex CPWL functions. Every CPWL function has infinitely many such decompositions, but for applications in optimization and neural network theory, it is crucial to find decompositions with as few linear pieces as possible. This is a highly challenging problem, as we further demonstrate by disproving a recently proposed approach by Tran and Wang [Minimal representations of tropical rational functions. Algebraic Statistics, 15(1):27-59, 2024]. To make the problem more tractable, we propose to fix an underlying polyhedral complex determining the possible locus of nonlinearity. Under this assumption, we prove that the set of decompositions forms a polyhedron that arises as intersection of two translated cones. We prove that irreducible decompositions correspond to the bounded faces of this polyhedron and minimal solutions must be vertices. We then identify cases with a unique minimal decomposition, and illustrate how our insights have consequences in the theory of submodular functions. Finally, we improve upon previous constructions of neural networks for a given convex CPWL function and apply our framework to obtain results in the nonconvex case.
Authors: Mitsuki Morita, Satoshi Yamamori, Satoshi Yagi, Norikazu Sugimoto, Jun Morimoto
Abstract: While MPC enables nonlinear feedback control by solving an optimal control problem at each timestep, the computational burden tends to be significantly large, making it difficult to optimize a policy within the control period. To address this issue, one possible approach is to utilize terminal value learning to reduce computational costs. However, the learned value cannot be used for other tasks in situations where the task dynamically changes in the original MPC setup. In this study, we develop an MPC framework with goal-conditioned terminal value learning to achieve multitask policy optimization while reducing computational time. Furthermore, by using a hierarchical control structure that allows the upper-level trajectory planner to output appropriate goal-conditioned trajectories, we demonstrate that a robot model is able to generate diverse motions. We evaluate the proposed method on a bipedal inverted pendulum robot model and confirm that combining goal-conditioned terminal value learning with an upper-level trajectory planner enables real-time control; thus, the robot successfully tracks a target trajectory on sloped terrain.
Authors: Jianan Zhou, Yaoxin Wu, Zhiguang Cao, Wen Song, Jie Zhang, Zhiqi Shen
Abstract: Despite enjoying desirable efficiency and reduced reliance on domain expertise, existing neural methods for vehicle routing problems (VRPs) suffer from severe robustness issues -- their performance significantly deteriorates on clean instances with crafted perturbations. To enhance robustness, we propose an ensemble-based Collaborative Neural Framework (CNF) w.r.t. the defense of neural VRP methods, which is crucial yet underexplored in the literature. Given a neural VRP method, we adversarially train multiple models in a collaborative manner to synergistically promote robustness against attacks, while boosting standard generalization on clean instances. A neural router is designed to adeptly distribute training instances among models, enhancing overall load balancing and collaborative efficacy. Extensive experiments verify the effectiveness and versatility of CNF in defending against various attacks across different neural VRP methods. Notably, our approach also achieves impressive out-of-distribution generalization on benchmark instances.
Authors: Sebastian Hirt, Andreas H\"ohl, Johannes Pohlodek, Joachim Schaeffer, Maik Pfefferkorn, Richard D. Braatz, Rolf Findeisen
Abstract: Model predictive control (MPC) is a powerful tool for controlling complex nonlinear systems under constraints, but often struggles with model uncertainties and the design of suitable cost functions. To address these challenges, we discuss an approach that integrates MPC with safe Bayesian optimization to optimize long-term closed-loop performance despite significant model-plant mismatches. By parameterizing the MPC stage cost function using a radial basis function network, we employ Bayesian optimization as a multi-episode learning strategy to tune the controller without relying on precise system models. This method mitigates conservativeness introduced by overly cautious soft constraints in the MPC cost function and provides probabilistic safety guarantees during learning, ensuring that safety-critical constraints are met with high probability. As a practical application, we apply our approach to fast charging of lithium-ion batteries, a challenging task due to the complicated battery dynamics and strict safety requirements, subject to the requirement to be implementable in real time. Simulation results demonstrate that, in the context of model-plant mismatch, our method reduces charging times compared to traditional MPC methods while maintaining safety. This work extends previous research by emphasizing closed-loop constraint satisfaction and offers a promising solution for enhancing performance in systems where model uncertainties and safety are critical concerns.
Authors: Ajay B. S., Phani Pavan K, Madhav Rao
Abstract: Long short-term memory (LSTM) has emerged as a definitive network for analyzing and inferring time series data. LSTM has the capability to extract spectral features and a mixture of temporal features. Due to this benefit, a similar feature extraction method is explored for the spiking counterparts targeting time-series data. Though LSTMs perform well in their spiking form, they tend to be compute and power intensive. Addressing this issue, this work proposes Multi-Compartment Leaky (MCLeaky) neuron as a viable alternative for efficient processing of time series data. The MCLeaky neuron, derived from the Leaky Integrate and Fire (LIF) neuron model, contains multiple memristive synapses interlinked to form a memory component, which emulates the human brain's Hippocampus region. The proposed MCLeaky neuron based Spiking Neural Network model and its quantized variant were benchmarked against state-of-the-art (SOTA) Spiking LSTMs to perform human stress detection, by comparing compute requirements, latency and real-world performances on unseen data with models derived through Neural Architecture Search (NAS). Results show that networks with MCLeaky activation neuron managed a superior accuracy of 98.8% to detect stress based on Electrodermal Activity (EDA) signals, better than any other investigated models, while using 20% less parameters on average. MCLeaky neuron was also tested for various signals including EDA Wrist and Chest, Temperature, ECG, and combinations of them. Quantized MCLeaky model was also derived and validated to forecast their performance on hardware architectures, which resulted in 91.84% accuracy. The neurons were evaluated for multiple modalities of data towards stress detection, which resulted in energy savings of 25.12x to 39.20x and EDP gains of 52.37x to 81.9x over ANNs, while offering a best accuracy of 98.8% when compared with the rest of the SOTA implementations.
Authors: Jin-Hong Du, Kathryn Roeder, Larry Wasserman
Abstract: Data integration has become increasingly common in aligning multiple heterogeneous datasets. With high-dimensional outcomes, data integration methods aim to extract low-dimensional embeddings of observations to remove unwanted variations, such as batch effects and unmeasured covariates, inherent in data collected from different sources. However, multiple hypothesis testing after data integration can be substantially biased due to the data-dependent integration processes. To address this challenge, we introduce a robust post-integrated inference (PII) method that adjusts for latent heterogeneity using negative control outcomes. By leveraging causal interpretations, we derive nonparametric identification conditions that form the basis of our PII approach. Our assumption-lean semiparametric inference method extends robustness and generality to projected direct effect estimands that account for mediators, confounders, and moderators. These estimands remain statistically meaningful under model misspecifications and with error-prone embeddings. We provide deterministic quantifications of the bias of target estimands induced by estimated embeddings and finite-sample linear expansions of the estimators with uniform concentration bounds on the residuals for all outcomes. The proposed doubly robust estimators are consistent and efficient under minimal assumptions, facilitating data-adaptive estimation with machine learning algorithms. Using random forests, we evaluate empirical statistical errors in simulations and analyze single-cell CRISPR perturbed datasets with potential unmeasured confounders.
Authors: Ibrahim Aldarmaki, Thamar Solorio, Bhiksha Raj, Hanan Aldarmaki
Abstract: Neural multi-channel speech enhancement models, in particular those based on the U-Net architecture, demonstrate promising performance and generalization potential. These models typically encode input channels independently, and integrate the channels during later stages of the network. In this paper, we propose a novel modification of these models by incorporating relative information from the outset, where each channel is processed in conjunction with a reference channel through stacking. This input strategy exploits comparative differences to adaptively fuse information between channels, thereby capturing crucial spatial information and enhancing the overall performance. The experiments conducted on the CHiME-3 dataset demonstrate improvements in speech enhancement metrics across various architectures.
Authors: Stacey D. Scott, Zayn J. Abbas, Feerass Ellid, Eli-Henry Dykhne, Muhammad Muhaiminul Islam, Weam Ayad, Kristina Kacmorova, Dan Tulpan, Minglun Gong
Abstract: Precision livestock farming (PLF) aims to improve the health and welfare of livestock animals and farming outcomes through the use of advanced technologies. Computer vision, combined with recent advances in machine learning and deep learning artificial intelligence approaches, offers a possible solution to the PLF ideal of 24/7 livestock monitoring that helps facilitate early detection of animal health and welfare issues. However, a significant number of livestock species are raised in large outdoor habitats that pose technological challenges for computer vision approaches. This review provides a comprehensive overview of computer vision methods and open challenges in outdoor animal monitoring. We include research from both the livestock and wildlife fields in the review because of the similarities in appearance, behaviour, and habitat for many livestock and wildlife. We focus on large terrestrial mammals, such as cattle, horses, deer, goats, sheep, koalas, giraffes, and elephants. We use an image processing pipeline to frame our discussion and highlight the current capabilities and open technical challenges at each stage of the pipeline. The review found a clear trend towards the use of deep learning approaches for animal detection, counting, and multi-species classification. We discuss in detail the applicability of current vision-based methods to PLF contexts and promising directions for future research.
Authors: Ziwen Yuan, Tianyi Zhang, Matthew Johnson-Roberson, Weiming Zhi
Abstract: Building accurate representations of the environment is critical for intelligent robots to make decisions during deployment. Advances in photorealistic environment models have enabled robots to develop hyper-realistic reconstructions, which can be used to generate images that are intuitive for human inspection. In particular, the recently introduced \ac{3DGS}, which describes the scene with up to millions of primitive ellipsoids, can be rendered in real time. \ac{3DGS} has rapidly gained prominence. However, a critical unsolved problem persists: how can we fuse multiple \ac{3DGS} into a single coherent model? Solving this problem will enable robot teams to jointly build \ac{3DGS} models of their surroundings. A key insight of this work is to leverage the {duality} between photorealistic reconstructions, which render realistic 2D images from 3D structure, and \emph{3D foundation models}, which predict 3D structure from image pairs. To this end, we develop PhotoReg, a framework to register multiple photorealistic \ac{3DGS} models with 3D foundation models. As \ac{3DGS} models are generally built from monocular camera images, they have \emph{arbitrary scale}. To resolve this, PhotoReg actively enforces scale consistency among the different \ac{3DGS} models by considering depth estimates within these models. Then, the alignment is iteratively refined with fine-grained photometric losses to produce high-quality fused \ac{3DGS} models. We rigorously evaluate PhotoReg on both standard benchmark datasets and our custom-collected datasets, including with two quadruped robots. The code is released at \url{ziweny11.github.io/photoreg}.
Authors: Benjamin Feuer, Jiawei Xu, Niv Cohen, Patrick Yubeaton, Govind Mittal, Chinmay Hegde
Abstract: Data curation is the problem of how to collect and organize samples into a dataset that supports efficient learning. Despite the centrality of the task, little work has been devoted towards a large-scale, systematic comparison of various curation methods. In this work, we take steps towards a formal evaluation of data curation strategies and introduce SELECT, the first large-scale benchmark of curation strategies for image classification. In order to generate baseline methods for the SELECT benchmark, we create a new dataset, ImageNet++, which constitutes the largest superset of ImageNet-1K to date. Our dataset extends ImageNet with 5 new training-data shifts, each approximately the size of ImageNet-1K itself, and each assembled using a distinct curation strategy. We evaluate our data curation baselines in two ways: (i) using each training-data shift to train identical image classification models from scratch (ii) using the data itself to fit a pretrained self-supervised representation. Our findings show interesting trends, particularly pertaining to recent methods for data curation such as synthetic data generation and lookup based on CLIP embeddings. We show that although these strategies are highly competitive for certain tasks, the curation strategy used to assemble the original ImageNet-1K dataset remains the gold standard. We anticipate that our benchmark can illuminate the path for new methods to further reduce the gap. We release our checkpoints, code, documentation, and a link to our dataset at https://github.com/jimmyxu123/SELECT.
Authors: Ziru Chen, Shijie Chen, Yuting Ning, Qianheng Zhang, Boshi Wang, Botao Yu, Yifei Li, Zeyi Liao, Chen Wei, Zitong Lu, Vishal Dey, Mingyi Xue, Frazier N. Baker, Benjamin Burns, Daniel Adu-Ampratwum, Xuhui Huang, Xia Ning, Song Gao, Yu Su, Huan Sun
Abstract: The advancements of language language models (LLMs) have piqued growing interest in developing LLM-based language agents to automate scientific discovery end-to-end, which has sparked both excitement and skepticism about the true capabilities of such agents. In this work, we argue that for an agent to fully automate scientific discovery, it must be able to complete all essential tasks in the workflow. Thus, we call for rigorous assessment of agents on individual tasks in a scientific workflow before making bold claims on end-to-end automation. To this end, we present ScienceAgentBench, a new benchmark for evaluating language agents for data-driven scientific discovery. To ensure the scientific authenticity and real-world relevance of our benchmark, we extract 102 tasks from 44 peer-reviewed publications in four disciplines and engage nine subject matter experts to validate them. We unify the target output for every task to a self-contained Python program file and employ an array of evaluation metrics to examine the generated programs, execution results, and costs. Each task goes through multiple rounds of manual validation by annotators and subject matter experts to ensure its annotation quality and scientific plausibility. We also propose two effective strategies to mitigate data contamination concerns. Using our benchmark, we evaluate five open-weight and proprietary LLMs, each with three frameworks: direct prompting, OpenHands, and self-debug. Given three attempts for each task, the best-performing agent can only solve 32.4% of the tasks independently and 34.3% with expert-provided knowledge. These results underscore the limited capacities of current language agents in generating code for data-driven discovery, let alone end-to-end automation for scientific research.
Authors: Nidhi Mathihalli, Audrey Wei, Giovanni Lavezzi, Peng Mun Siew, Victor Rodriguez-Fernandez, Hodei Urrutxua, Richard Linares
Abstract: Novel view synthesis (NVS) enables to generate new images of a scene or convert a set of 2D images into a comprehensive 3D model. In the context of Space Domain Awareness, since space is becoming increasingly congested, NVS can accurately map space objects and debris, improving the safety and efficiency of space operations. Similarly, in Rendezvous and Proximity Operations missions, 3D models can provide details about a target object's shape, size, and orientation, allowing for better planning and prediction of the target's behavior. In this work, we explore the generalization abilities of these reconstruction techniques, aiming to avoid the necessity of retraining for each new scene, by presenting a novel approach to 3D spacecraft reconstruction from single-view images, DreamSat, by fine-tuning the Zero123 XL, a state-of-the-art single-view reconstruction model, on a high-quality dataset of 190 high-quality spacecraft models and integrating it into the DreamGaussian framework. We demonstrate consistent improvements in reconstruction quality across multiple metrics, including Contrastive Language-Image Pretraining (CLIP) score (+0.33%), Peak Signal-to-Noise Ratio (PSNR) (+2.53%), Structural Similarity Index (SSIM) (+2.38%), and Learned Perceptual Image Patch Similarity (LPIPS) (+0.16%) on a test set of 30 previously unseen spacecraft images. Our method addresses the lack of domain-specific 3D reconstruction tools in the space industry by leveraging state-of-the-art diffusion models and 3D Gaussian splatting techniques. This approach maintains the efficiency of the DreamGaussian framework while enhancing the accuracy and detail of spacecraft reconstructions. The code for this work can be accessed on GitHub (https://github.com/ARCLab-MIT/space-nvs).
Authors: Zengwei Yao, Wei Kang, Xiaoyu Yang, Fangjun Kuang, Liyong Guo, Han Zhu, Zengrui Jin, Zhaoqing Li, Long Lin, Daniel Povey
Abstract: Connectionist Temporal Classification (CTC) is a widely used method for automatic speech recognition (ASR), renowned for its simplicity and computational efficiency. However, it often falls short in recognition performance compared to transducer or systems combining CTC and attention-based encoder-decoder (CTC/AED). In this work, we propose the Consistency-Regularized CTC (CR-CTC), which enforces consistency between two CTC distributions obtained from different augmented views of the input speech mel-spectrogram. We provide in-depth insights into its essential behaviors from three perspectives: 1) it conducts self-distillation between random pairs of sub-models that process different augmented views; 2) it learns contextual representation through masked prediction for positions within time-masked regions, especially when we increase the amount of time masking; 3) it suppresses the extremely peaky CTC distributions, thereby reducing overfitting and improving the generalization ability. Extensive experiments on LibriSpeech, Aishell-1, and GigaSpeech datasets demonstrate the effectiveness of our CR-CTC, which achieves performance comparable to, or even slightly better than, that of transducer and CTC/AED.
Authors: Fenia Christopoulou, Ronald Cardenas, Gerasimos Lampouras, Haitham Bou-Ammar, Jun Wang
Abstract: Preference Optimization (PO) has proven an effective step for aligning language models to human-desired behaviors. Current variants, following the offline Direct Preference Optimization objective, have focused on a strict setting where all tokens are contributing signals of KL divergence and rewards to the loss function. However, human preference is not affected by each word in a sequence equally but is often dependent on specific words or phrases, e.g. existence of toxic terms leads to non-preferred responses. Based on this observation, we argue that not all tokens should be weighted equally during PO and propose a flexible objective termed SparsePO, that aims to automatically learn to weight the KL divergence and reward corresponding to each token during PO training. We propose two different variants of weight-masks that can either be derived from the reference model itself or learned on the fly. Notably, our method induces sparsity in the learned masks, allowing the model to learn how to best weight reward and KL divergence contributions at the token level, learning an optimal level of mask sparsity. Extensive experiments on multiple domains, including sentiment control, dialogue, text summarization and text-to-code generation, illustrate that our approach assigns meaningful weights to tokens according to the target task, generates more responses with the desired preference and improves reasoning tasks by up to 2 percentage points compared to other token- and response-level PO methods.
Authors: Marina Sheshukova, Denis Belomestny, Alain Durmus, Eric Moulines, Alexey Naumov, Sergey Samsonov
Abstract: We address the problem of solving strongly convex and smooth minimization problems using stochastic gradient descent (SGD) algorithm with a constant step size. Previous works suggested to combine the Polyak-Ruppert averaging procedure with the Richardson-Romberg extrapolation technique to reduce the asymptotic bias of SGD at the expense of a mild increase of the variance. We significantly extend previous results by providing an expansion of the mean-squared error of the resulting estimator with respect to the number of iterations $n$. More precisely, we show that the mean-squared error can be decomposed into the sum of two terms: a leading one of order $\mathcal{O}(n^{-1/2})$ with explicit dependence on a minimax-optimal asymptotic covariance matrix, and a second-order term of order $\mathcal{O}(n^{-3/4})$ where the power $3/4$ can not be improved in general. We also extend this result to the $p$-th moment bound keeping optimal scaling of the remainders with respect to $n$. Our analysis relies on the properties of the SGD iterates viewed as a time-homogeneous Markov chain. In particular, we establish that this chain is geometrically ergodic with respect to a suitably defined weighted Wasserstein semimetric.
Authors: Mo\"ise Blanchard
Abstract: Classical results in statistical learning typically consider two extreme data-generating models: i.i.d. instances from an unknown distribution, or fully adversarial instances, often much more challenging statistically. To bridge the gap between these models, recent work introduced the smoothed framework, in which at each iteration an adversary generates instances from a distribution constrained to have density bounded by $\sigma^{-1}$ compared to some fixed base measure $\mu$. This framework interpolates between the i.i.d. and adversarial cases, depending on the value of $\sigma$. For the classical online prediction problem, most prior results in smoothed online learning rely on the arguably strong assumption that the base measure $\mu$ is known to the learner, contrasting with standard settings in the PAC learning or consistency literature. We consider the general agnostic problem in which the base measure is unknown and values are arbitrary. Along this direction, Block et al. showed that empirical risk minimization has sublinear regret under the well-specified assumption. We propose an algorithm R-Cover based on recursive coverings which is the first to guarantee sublinear regret for agnostic smoothed online learning without prior knowledge of $\mu$. For classification, we prove that R-Cover has adaptive regret $\tilde O(\sqrt{dT/\sigma})$ for function classes with VC dimension $d$, which is optimal up to logarithmic factors. For regression, we establish that R-Cover has sublinear oblivious regret for function classes with polynomial fat-shattering dimension growth.
Authors: Wesley Brewer, Matthias Maiterth, Vineet Kumar, Rafal Wojda, Sedrick Bouknight, Jesse Hines, Woong Shin, Scott Greenwood, David Grant, Wesley Williams, Feiyi Wang
Abstract: We present ExaDigiT, an open-source framework for developing comprehensive digital twins of liquid-cooled supercomputers. It integrates three main modules: (1) a resource allocator and power simulator, (2) a transient thermo-fluidic cooling model, and (3) an augmented reality model of the supercomputer and central energy plant. The framework enables the study of "what-if" scenarios, system optimizations, and virtual prototyping of future systems. Using Frontier as a case study, we demonstrate the framework's capabilities by replaying six months of system telemetry for systematic verification and validation. Such a comprehensive analysis of a liquid-cooled exascale supercomputer is the first of its kind. ExaDigiT elucidates complex transient cooling system dynamics, runs synthetic or real workloads, and predicts energy losses due to rectification and voltage conversion. Throughout our paper, we present lessons learned to benefit HPC practitioners developing similar digital twins. We envision the digital twin will be a key enabler for sustainable, energy-efficient supercomputing.
Authors: Jihoon Yun, Chengzhang Li, Anish Arora
Abstract: Achieving low duty cycle operation in low-power wireless networks in urban environments is complicated by the complex and variable dynamics of external interference and fading. We explore the use of reinforcement learning for achieving low power consumption for the task of optimal selection of channels. The learning relies on a hybrid of passive channel sampling for dealing with external interference and active channel sampling for dealing with fading. Our solution, Passive-Active Multi-armed bandit for LoRa (PAMLR, pronounced "Pamela"), balances the two types of samples to achieve energy-efficient channel selection: active channel measurements are tuned to an appropriately low level to update noise thresholds, and to compensate passive channel measurements are tuned to an appropriately high level for selecting the top-most channels from channel exploration using the noise thresholds. The rates of both types of samples are adapted in response to channel dynamics. Based on extensive testing in multiple environments in different cities, we validate that PAMLR can maintain excellent communication quality, as demonstrated by a low SNR regret compared to the optimal channel allocation policy, while substantially minimizing the energy cost associated with channel measurements.
Authors: Zachary Novack, Ge Zhu, Jonah Casebeer, Julian McAuley, Taylor Berg-Kirkpatrick, Nicholas J. Bryan
Abstract: Despite advances in diffusion-based text-to-music (TTM) methods, efficient, high-quality generation remains a challenge. We introduce Presto!, an approach to inference acceleration for score-based diffusion transformers via reducing both sampling steps and cost per step. To reduce steps, we develop a new score-based distribution matching distillation (DMD) method for the EDM-family of diffusion models, the first GAN-based distillation method for TTM. To reduce the cost per step, we develop a simple, but powerful improvement to a recent layer distillation method that improves learning via better preserving hidden state variance. Finally, we combine our step and layer distillation methods together for a dual-faceted approach. We evaluate our step and layer distillation methods independently and show each yield best-in-class performance. Our combined distillation method can generate high-quality outputs with improved diversity, accelerating our base model by 10-18x (230/435ms latency for 32 second mono/stereo 44.1kHz, 15x faster than comparable SOTA) -- the fastest high-quality TTM to our knowledge. Sound examples can be found at https://presto-music.github.io/web/.
Authors: Sherly Alfonso-S\'anchez, Kristina P. Sendova, Cristi\'an Bravo
Abstract: When making treatment selection decisions, it is essential to include a causal effect estimation analysis to compare potential outcomes under different treatments or controls, assisting in optimal selection. However, merely estimating individual treatment effects may not suffice for truly optimal decisions. Our study addressed this issue by incorporating additional criteria, such as the estimations' uncertainty, measured by the conditional value-at-risk, commonly used in portfolio and insurance management. For continuous outcomes observable before and after treatment, we incorporated a specific prediction condition. We prioritized treatments that could yield optimal treatment effect results and lead to post-treatment outcomes more desirable than pretreatment levels, with the latter condition being called the prediction criterion. With these considerations, we propose a comprehensive methodology for multitreatment selection. Our approach ensures satisfaction of the overlap assumption, crucial for comparing outcomes for treated and control groups, by training propensity score models as a preliminary step before employing traditional causal models. To illustrate a practical application of our methodology, we applied it to the credit card limit adjustment problem. Analyzing a fintech company's historical data, we found that relying solely on counterfactual predictions was inadequate for appropriate credit line modifications. Incorporating our proposed additional criteria significantly enhanced policy performance.
Authors: Timothy Chase Jr, Karthik Dantu
Abstract: The visual detection and tracking of surface terrain is required for spacecraft to safely land on or navigate within close proximity to celestial objects. Current approaches rely on template matching with pre-gathered patch-based features, which are expensive to obtain and a limiting factor in perceptual capability. While recent literature has focused on in-situ detection methods to enhance navigation and operational autonomy, robust description is still needed. In this work, we explore metric learning as the lightweight feature description mechanism and find that current solutions fail to address inter-class similarity and multi-view observational geometry. We attribute this to the view-unaware attention mechanism and introduce Multi-view Attention Regularizations (MARs) to constrain the channel and spatial attention across multiple feature views, regularizing the what and where of attention focus. We thoroughly analyze many modern metric learning losses with and without MARs and demonstrate improved terrain-feature recognition performance by upwards of 85%. We additionally introduce the Luna-1 dataset, consisting of Moon crater landmarks and reference navigation frames from NASA mission data to support future research in this difficult task. Luna-1 and source code are publicly available at https://droneslab.github.io/mars/.
Authors: Yu Tian, Sadamori Kojaku, Hiroki Sayama, Renaud Lambiotte
Abstract: Networks are powerful tools for modeling interactions in complex systems. While traditional networks use scalar edge weights, many real-world systems involve multidimensional interactions. For example, in social networks, individuals often have multiple interconnected opinions that can affect different opinions of other individuals, which can be better characterized by matrices. We propose a novel, general framework for modeling such multidimensional interacting dynamics: matrix-weighted networks (MWNs). We present the mathematical foundations of MWNs and examine consensus dynamics and random walks within this context. Our results reveal that the coherence of MWNs gives rise to non-trivial steady states that generalize the notions of communities and structural balance in traditional networks.
Authors: Ge Ya (Olga), Luo, Gian Favero, Zhi Hao Luo, Alexia Jolicoeur-Martineau, Christopher Pal
Abstract: The Fr\'echet Video Distance (FVD) is a widely adopted metric for evaluating video generation distribution quality. However, its effectiveness relies on critical assumptions. Our analysis reveals three significant limitations: (1) the non-Gaussianity of the Inflated 3D Convnet (I3D) feature space; (2) the insensitivity of I3D features to temporal distortions; (3) the impractical sample sizes required for reliable estimation. These findings undermine FVD's reliability and show that FVD falls short as a standalone metric for video generation evaluation. After extensive analysis of a wide range of metrics and backbone architectures, we propose JEDi, the JEPA Embedding Distance, based on features derived from a Joint Embedding Predictive Architecture, measured using Maximum Mean Discrepancy with polynomial kernel. Our experiments on multiple open-source datasets show clear evidence that it is a superior alternative to the widely used FVD metric, requiring only 16% of the samples to reach its steady value, while increasing alignment with human evaluation by 34%, on average.
Authors: Avanika Narayan, Mayee F. Chen, Kush Bhatia, Christopher R\'e
Abstract: Fine-tuning large language models (LLMs) on instruction datasets is a common way to improve their generative capabilities. However, instruction datasets can be expensive and time-consuming to manually curate, and while LLM-generated data is less labor-intensive, it may violate user privacy agreements or terms of service of LLM providers. Therefore, we seek a way of constructing instruction datasets with samples that are not generated by humans or LLMs but still improve LLM generative capabilities. In this work, we introduce Cookbook, a framework that programmatically generates training data consisting of simple patterns over random tokens, resulting in a scalable, cost-effective approach that avoids legal and privacy issues. First, Cookbook uses a template -- a data generating Python function -- to produce training data that encourages the model to learn an explicit pattern-based rule that corresponds to a desired task. We find that fine-tuning on Cookbook-generated data is able to improve performance on its corresponding task by up to 52.7 accuracy points. Second, since instruction datasets improve performance on multiple downstream tasks simultaneously, Cookbook algorithmically learns how to mix data from various templates to optimize performance on multiple tasks. On the standard multi-task GPT4ALL evaluation suite, Mistral-7B fine-tuned using a Cookbook-generated dataset attains the best accuracy on average compared to other 7B parameter instruction-tuned models and is the best performing model on 3 out of 8 tasks. Finally, we analyze when and why Cookbook improves performance and present a metric that allows us to verify that the improvement is largely explained by the model's generations adhering better to template rules.
Authors: Yuxin Xiao, Shujian Zhang, Wenxuan Zhou, Marzyeh Ghassemi, Sanqiang Zhao
Abstract: To induce desired behaviors in large language models (LLMs) for interaction-driven tasks, the instruction-tuning stage typically trains LLMs on instruction-response pairs using the next-token prediction (NTP) loss. Previous work aiming to improve instruction-tuning performance often emphasizes the need for higher-quality supervised fine-tuning (SFT) datasets, which typically involves expensive data filtering with proprietary LLMs or labor-intensive data generation by human annotators. However, these approaches do not fully leverage the datasets' intrinsic properties, resulting in high computational and labor costs, thereby limiting scalability and performance gains. In this paper, we propose SFTMix, a novel recipe that elevates instruction-tuning performance beyond the conventional NTP paradigm, without the need for well-curated datasets. Observing that LLMs exhibit uneven confidence across the semantic representation space, we argue that examples with different confidence levels should play distinct roles during the instruction-tuning process. Based on this insight, SFTMix leverages training dynamics to identify examples with varying confidence levels, then applies a Mixup-based regularization to mitigate overfitting on confident examples while propagating supervision signals to improve learning on relatively unconfident ones. This approach enables SFTMix to significantly outperform NTP across a wide range of instruction-following and healthcare domain-specific SFT tasks, demonstrating its adaptability to diverse LLM families and scalability to datasets of any size. Comprehensive ablation studies further verify the robustness of SFTMix's design choices, underscoring its versatility in consistently enhancing performance across different LLMs and datasets in broader natural language processing applications.
Authors: Mourad Heddaya, Qingcheng Zeng, Chenhao Tan, Rob Voigt, Alexander Zentefis
Abstract: We present a novel approach to classify causal micro-narratives from text. These narratives are sentence-level explanations of the cause(s) and/or effect(s) of a target subject. The approach requires only a subject-specific ontology of causes and effects, and we demonstrate it with an application to inflation narratives. Using a human-annotated dataset spanning historical and contemporary US news articles for training, we evaluate several large language models (LLMs) on this multi-label classification task. The best-performing model--a fine-tuned Llama 3.1 8B--achieves F1 scores of 0.87 on narrative detection and 0.71 on narrative classification. Comprehensive error analysis reveals challenges arising from linguistic ambiguity and highlights how model errors often mirror human annotator disagreements. This research establishes a framework for extracting causal micro-narratives from real-world data, with wide-ranging applications to social science research.
Authors: Eilam Shapira, Omer Madmon, Itamar Reinman, Samuel Joseph Amouyal, Roi Reichart, Moshe Tennenholtz
Abstract: Large Language Models (LLMs) show significant potential in economic and strategic interactions, where communication via natural language is often prevalent. This raises key questions: Do LLMs behave rationally? Can they mimic human behavior? Do they tend to reach an efficient and fair outcome? What is the role of natural language in the strategic interaction? How do characteristics of the economic environment influence these dynamics? These questions become crucial concerning the economic and societal implications of integrating LLM-based agents into real-world data-driven systems, such as online retail platforms and recommender systems. While the ML community has been exploring the potential of LLMs in such multi-agent setups, varying assumptions, design choices and evaluation criteria across studies make it difficult to draw robust and meaningful conclusions. To address this, we introduce a benchmark for standardizing research on two-player, sequential, language-based games. Inspired by the economic literature, we define three base families of games with consistent parameterization, degrees of freedom and economic measures to evaluate agents' performance (self-gain), as well as the game outcome (efficiency and fairness). We develop an open-source framework for interaction simulation and analysis, and utilize it to collect a dataset of LLM vs. LLM interactions across numerous game configurations and an additional dataset of human vs. LLM interactions. Through extensive experimentation, we demonstrate how our framework and dataset can be used to: (i) compare the behavior of LLM-based agents to human players in various economic contexts; (ii) evaluate agents in both individual and collective performance measures; and (iii) quantify the effect of the economic characteristics of the environments on the behavior of agents.
Authors: Daoan Zhang, Guangchen Lan, Dong-Jun Han, Wenlin Yao, Xiaoman Pan, Hongming Zhang, Mingxiao Li, Pengcheng Chen, Yu Dong, Christopher Brinton, Jiebo Luo
Abstract: Reinforcement learning from human feedback (RLHF) methods are emerging as a way to fine-tune diffusion models (DMs) for visual generation. However, commonly used on-policy strategies are limited by the generalization capability of the reward model, while off-policy approaches require large amounts of difficult-to-obtain paired human-annotated data, particularly in visual generation tasks. To address the limitations of both on- and off-policy RLHF, we propose a preference optimization method that aligns DMs with preferences without relying on reward models or paired human-annotated data. Specifically, we introduce a Semi-Policy Preference Optimization (SePPO) method. SePPO leverages previous checkpoints as reference models while using them to generate on-policy reference samples, which replace "losing images" in preference pairs. This approach allows us to optimize using only off-policy "winning images." Furthermore, we design a strategy for reference model selection that expands the exploration in the policy space. Notably, we do not simply treat reference samples as negative examples for learning. Instead, we design an anchor-based criterion to assess whether the reference samples are likely to be winning or losing images, allowing the model to selectively learn from the generated reference samples. This approach mitigates performance degradation caused by the uncertainty in reference sample quality. We validate SePPO across both text-to-image and text-to-video benchmarks. SePPO surpasses all previous approaches on the text-to-image benchmarks and also demonstrates outstanding performance on the text-to-video benchmarks. Code will be released in https://github.com/DwanZhang-AI/SePPO.
Authors: Tianzhu Ye, Li Dong, Yuqing Xia, Yutao Sun, Yi Zhu, Gao Huang, Furu Wei
Abstract: Transformer tends to overallocate attention to irrelevant context. In this work, we introduce Diff Transformer, which amplifies attention to the relevant context while canceling noise. Specifically, the differential attention mechanism calculates attention scores as the difference between two separate softmax attention maps. The subtraction cancels noise, promoting the emergence of sparse attention patterns. Experimental results on language modeling show that Diff Transformer outperforms Transformer in various settings of scaling up model size and training tokens. More intriguingly, it offers notable advantages in practical applications, such as long-context modeling, key information retrieval, hallucination mitigation, in-context learning, and reduction of activation outliers. By being less distracted by irrelevant context, Diff Transformer can mitigate hallucination in question answering and text summarization. For in-context learning, Diff Transformer not only enhances accuracy but is also more robust to order permutation, which was considered as a chronic robustness issue. The results position Diff Transformer as a highly effective and promising architecture to advance large language models.
Authors: Matt Y. Cheung, Tucker J. Netherton, Laurence E. Court, Ashok Veeraraghavan, Guha Balakrishnan
Abstract: Uncertainty quantification is crucial to account for the imperfect predictions of machine learning algorithms for high-impact applications. Conformal prediction (CP) is a powerful framework for uncertainty quantification that generates calibrated prediction intervals with valid coverage. In this work, we study how CP intervals are affected by bias - the systematic deviation of a prediction from ground truth values - a phenomenon prevalent in many real-world applications. We investigate the influence of bias on interval lengths of two different types of adjustments -- symmetric adjustments, the conventional method where both sides of the interval are adjusted equally, and asymmetric adjustments, a more flexible method where the interval can be adjusted unequally in positive or negative directions. We present theoretical and empirical analyses characterizing how symmetric and asymmetric adjustments impact the "tightness" of CP intervals for regression tasks. Specifically for absolute residual and quantile-based non-conformity scores, we prove: 1) the upper bound of symmetrically adjusted interval lengths increases by $2|b|$ where $b$ is a globally applied scalar value representing bias, 2) asymmetrically adjusted interval lengths are not affected by bias, and 3) conditions when asymmetrically adjusted interval lengths are guaranteed to be smaller than symmetric ones. Our analyses suggest that even if predictions exhibit significant drift from ground truth values, asymmetrically adjusted intervals are still able to maintain the same tightness and validity of intervals as if the drift had never happened, while symmetric ones significantly inflate the lengths. We demonstrate our theoretical results with two real-world prediction tasks: sparse-view computed tomography (CT) reconstruction and time-series weather forecasting. Our work paves the way for more bias-robust machine learning systems.
Authors: Fei Wang, Ninareh Mehrabi, Palash Goyal, Rahul Gupta, Kai-Wei Chang, Aram Galstyan
Abstract: Data is a crucial element in large language model (LLM) alignment. Recent studies have explored using LLMs for efficient data collection. However, LLM-generated data often suffers from quality issues, with underrepresented or absent aspects and low-quality datapoints. To address these problems, we propose Data Advisor, an enhanced LLM-based method for generating data that takes into account the characteristics of the desired dataset. Starting from a set of pre-defined principles in hand, Data Advisor monitors the status of the generated data, identifies weaknesses in the current dataset, and advises the next iteration of data generation accordingly. Data Advisor can be easily integrated into existing data generation methods to enhance data quality and coverage. Experiments on safety alignment of three representative LLMs (i.e., Mistral, Llama2, and Falcon) demonstrate the effectiveness of Data Advisor in enhancing model safety against various fine-grained safety issues without sacrificing model utility.
Authors: Jan Achterhold, Joerg Stueckler
Abstract: In this paper, we learn dynamics models for parametrized families of dynamical systems with varying properties. The dynamics models are formulated as stochastic processes conditioned on a latent context variable which is inferred from observed transitions of the respective system. The probabilistic formulation allows us to compute an action sequence which, for a limited number of environment interactions, optimally explores the given system within the parametrized family. This is achieved by steering the system through transitions being most informative for the context variable. We demonstrate the effectiveness of our method for exploration on a non-linear toy-problem and two well-known reinforcement learning environments.
Authors: Ilya Levin, Denis Belomestny, Alexey Naumov, Sergey Samsonov
Abstract: Policy evaluation is an important instrument for the comparison of different algorithms in Reinforcement Learning (RL). Yet even a precise knowledge of the value function $V^{\pi}$ corresponding to a policy $\pi$ does not provide reliable information on how far is the policy $\pi$ from the optimal one. We present a novel model-free upper value iteration procedure $({\sf UVIP})$ that allows us to estimate the suboptimality gap $V^{\star}(x) - V^{\pi}(x)$ from above and to construct confidence intervals for $V^\star$. Our approach relies on upper bounds to the solution of the Bellman optimality equation via martingale approach. We provide theoretical guarantees for ${\sf UVIP}$ under general assumptions and illustrate its performance on a number of benchmark RL problems.
Authors: Beomseok Seo, Lin Lin, Jia Li
Abstract: Deep neural network (DNN) models have achieved phenomenal success for applications in many domains, ranging from academic research in science and engineering to industry and business. The modeling power of DNN is believed to have come from the complexity and over-parameterization of the model, which on the other hand has been criticized for the lack of interpretation. Although certainly not true for every application, in some applications, especially in economics, social science, healthcare industry, and administrative decision making, scientists or practitioners are resistant to use predictions made by a black-box system for multiple reasons. One reason is that a major purpose of a study can be to make discoveries based upon the prediction function, e.g., to reveal the relationships between measurements. Another reason can be that the training dataset is not large enough to make researchers feel completely sure about a purely data-driven result. Being able to examine and interpret the prediction function will enable researchers to connect the result with existing knowledge or gain insights about new directions to explore. Although classic statistical models are much more explainable, their accuracy often falls considerably below DNN. In this paper, we propose an approach to fill the gap between relatively simple explainable models and DNN such that we can more flexibly tune the trade-off between interpretability and accuracy. Our main idea is a mixture of discriminative models that is trained with the guidance from a DNN. Although mixtures of discriminative models have been studied before, our way of generating the mixture is quite different.
Authors: Akim Kotelnikov, Dmitry Baranchuk, Ivan Rubachev, Artem Babenko
Abstract: Denoising diffusion probabilistic models are currently becoming the leading paradigm of generative modeling for many important data modalities. Being the most prevalent in the computer vision community, diffusion models have also recently gained some attention in other domains, including speech, NLP, and graph-like data. In this work, we investigate if the framework of diffusion models can be advantageous for general tabular problems, where datapoints are typically represented by vectors of heterogeneous features. The inherent heterogeneity of tabular data makes it quite challenging for accurate modeling, since the individual features can be of completely different nature, i.e., some of them can be continuous and some of them can be discrete. To address such data types, we introduce TabDDPM -- a diffusion model that can be universally applied to any tabular dataset and handles any type of feature. We extensively evaluate TabDDPM on a wide set of benchmarks and demonstrate its superiority over existing GAN/VAE alternatives, which is consistent with the advantage of diffusion models in other fields. Additionally, we show that TabDDPM is eligible for privacy-oriented setups, where the original datapoints cannot be publicly shared.
Authors: Grace Zhang, Ayush Jain, Injune Hwang, Shao-Hua Sun, Joseph J. Lim
Abstract: Multi-task reinforcement learning (MTRL) aims to learn several tasks simultaneously for better sample efficiency than learning them separately. Traditional methods achieve this by sharing parameters or relabeled data between tasks. In this work, we introduce a new framework for sharing behavioral policies across tasks, which can be used in addition to existing MTRL methods. The key idea is to improve each task's off-policy data collection by employing behaviors from other task policies. Selectively sharing helpful behaviors acquired in one task to collect training data for another task can lead to higher-quality trajectories, leading to more sample-efficient MTRL. Thus, we introduce a simple and principled framework called Q-switch mixture of policies (QMP) that selectively shares behavior between different task policies by using the task's Q-function to evaluate and select useful shareable behaviors. We theoretically analyze how QMP improves the sample efficiency of the underlying RL algorithm. Our experiments show that QMP's behavioral policy sharing provides complementary gains over many popular MTRL algorithms and outperforms alternative ways to share behaviors in various manipulation, locomotion, and navigation environments. Videos are available at https://qmp-mtrl.github.io.
Authors: Takuya Kanazawa, Chetan Gupta
Abstract: Sequential decision making in the real world often requires finding a good balance of conflicting objectives. In general, there exist a plethora of Pareto-optimal policies that embody different patterns of compromises between objectives, and it is technically challenging to obtain them exhaustively using deep neural networks. In this work, we propose a novel multi-objective reinforcement learning (MORL) algorithm that trains a single neural network via policy gradient to approximately obtain the entire Pareto set in a single run of training, without relying on linear scalarization of objectives. The proposed method works in both continuous and discrete action spaces with no design change of the policy network. Numerical experiments in benchmark environments demonstrate the practicality and efficacy of our approach in comparison to standard MORL baselines.
Authors: Ahmad Faraz Khan, Xinran Wang, Qi Le, Zain ul Abdeen, Azal Ahmad Khan, Haider Ali, Ming Jin, Jie Ding, Ali R. Butt, Ali Anwar
Abstract: Existing incentive solutions for traditional Federated Learning (FL) focus on individual contributions to a single global objective, neglecting the nuances of clustered personalization with multiple cluster-level models and the non-monetary incentives such as personalized model appeal for clients. In this paper, we first propose to treat incentivization and personalization as interrelated challenges and solve them with an incentive mechanism that fosters personalized learning. Additionally, current methods depend on an aggregator for client clustering, which is limited by a lack of access to clients' confidential information due to privacy constraints, leading to inaccurate clustering. To overcome this, we propose direct client involvement, allowing clients to indicate their cluster membership preferences based on data distribution and incentive-driven feedback. Our approach enhances the personalized model appeal for self-aware clients with high-quality data leading to their active and consistent participation. Our evaluation demonstrates significant improvements in test accuracy (8-45%), personalized model appeal (3-38%), and participation rates (31-100%) over existing FL models, including those addressing data heterogeneity and personalization.
Authors: Tae Yeob Kang, Haebom Lee, Sungho Suh
Abstract: Fault detection and diagnosis of the interconnects are crucial for prognostics and health management (PHM) of electronics. Traditional methods, which rely on electronic signals as prognostic factors, often struggle to accurately identify the root causes of defects without resorting to destructive testing. Furthermore, these methods are vulnerable to noise interference, which can result in false alarms. To address these limitations, in this paper, we propose a novel, non-destructive approach for early fault detection and accurate diagnosis of interconnect defects, with improved noise resilience. Our approach uniquely utilizes the signal patterns of the reflection coefficient across a range of frequencies, enabling both root cause identification and severity assessment. This approach departs from conventional time-series analysis and effectively transforms the signal data into a format suitable for advanced learning algorithms. Additionally, we introduce a novel severity rating ensemble learning (SREL) approach, which enhances diagnostic accuracy and robustness in noisy environments. Experimental results demonstrate that the proposed method is effective for fault detection and diagnosis and has the potential to extend to real-world industrial applications.
Authors: Xiaolong Li, Zhi-Qin John Xu, Zhongwang Zhang
Abstract: In this work, we investigate the mechanism underlying loss spikes observed during neural network training. When the training enters a region with a lower-loss-as-sharper (LLAS) structure, the training becomes unstable, and the loss exponentially increases once the loss landscape is too sharp, resulting in the rapid ascent of the loss spike. The training stabilizes when it finds a flat region. From a frequency perspective, we explain the rapid descent in loss as being primarily influenced by low-frequency components. We observe a deviation in the first eigendirection, which can be reasonably explained by the frequency principle, as low-frequency information is captured rapidly, leading to the rapid descent. Inspired by our analysis of loss spikes, we revisit the link between the maximum eigenvalue of the loss Hessian ($\lambda_{\mathrm{max}}$), flatness and generalization. We suggest that $\lambda_{\mathrm{max}}$ is a good measure of sharpness but not a good measure for generalization. Furthermore, we experimentally observe that loss spikes can facilitate condensation, causing input weights to evolve towards the same direction. And our experiments show that there is a correlation (similar trend) between $\lambda_{\mathrm{max}}$ and condensation. This observation may provide valuable insights for further theoretical research on the relationship between loss spikes, $\lambda_{\mathrm{max}}$, and generalization.
Authors: Chanyoung Chung, Jaejun Lee, Joyce Jiyoung Whang
Abstract: A hyper-relational knowledge graph has been recently studied where a triplet is associated with a set of qualifiers; a qualifier is composed of a relation and an entity, providing auxiliary information for a triplet. While existing hyper-relational knowledge graph embedding methods assume that the entities are discrete objects, some information should be represented using numeric values, e.g., (J.R.R., was born in, 1892). Also, a triplet (J.R.R., educated at, Oxford Univ.) can be associated with a qualifier such as (start time, 1911). In this paper, we propose a unified framework named HyNT that learns representations of a hyper-relational knowledge graph containing numeric literals in either triplets or qualifiers. We define a context transformer and a prediction transformer to learn the representations based not only on the correlations between a triplet and its qualifiers but also on the numeric information. By learning compact representations of triplets and qualifiers and feeding them into the transformers, we reduce the computation cost of using transformers. Using HyNT, we can predict missing numeric values in addition to missing entities or relations in a hyper-relational knowledge graph. Experimental results show that HyNT significantly outperforms state-of-the-art methods on real-world datasets.
Authors: Duy A. Nguyen, Trang H. Tran, Huy Hieu Pham, Phi Le Nguyen, Lam M. Nguyen
Abstract: In this work, we investigate the time series representation learning problem using self-supervised techniques. Contrastive learning is well-known in this area as it is a powerful method for extracting information from the series and generating task-appropriate representations. Despite its proficiency in capturing time series characteristics, these techniques often overlook a critical factor - the inherent noise in this type of data, a consideration usually emphasized in general time series analysis. Moreover, there is a notable absence of attention to developing efficient yet lightweight encoder architectures, with an undue focus on delivering contrastive losses. Our work address these gaps by proposing an innovative training strategy that promotes consistent representation learning, accounting for the presence of noise-prone signals in natural time series. Furthermore, we propose an encoder architecture that incorporates dilated convolution within the Inception block, resulting in a scalable and robust network with a wide receptive field. Experimental findings underscore the effectiveness of our method, consistently outperforming state-of-the-art approaches across various tasks, including forecasting, classification, and abnormality detection. Notably, our method attains the top rank in over two-thirds of the classification UCR datasets, utilizing only 40% of the parameters compared to the second-best approach. Our source code for CoInception framework is accessible at https://github.com/anhduy0911/CoInception.
Authors: Bavo D. C. Campo, Katrien Antonio
Abstract: Traditionally, the detection of fraudulent insurance claims relies on business rules and expert judgement which makes it a time-consuming and expensive process (\'Oskarsd\'ottir et al., 2022). Consequently, researchers have been examining ways to develop efficient and accurate analytic strategies to flag suspicious claims. Feeding learning methods with features engineered from the social network of parties involved in a claim is a particularly promising strategy (see for example Van Vlasselaer et al. (2016); Tumminello et al. (2023)). When developing a fraud detection model, however, we are confronted with several challenges. The uncommon nature of fraud, for example, creates a high class imbalance which complicates the development of well performing analytic classification models. In addition, only a small number of claims are investigated and get a label, which results in a large corpus of unlabeled data. Yet another challenge is the lack of publicly available data. This hinders not only the development of new methods, but also the validation of existing techniques. We therefore design a simulation machine that is engineered to create synthetic data with a network structure and available covariates similar to the real life insurance fraud data set analyzed in \'Oskarsd\'ottir et al. (2022). Further, the user has control over several data-generating mechanisms. We can specify the total number of policyholders and parties, the desired level of imbalance and the (effect size of the) features in the fraud generating model. As such, the simulation engine enables researchers and practitioners to examine several methodological challenges as well as to test their (development strategy of) insurance fraud detection models in a range of different settings. Moreover, large synthetic data sets can be generated to evaluate the predictive performance of (advanced) machine learning techniques.
Authors: Yahong Yang, Qipin Chen, Wenrui Hao
Abstract: In this paper, we present a novel training approach called the Homotopy Relaxation Training Algorithm (HRTA), aimed at accelerating the training process in contrast to traditional methods. Our algorithm incorporates two key mechanisms: one involves building a homotopy activation function that seamlessly connects the linear activation function with the ReLU activation function; the other technique entails relaxing the homotopy parameter to enhance the training refinement process. We have conducted an in-depth analysis of this novel method within the context of the neural tangent kernel (NTK), revealing significantly improved convergence rates. Our experimental results, especially when considering networks with larger widths, validate the theoretical conclusions. This proposed HRTA exhibits the potential for other activation functions and deep neural networks.
Authors: Zenan Li, Fan Nie, Qiao Sun, Fang Da, Hang Zhao
Abstract: Offline Reinforcement Learning (RL) enables policy learning without active interactions, making it especially appealing for self-driving tasks. Recent successes of Transformers inspire casting offline RL as sequence modeling, which, however, fails in stochastic environments with incorrect assumptions that identical actions can consistently achieve the same goal. In this paper, we introduce an UNcertainty-awaRE deciSion Transformer (UNREST) for planning in stochastic driving environments without introducing additional transition or complex generative models. Specifically, UNREST estimates uncertainties by conditional mutual information between transitions and returns. Discovering 'uncertainty accumulation' and 'temporal locality' properties of driving environments, we replace the global returns in decision transformers with truncated returns less affected by environments to learn from actual outcomes of actions rather than environment transitions. We also dynamically evaluate uncertainty at inference for cautious planning. Extensive experiments demonstrate UNREST's superior performance in various driving scenarios and the power of our uncertainty estimation strategy.
Authors: Razan Baltaji, Sourya Basu, Lav R. Varshney
Abstract: Constructing model-agnostic group equivariant networks, such as equitune (Basu et al., 2023b) and its generalizations (Kim et al., 2023), can be computationally expensive for large product groups. We address this problem by providing efficient model-agnostic equivariant designs for two related problems: one where the network has multiple inputs each with potentially different groups acting on them, and another where there is a single input but the group acting on it is a large product group. For the first design, we initially consider a linear model and characterize the entire equivariant space that satisfies this constraint. This characterization gives rise to a novel fusion layer between different channels that satisfies an invariance-symmetry (IS) constraint, which we call an IS layer. We then extend this design beyond linear models, similar to equitune, consisting of equivariant and IS layers. We also show that the IS layer is a universal approximator of invariant-symmetric functions. Inspired by the first design, we use the notion of the IS property to design a second efficient model-agnostic equivariant design for large product groups acting on a single input. For the first design, we provide experiments on multi-image classification where each view is transformed independently with transformations such as rotations. We find equivariant models are robust to such transformations and perform competitively otherwise. For the second design, we consider three applications: language compositionality on the SCAN dataset to product groups; fairness in natural language generation from GPT-2 to address intersectionality; and robust zero-shot image classification with CLIP. Overall, our methods are simple and general, competitive with equitune and its variants, while also being computationally more efficient.
Authors: Sulin Liu, Peter J. Ramadge, Ryan P. Adams
Abstract: We introduce marginalization models (MAMs), a new family of generative models for high-dimensional discrete data. They offer scalable and flexible generative modeling by explicitly modeling all induced marginal distributions. Marginalization models enable fast approximation of arbitrary marginal probabilities with a single forward pass of the neural network, which overcomes a major limitation of arbitrary marginal inference models, such as any-order autoregressive models. MAMs also address the scalability bottleneck encountered in training any-order generative models for high-dimensional problems under the context of energy-based training, where the goal is to match the learned distribution to a given desired probability (specified by an unnormalized log-probability function such as energy or reward function). We propose scalable methods for learning the marginals, grounded in the concept of "marginalization self-consistency". We demonstrate the effectiveness of the proposed model on a variety of discrete data distributions, including images, text, physical systems, and molecules, for maximum likelihood and energy-based training settings. MAMs achieve orders of magnitude speedup in evaluating the marginal probabilities on both settings. For energy-based training tasks, MAMs enable any-order generative modeling of high-dimensional problems beyond the scale of previous methods. Code is available at https://github.com/PrincetonLIPS/MaM.
Authors: Evgenii Dzhivelikian, Petr Kuderov, Aleksandr I. Panov
Abstract: This paper presents a novel approach to address the challenge of online temporal memory learning for decision-making under uncertainty in non-stationary, partially observable environments. The proposed algorithm, Distributed Hebbian Temporal Memory (DHTM), is based on factor graph formalism and a multicomponent neuron model. DHTM aims to capture sequential data relationships and make cumulative predictions about future observations, forming Successor Features (SF). Inspired by neurophysiological models of the neocortex, the algorithm utilizes distributed representations, sparse transition matrices, and local Hebbian-like learning rules to overcome the instability and slow learning process of traditional temporal memory algorithms like RNN and HMM. Experimental results demonstrate that DHTM outperforms LSTM and a biologically inspired HMM-like algorithm, CSCG, in the case of non-stationary datasets. Our findings suggest that DHTM is a promising approach for addressing the challenges of online sequence learning and planning in dynamic environments.
Authors: Aysin Tumay, Mustafa E. Aydin, Ali T. Koc, Suleyman S. Kozat
Abstract: We introduce a novel ensemble approach for feature selection based on hierarchical stacking for non-stationarity and/or a limited number of samples with a large number of features. Our approach exploits the co-dependency between features using a hierarchical structure. Initially, a machine learning model is trained using a subset of features, and then the output of the model is updated using other algorithms in a hierarchical manner with the remaining features to minimize the target loss. This hierarchical structure allows for flexible depth and feature selection. By exploiting feature co-dependency hierarchically, our proposed approach overcomes the limitations of traditional feature selection methods and feature importance scores. The effectiveness of the approach is demonstrated on synthetic and well-known real-life datasets, providing significant scalable and stable performance improvements compared to the traditional methods and the state-of-the-art approaches. We also provide the source code of our approach to facilitate further research and replicability of our results.
Authors: Surojit Saha, Sarang Joshi, Ross Whitaker
Abstract: The variational autoencoder (VAE) is a well-studied, deep, latent-variable model (DLVM) that efficiently optimizes the variational lower bound of the log marginal data likelihood and has a strong theoretical foundation. However, the VAE's known failure to match the aggregate posterior often results in \emph{pockets/holes} in the latent distribution (i.e., a failure to match the prior) and/or \emph{posterior collapse}, which is associated with a loss of information in the latent space. This paper addresses these shortcomings in VAEs by reformulating the objective function associated with VAEs in order to match the aggregate/marginal posterior distribution to the prior. We use kernel density estimate (KDE) to model the aggregate posterior in high dimensions. The proposed method is named the \emph{aggregate variational autoencoder} (AVAE) and is built on the theoretical framework of the VAE. Empirical evaluation of the proposed method on multiple benchmark data sets demonstrates the effectiveness of the AVAE relative to state-of-the-art (SOTA) methods.
Authors: Nicholas E. Corrado, Josiah P. Hanna
Abstract: On-policy reinforcement learning (RL) algorithms perform policy updates using i.i.d. trajectories collected by the current policy. However, after observing only a finite number of trajectories, on-policy sampling may produce data that fails to match the expected on-policy data distribution. This sampling error leads to noisy updates and data inefficient on-policy learning. Recent work in the policy evaluation setting has shown that non-i.i.d., off-policy sampling can produce data with lower sampling error than on-policy sampling can produce (Zhong et. al, 2022). Motivated by this observation, we introduce an adaptive, off-policy sampling method to improve the data efficiency of on-policy policy gradient algorithms. Our method, Proximal Robust On-Policy Sampling (PROPS), reduces sampling error by collecting data with a behavior policy that increases the probability of sampling actions that are under-sampled with respect to the current policy. We empirically evaluate PROPS on both continuous-action MuJoCo benchmark tasks as well discrete-action tasks and demonstrate that (1) PROPS decreases sampling error throughout training and (2) improves the data efficiency of on-policy policy gradient algorithms.
Authors: Ian Berlot-Attwell, Kumar Krishna Agrawal, A. Michael Carrell, Yash Sharma, Naomi Saphra
Abstract: Although modern neural networks often generalize to new combinations of familiar concepts, the conditions that enable such compositionality have long been an open question. In this work, we study the systematicity gap in visual question answering: the performance difference between reasoning on previously seen and unseen combinations of object attributes. To test, we introduce a novel diagnostic dataset, CLEVR-HOPE. We find that the systematicity gap is not reduced by increasing the quantity of training data, but is reduced by increasing the diversity of training data. In particular, our experiments suggest that the more distinct attribute type combinations are seen during training, the more systematic we can expect the resulting model to be.
Authors: Max Milkert, David Hyde, Forrest Laine
Abstract: A neural network with ReLU activations may be viewed as a composition of piecewise linear functions. For such networks, the number of distinct linear regions expressed over the input domain has the potential to scale exponentially with depth, but it is not expected to do so when the initial parameters are chosen randomly. Therefore, randomly initialized models are often unnecessarily large, even when approximating simple functions. To address this issue, we introduce a novel training strategy: we first reparameterize the network weights in a manner that forces the network to exhibit a number of linear regions exponential in depth. Training first on our derived parameters provides an initial solution that can later be refined by directly updating the underlying model weights. This approach allows us to learn approximations of convex, one-dimensional functions that are several orders of magnitude more accurate than their randomly initialized counterparts. We further demonstrate how to extend our approach to multidimensional and non convex functions, with similar benefits observed.
Authors: Alex Lewandowski, Haruto Tanaka, Dale Schuurmans, Marlos C. Machado
Abstract: Loss of plasticity is a phenomenon in which neural networks lose their ability to learn from new experience. Despite being empirically observed in several problem settings, little is understood about the mechanisms that lead to loss of plasticity. In this paper, we offer a consistent explanation for loss of plasticity: Neural networks lose directions of curvature during training and that loss of plasticity can be attributed to this reduction in curvature. To support such a claim, we provide a systematic investigation of loss of plasticity across continual learning tasks using MNIST, CIFAR-10 and ImageNet. Our findings illustrate that loss of curvature directions coincides with loss of plasticity, while also showing that previous explanations are insufficient to explain loss of plasticity in all settings. Lastly, we show that regularizers which mitigate loss of plasticity also preserve curvature, motivating a simple distributional regularizer that proves to be effective across the problem settings we considered.
Authors: Lingfan Yu, Jinkun Lin, Jinyang Li
Abstract: Large Language Models (LLMs) are wildly popular today and it is important to serve them efficiently. Existing LLM serving systems are stateless across requests. Consequently, when LLMs are used in the common setting of multi-turn conversations, a growing log of the conversation history must be processed alongside any request by the serving system at each turn, resulting in repeated processing. In this paper, we design $Pensieve$, a system optimized for multi-turn conversation LLM serving. $Pensieve$ maintains the conversation state across requests by caching previously processed history to avoid duplicate processing. $Pensieve$'s multi-tier caching strategy can utilize both GPU and CPU memory to efficiently store and retrieve cached data. $Pensieve$ also generalizes the recent PagedAttention kernel to support attention between multiple input tokens with a GPU cache spread over non-contiguous memory. Our evaluation shows that $Pensieve$ can achieve $1.14$-$3.0\times$ the throughput of vLLM and TensorRT-LLM and significantly reduce latency.
Authors: Rebecca M. Neeser, Bruno Correia, Philippe Schwaller
Abstract: Determining whether a molecule can be synthesized is crucial in chemistry and drug discovery, as it guides experimental prioritization and molecule ranking in de novo design tasks. Existing scoring approaches to assess synthetic feasibility struggle to extrapolate to new chemical spaces or fail to discriminate based on subtle differences such as chirality. This work addresses these limitations by introducing the Focused Synthesizability score~(FSscore), which uses machine learning to rank structures based on their relative ease of synthesis. First, a baseline trained on an extensive set of reactant-product pairs is established, which is then refined with expert human feedback tailored to specific chemical spaces. This targeted fine-tuning improves performance on these chemical scopes, enabling more accurate differentiation between molecules that are hard and easy to synthesize. The FSscore showcases how a human-in-the-loop framework can be utilized to optimize the assessment of synthetic feasibility for various chemical applications.
Authors: Daniel Gei{\ss}ler, Bo Zhou, Mengxi Liu, Sungho Suh, Paul Lukowicz
Abstract: This work offers a heuristic evaluation of the effects of variations in machine learning training regimes and learning paradigms on the energy consumption of computing, especially HPC hardware with a life-cycle aware perspective. While increasing data availability and innovation in high-performance hardware fuels the training of sophisticated models, it also fosters the fading perception of energy consumption and carbon emission. Therefore, the goal of this work is to raise awareness about the energy impact of general training parameters and processes, from learning rate over batch size to knowledge transfer. Multiple setups with different hyperparameter configurations are evaluated on three different hardware systems. Among many results, we have found out that even with the same model and hardware to reach the same accuracy, improperly set training hyperparameters consume up to 5 times the energy of the optimal setup. We also extensively examined the energy-saving benefits of learning paradigms including recycling knowledge through pretraining and sharing knowledge through multitask training.
Authors: Yunkun Zhang, Jin Gao, Zheling Tan, Lingfeng Zhou, Kexin Ding, Mu Zhou, Shaoting Zhang, Dequan Wang
Abstract: The advent of foundation models (FMs) as an emerging suite of AI techniques has struck a wave of opportunities in computational healthcare. The interactive nature of these models, guided by pre-training data and human instructions, has ignited a data-centric AI paradigm that emphasizes better data characterization, quality, and scale. In healthcare AI, obtaining and processing high-quality clinical data records has been a longstanding challenge, ranging from data quantity, annotation, patient privacy, and ethics. In this survey, we investigate a wide range of data-centric approaches in the FM era (from model pre-training to inference) towards improving the healthcare workflow. We discuss key perspectives in AI security, assessment, and alignment with human values. Finally, we offer a promising outlook of FM-based analytics to enhance the performance of patient outcome and clinical workflow in the evolving landscape of healthcare and medicine. We provide an up-to-date list of healthcare-related foundation models and datasets at https://github.com/Yunkun-Zhang/Data-Centric-FM-Healthcare .
URLs: https://github.com/Yunkun-Zhang/Data-Centric-FM-Healthcare
Authors: Sungdong Kim, Minjoon Seo
Abstract: Learning from human feedback via proxy reward modeling has been studied to align Large Language Models (LLMs) with human values. However, achieving reliable training through that proxy reward model (RM) is not a trivial problem, and its behavior remained as a black-box. In this paper, we study the role of proxy rewards in the LLM alignment via `reverse reward engineering' by composing interpretable features as a white-box reward function. We aim to replicate the ground truth (gold) reward signal by achieving a monotonic relationship between the proxy and gold reward signals after training the model using the proxy reward in reinforcement learning (RL). Our findings indicate that successfully emulating the gold reward requires generating responses that are relevant with enough length to open-ended questions, while also ensuring response consistency in closed-ended questions. Furthermore, resulting models optimizing our devised white-box reward show competitive performances with strong open-source RMs in alignment benchmarks. We highlight its potential usage as a simple but strong reward baseline for the LLM alignment, not requiring explicit human feedback dataset and RM training. Our code is available at https://github.com/naver-ai/rethinking-proxy-reward.
Authors: Zachary Ankner, Rishab Parthasarathy, Aniruddha Nrusimha, Christopher Rinard, Jonathan Ragan-Kelley, William Brandon
Abstract: To combat the memory bandwidth-bound nature of autoregressive LLM inference, previous research has proposed the speculative decoding frame-work. To perform speculative decoding, a small draft model proposes candidate continuations of the input sequence that are then verified in parallel by the base model. One way to specify the draft model, as used in the recent Medusa decoding framework, is as a collection of lightweight heads, called draft heads, that operate on the base model's hidden states. To date, all existing draft heads have been sequentially independent, meaning that they speculate tokens in the candidate continuation independently of any preceding tokens in the candidate continuation. In this work, we propose Hydra heads: a sequentially-dependent drop-in replacement for standard draft heads that significantly improves the accuracy of draft head speculation. We further explore the design space of Hydra head training objectives and architectures, and propose a carefully tuned Hydra head recipe, which we call Hydra++, that improves decoding throughput by up to 1.31x and 2.70x compared to Medusa decoding and autoregressive de-coding respectively. Overall, Hydra heads are a simple and well-motivated intervention on standard draft heads that significantly improve the end-to-end speed of draft head-based speculative decoding. We make our code publicly available at https://github.com/zankner/Hydra.
Authors: Siyuan Li, Zicheng Liu, Juanxi Tian, Ge Wang, Zedong Wang, Weiyang Jin, Di Wu, Cheng Tan, Tao Lin, Yang Liu, Baigui Sun, Stan Z. Li
Abstract: Exponential Moving Average (EMA) is a widely used weight averaging (WA) regularization to learn flat optima for better generalizations without extra cost in deep neural network (DNN) optimization. Despite achieving better flatness, existing WA methods might fall into worse final performances or require extra test-time computations. This work unveils the full potential of EMA with a single line of modification, i.e., switching the EMA parameters to the original model after each epoch, dubbed as Switch EMA (SEMA). From both theoretical and empirical aspects, we demonstrate that SEMA can help DNNs to reach generalization optima that better trade-off between flatness and sharpness. To verify the effectiveness of SEMA, we conduct comparison experiments with discriminative, generative, and regression tasks on vision and language datasets, including image classification, self-supervised learning, object detection and segmentation, image generation, video prediction, attribute regression, and language modeling. Comprehensive results with popular optimizers and networks show that SEMA is a free lunch for DNN training by improving performances and boosting convergence speeds.
Authors: Eshwar Ram Arunachaleswaran, Natalie Collina, Aaron Roth, Mirah Shi
Abstract: Blasiok et al. [2023] proposed distance to calibration as a natural measure of calibration error that unlike expected calibration error (ECE) is continuous. Recently, Qiao and Zheng [2024] gave a non-constructive argument establishing the existence of an online predictor that can obtain $O(\sqrt{T})$ distance to calibration in the adversarial setting, which is known to be impossible for ECE. They leave as an open problem finding an explicit, efficient algorithm. We resolve this problem and give an extremely simple, efficient, deterministic algorithm that obtains distance to calibration error at most $2\sqrt{T}+1$.
Authors: Haoran He, Chenjia Bai, Ling Pan, Weinan Zhang, Bin Zhao, Xuelong Li
Abstract: Learning a generalist embodied agent capable of completing multiple tasks poses challenges, primarily stemming from the scarcity of action-labeled robotic datasets. In contrast, a vast amount of human videos exist, capturing intricate tasks and interactions with the physical world. Promising prospects arise for utilizing actionless human videos for pre-training and transferring the knowledge to facilitate robot policy learning through limited robot demonstrations. However, it remains a challenge due to the domain gap between humans and robots. Moreover, it is difficult to extract useful information representing the dynamic world from human videos, because of its noisy and multimodal data structure. In this paper, we introduce a novel framework to tackle these challenges, which leverages a unified discrete diffusion to combine generative pre-training on human videos and policy fine-tuning on a small number of action-labeled robot videos. We start by compressing both human and robot videos into unified video tokens. In the pre-training stage, we employ a discrete diffusion model with a mask-and-replace diffusion strategy to predict future video tokens in the latent space. In the fine-tuning stage, we harness the imagined future videos to guide low-level action learning with a limited set of robot data. Experiments demonstrate that our method generates high-fidelity future videos for planning and enhances the fine-tuned policies compared to previous state-of-the-art approaches with superior performance. Our project website is available at https://video-diff.github.io/.
Authors: Tongyi Liang, Han-Xiong Li
Abstract: Existing models encounter bottlenecks in balancing performance and computational efficiency when modeling long sequences. Although the state space model (SSM) has achieved remarkable success in handling long sequence tasks, it still faces the problem of large number of parameters. In order to further improve the efficiency of SSM, we propose a new state space layer based on multiple-input multiple-output SSM, called efficient SSM (eSSM). Our eSSM is built on the convolutional representation of multi-input and multi-input (MIMO) SSM. We propose a variety of effective strategies to improve the computational efficiency. The diagonalization of the system matrix first decouples the original system. Then a fast tensor convolution is proposed based on the fast Fourier transform. In addition, the block diagonalization of the SSM further reduces the model parameters and improves the model flexibility. Extensive experimental results show that the performance of the proposed model on multiple databases matches the performance of state-of-the-art models, such as S4, and is significantly better than Transformers and LSTM. In the model efficiency benchmark, the parameters of eSSM are only 12.89\% of LSTM and 13.24\% of Mamba. The training speed of eSSM is 3.94 times faster than LSTM and 1.35 times faster than Mamba. Code is available at: \href{https://github.com/leonty1/essm}{https://github.com/leonty1/essm}.
URLs: https://github.com/leonty1/essm, https://github.com/leonty1/essm
Authors: Juan L. Valerdi
Abstract: A characterization of the representability of neural networks is relevant to comprehend their success in artificial intelligence. This study investigate two topics on ReLU neural network expressivity and their connection with a conjecture related to the minimum depth required for representing any continuous piecewise linear (CPWL) function. The topics are the minimal depth representation of the sum and max operations, as well as the exploration of polytope neural networks. For the sum operation, we establish a sufficient condition on the minimal depth of the operands to find the minimal depth of the operation. In contrast, regarding the max operation, a comprehensive set of examples is presented, demonstrating that no sufficient conditions, depending solely on the depth of the operands, would imply a minimal depth for the operation. The study also examine the minimal depth relationship between convex CPWL functions. On polytope neural networks, we investigate basic depth properties from Minkowski sums, convex hulls, number of vertices, faces, affine transformations, and indecomposable polytopes. More significant findings include depth characterization of polygons; identification of polytopes with an increasing number of vertices, exhibiting small depth and others with arbitrary large depth; and most notably, the minimal depth of simplices, which is strictly related to the minimal depth conjecture in ReLU networks.
Authors: Ali Ebrahimpour Boroojeny, Matus Telgarsky, Hari Sundaram
Abstract: We show the effectiveness of automatic differentiation in efficiently and correctly computing and controlling the spectrum of implicitly linear operators, a rich family of layer types including all standard convolutional and dense layers. We provide the first clipping method which is correct for general convolution layers, and illuminate the representational limitation that caused correctness issues in prior work. We study the effect of the batch normalization layers when concatenated with convolutional layers and show how our clipping method can be applied to their composition. By comparing the accuracy and performance of our algorithms to the state-of-the-art methods, using various experiments, we show they are more precise and efficient and lead to better generalization and adversarial robustness. We provide the code for using our methods at https://github.com/Ali-E/FastClip.
Authors: Antonina Krajewska, Ewa Niewiadomska-Szynkiewicz
Abstract: We introduce a two-step method for the matrix recovery problem. Our approach combines the theoretical foundations of the Column Subset Selection and Low-rank Matrix Completion problems. The proposed method, in each step, solves a convex optimization task. We present two algorithms that implement our Columns Selected Matrix Completion (CSMC) method, each dedicated to a different size problem. We performed a formal analysis of the presented method, in which we formulated the necessary assumptions and the probability of finding a correct solution. In the second part of the paper, we present the results of the experimental work. Numerical experiments verified the correctness and performance of the algorithms. To study the influence of the matrix size, rank, and the proportion of missing elements on the quality of the solution and the computation time, we performed experiments on synthetic data. The presented method was applied to two real-life problems problems: prediction of movie rates in a recommendation system and image inpainting. Our thorough analysis shows that CSMC provides solutions of comparable quality to matrix completion algorithms, which are based on convex optimization. However, CSMC offers notable savings in terms of runtime.
Authors: Yongshuo Zong, Ondrej Bohdal, Timothy Hospedales
Abstract: Large language models (LLMs) famously exhibit emergent in-context learning (ICL) -- the ability to rapidly adapt to new tasks using few-shot examples provided as a prompt, without updating the model's weights. Built on top of LLMs, vision large language models (VLLMs) have advanced significantly in areas such as recognition, reasoning, and grounding. However, investigations into \emph{multimodal ICL} have predominantly focused on few-shot visual question answering (VQA), and image captioning, which we will show neither exploit the strengths of ICL, nor test its limitations. The broader capabilities and limitations of multimodal ICL remain under-explored. In this study, we introduce a comprehensive benchmark VL-ICL Bench for multimodal in-context learning, encompassing a broad spectrum of tasks that involve both images and text as inputs and outputs, and different types of challenges, from {perception to reasoning and long context length}. We evaluate the abilities of state-of-the-art VLLMs against this benchmark suite, revealing their diverse strengths and weaknesses, and showing that even the most advanced models, such as GPT-4, find the tasks challenging. By highlighting a range of new ICL tasks, and the associated strengths and limitations of existing models, we hope that our dataset will inspire future work on enhancing the in-context learning capabilities of VLLMs, as well as inspire new applications that leverage VLLM ICL. The code and dataset are available at https://github.com/ys-zong/VL-ICL.
Authors: Huanran Li, Daniel Pimentel-Alarc\'on
Abstract: Contrastive Learning (CL) has emerged as a powerful method for training feature extraction models using unlabeled data. Recent studies suggest that incorporating a linear projection head post-backbone significantly enhances model performance. In this work, we investigate the use of a transformer model as a projection head within the CL framework, aiming to exploit the transformer's capacity for capturing long-range dependencies across embeddings to further improve performance. Our key contributions are fourfold: First, we introduce a novel application of transformers in the projection head role for contrastive learning, marking the first endeavor of its kind. Second, our experiments reveal a compelling "Deep Fusion" phenomenon where the attention mechanism progressively captures the correct relational dependencies among samples from the same class in deeper layers. Third, we provide a theoretical framework that explains and supports this "Deep Fusion" behavior. Finally, we demonstrate through experimental results that our model achieves superior performance compared to the existing approach of using a feed-forward layer.
Authors: Huanran Li, Manh Nguyen, Daniel Pimentel-Alarc\'on
Abstract: Contrastive learning has emerged as a powerful method in deep learning, excelling at learning effective representations through contrasting samples from different distributions. However, neural collapse, where embeddings converge into a lower-dimensional space, poses a significant challenge, especially in semi-supervised and self-supervised setups. In this paper, we first theoretically analyze the effect of large learning rates on contrastive losses that solely rely on the cosine similarity metric, and derive a theoretical bound to mitigate this collapse. {Building on these insights, we propose CLOP, a novel semi-supervised loss function designed to prevent neural collapse by promoting the formation of orthogonal linear subspaces among class embeddings.} Unlike prior approaches that enforce a simplex ETF structure, CLOP focuses on subspace separation, leading to more distinguishable embeddings. Through extensive experiments on real and synthetic datasets, we demonstrate that CLOP enhances performance, providing greater stability across different learning rates and batch sizes.
Authors: Khadija Zanna, Akane Sano
Abstract: Bias originates from both data and algorithmic design, often exacerbated by traditional fairness methods that fail to address the subtle impacts of protected attributes. This study introduces an approach to mitigate bias in machine learning by leveraging model uncertainty. Our approach utilizes a multi-task learning (MTL) framework combined with Monte Carlo (MC) Dropout to assess and mitigate uncertainty in predictions related to protected labels. By incorporating MC Dropout, our framework quantifies prediction uncertainty, which is crucial in areas with vague decision boundaries, thereby enhancing model fairness. Our methodology integrates multi-objective learning through pareto-optimality to balance fairness and performance across various applications. We demonstrate the effectiveness and transferability of our approach across multiple datasets and enhance model explainability through saliency maps to interpret how input features influence predictions, thereby enhancing the interpretability of machine learning models in practical applications.
Authors: Junfan Li, Zheshun Wu, Zenglin Xu, Irwin King
Abstract: We consider online model selection with decentralized data over $M$ clients, and study the necessity of collaboration among clients. Previous work proposed various federated algorithms without demonstrating their necessity,while we answer the question from a novel perspective of computational constraints. We prove lower bounds on the regret, and propose a federated algorithm and analyze the upper bound.Our results show (i) collaboration is unnecessary in the absence of computational constraints on clients; (ii) collaboration is necessary if the computational cost on each client is limited to $o(K)$, where $K$ is the number of candidate hypothesis spaces. We clarify the unnecessary nature of collaboration in previous federated algorithms for distributed online multi-kernel learning,and improve the regret bounds at a smaller computational and communication cost. Our algorithm relies on three new techniques including an improved Bernstein's inequality for martingale, a federated online mirror descent framework, and decoupling model selection and prediction, which might be of independent interest.
Authors: Jose Florido, He Wang, Amirul Khan, Peter K. Jimack
Abstract: Physics-informed neural networks (PINNs) provide a means of obtaining approximate solutions of partial differential equations and systems through the minimisation of an objective function which includes the evaluation of a residual function at a set of collocation points within the domain. The quality of a PINNs solution depends upon numerous parameters, including the number and distribution of these collocation points. In this paper we consider a number of strategies for selecting these points and investigate their impact on the overall accuracy of the method. In particular, we suggest that no single approach is likely to be "optimal" but we show how a number of important metrics can have an impact in improving the quality of the results obtained when using a fixed number of residual evaluations. We illustrate these approaches through the use of two benchmark test problems: Burgers' equation and the Allen-Cahn equation.
Authors: Shitong Shao, Zikai Zhou, Huanran Chen, Zhiqiang Shen
Abstract: Dataset condensation, a concept within data-centric learning, efficiently transfers critical attributes from an original dataset to a synthetic version, maintaining both diversity and realism. This approach significantly improves model training efficiency and is adaptable across multiple application areas. Previous methods in dataset condensation have faced challenges: some incur high computational costs which limit scalability to larger datasets (e.g., MTT, DREAM, and TESLA), while others are restricted to less optimal design spaces, which could hinder potential improvements, especially in smaller datasets (e.g., SRe2L, G-VBSM, and RDED). To address these limitations, we propose a comprehensive design framework that includes specific, effective strategies like implementing soft category-aware matching and adjusting the learning rate schedule. These strategies are grounded in empirical evidence and theoretical backing. Our resulting approach, Elucidate Dataset Condensation (EDC), establishes a benchmark for both small and large-scale dataset condensation. In our testing, EDC achieves state-of-the-art accuracy, reaching 48.6% on ImageNet-1k with a ResNet-18 model at an IPC of 10, which corresponds to a compression ratio of 0.78%. This performance exceeds those of SRe2L, G-VBSM, and RDED by margins of 27.3%, 17.2%, and 6.6%, respectively.
Authors: Yue Wu, Zhiqing Sun, Huizhuo Yuan, Kaixuan Ji, Yiming Yang, Quanquan Gu
Abstract: Standard reinforcement learning from human feedback (RLHF) approaches relying on parametric models like the Bradley-Terry model fall short in capturing the intransitivity and irrationality in human preferences. Recent advancements suggest that directly working with preference probabilities can yield a more accurate reflection of human preferences, enabling more flexible and accurate language model alignment. In this paper, we propose a self-play-based method for language model alignment, which treats the problem as a constant-sum two-player game aimed at identifying the Nash equilibrium policy. Our approach, dubbed Self-Play Preference Optimization (SPPO), utilizes iterative policy updates to provably approximate the Nash equilibrium. Additionally, we propose a new SPPO objective which is both strongly motivated by theory and is simple and effective in practice. In our experiments, using only 60k prompts (without responses) from the UltraFeedback dataset and without any prompt augmentation, by leveraging a pre-trained preference model PairRM with only 0.4B parameters, SPPO can obtain a model from fine-tuning Mistral-7B-Instruct-v0.2 that achieves the state-of-the-art length-controlled win-rate of 28.53% against GPT-4-Turbo on AlpacaEval 2.0. It also outperforms the (iterative) DPO and IPO on MT-Bench, Arena-Hard, and the Open LLM Leaderboard. Starting from a stronger base model Llama-3-8B-Instruct, we are able to achieve a length-controlled win rate of 38.77%. Notably, the strong performance of SPPO is achieved without additional external supervision (e.g., responses, preferences, etc.) from GPT-4 or other stronger language models. Codes are available at https://github.com/uclaml/SPPO.
Authors: Zhongwang Zhang, Pengxiao Lin, Zhiwei Wang, Yaoyu Zhang, Zhi-Qin John Xu
Abstract: Transformers have shown impressive capabilities across various tasks, but their performance on compositional problems remains a topic of debate. In this work, we investigate the mechanisms of how transformers behave on unseen compositional tasks. We discover that the parameter initialization scale plays a critical role in determining whether the model learns inferential solutions, which capture the underlying compositional primitives, or symmetric solutions, which simply memorize mappings without understanding the compositional structure. By analyzing the information flow and vector representations within the model, we reveal the distinct mechanisms underlying these solution types. We further find that inferential solutions exhibit low complexity bias, which we hypothesize is a key factor enabling them to learn individual mappings for single anchors. We validate our conclusions on various real-world datasets. Our findings provide valuable insights into the role of initialization scale in shaping the type of solution learned by transformers and their ability to learn and generalize compositional tasks.
Authors: Ke Liu, Kaijing Ding, Xi Cheng, Guanhao Xu, Xin Hu, Tong Liu, Siyuan Feng, Binze Cai, Jianan Chen, Hui Lin, Jilin Song, Chen Zhu
Abstract: Since flight delay hurts passengers, airlines, and airports, its prediction becomes crucial for the decision-making of all stakeholders in the aviation industry and thus has been attempted by various previous research. However, previous delay predictions are often categorical and at a highly aggregated level. To improve that, this study proposes to apply the novel Temporal Fusion Transformer model and predict numerical airport arrival delays at quarter hour level for U.S. top 30 airports. Inputs to our model include airport demand and capacity forecasts, historic airport operation efficiency information, airport wind and visibility conditions, as well as enroute weather and traffic conditions. The results show that our model achieves satisfactory performance measured by small prediction errors on the test set. In addition, the interpretability analysis of the model outputs identifies the important input factors for delay prediction.
Authors: Yixing Jiang, Jeremy Irvin, Ji Hun Wang, Muhammad Ahmed Chaudhry, Jonathan H. Chen, Andrew Y. Ng
Abstract: Large language models are effective at few-shot in-context learning (ICL). Recent advancements in multimodal foundation models have enabled unprecedentedly long context windows, presenting an opportunity to explore their capability to perform ICL with many more demonstrating examples. In this work, we evaluate the performance of multimodal foundation models scaling from few-shot to many-shot ICL. We benchmark GPT-4o and Gemini 1.5 Pro across 14 datasets spanning multiple domains (natural imagery, medical imagery, remote sensing, and molecular imagery) and tasks (image classification, visual QA, and object localization). We observe that many-shot ICL, including up to almost 2,000 demonstrating examples, leads to substantial improvements compared to few-shot (<100 examples) ICL across all of the datasets. Further, Gemini 1.5 Pro performance continues to improve log-linearly up to the maximum number of tested examples on many datasets. We also find open-weights multimodal foundation models like Llama 3.2-Vision do not benefit from the demonstrating examples, highlighting an important gap between open and closed multimodal foundation models. Given the high inference costs required for many-shot ICL, we also explore the impact of batching multiple queries in a single API call. We show that batching up to 50 queries can lead to performance improvements under zero-shot and many-shot ICL, with substantial gains in the zero-shot setting on multiple datasets, while drastically reducing per-query cost and latency. Finally, while GPT-4o and Gemini 1.5 Pro achieve similar zero-shot performance across the datasets, Gemini 1.5 Pro learns more quickly than GPT-4o on most datasets. Our results suggest that many-shot ICL could enable users to efficiently adapt multimodal foundation models to new applications and domains. Our codebase is publicly available at https://github.com/stanfordmlgroup/ManyICL .
Authors: Tom S\"uhr, Samira Samadi, Chiara Farronato
Abstract: Machine learning (ML) models are increasingly used in various applications, from recommendation systems in e-commerce to diagnosis prediction in healthcare. In this paper, we present a novel dynamic framework for thinking about the deployment of ML models in a performative, human-ML collaborative system. In our framework, the introduction of ML recommendations changes the data-generating process of human decisions, which are only a proxy to the ground truth and which are then used to train future versions of the model. We show that this dynamic process in principle can converge to different stable points, i.e. where the ML model and the Human+ML system have the same performance. Some of these stable points are suboptimal with respect to the actual ground truth. As a proof of concept, we conduct an empirical user study with 1,408 participants. In the study, humans solve instances of the knapsack problem with the help of machine learning predictions of varying performance. This is an ideal setting because we can identify the actual ground truth, and evaluate the performance of human decisions supported by ML recommendations. We find that for many levels of ML performance, humans can improve upon the ML predictions. We also find that the improvement could be even higher if humans rationally followed the ML recommendations. Finally, we test whether monetary incentives can increase the quality of human decisions, but we fail to find any positive effect. Using our empirical data to approximate our collaborative system suggests that the learning process would dynamically reach an equilibrium performance that is around 92% of the maximum knapsack value. Our results have practical implications for the deployment of ML models in contexts where human decisions may deviate from the indisputable ground truth.
Authors: Zechun Liu, Changsheng Zhao, Igor Fedorov, Bilge Soran, Dhruv Choudhary, Raghuraman Krishnamoorthi, Vikas Chandra, Yuandong Tian, Tijmen Blankevoort
Abstract: Post-training quantization (PTQ) techniques applied to weights, activations, and the KV cache greatly reduce memory usage, latency, and power consumption of Large Language Models (LLMs), but may lead to large quantization errors when outliers are present. Rotating activation or weight matrices helps remove outliers and benefits quantization. In this work, we identify a collection of applicable rotation parameterizations that lead to identical outputs in full-precision Transformer architectures while enhancing quantization accuracy. In addition, we find that some random rotations lead to much better quantization than others, with an up to 13 points difference in downstream zero-shot reasoning performance. As a result, we propose SpinQuant, a novel approach that incorporates learned rotation matrices for optimal quantized network accuracy. With 4-bit quantization of weight, activation, and KV-cache, SpinQuant narrows the accuracy gap on zero-shot reasoning tasks with full precision to merely 2.9 points on the LLaMA-2 7B model, surpassing LLM-QAT by 19.1 points and SmoothQuant by 25.0 points. Furthermore, SpinQuant also outperforms concurrent work QuaRot, which applies random rotations to remove outliers. In particular, for LLaMA-3 8B models that are hard to quantize, SpinQuant reduces the gap to full precision by up to 45.1% relative to QuaRot.
Authors: Itai Shufaro, Nadav Merlis, Nir Weinberger, Shie Mannor
Abstract: In many sequential decision problems, an agent performs a repeated task. He then suffers regret and obtains information that he may use in the following rounds. However, sometimes the agent may also obtain information and avoid suffering regret by querying external sources. We study the trade-off between the information an agent accumulates and the regret it suffers. We invoke information-theoretic methods for obtaining regret lower bounds, that also allow us to easily re-derive several known lower bounds. We introduce the first Bayesian regret lower bounds that depend on the information an agent accumulates. We also prove regret upper bounds using the amount of information the agent accumulates. These bounds show that information measured in bits, can be traded off for regret, measured in reward. Finally, we demonstrate the utility of these bounds in improving the performance of a question-answering task with large language models, allowing us to obtain valuable insights.
Authors: Nikola Zubi\'c, Federico Sold\'a, Aurelio Sulser, Davide Scaramuzza
Abstract: Despite their successes, deep learning models struggle with tasks requiring complex reasoning and function composition. We present a theoretical and empirical investigation into the limitations of Structured State Space Models (SSMs) and Transformers in such tasks. We prove that one-layer SSMs cannot efficiently perform function composition over large domains without impractically large state sizes, and even with Chain-of-Thought prompting, they require a number of steps that scale unfavorably with the complexity of the function composition. Multi-layer SSMs are constrained by log-space computational capacity, limiting their reasoning abilities. Our experiments corroborate these theoretical findings. Evaluating models on tasks including various function composition settings, multi-digit multiplication, dynamic programming, and Einstein's puzzle, we find significant performance degradation even with advanced prompting techniques. Models often resort to shortcuts, leading to compounding errors. These findings highlight fundamental barriers within current deep learning architectures rooted in their computational capacities. We underscore the need for innovative solutions to transcend these constraints and achieve reliable multi-step reasoning and compositional task-solving, which is critical for advancing toward general artificial intelligence.
Authors: Kiyoung Seong, Seonghyun Park, Seonghwan Kim, Woo Youn Kim, Sungsoo Ahn
Abstract: Understanding transition pathways between meta-stable states in molecular systems is crucial to advance material design and drug discovery. However, unbiased molecular dynamics simulations are computationally infeasible due to the high energy barriers separating these states. Although recent machine learning techniques offer potential solutions, they are often limited to simple systems or rely on collective variables (CVs) derived from costly domain expertise. In this paper, we introduce a novel approach that trains diffusion path samplers (DPS) for transition path sampling (TPS) without the need for CVs. We recast the problem as an amortized sampling of the target path measure, minimizing the log-variance divergence between the path measure induced by our DPS and the target path measure. To ensure scalability for high-dimensional tasks, we introduce (1) a new off-policy training objective based on learning control variates with replay buffers and (2) a scale-based equivariant parameterization of the bias forces. We evaluate our approach, coined TPS-DPS, on a synthetic double-well potential and three peptides: Alanine Dipeptide, Polyproline Helix, and Chignolin. Results show that our approach produces more realistic and diverse transition pathways compared to existing baselines.
Authors: Jieun Yook, Junpyo Seo, Joon Huh, Han Joon Byun, Byung-ro Moon
Abstract: We propose a new transformer model for the Traveling Salesman Problem (TSP) called CycleFormer. We identified distinctive characteristics that need to be considered when applying a conventional transformer model to TSP and aimed to fully incorporate these elements into the TSP-specific transformer. Unlike the token sets in typical language models, which are limited and static, the token (node) set in TSP is unlimited and dynamic. To exploit this fact to the fullest, we equated the encoder output with the decoder linear layer and directly connected the context vector of the encoder to the decoder encoding. Additionally, we added a positional encoding to the encoder tokens that reflects the two-dimensional nature of TSP, and devised a circular positional encoding for the decoder tokens that considers the cyclic properties of a tour. By incorporating these ideas, CycleFormer outperforms state-of-the-art (SOTA) transformer models for TSP from TSP-50 to TSP-500. Notably, on TSP-500, the optimality gap was reduced by approximately 2.8 times, from 3.09% to 1.10%, compared to the existing SOTA. The code will be made available at https://github.com/Giventicket/CycleFormer.
Authors: Byoungwoo Park, Jungwon Choi, Sungbin Lim, Juho Lee
Abstract: Recent advancements in diffusion models and diffusion bridges primarily focus on finite-dimensional spaces, yet many real-world problems necessitate operations in infinite-dimensional function spaces for more natural and interpretable formulations. In this paper, we present a theory of stochastic optimal control (SOC) tailored to infinite-dimensional spaces, aiming to extend diffusion-based algorithms to function spaces. Specifically, we demonstrate how Doob's $h$-transform, the fundamental tool for constructing diffusion bridges, can be derived from the SOC perspective and expanded to infinite dimensions. This expansion presents a challenge, as infinite-dimensional spaces typically lack closed-form densities. Leveraging our theory, we establish that solving the optimal control problem with a specific objective function choice is equivalent to learning diffusion-based generative models. We propose two applications: (1) learning bridges between two infinite-dimensional distributions and (2) generative models for sampling from an infinite-dimensional distribution. Our approach proves effective for diverse problems involving continuous function space representations, such as resolution-free images, time-series data, and probability density functions.
Authors: Fengdi Che, Chenjun Xiao, Jincheng Mei, Bo Dai, Ramki Gummadi, Oscar A Ramirez, Christopher K Harris, A. Rupam Mahmood, Dale Schuurmans
Abstract: We prove that the combination of a target network and over-parameterized linear function approximation establishes a weaker convergence condition for bootstrapped value estimation in certain cases, even with off-policy data. Our condition is naturally satisfied for expected updates over the entire state-action space or learning with a batch of complete trajectories from episodic Markov decision processes. Notably, using only a target network or an over-parameterized model does not provide such a convergence guarantee. Additionally, we extend our results to learning with truncated trajectories, showing that convergence is achievable for all tasks with minor modifications, akin to value truncation for the final states in trajectories. Our primary result focuses on temporal difference estimation for prediction, providing high-probability value estimation error bounds and empirical analysis on Baird's counterexample and a Four-room task. Furthermore, we explore the control setting, demonstrating that similar convergence conditions apply to Q-learning.
Authors: Sushant Agarwal, Amit Deshpande
Abstract: Fair classification and fair representation learning are two important problems in supervised and unsupervised fair machine learning, respectively. Fair classification asks for a classifier that maximizes accuracy on a given data distribution subject to fairness constraints. Fair representation maps a given data distribution over the original feature space to a distribution over a new representation space such that all classifiers over the representation satisfy fairness. In this paper, we examine the power of randomization in both these problems to minimize the loss of accuracy that results when we impose fairness constraints. Previous work on fair classification has characterized the optimal fair classifiers on a given data distribution that maximize accuracy subject to fairness constraints, e.g., Demographic Parity (DP), Equal Opportunity (EO), and Predictive Equality (PE). We refine these characterizations to demonstrate when the optimal randomized fair classifiers can surpass their deterministic counterparts in accuracy. We also show how the optimal randomized fair classifier that we characterize can be obtained as a solution to a convex optimization problem. Recent work has provided techniques to construct fair representations for a given data distribution such that any classifier over this representation satisfies DP. However, the classifiers on these fair representations either come with no or weak accuracy guarantees when compared to the optimal fair classifier on the original data distribution. Extending our ideas for randomized fair classification, we improve on these works, and construct DP-fair, EO-fair, and PE-fair representations that have provably optimal accuracy and suffer no accuracy loss compared to the optimal DP-fair, EO-fair, and PE-fair classifiers respectively on the original data distribution.
Authors: Dexiong Chen, Till Hendrik Schulz, Karsten Borgwardt
Abstract: Message-passing graph neural networks (GNNs) excel at capturing local relationships but struggle with long-range dependencies in graphs. In contrast, graph transformers (GTs) enable global information exchange but often oversimplify the graph structure by representing graphs as sets of fixed-length vectors. This work introduces a novel architecture that overcomes the shortcomings of both approaches by combining the long-range information of random walks with local message passing. By treating random walks as sequences, our architecture leverages recent advances in sequence models to effectively capture long-range dependencies within these walks. Based on this concept, we propose a framework that offers (1) more expressive graph representations through random walk sequences, (2) the ability to utilize any sequence model for capturing long-range dependencies, and (3) the flexibility by integrating various GNN and GT architectures. Our experimental evaluations demonstrate that our approach achieves significant performance improvements on 19 graph and node benchmark datasets, notably outperforming existing methods by up to 13\% on the PascalVoc-SP and COCO-SP datasets. The code is available at https://github.com/BorgwardtLab/NeuralWalker.
Authors: Binchuan Qi
Abstract: In this paper, we explore bounds on the expected risk when using deep neural networks for supervised classification from an information theoretic perspective. Firstly, we introduce model risk and fitting error, which are derived from further decomposing the empirical risk. Model risk represents the expected value of the loss under the model's predicted probabilities and is exclusively dependent on the model. Fitting error measures the disparity between the empirical risk and model risk. Then, we derive the upper bound on fitting error, which links the back-propagated gradient and the model's parameter count with the fitting error. Furthermore, we demonstrate that the generalization errors are bounded by the classification uncertainty, which is characterized by both the smoothness of the distribution and the sample size. Based on the bounds on fitting error and generalization, by utilizing the triangle inequality, we establish an upper bound on the expected risk. This bound is applied to provide theoretical explanations for overparameterization, non-convex optimization and flat minima in deep learning. Finally, empirical verification confirms a significant positive correlation between the derived theoretical bounds and the practical expected risk, thereby affirming the practical relevance of the theoretical findings.
Authors: Roman Belaire, Arunesh Sinha, Pradeep Varakantham
Abstract: Deep Reinforcement Learning (DRL) policies are highly susceptible to adversarial noise in observations, which poses significant risks in safety-critical scenarios. For instance, a self-driving car could experience catastrophic consequences if its sensory inputs about traffic signs are manipulated by an adversary. The core challenge in such situations is that the true state of the environment becomes only partially observable due to these adversarial manipulations. Two key strategies have so far been employed in the literature; the first set of methods focuses on increasing the likelihood that nearby states--those close to the true state--share the same robust actions. The second set of approaches maximize the value for the worst possible true state within the range of adversarially perturbed observations. Although these approaches provide strong robustness against attacks, they tend to be either overly conservative or not generalizable. We hypothesize that the shortcomings of these approaches stem from their failure to explicitly account for partial observability. By making decisions that directly consider this partial knowledge of the true state, we believe it is possible to achieve a better balance between robustness and performance, particularly in adversarial settings. To achieve this, we introduce a novel objective called Adversarial Counterfactual Error (ACoE), which is defined on the beliefs about the underlying true state and naturally balances value optimization with robustness against adversarial attacks, and a theoretically-grounded, scalable surrogate objective Cumulative-ACoE (C-ACoE). Our empirical evaluations demonstrate that our method significantly outperforms current state-of-the-art approaches for addressing adversarial RL challenges, offering a promising direction for better DRL under adversarial conditions.
Authors: Xinyang Liu, Yilin He, Bo Chen, Mingyuan Zhou
Abstract: Diffusion models have excelled in generating natural images and are now being adapted to a variety of data types, including graphs. However, conventional models often rely on Gaussian or categorical diffusion processes, which can struggle to accommodate the mixed discrete and continuous components characteristic of graph data. Graphs typically feature discrete structures and continuous node attributes that often exhibit rich statistical patterns, including sparsity, bounded ranges, skewed distributions, and long-tailed behavior. To address these challenges, we introduce Graph Beta Diffusion (GBD), a generative model specifically designed to handle the diverse nature of graph data. GBD leverages a beta diffusion process, effectively modeling both continuous and discrete elements. Additionally, we propose a modulation technique that enhances the realism of generated graphs by stabilizing critical graph topology while maintaining flexibility for other components. GBD competes strongly with existing models across multiple general and biochemical graph benchmarks, showcasing its ability to capture the intricate balance between discrete and continuous features inherent in real-world graph data. The PyTorch code is available on GitHub.
Authors: Soumen Sikder Shuvo, Ali Aghdaei, Zhuo Feng
Abstract: This paper presents a spectral framework for assessing the generalization and stability of Graph Neural Networks (GNNs) by introducing a Graph Geodesic Distance (GGD) metric. For two different graphs with the same number of nodes, our framework leverages a spectral graph matching procedure to find node correspondence so that the geodesic distance between them can be subsequently computed by solving a generalized eigenvalue problem associated with their Laplacian matrices. For graphs with different sizes, a resistance-based spectral graph coarsening scheme is introduced to reduce the size of the bigger graph while preserving the original spectral properties. We show that the proposed GGD metric can effectively quantify dissimilarities between two graphs by encapsulating their differences in key structural (spectral) properties, such as effective resistances between nodes, cuts, the mixing time of random walks, etc. Through extensive experiments comparing with the state-of-the-art metrics, such as the latest Tree-Mover's Distance (TMD) metric, the proposed GGD metric shows significantly improved performance for stability evaluation of GNNs especially when only partial node features are available.
Authors: Taiqiang Wu, Jiahao Wang, Zhe Zhao, Ngai Wong
Abstract: In this paper, we introduce a subspace-inspired Low-Rank Adaptation (LoRA) method, which is computationally efficient, easy to implement, and readily applicable to large language, multimodal, and diffusion models. Initially, we equivalently decompose the weights of LoRA into two subspaces, and find that simply mixing them can enhance performance. To study such a phenomenon, we revisit it through a fine-grained subspace lens, showing that such modification is equivalent to employing a fixed mixer to fuse the subspaces. To be more flexible, we jointly learn the mixer with the original LoRA weights, and term the method Mixture-of-Subspaces LoRA (MoSLoRA). MoSLoRA consistently outperforms LoRA on tasks in different modalities, including commonsense reasoning, visual instruction tuning, and subject-driven text-to-image generation, demonstrating its effectiveness and robustness. Codes are available at https://github.com/wutaiqiang/MoSLoRA.
Authors: Shengyuan Hu, Yiwei Fu, Zhiwei Steven Wu, Virginia Smith
Abstract: Machine unlearning is a promising approach to mitigate undesirable memorization of training data in LLMs. However, in this work we show that existing approaches for unlearning in LLMs are surprisingly susceptible to a simple set of targeted relearning attacks. With access to only a small and potentially loosely related set of data, we find that we can "jog" the memory of unlearned models to reverse the effects of unlearning. For example, we show that relearning on public medical articles can lead an unlearned LLM to output harmful knowledge about bioweapons, and relearning general wiki information about the book series Harry Potter can force the model to output verbatim memorized text. We formalize this unlearning-relearning pipeline, explore the attack across three popular unlearning benchmarks, and discuss future directions and guidelines that result from our study.
Authors: Thomas Hehn, Markus Peschl, Tribhuvanesh Orekondy, Arash Behboodi, Johann Brehmer
Abstract: Modelling the propagation of electromagnetic wireless signals is critical for designing modern communication systems. Wireless ray tracing simulators model signal propagation based on the 3D geometry and other scene parameters, but their accuracy is fundamentally limited by underlying modelling assumptions and correctness of parameters. In this work, we introduce Wi-GATr, a fully-learnable neural simulation surrogate designed to predict the channel observations based on scene primitives (e.g., surface mesh, antenna position and orientation). Recognizing the inherently geometric nature of these primitives, Wi-GATr leverages an equivariant Geometric Algebra Transformer that operates on a tokenizer specifically tailored for wireless simulation. We evaluate our approach on a range of tasks (i.e., signal strength and delay spread prediction, receiver localization, and geometry reconstruction) and find that Wi-GATr is accurate, fast, sample-efficient, and robust to symmetry-induced transformations. Remarkably, we find our results also translate well to the real world: Wi-GATr demonstrates more than 35% lower error than hybrid techniques, and 70% lower error than a calibrated wireless tracer.
Authors: Shengbo Gong, Juntong Ni, Noveen Sachdeva, Carl Yang, Wei Jin
Abstract: Graph condensation (GC) is an emerging technique designed to learn a significantly smaller graph that retains the essential information of the original graph. This condensed graph has shown promise in accelerating graph neural networks while preserving performance comparable to those achieved with the original, larger graphs. Additionally, this technique facilitates downstream applications like neural architecture search and deepens our understanding of redundancies in large graphs. Despite the rapid development of GC methods, particularly for node classification, a unified evaluation framework is still lacking to systematically compare different GC methods or clarify key design choices for improving their effectiveness. To bridge these gaps, we introduce \textbf{GC4NC}, a comprehensive framework for evaluating diverse GC methods on node classification across multiple dimensions including performance, efficiency, privacy preservation, denoising ability, NAS effectiveness, and transferability. Our systematic evaluation offers novel insights into how condensed graphs behave and the critical design choices that drive their success. These findings pave the way for future advancements in GC methods, enhancing both performance and expanding their real-world applications. Our code is available at \url{https://github.com/Emory-Melody/GraphSlim/tree/main/benchmark}.
URLs: https://github.com/Emory-Melody/GraphSlim/tree/main/benchmark
Authors: Wenyu Du, Shuang Cheng, Tongxu Luo, Zihan Qiu, Zeyu Huang, Ka Chun Cheung, Reynold Cheng, Jie Fu
Abstract: Language models (LMs) exhibit impressive performance and generalization capabilities. However, LMs struggle with the persistent challenge of catastrophic forgetting, which undermines their long-term sustainability in continual learning (CL). Existing approaches usually address the issue by incorporating old task data or task-wise inductive bias into LMs. However, old data and accurate task information are often unavailable or costly to collect, hindering the availability of current CL approaches for LMs. To address this limitation, we introduce $\textbf{MIGU}$ ($\textbf{M}$agn$\textbf{I}$tude-based $\textbf{G}$radient $\textbf{U}$pdating for continual learning), a rehearsal-free and task-label-free method that only updates the model parameters with large magnitudes of output in LMs' linear layers. MIGU is based on our observation that the L1-normalized magnitude distribution of the output in LMs' linear layers is different when the LM models deal with different task data. By imposing this simple constraint on the gradient update process, we can leverage the inherent behaviors of LMs, thereby unlocking their innate CL abilities. Our experiments demonstrate that MIGU is universally applicable to all three LM architectures (T5, RoBERTa, and Llama2), delivering state-of-the-art or on-par performance across continual finetuning and continual pre-training settings on four CL benchmarks. For example, MIGU brings a 15.2% average accuracy improvement over conventional parameter-efficient finetuning baselines in a 15-task CL benchmark. MIGU can also seamlessly integrate with all three existing CL types to further enhance performance. Code is available at https://github.com/wenyudu/MIGU.
Authors: Yifan Chen, Daniel Zhengyu Huang, Jiaoyang Huang, Sebastian Reich, Andrew M. Stuart
Abstract: In this paper, we study efficient approximate sampling for probability distributions known up to normalization constants. We specifically focus on a problem class arising in Bayesian inference for large-scale inverse problems in science and engineering applications. The computational challenges we address with the proposed methodology are: (i) the need for repeated evaluations of expensive forward models; (ii) the potential existence of multiple modes; and (iii) the fact that gradient of, or adjoint solver for, the forward model might not be feasible. While existing Bayesian inference methods meet some of these challenges individually, we propose a framework that tackles all three systematically. Our approach builds upon the Fisher-Rao gradient flow in probability space, yielding a dynamical system for probability densities that converges towards the target distribution at a uniform exponential rate. This rapid convergence is advantageous for the computational burden outlined in (i). We apply Gaussian mixture approximations with operator splitting techniques to simulate the flow numerically; the resulting approximation can capture multiple modes thus addressing (ii). Furthermore, we employ the Kalman methodology to facilitate a derivative-free update of these Gaussian components and their respective weights, addressing the issue in (iii). The proposed methodology results in an efficient derivative-free sampler flexible enough to handle multi-modal distributions: Gaussian Mixture Kalman Inversion (GMKI). The effectiveness of GMKI is demonstrated both theoretically and numerically in several experiments with multimodal target distributions, including proof-of-concept and two-dimensional examples, as well as a large-scale application: recovering the Navier-Stokes initial condition from solution data at positive times.
Authors: Yun Dai, Tejas Dharamsi, Byron Hsu, Tao Song, Hamed Firooz
Abstract: Training extremely large language models (LLMs) with billions of parameters is a computationally intensive task that pushes the limits of current data parallel training systems. While techniques like ZeRO++ have enabled efficient distributed training of such giant models on inexpensive low-bandwidth clusters, they can suffer from convergence issues due to potential race conditions in the hierarchical partitioning (hpZ) scheme employed to reduce cross-machine communication. In this work, we first show how these race conditions cause instability when training models with billions of parameters. We then propose a modification to the partitioning algorithm that addresses these convergence challenges while maintaining competitive training efficiency. Empirical evaluation on training the multi-billion parameters Falcon Models and Llama-2 models demonstrates the updated algorithm's ability to achieve reliable convergence on these massive models, where stock ZeRO++ hpZ fails to converge. The updated algorithm enables robust training of larger models with 98\% throughput and model training speed improvement without sacrificing the quality of convergence.
Authors: Yi-Chen Li, Fuxiang Zhang, Wenjie Qiu, Lei Yuan, Chengxing Jia, Zongzhang Zhang, Yang Yu, Bo An
Abstract: Large Language Models (LLMs), trained on a large amount of corpus, have demonstrated remarkable abilities. However, it may not be sufficient to directly apply open-source LLMs like Llama to certain real-world scenarios, since most of them are trained for \emph{general} purposes. Thus, the demands for customizing publicly available LLMs emerge, but are currently under-studied. In this work, we consider customizing pre-trained LLMs with new human preferences. Specifically, the LLM should not only meet the new preference but also preserve its original capabilities after customization. Drawing inspiration from the observation that human preference can be expressed as a reward model, we propose to cast LLM customization as optimizing the sum of two reward functions, one of which (denoted as $r_1$) was used to pre-train the LLM while the other (denoted as $r_2$) characterizes the new human preference. The obstacle here is that both reward functions are unknown, making the application of modern reinforcement learning methods infeasible. Thanks to the residual Q-learning framework, we can restore the customized LLM with the pre-trained LLM and the \emph{residual Q-function} without the reward function $r_1$. Moreover, we find that for a fixed pre-trained LLM, the reward function $r_2$ can be derived from the residual Q-function, enabling us to directly learn the residual Q-function from the new human preference data upon the Bradley-Terry model. We name our method Q-Adapter as it introduces an adapter module to approximate the residual Q-function for customizing the pre-trained LLM towards the new preference. Experiments based on the Llama-3.1 model on the DSP dataset and HH-RLHF dataset illustrate the superior effectiveness of Q-Adapter on both retaining existing knowledge and learning new preferences. Code is available at \url{https://github.com/mansicer/Q-Adapter}.
Authors: Kai Biegun, Rares Dolga, Jake Cunningham, David Barber
Abstract: Linear recurrent neural networks, such as State Space Models (SSMs) and Linear Recurrent Units (LRUs), have recently shown state-of-the-art performance on long sequence modelling benchmarks. Despite their success, their empirical performance is not well understood and they come with a number of drawbacks, most notably their complex initialisation and normalisation schemes. In this work, we address some of these issues by proposing RotRNN -- a linear recurrent model which utilises the convenient properties of rotation matrices. We show that RotRNN provides a simple and efficient model with a robust normalisation procedure, and a practical implementation that remains faithful to its theoretical derivation. RotRNN also achieves competitive performance to state-of-the-art linear recurrent models on several long sequence modelling datasets.
Authors: Guoqiang Zhang, J. P. Lewis, W. B. Kleijn
Abstract: Various reversible deep neural networks (DNN) models have been proposed to reduce memory consumption in the training process. However, almost all existing reversible DNNs either require special non-standard architectures or are constructed by modifying existing DNN architectures considerably to enable reversibility. In this work we present the BDIA-transformer, which is an exact bit-level reversible transformer that uses an unchanged standard architecture for inference. The basic idea is to first treat each transformer block as the Euler integration approximation for solving an ordinary differential equation (ODE) and then incorporate the technique of bidirectional integration approximation (BDIA) into the neural architecture, together with activation quantization to make it exactly bit-level reversible. In the training process, we let a hyper-parameter $\gamma$ in BDIA-transformer randomly take one of the two values $\{0.5, -0.5\}$ per training sample per transformer block for averaging every two consecutive integration approximations. As a result, BDIA-transformer can be viewed as training an ensemble of ODE solvers parameterized by a set of binary random variables, which regularizes the model and results in improved validation accuracy. Lightweight side information per transformer block is required to be stored in the forward process to account for binary quantization loss to enable exact bit-level reversibility. In the inference procedure, the expectation $\mathbb{E}(\gamma)=0$ is taken to make the resulting architectures of BDIA-transformer identical to transformers up to activation quantization. Our experiments in both image classification and language translation show that BDIA-transformers outperform their conventional counterparts significantly in terms of validation performance while also requiring considerably less training memory.
Authors: Zhicheng Yang, Yiwei Wang, Yinya Huang, Zhijiang Guo, Wei Shi, Xiongwei Han, Liang Feng, Linqi Song, Xiaodan Liang, Jing Tang
Abstract: Large language models (LLMs) have exhibited their problem-solving abilities in mathematical reasoning. Solving realistic optimization (OPT) problems in application scenarios requires advanced and applied mathematics ability. However, current OPT benchmarks that merely solve linear programming are far from complex realistic situations. In this work, we propose OptiBench, a benchmark for End-to-end optimization problem-solving with human-readable inputs and outputs. OptiBench contains rich optimization problems, including linear and nonlinear programming with or without tabular data, which can comprehensively evaluate LLMs' solving ability. In our benchmark, LLMs are required to call a code solver to provide precise numerical answers. Furthermore, to alleviate the data scarcity for optimization problems, and to bridge the gap between open-source LLMs on a small scale (e.g., Llama-3-8b) and closed-source LLMs (e.g., GPT-4), we further propose a data synthesis method namely ReSocratic. Unlike general data synthesis methods that proceed from questions to answers, \ReSocratic first incrementally synthesizes formatted optimization demonstration with mathematical formulations step by step and then back-translates the generated demonstrations into questions. Based on this, we synthesize the ReSocratic-29k dataset. We further conduct supervised fine-tuning with ReSocratic-29k on multiple open-source models. Experimental results show that ReSocratic-29k significantly improves the performance of open-source models.
Authors: Kaifu Wang, Efthymia Tsamoura, Dan Roth
Abstract: *Multi-Instance Partial Label Learning* (MI-PLL) is a weakly-supervised learning setting encompassing *partial label learning*, *latent structural learning*, and *neurosymbolic learning*. Unlike supervised learning, in MI-PLL, the inputs to the classifiers at training-time are tuples of instances $\mathbf{x}$. At the same time, the supervision signal is generated by a function $\sigma$ over the (hidden) gold labels of $\mathbf{x}$. In this work, we make multiple contributions towards addressing a problem that hasn't been studied so far in the context of MI-PLL: that of characterizing and mitigating *learning imbalances*, i.e., major differences in the errors occurring when classifying instances of different classes (aka *class-specific risks*). In terms of theory, we derive class-specific risk bounds for MI-PLL, while making minimal assumptions. Our theory reveals a unique phenomenon: that $\sigma$ can greatly impact learning imbalances. This result is in sharp contrast with previous research on supervised and weakly-supervised learning, which only studies learning imbalances under the prism of data imbalances. On the practical side, we introduce a technique for estimating the marginal of the hidden labels using only MI-PLL data. Then, we introduce algorithms that mitigate imbalances at training- and testing-time, by treating the marginal of the hidden labels as a constraint. We demonstrate the effectiveness of our techniques using strong baselines from neurosymbolic and long-tail learning, suggesting performance improvements of up to 14\%.
Authors: Jiawei Huang, Vinzenz Thoma, Zebang Shen, Heinrich H. Nax, Niao He
Abstract: Designing incentives for an adapting population is a ubiquitous problem in a wide array of economic applications and beyond. In this work, we study how to design additional rewards to steer multi-agent systems towards desired policies \emph{without} prior knowledge of the agents' underlying learning dynamics. Motivated by the limitation of existing works, we consider a new and general category of learning dynamics called \emph{Markovian agents}. We introduce a model-based non-episodic Reinforcement Learning (RL) formulation for our steering problem. Importantly, we focus on learning a \emph{history-dependent} steering strategy to handle the inherent model uncertainty about the agents' learning dynamics. We introduce a novel objective function to encode the desiderata of achieving a good steering outcome with reasonable cost. Theoretically, we identify conditions for the existence of steering strategies to guide agents to the desired policies. Complementing our theoretical contributions, we provide empirical algorithms to approximately solve our objective, which effectively tackles the challenge in learning history-dependent strategies. We demonstrate the efficacy of our algorithms through empirical evaluations.
Authors: Ayush Kaushal, Tejas Vaidhya, Arnab Kumar Mondal, Tejas Pandey, Aaryan Bhagat, Irina Rish
Abstract: Rapid advancements in GPU computational power has outpaced memory capacity and bandwidth growth, creating bottlenecks in Large Language Model (LLM) inference. Post-training quantization is the leading method for addressing memory-related bottlenecks in LLM inference, but it suffers from significant performance degradation below 4-bit precision. This paper addresses these challenges by investigating the pretraining of low-bitwidth models specifically Ternary Language Models (TriLMs) as an alternative to traditional floating-point models (FloatLMs) and their post-training quantized versions (QuantLMs). We present Spectra LLM suite, the first open suite of LLMs spanning multiple bit-widths, including FloatLMs, QuantLMs, and TriLMs, ranging from 99M to 3.9B parameters trained on 300B tokens. Our comprehensive evaluation demonstrates that TriLMs offer superior scaling behavior in terms of model size (in bits). Surprisingly, at scales exceeding one billion parameters, TriLMs consistently outperform their QuantLM and FloatLM counterparts for a given bit size across various benchmarks. Notably, the 3.9B parameter TriLM matches the performance of the FloatLM 3.9B across all benchmarks, despite having fewer bits than FloatLM 830M. Overall, this research provides valuable insights into the feasibility and scalability of low-bitwidth language models, paving the way for the development of more efficient LLMs. To enhance understanding of low-bitwidth models, we are releasing 500+ intermediate checkpoints of the Spectra suite at \href{https://github.com/NolanoOrg/SpectraSuite}{https://github.com/NolanoOrg/SpectraSuite}.
URLs: https://github.com/NolanoOrg/SpectraSuite, https://github.com/NolanoOrg/SpectraSuite
Authors: Georgii Novikov, Ivan Oseledets
Abstract: The scaling of neural networks with increasing data and model sizes necessitates the development of more efficient deep learning algorithms. A significant challenge in neural network training is the memory footprint associated with activation tensors, particularly in pointwise nonlinearity layers that traditionally save the entire input tensor for the backward pass, leading to substantial memory consumption. In this paper, we propose a modification to the handling of activation tensors in pointwise nonlinearity layers. Our method involves saving the output tensor instead of the input tensor during the forward pass. Since the subsequent layer typically also saves its input tensor, this approach reduces the total memory required by storing only one tensor between layers instead of two. This optimization is especially beneficial for transformer-based architectures like GPT, BERT, Mistral, and Llama. To enable this approach, we utilize the inverse function of the nonlinearity during the backward pass. As the inverse cannot be computed analytically for most nonlinearities, we construct accurate approximations using simpler functions. Experimental results demonstrate that our method significantly reduces memory usage without affecting training accuracy or computational performance. Our implementation is provided as a drop-in replacement for standard nonlinearity layers in the PyTorch framework, facilitating easy adoption without requiring architectural modifications.
Authors: M. Sajid, A. Quadir, M. Tanveer
Abstract: The random vector functional link (RVFL) network is well-regarded for its strong generalization capabilities in the field of machine learning. However, its inherent dependencies on the square loss function make it susceptible to noise and outliers. Furthermore, the calculation of RVFL's unknown parameters necessitates matrix inversion of the entire training sample, which constrains its scalability. To address these challenges, we propose the Wave-RVFL, an RVFL model incorporating the wave loss function. We formulate and solve the proposed optimization problem of the Wave-RVFL using the adaptive moment estimation (Adam) algorithm in a way that successfully eliminates the requirement for matrix inversion and significantly enhances scalability. The Wave-RVFL exhibits robustness against noise and outliers by preventing over-penalization of deviations, thereby maintaining a balanced approach to managing noise and outliers. The proposed Wave-RVFL model is evaluated on multiple UCI datasets, both with and without the addition of noise and outliers, across various domains and sizes. Empirical results affirm the superior performance and robustness of the Wave-RVFL compared to baseline models, establishing it as a highly effective and scalable classification solution. The source codes and the Supplementary Material are available at https://github.com/mtanveer1/Wave-RVFL.
Authors: Avisek Naug, Antonio Guillen, Ricardo Luna, Vineet Gundecha, Desik Rengarajan, Sahand Ghorbanpour, Sajad Mousavi, Ashwin Ramesh Babu, Dejan Markovikj, Lekhapriya D Kashyap, Soumyendu Sarkar
Abstract: Machine learning has driven an exponential increase in computational demand, leading to massive data centers that consume significant amounts of energy and contribute to climate change. This makes sustainable data center control a priority. In this paper, we introduce SustainDC, a set of Python environments for benchmarking multi-agent reinforcement learning (MARL) algorithms for data centers (DC). SustainDC supports custom DC configurations and tasks such as workload scheduling, cooling optimization, and auxiliary battery management, with multiple agents managing these operations while accounting for the effects of each other. We evaluate various MARL algorithms on SustainDC, showing their performance across diverse DC designs, locations, weather conditions, grid carbon intensity, and workload requirements. Our results highlight significant opportunities for improvement of data center operations using MARL algorithms. Given the increasing use of DC due to AI, SustainDC provides a crucial platform for the development and benchmarking of advanced algorithms essential for achieving sustainable computing and addressing other heterogeneous real-world challenges.
Authors: Zeju Qiu, Weiyang Liu, Haiwen Feng, Zhen Liu, Tim Z. Xiao, Katherine M. Collins, Joshua B. Tenenbaum, Adrian Weller, Michael J. Black, Bernhard Sch\"olkopf
Abstract: Against the backdrop of enthusiasm for large language models (LLMs), there is an urgent need to scientifically assess their capabilities and shortcomings. This is nontrivial in part because it is difficult to find tasks which the models have not encountered during training. Utilizing symbolic graphics programs, we propose a domain well-suited to test multiple spatial-semantic reasoning skills of LLMs. Popular in computer graphics, these programs procedurally generate visual data. While LLMs exhibit impressive skills in general program synthesis and analysis, symbolic graphics programs offer a new layer of evaluation: they allow us to test an LLM's ability to answer different-grained semantic-level questions of the images or 3D geometries without a vision encoder. To semantically understand the symbolic programs, LLMs would need to possess the ability to "imagine" and reason how the corresponding graphics content would look with only the symbolic description. We use this task to evaluate LLMs by creating a large benchmark for the semantic visual understanding of symbolic graphics programs, built procedurally with minimal human effort. Particular emphasis is placed on transformations of images that leave the image level semantics invariant while introducing significant changes to the underlying program. We evaluate commercial and open-source LLMs on our benchmark to assess their ability to reason about visual output of programs, finding that LLMs considered stronger at reasoning generally perform better. Lastly, we introduce a novel method to improve this ability -- Symbolic Instruction Tuning (SIT), in which the LLM is finetuned with pre-collected instruction data on symbolic graphics programs. Interestingly, we find that SIT not only improves LLM's understanding on symbolic programs, but it also improves general reasoning ability on various other benchmarks.
Authors: Gus Kristiansen, Mark Sandler, Andrey Zhmoginov, Nolan Miller, Anirudh Goyal, Jihwan Lee, Max Vladymyrov
Abstract: In modern deep learning, the models are learned by applying gradient updates using an optimizer, which transforms the updates based on various statistics. Optimizers are often hand-designed and tuning their hyperparameters is a big part of the training process. Learned optimizers have shown some initial promise, but are generally unsuccessful as a general optimization mechanism applicable to every problem. In this work we explore a different direction: instead of learning general optimizers, we instead specialize them to a specific training environment. We propose a novel optimizer technique that learns a layer-specific linear combination of update directions provided by a set of base optimizers, effectively adapting its strategy to the specific model and dataset. When evaluated on image classification tasks, this specialized optimizer significantly outperforms both traditional off-the-shelf methods such as Adam, as well as existing general learned optimizers. Moreover, it demonstrates robust generalization with respect to model initialization, evaluating on unseen datasets, and training durations beyond its meta-training horizon.
Authors: Lucia Pezzetti, Stefano Favaro, Stefano Peluchetti
Abstract: Bayesian Neural Networks represent a fascinating confluence of deep learning and probabilistic reasoning, offering a compelling framework for understanding uncertainty in complex predictive models. In this paper, we investigate the use of the preconditioned Crank-Nicolson algorithm and its Langevin version to sample from the reparametrised posterior distribution of the weights as the widths of Bayesian Neural Networks grow larger. In addition to being robust in the infinite-dimensional setting, we prove that the acceptance probabilities of the proposed methods approach 1 as the width of the network increases, independently of any stepsize tuning. Moreover, we examine and compare how the mixing speeds of the underdamped Langevin Monte Carlo, the preconditioned Crank-Nicolson and the preconditioned Crank-Nicolson Langevin samplers are influenced by changes in the network width in some real-world cases. Our findings suggest that, in wide Bayesian Neural Networks configurations, the preconditioned Crank-Nicolson method allows for more efficient sampling of the reparametrised posterior distribution, as evidenced by a higher effective sample size and improved diagnostic results compared with the other analysed algorithms.
Authors: Ruoyu Wang, Lina Yao
Abstract: Disentangled Representation Learning aims to improve the explainability of deep learning methods by training a data encoder that identifies semantically meaningful latent variables in the data generation process. Nevertheless, there is no consensus regarding a universally accepted definition for the objective of disentangled representation learning. In particular, there is a considerable amount of discourse regarding whether should the latent variables be mutually independent or not. In this paper, we first investigate these arguments on the interrelationships between latent variables by establishing a conceptual bridge between Epistemology and Disentangled Representation Learning. Then, inspired by these interdisciplinary concepts, we introduce a two-level latent space framework to provide a general solution to the prior arguments on this issue. Finally, we propose a novel method for disentangled representation learning by employing an integration of mutual information constraint and independence constraint within the Generative Adversarial Network (GAN) framework. Experimental results demonstrate that our proposed method consistently outperforms baseline approaches in both quantitative and qualitative evaluations. The method exhibits strong performance across multiple commonly used metrics and demonstrates a great capability in disentangling various semantic factors, leading to an improved quality of controllable generation, which consequently benefits the explainability of the algorithm.
Authors: Matthias Pirlet, Adrien Bolland, Gilles Louppe, Damien Ernst
Abstract: The Unit Commitment (UC) problem is a key optimization task in power systems to forecast the generation schedules of power units over a finite time period by minimizing costs while meeting demand and technical constraints. However, many parameters required by the UC problem are unknown, such as the costs. In this work, we estimate these unknown costs using simulation-based inference on an illustrative UC problem, which provides an approximated posterior distribution of the parameters given observed generation schedules and demands. Our results highlight that the learned posterior distribution effectively captures the underlying distribution of the data, providing a range of possible values for the unknown parameters given a past observation. This posterior allows for the estimation of past costs using observed past generation schedules, enabling operators to better forecast future costs and make more robust generation scheduling forecasts. We present avenues for future research to address overconfidence in posterior estimation, enhance the scalability of the methodology and apply it to more complex UC problems modeling the network constraints and renewable energy sources.
Authors: Tim Lawson, Lucy Farnik, Conor Houghton, Laurence Aitchison
Abstract: Sparse autoencoders (SAEs) are a promising approach to interpreting the internal representations of transformer language models. However, SAEs are usually trained separately on each transformer layer, making it difficult to use them to study how information flows across layers. To solve this problem, we introduce the multi-layer SAE (MLSAE): a single SAE trained on the residual stream activation vectors from every transformer layer. Given that the residual stream is understood to preserve information across layers, we expected MLSAE latents to `switch on' at a token position and remain active at later layers. Interestingly, we find that individual latents are often active at a single layer for a given token or prompt, but this layer may differ for different tokens or prompts. We quantify these phenomena by defining a distribution over layers and considering its variance. We find that the variance of the distributions of latent activations over layers is about two orders of magnitude greater when aggregating over tokens compared with a single token. For larger underlying models, the degree to which latents are active at multiple layers increases, which is consistent with the fact that the residual stream activation vectors at adjacent layers become more similar. Finally, we relax the assumption that the residual stream basis is the same at every layer by applying pre-trained tuned-lens transformations, but our findings remain qualitatively similar. Our results represent a new approach to understanding how representations change as they flow through transformers. We release our code to train and analyze MLSAEs at https://github.com/tim-lawson/mlsae.
Authors: Chhavi Yadav, Ruihan Wu, Kamalika Chaudhuri
Abstract: Influence Functions are a standard tool for attributing predictions to training data in a principled manner and are widely used in applications such as data valuation and fairness. In this work, we present realistic incentives to manipulate influence-based attributions and investigate whether these attributions can be \textit{systematically} tampered by an adversary. We show that this is indeed possible for logistic regression models trained on ResNet feature embeddings and standard tabular fairness datasets and provide efficient attacks with backward-friendly implementations. Our work raises questions on the reliability of influence-based attributions in adversarial circumstances. Code is available at : \url{https://github.com/infinite-pursuits/influence-based-attributions-can-be-manipulated}
URLs: https://github.com/infinite-pursuits/influence-based-attributions-can-be-manipulated
Authors: Xinhe Wang, Pingbang Hu, Junwei Deng, Jiaqi W. Ma
Abstract: Data attribution aims to quantify the contribution of individual training data points to the outputs of an AI model, which has been used to measure the value of training data and compensate data providers. Given the impact on financial decisions and compensation mechanisms, a critical question arises concerning the adversarial robustness of data attribution methods. However, there has been little to no systematic research addressing this issue. In this work, we aim to bridge this gap by detailing a threat model with clear assumptions about the adversary's goal and capabilities and proposing principled adversarial attack methods on data attribution. We present two methods, Shadow Attack and Outlier Attack, which generate manipulated datasets to inflate the compensation adversarially. The Shadow Attack leverages knowledge about the data distribution in the AI applications, and derives adversarial perturbations through "shadow training", a technique commonly used in membership inference attacks. In contrast, the Outlier Attack does not assume any knowledge about the data distribution and relies solely on black-box queries to the target model's predictions. It exploits an inductive bias present in many data attribution methods - outlier data points are more likely to be influential - and employs adversarial examples to generate manipulated datasets. Empirically, in image classification and text generation tasks, the Shadow Attack can inflate the data-attribution-based compensation by at least 200%, while the Outlier Attack achieves compensation inflation ranging from 185% to as much as 643%.
Authors: Hongpeng Cao, Yanbing Mao, Yihao Cai, Lui Sha, Marco Caccamo
Abstract: This paper proposes the SeC-Learning Machine: Simplex-enabled safe continual learning for safety-critical autonomous systems. The SeC-learning machine is built on Simplex logic (that is, ``using simplicity to control complexity'') and physics-regulated deep reinforcement learning (Phy-DRL). The SeC-learning machine thus constitutes HP (high performance)-Student, HA (high assurance)-Teacher, and Coordinator. Specifically, the HP-Student is a pre-trained high-performance but not fully verified Phy-DRL, continuing to learn in a real plant to tune the action policy to be safe. In contrast, the HA-Teacher is a mission-reduced, physics-model-based, and verified design. As a complementary, HA-Teacher has two missions: backing up safety and correcting unsafe learning. The Coordinator triggers the interaction and the switch between HP-Student and HA-Teacher. Powered by the three interactive components, the SeC-learning machine can i) assure lifetime safety (i.e., safety guarantee in any continual-learning stage, regardless of HP-Student's success or convergence), ii) address the Sim2Real gap, and iii) learn to tolerate unknown unknowns in real plants. The experiments on a cart-pole system and a real quadruped robot demonstrate the distinguished features of the SeC-learning machine, compared with continual learning built on state-of-the-art safe DRL frameworks with approaches to addressing the Sim2Real gap.
Authors: Mohammad Shojaeifard, Matteo Ferraresso, Alessandro Lucantonio, Mattia Bacca
Abstract: Fibrillar adhesion, observed in animals like beetles, spiders, and geckos, relies on nanoscopic or microscopic fibrils to enhance surface adhesion via 'contact splitting.' This concept has inspired engineering applications across robotics, transportation, and medicine. Recent studies suggest that functional grading of fibril properties can improve adhesion, but this is a complex design challenge that has only been explored in simplified geometries. While machine learning (ML) has gained traction in adhesive design, no previous attempts have targeted fibril-array scale optimization. In this study, we propose an ML-based tool that optimizes the distribution of fibril compliance to maximize adhesive strength. Our tool, featuring two deep neural networks (DNNs), recovers previous design results for simple geometries and introduces novel solutions for complex configurations. The Predictor DNN estimates adhesive strength based on random compliance distributions, while the Designer DNN optimizes compliance for maximum strength using gradient-based optimization. Our method significantly reduces test error and accelerates the optimization process, offering a high-performance solution for designing fibrillar adhesives and micro-architected materials aimed at fracture resistance by achieving equal load sharing (ELS).
Authors: Christopher M. Ackerman
Abstract: Activation engineering is becoming increasingly popular as a means of online control of large language models (LLMs). In this work, I extend the idea of active steering with vectors that represent a behavioral direction of interest to tuning those vectors directly into the model, obviating the need for online control. First, I identify activation vectors related to honesty in an open-source LLM (Llama- 2-13b-chat). Next, I demonstrate that model output can be made more or less honest by adding positive or negative multiples of these vectors to residual stream activations during generation. Then, I show that a similar effect can be achieved by fine-tuning the vectors directly into the model, by use of a dual loss function based on the cosine similarity of residual stream activations to the vectors combined with a standard token-based loss ("representation tuning"). Finally, I compare the generations in response to honesty-probing prompts from the resulting models to those from models fine-tuned with a token-based loss alone, and to those from the untuned model subjected to online steering. Overall, fine-tuning the vectors into the models using the cosine similarity plus token loss showed a stronger effect than online steering, and generalized better than using the standard loss, suggesting the potential utility of this approach as a safety measure. Code and data are available at https://github.com/cma1114/representation_tuning; tuned models are available at https://huggingface.co/collections/cackerman/ representation-tuning-66da1e5ab41cd1b824687d9f.
URLs: https://github.com/cma1114/representation_tuning;, https://huggingface.co/collections/cackerman/
Authors: Wayne Chi, Ameet Talwalkar, Chris Donahue
Abstract: There has been a surge of interest in language model agents that can navigate virtual environments such as the web or desktop. To navigate such environments, agents benefit from information on the various elements (e.g., buttons, text, or images) present. It remains unclear which element attributes have the greatest impact on agent performance, especially in environments that only provide a graphical representation (i.e., pixels). Here we find that the ordering in which elements are presented to the language model is surprisingly impactful--randomizing element ordering in a webpage degrades agent performance comparably to removing all visible text from an agent's state representation. While a webpage provides a hierarchical ordering of elements, there is no such ordering when parsing elements directly from pixels. Moreover, as tasks become more challenging and models more sophisticated, our experiments suggest that the impact of ordering increases. Finding an effective ordering is non-trivial. We investigate the impact of various element ordering methods in web and desktop environments. We find that dimensionality reduction provides a viable ordering for pixel-only environments. We train a UI element detection model to derive elements from pixels and apply our findings to an agent benchmark--OmniACT--where we only have access to pixels. Our method completes more than two times as many tasks on average relative to the previous state-of-the-art.
Authors: Yao Ni, Shan Zhang, Piotr Koniusz
Abstract: Parameter-Efficient Fine-Tuning (PEFT) effectively adapts pre-trained vision transformers to downstream tasks. However, the optimization for tasks performance often comes at the cost of generalizability in fine-tuned models. To address this issue, we theoretically connect smaller weight gradient norms during training and larger datasets to the improved model generalization. Motivated by this connection, we propose reducing gradient norms for enhanced generalization and aligning fine-tuned model with the pre-trained counterpart to retain knowledge from large-scale pre-training data. Yet, naive alignment does not guarantee gradient reduction and can potentially cause gradient explosion, complicating efforts to manage gradients. To address such issues, we propose PACE, marrying generalization of PArameter-efficient fine-tuning with Consistency rEgularization. We perturb features learned from the adapter with the multiplicative noise and ensure the fine-tuned model remains consistent for same sample under different perturbations. Theoretical analysis shows that PACE not only implicitly regularizes gradients for enhanced generalization, but also implicitly aligns the fine-tuned and pre-trained models to retain knowledge. Experimental evidence supports our theories. PACE outperforms existing PEFT methods in four visual adaptation tasks: VTAB-1k, FGVC, few-shot learning and domain adaptation. Code will be available at https://github.com/MaxwellYaoNi/PACE
Authors: Jakub {\L}ucki, Boyi Wei, Yangsibo Huang, Peter Henderson, Florian Tram\`er, Javier Rando
Abstract: Large language models are finetuned to refuse questions about hazardous knowledge, but these protections can often be bypassed. Unlearning methods aim at completely removing hazardous capabilities from models and make them inaccessible to adversaries. This work challenges the fundamental differences between unlearning and traditional safety post-training from an adversarial perspective. We demonstrate that existing jailbreak methods, previously reported as ineffective against unlearning, can be successful when applied carefully. Furthermore, we develop a variety of adaptive methods that recover most supposedly unlearned capabilities. For instance, we show that finetuning on 10 unrelated examples or removing specific directions in the activation space can recover most hazardous capabilities for models edited with RMU, a state-of-the-art unlearning method. Our findings challenge the robustness of current unlearning approaches and question their advantages over safety training.
Authors: Yueyang Liu, Lance Kennedy, Hossein Amiri, Andreas Z\"ufle
Abstract: Human trajectory anomaly detection has become increasingly important across a wide range of applications, including security surveillance and public health. However, existing trajectory anomaly detection methods are primarily focused on vehicle-level traffic, while human-level trajectory anomaly detection remains under-explored. Since human trajectory data is often very sparse, machine learning methods have become the preferred approach for identifying complex patterns. However, concerns regarding potential biases and the robustness of these models have intensified the demand for more transparent and explainable alternatives. In response to these challenges, our research focuses on developing a lightweight anomaly detection model specifically designed to detect anomalies in human trajectories. We propose a Neural Collaborative Filtering approach to model and predict normal mobility. Our method is designed to model users' daily patterns of life without requiring prior knowledge, thereby enhancing performance in scenarios where data is sparse or incomplete, such as in cold start situations. Our algorithm consists of two main modules. The first is the collaborative filtering module, which applies collaborative filtering to model normal mobility of individual humans to places of interest. The second is the neural module, responsible for interpreting the complex spatio-temporal relationships inherent in human trajectory data. To validate our approach, we conducted extensive experiments using simulated and real-world datasets comparing to numerous state-of-the-art trajectory anomaly detection approaches.
Authors: Daniel Jiwoong Im, Kevin Zhang, Nakul Verma, Kyunghyun Cho
Abstract: Existing causal inference (CI) models are limited to primarily handling low-dimensional confounders and singleton actions. We propose an autoregressive (AR) CI framework capable of handling complex confounders and sequential actions common in modern applications. We accomplish this by {\em sequencification}, transforming data from an underlying causal diagram into a sequence of tokens. This approach not only enables training with data generated from any DAG but also extends existing CI capabilities to accommodate estimating several statistical quantities using a {\em single} model. We can directly predict interventional probabilities, simplifying inference and enhancing outcome prediction accuracy. We demonstrate that an AR model adapted for CI is efficient and effective in various complex applications such as navigating mazes, playing chess endgames, and evaluating the impact of certain keywords on paper acceptance rates.
Authors: Ruizhe Shi, Runlong Zhou, Simon S. Du
Abstract: Direct Preference Optimization (DPO) has emerged as a stable, scalable, and efficient solution for language model alignment. Despite its empirical success, the $\textit{optimization}$ properties, particularly the impact of samplers on its convergence rates, remain underexplored. In this paper, we provide a rigorous analysis of DPO's $\textit{convergence rates}$ with different sampling strategies under the exact gradient setting, revealing a surprising separation: uniform sampling achieves $\textit{linear}$ convergence, while our proposed online sampler achieves $\textit{quadratic}$ convergence. We further adapt the sampler to practical settings by incorporating posterior distributions and $\textit{logit mixing}$, demonstrating significant improvements over previous approaches. On Safe-RLHF dataset, our method exhibits a $4.5$% improvement over vanilla DPO and a $3.0$% improvement over on-policy DPO; on Iterative-Prompt, our approach outperforms vanilla DPO, on-policy DPO, and Hybrid GSHF by over $4.2$%. Our results not only offer insights into the theoretical standing of DPO but also pave the way for potential algorithm designs in the future.
Authors: Momin Ahmad Khan, Yasra Chandio, Fatima Muhammad Anwar
Abstract: Data heterogeneity among Federated Learning (FL) users poses a significant challenge, resulting in reduced global model performance. The community has designed various techniques to tackle this issue, among which Knowledge Distillation (KD)-based techniques are common. While these techniques effectively improve performance under high heterogeneity, they inadvertently cause higher accuracy degradation under model poisoning attacks (known as attack amplification). This paper presents a case study to reveal this critical vulnerability in KD-based FL systems. We show why KD causes this issue through empirical evidence and use it as motivation to design a hybrid distillation technique. We introduce a novel algorithm, Hybrid Knowledge Distillation for Robust and Accurate FL (HYDRA-FL), which reduces the impact of attacks in attack scenarios by offloading some of the KD loss to a shallow layer via an auxiliary classifier. We model HYDRA-FL as a generic framework and adapt it to two KD-based FL algorithms, FedNTD and MOON. Using these two as case studies, we demonstrate that our technique outperforms baselines in attack settings while maintaining comparable performance in benign settings.
Authors: Petar Veli\v{c}kovi\'c, Christos Perivolaropoulos, Federico Barbero, Razvan Pascanu
Abstract: A key property of reasoning systems is the ability to make sharp decisions on their input data. For contemporary AI systems, a key carrier of sharp behaviour is the softmax function, with its capability to perform differentiable query-key lookups. It is a common belief that the predictive power of networks leveraging softmax arises from "circuits" which sharply perform certain kinds of computations consistently across many diverse inputs. However, for these circuits to be robust, they would need to generalise well to arbitrary valid inputs. In this paper, we dispel this myth: even for tasks as simple as finding the maximum key, any learned circuitry must disperse as the number of items grows at test time. We attribute this to a fundamental limitation of the softmax function to robustly approximate sharp functions, prove this phenomenon theoretically, and propose adaptive temperature as an ad-hoc technique for improving the sharpness of softmax at inference time.
Authors: Ioannis Mavromatis, Stefano De Feo, Aftab Khan
Abstract: This paper presents Federated Learning with Adaptive Monitoring and Elimination (FLAME), a novel solution capable of detecting and mitigating concept drift in Federated Learning (FL) Internet of Things (IoT) environments. Concept drift poses significant challenges for FL models deployed in dynamic and real-world settings. FLAME leverages an FL architecture, considers a real-world FL pipeline, and proves capable of maintaining model performance and accuracy while addressing bandwidth and privacy constraints. Introducing various features and extensions on previous works, FLAME offers a robust solution to concept drift, significantly reducing computational load and communication overhead. Compared to well-known lightweight mitigation methods, FLAME demonstrates superior performance in maintaining high F1 scores and reducing resource utilisation in large-scale IoT deployments, making it a promising approach for real-world applications.
Authors: Shengyu Feng, Xiang Kong, Shuang Ma, Aonan Zhang, Dong Yin, Chong Wang, Ruoming Pang, Yiming Yang
Abstract: Augmenting the multi-step reasoning abilities of Large Language Models (LLMs) has been a persistent challenge. Recently, verification has shown promise in improving solution consistency by evaluating generated outputs. However, current verification approaches suffer from sampling inefficiencies, requiring a large number of samples to achieve satisfactory performance. Additionally, training an effective verifier often depends on extensive process supervision, which is costly to acquire. In this paper, we address these limitations by introducing a novel verification method based on Twisted Sequential Monte Carlo (TSMC). TSMC sequentially refines its sampling effort to focus exploration on promising candidates, resulting in more efficient generation of high-quality solutions. We apply TSMC to LLMs by estimating the expected future rewards at partial solutions. This approach results in a more straightforward training target that eliminates the need for step-wise human annotations. We empirically demonstrate the advantages of our method across multiple math benchmarks, and also validate our theoretical analysis of both our approach and existing verification methods.
Authors: Thao Nguyen, Kuan-Hao Huang, Ge Liu, Martin D. Burke, Ying Diao, Heng Ji
Abstract: We introduce Functional Group-Aware Representations for Small Molecules (FARM), a novel foundation model designed to bridge the gap between SMILES, natural language, and molecular graphs. The key innovation of FARM lies in its functional group-aware tokenization, which directly incorporates functional group information into the representations. This strategic reduction in tokenization granularity is intentionally aligned with key drivers of functional properties (i.e., functional groups), enhancing the model's understanding of chemical language. By expanding the chemical lexicon, FARM more effectively bridges SMILES and natural language, ultimately advancing the model's capacity to predict molecular properties. FARM also represents molecules from two perspectives: by using masked language modeling to capture atom-level features and by employing graph neural networks to encode the whole molecule topology. By leveraging contrastive learning, FARM aligns these two views of representations into a unified molecular embedding. We rigorously evaluate FARM on the MoleculeNet dataset, where it achieves state-of-the-art performance on 10 out of 12 tasks. These results highlight FARM's potential to improve molecular representation learning, with promising applications in drug discovery and pharmaceutical research.
Authors: Hongkang Li, Meng Wang, Songtao Lu, Xiaodong Cui, Pin-Yu Chen
Abstract: Chain-of-Thought (CoT) is an efficient prompting method that enables the reasoning ability of large language models by augmenting the query using multiple examples with multiple intermediate steps. Despite the empirical success, the theoretical understanding of how to train a Transformer to achieve the CoT ability remains less explored. This is primarily due to the technical challenges involved in analyzing the nonconvex optimization on nonlinear attention models. To the best of our knowledge, this work provides the first theoretical study of training Transformers with nonlinear attention to obtain the CoT generalization capability so that the resulting model can inference on unseen tasks when the input is augmented by examples of the new task. We first quantify the required training samples and iterations to train a Transformer model towards CoT ability. We then prove the success of its CoT generalization on unseen tasks with distribution-shifted testing data. Moreover, we theoretically characterize the conditions for an accurate reasoning output by CoT even when the provided reasoning examples contain noises and are not always accurate. In contrast, in-context learning (ICL), which can be viewed as one-step CoT without intermediate steps, may fail to provide an accurate output when CoT does. These theoretical findings are justified through experiments.
Authors: Tianchi Xie, Jiangning Zhu, Guozu Ma, Minzhi Lin, Wei Chen, Weikai Yang, Shixia Liu
Abstract: Sample selection improves the efficiency and effectiveness of machine learning models by providing informative and representative samples. Typically, samples can be modeled as a sample graph, where nodes are samples and edges represent their similarities. Most existing methods are based on local information, such as the training difficulty of samples, thereby overlooking global information, such as connectivity patterns. This oversight can result in suboptimal selection because global information is crucial for ensuring that the selected samples well represent the structural properties of the graph. To address this issue, we employ structural entropy to quantify global information and losslessly decompose it from the whole graph to individual nodes using the Shapley value. Based on the decomposition, we present $\textbf{S}$tructural-$\textbf{E}$ntropy-based sample $\textbf{S}$election ($\textbf{SES}$), a method that integrates both global and local information to select informative and representative samples. SES begins by constructing a $k$NN-graph among samples based on their similarities. It then measures sample importance by combining structural entropy (global metric) with training difficulty (local metric). Finally, SES applies importance-biased blue noise sampling to select a set of diverse and representative samples. Comprehensive experiments on three learning scenarios -- supervised learning, active learning, and continual learning -- clearly demonstrate the effectiveness of our method.
Authors: Anthony Costarelli, Mat Allen, Severin Field
Abstract: As Large Language Models (LLMs) become increasingly integrated into our daily lives, the potential harms from deceptive behavior underlie the need for faithfully interpreting their decision-making. While traditional probing methods have shown some effectiveness, they remain best for narrowly scoped tasks while more comprehensive explanations are still necessary. To this end, we investigate meta-models-an architecture using a "meta-model" that takes activations from an "input-model" and answers natural language questions about the input-model's behaviors. We evaluate the meta-model's ability to generalize by training them on selected task types and assessing their out-of-distribution performance in deceptive scenarios. Our findings show that meta-models generalize well to out-of-distribution tasks and point towards opportunities for future research in this area.
Authors: Zakaria Mhammedi
Abstract: In this paper, we introduce a new projection-free algorithm for Online Convex Optimization (OCO) with a state-of-the-art regret guarantee among separation-based algorithms. Existing projection-free methods based on the classical Frank-Wolfe algorithm achieve a suboptimal regret bound of $O(T^{3/4})$, while more recent separation-based approaches guarantee a regret bound of $O(\kappa \sqrt{T})$, where $\kappa$ denotes the asphericity of the feasible set, defined as the ratio of the radii of the containing and contained balls. However, for ill-conditioned sets, $\kappa$ can be arbitrarily large, potentially leading to poor performance. Our algorithm achieves a regret bound of $\widetilde{O}(\sqrt{dT} + \kappa d)$, while requiring only $\widetilde{O}(1)$ calls to a separation oracle per round. Crucially, the main term in the bound, $\widetilde{O}(\sqrt{d T})$, is independent of $\kappa$, addressing the limitations of previous methods. Additionally, as a by-product of our analysis, we recover the $O(\kappa \sqrt{T})$ regret bound of existing OCO algorithms with a more straightforward analysis and improve the regret bound for projection-free online exp-concave optimization. Finally, for constrained stochastic convex optimization, we achieve a state-of-the-art convergence rate of $\widetilde{O}(\sigma/\sqrt{T} + \kappa d/T)$, where $\sigma$ represents the noise in the stochastic gradients, while requiring only $\widetilde{O}(1)$ calls to a separation oracle per iteration.
Authors: Duy M. H. Nguyen, Nghiem T. Diep, Trung Q. Nguyen, Hoang-Bao Le, Tai Nguyen, Tien Nguyen, TrungTin Nguyen, Nhat Ho, Pengtao Xie, Roger Wattenhofer, James Zhou, Daniel Sonntag, Mathias Niepert
Abstract: State-of-the-art medical multi-modal large language models (med-MLLM), like LLaVA-Med or BioMedGPT, leverage instruction-following data in pre-training. However, those models primarily focus on scaling the model size and data volume to boost performance while mainly relying on the autoregressive learning objectives. Surprisingly, we reveal that such learning schemes might result in a weak alignment between vision and language modalities, making these models highly reliant on extensive pre-training datasets - a significant challenge in medical domains due to the expensive and time-consuming nature of curating high-quality instruction-following instances. We address this with LoGra-Med, a new multi-graph alignment algorithm that enforces triplet correlations across image modalities, conversation-based descriptions, and extended captions. This helps the model capture contextual meaning, handle linguistic variability, and build cross-modal associations between visuals and text. To scale our approach, we designed an efficient end-to-end learning scheme using black-box gradient estimation, enabling faster LLaMa 7B training. Our results show LoGra-Med matches LLAVA-Med performance on 600K image-text pairs for Medical VQA and significantly outperforms it when trained on 10% of the data. For example, on VQA-RAD, we exceed LLAVA-Med by 20.13% and nearly match the 100% pre-training score (72.52% vs. 72.64%). We also surpass SOTA methods like BiomedGPT on visual chatbots and RadFM on zero-shot image classification with VQA, highlighting the effectiveness of multi-graph alignment.
Authors: Jinghao Shi, Xiang Shen, Kaili Zhao, Xuedong Wang, Vera Wen, Zixuan Wang, Yifan Wu, Zhixin Zhang
Abstract: Dense features, customized for different business scenarios, are essential in short video classification. However, their complexity, specific adaptation requirements, and high computational costs make them resource-intensive and less accessible during online inference. Consequently, these dense features are categorized as `Privileged Dense Features'.Meanwhile, end-to-end multi-modal models have shown promising results in numerous computer vision tasks. In industrial applications, prioritizing end-to-end multi-modal features, can enhance efficiency but often leads to the loss of valuable information from historical privileged dense features. To integrate both features while maintaining efficiency and manageable resource costs, we present Confidence-aware Privileged Feature Distillation (CPFD), which empowers features of an end-to-end multi-modal model by adaptively distilling privileged features during training. Unlike existing privileged feature distillation (PFD) methods, which apply uniform weights to all instances during distillation, potentially causing unstable performance across different business scenarios and a notable performance gap between teacher model (Dense Feature enhanced multimodal-model DF-X-VLM) and student model (multimodal-model only X-VLM), our CPFD leverages confidence scores derived from the teacher model to adaptively mitigate the performance variance with the student model. We conducted extensive offline experiments on five diverse tasks demonstrating that CPFD improves the video classification F1 score by 6.76% compared with end-to-end multimodal-model (X-VLM) and by 2.31% with vanilla PFD on-average. And it reduces the performance gap by 84.6% and achieves results comparable to teacher model DF-X-VLM. The effectiveness of CPFD is further substantiated by online experiments, and our framework has been deployed in production systems for over a dozen models.
Authors: Ethan Kane Waters, Carla Chia-ming Chen, Mostafa Rahimi Azghadi
Abstract: Disease detection in sugarcane, particularly the identification of asymptomatic infectious diseases such as Ratoon Stunting Disease (RSD), is critical for effective crop management. This study employed various machine learning techniques to detect the presence of RSD in different sugarcane varieties, using vegetation indices derived from freely available satellite-based spectral data. Our results show that the Support Vector Machine with a Radial Basis Function Kernel (SVM-RBF) was the most effective algorithm, achieving classification accuracy between 85.64% and 96.55%, depending on the variety. Gradient Boosting and Random Forest also demonstrated high performance achieving accuracy between 83.33% to 96.55%, while Logistic Regression and Quadratic Discriminant Analysis showed variable results across different varieties. The inclusion of sugarcane variety and vegetation indices was important in the detection of RSD. This agreed with what was identified in the current literature. Our study highlights the potential of satellite-based remote sensing as a cost-effective and efficient method for large-scale sugarcane disease detection alternative to traditional manual laboratory testing methods.
Authors: Yizhen Xu, Numair Sani, AmirEmad Ghassami, Ilya Shpitser
Abstract: In many applications, researchers are interested in the direct and indirect causal effects of a treatment or exposure on an outcome of interest. Mediation analysis offers a rigorous framework for identifying and estimating these causal effects. For binary treatments, efficient estimators for the direct and indirect effects are presented by Tchetgen Tchetgen and Shpitser (2012) based on the influence function of the parameter of interest. These estimators possess desirable properties such as multiple-robustness and asymptotic normality while allowing for slower than root-n rates of convergence for the nuisance parameters. However, in settings involving continuous treatments, these influence function-based estimators are not readily applicable without making strong parametric assumptions. In this work, utilizing a kernel-smoothing approach, we propose an estimator suitable for settings with continuous treatments inspired by the influence function-based estimator of Tchetgen Tchetgen and Shpitser (2012). Our proposed approach employs cross-fitting, relaxing the smoothness requirements on the nuisance functions and allowing them to be estimated at slower rates than the target parameter. Additionally, similar to influence function-based estimators, our proposed estimator is multiply robust and asymptotically normal, allowing for inference in settings where parametric assumptions may not be justified.
Authors: Jiachen Lian, Chunlei Zhang, Gopala Krishna Anumanchipalli, Dong Yu
Abstract: In this paper, we propose a novel unsupervised text-to-speech acoustic model training scheme, named UTTS, which does not require text-audio pairs. UTTS is a multi-speaker speech synthesizer that supports zero-shot voice cloning, it is developed from a perspective of disentangled speech representation learning. The framework offers a flexible choice of a speaker's duration model, timbre feature (identity) and content for TTS inference. We leverage recent advancements in self-supervised speech representation learning as well as speech synthesis front-end techniques for system development. Specifically, we employ our recently formulated Conditional Disentangled Sequential Variational Auto-encoder (C-DSVAE) as the backbone UTTS AM, which offers well-structured content representations given unsupervised alignment (UA) as condition during training. For UTTS inference, we utilize a lexicon to map input text to the phoneme sequence, which is expanded to the frame-level forced alignment (FA) with a speaker-dependent duration model. Then, we develop an alignment mapping module that converts FA to UA. Finally, the C-DSVAE, serving as the self-supervised TTS AM, takes the predicted UA and a target speaker embedding to generate the mel spectrogram, which is ultimately converted to waveform with a neural vocoder. We show how our method enables speech synthesis without using a paired TTS corpus in AM development stage. Experiments demonstrate that UTTS can synthesize speech of high naturalness and intelligibility measured by human and objective evaluations. Audio samples are available at our demo page https://neurtts.github.io/utts\_demo/.
Authors: Lijun Ding, Zhen Qin, Liwei Jiang, Jinxin Zhou, Zhihui Zhu
Abstract: This paper studies the problem of recovering a low-rank matrix from several noisy random linear measurements. We consider the setting where the rank of the ground-truth matrix is unknown a priori and use an objective function built from a rank-overspecified factored representation of the matrix variable, where the global optimal solutions overfit and do not correspond to the underlying ground truth. We then solve the associated nonconvex problem using gradient descent with small random initialization. We show that as long as the measurement operators satisfy the restricted isometry property (RIP) with its rank parameter scaling with the rank of the ground-truth matrix rather than scaling with the overspecified matrix rank, gradient descent iterations are on a particular trajectory towards the ground-truth matrix and achieve nearly information-theoretically optimal recovery when it is stopped appropriately. We then propose an efficient stopping strategy based on the common hold-out method and show that it detects a nearly optimal estimator provably. Moreover, experiments show that the proposed validation approach can also be efficiently used for image restoration with deep image prior, which over-parameterizes an image with a deep network.
Authors: Filippo Fabiani, Alberto Bemporad
Abstract: To identify a stationary action profile for a population of competitive agents, each executing private strategies, we introduce a novel active-learning scheme where a centralized external observer (or entity) can probe the agents' reactions and recursively update simple local parametric estimates of the action-reaction mappings. Under very general working assumptions (not even assuming that a stationary profile exists), sufficient conditions are established to assess the asymptotic properties of the proposed active learning methodology so that, if the parameters characterizing the action-reaction mappings converge, a stationary action profile is achieved. Such conditions hence act also as certificates for the existence of such a profile. Extensive numerical simulations involving typical competitive multi-agent control and decision-making problems illustrate the practical effectiveness of the proposed learning-based approach.
Authors: Ryan Cory-Wright, Andr\'es G\'omez
Abstract: Given a high-dimensional covariate matrix and a response vector, ridge-regularized sparse linear regression selects a subset of features that explains the relationship between covariates and the response in an interpretable manner. To select the sparsity and robustness of linear regressors, techniques like k-fold cross-validation are commonly used for hyperparameter tuning. However, cross-validation substantially increases the computational cost of sparse regression as it requires solving many mixed-integer optimization problems (MIOs). Additionally, validation metrics often serve as noisy estimators of test set errors, with different hyperparameter combinations leading to models with different noise levels. Therefore, optimizing over these metrics is vulnerable to out-of-sample disappointment, especially in underdetermined settings. To improve upon this state of affairs, we make two key contributions. First, motivated by the generalization theory literature, we propose selecting hyperparameters that minimize a weighted sum of a cross-validation metric and a model's output stability, thus reducing the risk of poor out-of-sample performance. Second, we leverage ideas from the mixed-integer optimization literature to obtain computationally tractable relaxations of k-fold cross-validation metrics and the output stability of regressors, facilitating hyperparameter selection after solving fewer MIOs. These relaxations result in an efficient cyclic coordinate descent scheme, achieving lower validation errors than via traditional methods such as grid search. On synthetic datasets, our confidence adjustment procedure improves out-of-sample performance by 2%-5% compared to minimizing the k-fold error alone. On 13 real-world datasets, our confidence adjustment procedure reduces test set error by 2%, on average.
Authors: Sully F. Chen, Zhicheng Guo, Cheng Ding, Xiao Hu, Cynthia Rudin
Abstract: Rapid, reliable, and accurate interpretation of medical time-series signals is crucial for high-stakes clinical decision-making. Deep learning methods offered unprecedented performance in medical signal processing but at a cost: they were compute-intensive and lacked interpretability. We propose Sparse Mixture of Learned Kernels (SMoLK), an interpretable architecture for medical time series processing. SMoLK learns a set of lightweight flexible kernels that form a single-layer sparse neural network, providing not only interpretability, but also efficiency, robustness, and generalization to unseen data distributions. We introduce a parameter reduction techniques to reduce the size of SMoLK's networks while maintaining performance. We test SMoLK on two important tasks common to many consumer wearables: photoplethysmography (PPG) artifact detection and atrial fibrillation detection from single-lead electrocardiograms (ECGs). We find that SMoLK matches the performance of models orders of magnitude larger. It is particularly suited for real-time applications using low-power devices, and its interpretability benefits high-stakes situations.
Authors: Suresh Guttikonda, Jan Achterhold, Haolong Li, Joschka Boedecker, Joerg Stueckler
Abstract: In autonomous navigation settings, several quantities can be subject to variations. Terrain properties such as friction coefficients may vary over time depending on the location of the robot. Also, the dynamics of the robot may change due to, e.g., different payloads, changing the system's mass, or wear and tear, changing actuator gains or joint friction. An autonomous agent should thus be able to adapt to such variations. In this paper, we develop a novel probabilistic, terrain- and robot-aware forward dynamics model, termed TRADYN, which is able to adapt to the above-mentioned variations. It builds on recent advances in meta-learning forward dynamics models based on Neural Processes. We evaluate our method in a simulated 2D navigation setting with a unicycle-like robot and different terrain layouts with spatially varying friction coefficients. In our experiments, the proposed model exhibits lower prediction error for the task of long-horizon trajectory prediction, compared to non-adaptive ablation models. We also evaluate our model on the downstream task of navigation planning, which demonstrates improved performance in planning control-efficient paths by taking robot and terrain properties into account.
Authors: Rabia Asghar, Sanjay Kumar, Paul Hynds, Abeera Mahfooz
Abstract: Human blood primarily comprises plasma, red blood cells, white blood cells, and platelets. It plays a vital role in transporting nutrients to different organs, where it stores essential health-related data about the human body. Blood cells are utilized to defend the body against diverse infections, including fungi, viruses, and bacteria. Hence, blood analysis can help physicians assess an individual's physiological condition. Blood cells have been sub-classified into eight groups: Neutrophils, eosinophils, basophils, lymphocytes, monocytes, immature granulocytes (promyelocytes, myelocytes, and metamyelocytes), erythroblasts, and platelets or thrombocytes on the basis of their nucleus, shape, and cytoplasm. Traditionally, pathologists and hematologists in laboratories have examined these blood cells using a microscope before manually classifying them. The manual approach is slower and more prone to human error. Therefore, it is essential to automate this process. In our paper, transfer learning with CNN pre-trained models. VGG16, VGG19, ResNet-50, ResNet-101, ResNet-152, InceptionV3, MobileNetV2, and DenseNet-20 applied to the PBC dataset's normal DIB. The overall accuracy achieved with these models lies between 91.375 and 94.72%. Hence, inspired by these pre-trained architectures, a model has been proposed to automatically classify the ten types of blood cells with increased accuracy. A novel CNN-based framework has been presented to improve accuracy. The proposed CNN model has been tested on the PBC dataset normal DIB. The outcomes of the experiments demonstrate that our CNN-based framework designed for blood cell classification attains an accuracy of 99.91% on the PBC dataset. Our proposed convolutional neural network model performs competitively when compared to earlier results reported in the literature.
Authors: Zaber Ibn Abdul Hakim, Rasman Mubtasim Swargo, Muhammad Abdullah Adnan
Abstract: Understanding intricate and fast-paced movements of body parts is essential for the recognition and translation of sign language. The inclusion of additional information intended to identify and locate the moving body parts has been an interesting research topic recently. However, previous works on using multi-modal information raise concerns such as sub-optimal multi-modal feature merging method, or the model itself being too computationally heavy. In our work, we have addressed such issues and used a plugin module based on cross-attention to properly attend to each modality with another. Moreover, we utilized 2-stage training to remove the dependency of separate feature extractors for additional modalities in an end-to-end approach, which reduces the concern about computational complexity. Besides, our additional cross-attention plugin module is very lightweight which doesn't add significant computational overhead on top of the original baseline. We have evaluated the performance of our approaches on the RWTH-PHOENIX-2014 dataset for sign language recognition and the RWTH-PHOENIX-2014T dataset for the sign language translation task. Our approach reduced the WER by 0.9 on the recognition task and increased the BLEU-4 scores by 0.8 on the translation task.
Authors: Leying Zhang, Yao Qian, Linfeng Yu, Heming Wang, Hemin Yang, Long Zhou, Shujie Liu, Yanmin Qian
Abstract: Diffusion models have gained attention in speech enhancement tasks, providing an alternative to conventional discriminative methods. However, research on target speech extraction under multi-speaker noisy conditions remains relatively unexplored. Moreover, the superior quality of diffusion methods typically comes at the cost of slower inference speed. In this paper, we introduce the Discriminative Diffusion model for Target Speech Extraction (DDTSE). We apply the same forward process as diffusion models and utilize the reconstruction loss similar to discriminative methods. Furthermore, we devise a two-stage training strategy to emulate the inference process during model training. DDTSE not only works as a standalone system, but also can further improve the performance of discriminative models without additional retraining. Experimental results demonstrate that DDTSE not only achieves higher perceptual quality but also accelerates the inference process by 3 times compared to the conventional diffusion model.
Authors: Felix Frederik Zimmermann, Andreas Kofler
Abstract: We present a novel learned image reconstruction method for accelerated cardiac MRI with multiple receiver coils based on deep convolutional neural networks (CNNs) and algorithm unrolling. In contrast to many existing learned MR image reconstruction techniques that necessitate coil-sensitivity map (CSM) estimation as a distinct network component, our proposed approach avoids explicit CSM estimation. Instead, it implicitly captures and learns to exploit the inter-coil relationships of the images. Our method consists of a series of novel learned image and k-space blocks with shared latent information and adaptation to the acquisition parameters by feature-wise modulation (FiLM), as well as coil-wise data-consistency (DC) blocks. Our method achieved PSNR values of 34.89 and 35.56 and SSIM values of 0.920 and 0.942 in the cine track and mapping track validation leaderboard of the MICCAI STACOM CMRxRecon Challenge, respectively, ranking 4th among different teams at the time of writing. Code will be made available at https://github.com/fzimmermann89/CMRxRecon
Authors: Charuta Pethe, Bach Pham, Felix D Childress, Yunting Yin, Steven Skiena
Abstract: Recent advances in text-to-speech have made it possible to generate natural-sounding audio from text. However, audiobook narrations involve dramatic vocalizations and intonations by the reader, with greater reliance on emotions, dialogues, and descriptions in the narrative. Using our dataset of 93 aligned book-audiobook pairs, we present improved models for prosody prediction properties (pitch, volume, and rate of speech) from narrative text using language modeling. Our predicted prosody attributes correlate much better with human audiobook readings than results from a state-of-the-art commercial TTS system: our predicted pitch shows a higher correlation with human reading for 22 out of the 24 books, while our predicted volume attribute proves more similar to human reading for 23 out of the 24 books. Finally, we present a human evaluation study to quantify the extent that people prefer prosody-enhanced audiobook readings over commercial text-to-speech systems.
Authors: Hongfu Liu, Hengguan Huang, Ye Wang
Abstract: Acoustic foundation models, fine-tuned for Automatic Speech Recognition (ASR), suffer from performance degradation in wild acoustic test settings when deployed in real-world scenarios. Stabilizing online Test-Time Adaptation (TTA) under these conditions remains an open and unexplored question. Existing wild vision TTA methods often fail to handle speech data effectively due to the unique characteristics of high-entropy speech frames, which are unreliably filtered out even when containing crucial semantic content. Furthermore, unlike static vision data, speech signals follow short-term consistency, requiring specialized adaptation strategies. In this work, we propose a novel wild acoustic TTA method tailored for ASR fine-tuned acoustic foundation models. Our method, Confidence-Enhanced Adaptation, performs frame-level adaptation using a confidence-aware weight scheme to avoid filtering out essential information in high-entropy frames. Additionally, we apply consistency regularization during test-time optimization to leverage the inherent short-term consistency of speech signals. Our experiments on both synthetic and real-world datasets demonstrate that our approach outperforms existing baselines under various wild acoustic test settings, including Gaussian noise, environmental sounds, accent variations, and sung speech.
Authors: Jiaming Ji, Borong Zhang, Jiayi Zhou, Xuehai Pan, Weidong Huang, Ruiyang Sun, Yiran Geng, Yifan Zhong, Juntao Dai, Yaodong Yang
Abstract: Artificial intelligence (AI) systems possess significant potential to drive societal progress. However, their deployment often faces obstacles due to substantial safety concerns. Safe reinforcement learning (SafeRL) emerges as a solution to optimize policies while simultaneously adhering to multiple constraints, thereby addressing the challenge of integrating reinforcement learning in safety-critical scenarios. In this paper, we present an environment suite called Safety-Gymnasium, which encompasses safety-critical tasks in both single and multi-agent scenarios, accepting vector and vision-only input. Additionally, we offer a library of algorithms named Safe Policy Optimization (SafePO), comprising 16 state-of-the-art SafeRL algorithms. This comprehensive library can serve as a validation tool for the research community. By introducing this benchmark, we aim to facilitate the evaluation and comparison of safety performance, thus fostering the development of reinforcement learning for safer, more reliable, and responsible real-world applications. The website of this project can be accessed at https://sites.google.com/view/safety-gymnasium.
Authors: Mingqian Zheng, Jiaxin Pei, Lajanugen Logeswaran, Moontae Lee, David Jurgens
Abstract: Prompting serves as the major way humans interact with Large Language Models (LLM). Commercial AI systems commonly define the role of the LLM in system prompts. For example, ChatGPT uses "You are a helpful assistant" as part of its default system prompt. Despite current practices of adding personas to system prompts, it remains unclear how different personas affect a model's performance on objective tasks. In this study, we present a systematic evaluation of personas in system prompts. We curate a list of 162 roles covering 6 types of interpersonal relationships and 8 domains of expertise. Through extensive analysis of 4 popular families of LLMs and 2,410 factual questions, we demonstrate that adding personas in system prompts does not improve model performance across a range of questions compared to the control setting where no persona is added. Nevertheless, further analysis suggests that the gender, type, and domain of the persona can all influence the resulting prediction accuracies. We further experimented with a list of persona search strategies and found that, while aggregating results from the best persona for each question significantly improves prediction accuracy, automatically identifying the best persona is challenging, with predictions often performing no better than random selection. Overall, our findings suggest that while adding a persona may lead to performance gains in certain settings, the effect of each persona can be largely random. Code and data are available at https://github.com/Jiaxin-Pei/Prompting-with-Social-Roles.
URLs: https://github.com/Jiaxin-Pei/Prompting-with-Social-Roles.
Authors: Zhicheng Yang, Yinya Huang, Jing Xiong, Liang Feng, Xiaodan Liang, Yiwei Wang, Jing Tang
Abstract: Large Language Models prompting, such as using in-context demonstrations, is a mainstream technique for invoking LLMs to perform high-performance and solid complex reasoning (e.g., mathematical reasoning, commonsense reasoning), and has the potential for further human-machine collaborative scientific findings. However, current LLMs are delicate and elusive in prompt words and styles. And there is an unseen gap between LLM understanding and human-written prompts. This paper introduces Alignedcot, an LLM-acquainted prompting technique that includes proficient ``native-speaking'' in in-context learning for the LLMs. Specifically, it achieves consistent and correct step-wise prompts in zero-shot scenarios by progressively probing, refining, and formatting the LLM chain of thoughts so that free from handcrafted few-shot demonstrations while maintaining the prompt quality. We conduct experiments on mathematical reasoning and commonsense reasoning. We find that LLMs with Alignedcot perform significantly superior to them with human-crafted demonstrations. We further apply Alignedcot for rewriting the GSM8K training set, resulting in a GSM8K-Align dataset. We observe its benefits for retrieval augmented generation. The code and data can be found at https://github.com/yangzhch6/AlignedCoT.
Authors: Bingzhi Zhang, Junyu Liu, Xiao-Chuan Wu, Liang Jiang, Quntao Zhuang
Abstract: Understanding the training dynamics of quantum neural networks is a fundamental task in quantum information science with wide impact in physics, chemistry and machine learning. In this work, we show that the late-time training dynamics of quantum neural networks with a quadratic loss function can be described by the generalized Lotka-Volterra equations, which lead to a transcritical bifurcation transition in the dynamics. When the targeted value of loss function crosses the minimum achievable value from above to below, the dynamics evolve from a frozen-kernel dynamics to a frozen-error dynamics, showing a duality between the quantum neural tangent kernel and the total error. In both regions, the convergence towards the fixed point is exponential, while at the critical point becomes polynomial. We provide a non-perturbative analytical theory to explain the transition via a restricted Haar ensemble at late time, when the output state approaches the steady state. Via mapping the Hessian to an effective Hamiltonian, we also identify a linearly vanishing gap at the transition point. Compared with the linear loss function, we show that a quadratic loss function within the frozen-error dynamics enables a speedup in the training convergence. The theory findings are verified experimentally on IBM quantum devices.
Authors: Chinmay Savadikar, Xi Song, Tianfu Wu
Abstract: We present Generative Parameter-Efficient Fine-Tuning (GIFT) for adapting pretrained Transformer backbones on downstream tasks. GIFT learns to generate the fine-tuned weights for a layer directly from its pretrained weights. The GIFT network is parameterized in a minimally-simple way by two linear layers (without bias terms), and is shared by different pretrained layers selected for fine-tuning (e.g., the Query layers), which result in significantly fewer trainable parameters compared to the layer-specific methods like Low-Rank Adapter (LoRA). We also show this formulation bridges parameter-efficient fine-tuning and representation fine-tuning. We perform comprehensive experiments on natural language tasks (commonsense and arithmetic reasoning, instruction tuning, and sequence classification) and computer vision tasks (fine-grained classification). We obtain the best performance and parameter efficiency among baselines on commonsense and arithmetic reasoning, and instruction following using the Llama family of models and on visual recognition benchmarks using Vision Transformers. Notably, compared to LoRA, we obtain 5.7% absolute increase in average accuracy with 14 times reduction of parameters on Commonsense170k using Llama-3 (8B), and 5.4% absolute increase in the win rate with 4 times reduction of parameters using Llama-2 (7B) during instruction tuning. Our GIFT also obtains a slightly higher win rate on instruction tuning than GPT 3.5 (Turbo 1106).
Authors: Matteo Gioele Collu, Tom Janssen-Groesbeek, Stefanos Koffas, Mauro Conti, Stjepan Picek
Abstract: Recently, we have witnessed a rise in the use of Large Language Models (LLMs), especially in applications like chatbots. Safety mechanisms are implemented to prevent improper responses from these chatbots. In this work, we bypass these measures for ChatGPT and Gemini by making them impersonate complex personas with personality characteristics that are not aligned with a truthful assistant. First, we create elaborate biographies of these personas, which we then use in a new session with the same chatbots. Our conversations then follow a role-play style to elicit prohibited responses. Using personas, we show that prohibited responses are provided, making it possible to obtain unauthorized, illegal, or harmful information in both ChatGPT and Gemini. We also introduce several ways of activating such adversarial personas, showing that both chatbots are vulnerable to this attack. With the same principle, we introduce two defenses that push the model to interpret trustworthy personalities and make it more robust against such attacks.
Authors: Yujian Liu, Yang Zhang, Tommi Jaakkola, Shiyu Chang
Abstract: Despite diffusion models' superior capabilities in modeling complex distributions, there are still non-trivial distributional discrepancies between generated and ground-truth images, which has resulted in several notable problems in image generation, including missing object errors in text-to-image generation and low image quality. Existing methods that attempt to address these problems mostly do not tend to address the fundamental cause behind these problems, which is the distributional discrepancies, and hence achieve sub-optimal results. In this paper, we propose a particle filtering framework that can effectively address both problems by explicitly reducing the distributional discrepancies. Specifically, our method relies on a set of external guidance, including a small set of real images and a pre-trained object detector, to gauge the distribution gap, and then design the resampling weight accordingly to correct the gap. Experiments show that our methods can effectively correct missing object errors and improve image quality in various image generation tasks. Notably, our method outperforms the existing strongest baseline by 5% in object occurrence and 1.0 in FID on MS-COCO. Our code is publicly available at https://github.com/UCSB-NLP-Chang/diffusion_resampling.git.
URLs: https://github.com/UCSB-NLP-Chang/diffusion_resampling.git.
Authors: Hisato Komatsu
Abstract: In recent years, simulations of pedestrians using the multi-agent reinforcement learning (MARL) have been studied. This study considered the roads on a grid-world environment, and implemented pedestrians as MARL agents using an echo-state network and the least squares policy iteration method. Under this environment, the ability of these agents to learn to move forward by avoiding other agents was investigated. Specifically, we considered two types of tasks: the choice between a narrow direct route and a broad detour, and the bidirectional pedestrian flow in a corridor. The simulations results indicated that the learning was successful when the density of the agents was not that high.
Authors: Lei Yang
Abstract: We introduce a novel generative model, the Discrete Distribution Networks (DDN), that approximates data distribution using hierarchical discrete distributions. We posit that since the features within a network inherently capture distributional information, enabling the network to generate multiple samples simultaneously, rather than a single output, may offer an effective way to represent distributions. Therefore, DDN fits the target distribution, including continuous ones, by generating multiple discrete sample points. To capture finer details of the target data, DDN selects the output that is closest to the Ground Truth (GT) from the coarse results generated in the first layer. This selected output is then fed back into the network as a condition for the second layer, thereby generating new outputs more similar to the GT. As the number of DDN layers increases, the representational space of the outputs expands exponentially, and the generated samples become increasingly similar to the GT. This hierarchical output pattern of discrete distributions endows DDN with unique property: more general zero-shot conditional generation. We demonstrate the efficacy of DDN and its intriguing properties through experiments on CIFAR-10 and FFHQ. The code is available at https://discrete-distribution-networks.github.io/
Authors: Nicolas Boull\'e, Matthew J. Colbrook
Abstract: We study the convergence of Hermitian Dynamic Mode Decomposition (DMD) to the spectral properties of self-adjoint Koopman operators. Hermitian DMD is a data-driven method that approximates the Koopman operator associated with an unknown nonlinear dynamical system, using discrete-time snapshots. This approach preserves the self-adjointness of the operator in its finite-dimensional approximations. \rev{We prove that, under suitably broad conditions, the spectral measures corresponding to the eigenvalues and eigenfunctions computed by Hermitian DMD converge to those of the underlying Koopman operator}. This result also applies to skew-Hermitian systems (after multiplication by $i$), applicable to generators of continuous-time measure-preserving systems. Along the way, we establish a general theorem on the convergence of spectral measures for finite sections of self-adjoint operators, including those that are unbounded, which is of independent interest to the wider spectral community. We numerically demonstrate our results by applying them to two-dimensional Schr\"odinger equations.
Authors: Rohin Manvi, Samar Khanna, Marshall Burke, David Lobell, Stefano Ermon
Abstract: Large Language Models (LLMs) inherently carry the biases contained in their training corpora, which can lead to the perpetuation of societal harm. As the impact of these foundation models grows, understanding and evaluating their biases becomes crucial to achieving fairness and accuracy. We propose to study what LLMs know about the world we live in through the lens of geography. This approach is particularly powerful as there is ground truth for the numerous aspects of human life that are meaningfully projected onto geographic space such as culture, race, language, politics, and religion. We show various problematic geographic biases, which we define as systemic errors in geospatial predictions. Initially, we demonstrate that LLMs are capable of making accurate zero-shot geospatial predictions in the form of ratings that show strong monotonic correlation with ground truth (Spearman's $\rho$ of up to 0.89). We then show that LLMs exhibit common biases across a range of objective and subjective topics. In particular, LLMs are clearly biased against locations with lower socioeconomic conditions (e.g. most of Africa) on a variety of sensitive subjective topics such as attractiveness, morality, and intelligence (Spearman's $\rho$ of up to 0.70). Finally, we introduce a bias score to quantify this and find that there is significant variation in the magnitude of bias across existing LLMs. Code is available on the project website: https://rohinmanvi.github.io/GeoLLM
Authors: Junjie Chu, Zeyang Sha, Michael Backes, Yang Zhang
Abstract: Significant advancements have recently been made in large language models represented by GPT models. Users frequently have multi-round private conversations with cloud-hosted GPT models for task optimization. Yet, this operational paradigm introduces additional attack surfaces, particularly in custom GPTs and hijacked chat sessions. In this paper, we introduce a straightforward yet potent Conversation Reconstruction Attack. This attack targets the contents of previous conversations between GPT models and benign users, i.e., the benign users' input contents during their interaction with GPT models. The adversary could induce GPT models to leak such contents by querying them with designed malicious prompts. Our comprehensive examination of privacy risks during the interactions with GPT models under this attack reveals GPT-4's considerable resilience. We present two advanced attacks targeting improved reconstruction of past conversations, demonstrating significant privacy leakage across all models under these advanced techniques. Evaluating various defense mechanisms, we find them ineffective against these attacks. Our findings highlight the ease with which privacy can be compromised in interactions with GPT models, urging the community to safeguard against potential abuses of these models' capabilities.
Authors: Ziqiao Shang, Bin Liu, Fengmao Lv, Fei Teng, Tianrui Li
Abstract: Facial action unit (AU) detection has long encountered the challenge of detecting subtle feature differences when AUs activate. Existing methods often rely on encoding pixel-level information of AUs, which not only encodes additional redundant information but also leads to increased model complexity and limited generalizability. Additionally, the accuracy of AU detection is negatively impacted by the class imbalance issue of each AU type, and the presence of noisy and false AU labels. In this paper, we introduce a novel contrastive learning framework aimed for AU detection that incorporates both self-supervised and supervised signals, thereby enhancing the learning of discriminative features for accurate AU detection. To tackle the class imbalance issue, we employ a negative sample re-weighting strategy that adjusts the step size of updating parameters for minority and majority class samples. Moreover, to address the challenges posed by noisy and false AU labels, we employ a sampling technique that encompasses three distinct types of positive sample pairs. This enables us to inject self-supervised signals into the supervised signal, effectively mitigating the adverse effects of noisy labels. Our experimental assessments, conducted on four widely-utilized benchmark datasets (BP4D, DISFA, GFT and Aff-Wild2), underscore the superior performance of our approach compared to state-of-the-art methods of AU detection. Our code is available at \url{https://github.com/Ziqiao-Shang/AUNCE}.
Authors: Jona te Lintelo, Stefanos Koffas, Stjepan Picek
Abstract: Sponge attacks aim to increase the energy consumption and computation time of neural networks. In this work, we present a novel sponge attack called SkipSponge. SkipSponge is the first sponge attack that is performed directly on the parameters of a pre-trained model using only a few data samples. Our experiments show that SkipSponge can successfully increase the energy consumption of image classification models, GANs, and autoencoders requiring fewer samples than the state-of-the-art (Sponge Poisoning). We show that poisoning defenses are ineffective if not adjusted specifically for the defense against SkipSponge (i.e., they decrease target layer bias values). Our work shows that SkipSponge is more effective on the GANs and the autoencoders than Sponge Poisoning. Additionally, SkipSponge is stealthier than Sponge Poisoning as it does not require significant changes in the victim model's weights. Our experiments indicate that SkipSponge can be performed even when an attacker has access to only 1% of the entire dataset and reaches up to 13% energy increase.
Authors: Wenjing Jiang, Vidya A. Chhabria, Sachin S. Sapatnekar
Abstract: Engineering change orders (ECOs) in late stages make minimal design fixes to recover from timing shifts due to excessive IR drops. This paper integrates IR-drop-aware timing analysis and ECO timing optimization using reinforcement learning (RL). The method operates after physical design and power grid synthesis, and rectifies IR-drop-induced timing degradation through gate sizing. It incorporates the Lagrangian relaxation (LR) technique into a novel RL framework, which trains a relational graph convolutional network (R-GCN) agent to sequentially size gates to fix timing violations. The R-GCN agent outperforms a classical LR-only algorithm: in an open 45nm technology, it (a) moves the Pareto front of the delay-power tradeoff curve to the left (b) saves runtime over the prior approaches by running fast inference using trained models, and (c) reduces the perturbation to placement by sizing fewer cells. The RL model is transferable across timing specifications and to unseen designs with fine tuning.
Authors: Kang He, Yinghan Long, Kaushik Roy
Abstract: Prompt-based learning is susceptible to intrinsic bias present in pre-trained language models (LMs), leading to sub-optimal performance in prompt-based zero/few-shot settings. In this work, we propose a null-input prompting method to calibrate intrinsic bias encoded in pre-trained LMs. Different from prior efforts that address intrinsic bias primarily for social fairness and often involve excessive computational cost, our objective is to explore enhancing LMs' performance in downstream zero/few-shot learning while emphasizing the efficiency of intrinsic bias calibration. Specifically, we leverage a diverse set of auto-selected null-meaning inputs generated from GPT-4 to probe intrinsic bias of pre-trained LMs. Utilizing the bias-reflected probability distribution, we formulate a distribution disparity loss for bias calibration, where we exclusively update bias parameters ($0.1\%$ of total parameters) of LMs towards equal probability distribution. Experimental results show that the calibration promotes an equitable starting point for LMs while preserving language modeling abilities. Across a wide range of datasets, including sentiment analysis and topic classification, our method significantly improves zero/few-shot learning performance of LMs for both in-context learning and prompt-based fine-tuning (on average $9\%$ and $2\%$, respectively).
Authors: Agam Shah, Arnav Hiray, Pratvi Shah, Arkaprabha Banerjee, Anushka Singh, Dheeraj Eidnani, Sahasra Chava, Bhaskar Chaudhury, Sudheer Chava
Abstract: In this paper, we investigate the influence of claims in analyst reports and earnings calls on financial market returns, considering them as significant quarterly events for publicly traded companies. To facilitate a comprehensive analysis, we construct a new financial dataset for the claim detection task in the financial domain. We benchmark various language models on this dataset and propose a novel weak-supervision model that incorporates the knowledge of subject matter experts (SMEs) in the aggregation function, outperforming existing approaches. We also demonstrate the practical utility of our proposed model by constructing a novel measure of optimism. Here, we observe the dependence of earnings surprise and return on our optimism measure. Our dataset, models, and code are publicly (under CC BY 4.0 license) available on GitHub.
Authors: Isha Chaudhary, Vedaant V. Jain, Gagandeep Singh
Abstract: Knowledge comprehension capability is an important aspect of human intelligence. As Large Language Models (LLMs) are being envisioned as superhuman agents, it is crucial for them to be proficient at knowledge comprehension. However, existing benchmarking studies do not provide consistent, generalizable, and formal guarantees on the knowledge comprehension capabilities of LLMs. In this work, we propose the first framework to certify knowledge comprehension in LLMs with formal probabilistic guarantees. Our certificates are quantitative -- they consist of high-confidence, tight bounds on the probability that a target LLM gives the correct answer on any knowledge comprehension prompt sampled from a distribution. We design and certify novel specifications that precisely represent distributions of knowledge comprehension prompts leveraging knowledge graphs. We certify SOTA LLMs for specifications over the Wikidata5m knowledge graph. We find that the knowledge comprehension capability improves significantly with scaling the size of the models.
Authors: Zhenting Qi, Hanlin Zhang, Eric Xing, Sham Kakade, Himabindu Lakkaraju
Abstract: Retrieval-Augmented Generation (RAG) improves pre-trained models by incorporating external knowledge at test time to enable customized adaptation. We study the risk of datastore leakage in Retrieval-In-Context RAG Language Models (LMs). We show that an adversary can exploit LMs' instruction-following capabilities to easily extract text data verbatim from the datastore of RAG systems built with instruction-tuned LMs via prompt injection. The vulnerability exists for a wide range of modern LMs that span Llama2, Mistral/Mixtral, Vicuna, SOLAR, WizardLM, Qwen1.5, and Platypus2, and the exploitability exacerbates as the model size scales up. We also study multiple effects of RAG setup on the extractability of data, indicating that following unexpected instructions to regurgitate data can be an outcome of failure in effectively utilizing contexts for modern LMs, and further show that such vulnerability can be greatly mitigated by position bias elimination strategies. Extending our study to production RAG models GPTs, we design an attack that can cause datastore leakage with a 100% success rate on 25 randomly selected customized GPTs with at most 2 queries, and we extract text data verbatim at a rate of 41% from a book of 77,000 words and 3% from a corpus of 1,569,000 words by prompting the GPTs with only 100 queries generated by themselves.
Authors: Noriaki Hirose, Dhruv Shah, Kyle Stachowicz, Ajay Sridhar, Sergey Levine
Abstract: Autonomous self-improving robots that interact and improve with experience are key to the real-world deployment of robotic systems. In this paper, we propose an online learning method, SELFI, that leverages online robot experience to rapidly fine-tune pre-trained control policies efficiently. SELFI applies online model-free reinforcement learning on top of offline model-based learning to bring out the best parts of both learning paradigms. Specifically, SELFI stabilizes the online learning process by incorporating the same model-based learning objective from offline pre-training into the Q-values learned with online model-free reinforcement learning. We evaluate SELFI in multiple real-world environments and report improvements in terms of collision avoidance, as well as more socially compliant behavior, measured by a human user study. SELFI enables us to quickly learn useful robotic behaviors with less human interventions such as pre-emptive behavior for the pedestrians, collision avoidance for small and transparent objects, and avoiding travel on uneven floor surfaces. We provide supplementary videos to demonstrate the performance of our fine-tuned policy on our project page.
Authors: Yuya Sasaki, Jing Tao, Yulong Wang
Abstract: Motivated by the empirical observation of power-law distributions in the credits (e.g., "likes") of viral social media posts, we introduce a high-dimensional tail index regression model and propose methods for estimation and inference of its parameters. First, we present a regularized estimator, establish its consistency, and derive its convergence rate. Second, we introduce a debiasing technique for the regularized estimator to facilitate inference and prove its asymptotic normality. Third, we extend our approach to handle large-scale online streaming data using stochastic gradient descent. Simulation studies corroborate our theoretical findings. We apply these methods to the text analysis of viral posts on X (formerly Twitter) related to LGBTQ+ topics.
Authors: Xiaohan Wang, Shengyu Mao, Ningyu Zhang, Shumin Deng, Yunzhi Yao, Yue Shen, Lei Liang, Jinjie Gu, Huajun Chen
Abstract: Recently, there has been a growing interest in knowledge editing for Large Language Models (LLMs). Current approaches and evaluations merely explore the instance-level editing, while whether LLMs possess the capability to modify concepts remains unclear. This paper pioneers the investigation of editing conceptual knowledge for LLMs, by constructing a novel benchmark dataset ConceptEdit and establishing a suite of new metrics for evaluation. The experimental results reveal that, although existing editing methods can efficiently modify concept-level definition to some extent, they also have the potential to distort the related instantial knowledge in LLMs, leading to poor performance. We anticipate this can inspire further progress in better understanding LLMs. Our project homepage is available at https://zjunlp.github.io/project/ConceptEdit.
Authors: Andrew Holliday, Gregory Dudek
Abstract: Planning a public transit network is a challenging optimization problem, but essential in order to realize the benefits of autonomous buses. We propose a novel algorithm for planning networks of routes for autonomous buses. We first train a graph neural net model as a policy for constructing route networks, and then use the policy as one of several mutation operators in a evolutionary algorithm. We evaluate this algorithm on a standard set of benchmarks for transit network design, and find that it outperforms the learned policy alone by up to 20% and a plain evolutionary algorithm approach by up to 53% on realistic benchmark instances.
Authors: Stephen R. Pfohl, Heather Cole-Lewis, Rory Sayres, Darlene Neal, Mercy Asiedu, Awa Dieng, Nenad Tomasev, Qazi Mamunur Rashid, Shekoofeh Azizi, Negar Rostamzadeh, Liam G. McCoy, Leo Anthony Celi, Yun Liu, Mike Schaekermann, Alanna Walton, Alicia Parrish, Chirag Nagpal, Preeti Singh, Akeiylah Dewitt, Philip Mansfield, Sushant Prakash, Katherine Heller, Alan Karthikesalingam, Christopher Semturs, Joelle Barral, Greg Corrado, Yossi Matias, Jamila Smith-Loud, Ivor Horn, Karan Singhal
Abstract: Large language models (LLMs) hold promise to serve complex health information needs but also have the potential to introduce harm and exacerbate health disparities. Reliably evaluating equity-related model failures is a critical step toward developing systems that promote health equity. We present resources and methodologies for surfacing biases with potential to precipitate equity-related harms in long-form, LLM-generated answers to medical questions and conduct a large-scale empirical case study with the Med-PaLM 2 LLM. Our contributions include a multifactorial framework for human assessment of LLM-generated answers for biases, and EquityMedQA, a collection of seven datasets enriched for adversarial queries. Both our human assessment framework and dataset design process are grounded in an iterative participatory approach and review of Med-PaLM 2 answers. Through our empirical study, we find that our approach surfaces biases that may be missed via narrower evaluation approaches. Our experience underscores the importance of using diverse assessment methodologies and involving raters of varying backgrounds and expertise. While our approach is not sufficient to holistically assess whether the deployment of an AI system promotes equitable health outcomes, we hope that it can be leveraged and built upon towards a shared goal of LLMs that promote accessible and equitable healthcare.
Authors: Beomsu Kim, Jaemin Kim, Jeongsol Kim, Jong Chul Ye
Abstract: Diffusion models (DMs) excel in unconditional generation, as well as on applications such as image editing and restoration. The success of DMs lies in the iterative nature of diffusion: diffusion breaks down the complex process of mapping noise to data into a sequence of simple denoising tasks. Moreover, we are able to exert fine-grained control over the generation process by injecting guidance terms into each denoising step. However, the iterative process is also computationally intensive, often taking from tens up to thousands of function evaluations. Although consistency trajectory models (CTMs) enable traversal between any time points along the probability flow ODE (PFODE) and score inference with a single function evaluation, CTMs only allow translation from Gaussian noise to data. This work aims to unlock the full potential of CTMs by proposing generalized CTMs (GCTMs), which translate between arbitrary distributions via ODEs. We discuss the design space of GCTMs and demonstrate their efficacy in various image manipulation tasks such as image-to-image translation, restoration, and editing.
Authors: Qianglin Wen, Chengchun Shi, Ying Yang, Niansheng Tang, Hongtu Zhu
Abstract: This paper offers a detailed investigation of switchback designs in A/B testing, which alternate between baseline and new policies over time. Our aim is to thoroughly evaluate the effects of these designs on the accuracy of their resulting average treatment effect (ATE) estimators. We propose a novel "weak signal analysis" framework, which substantially simplifies the calculations of the mean squared errors (MSEs) of these ATEs in Markov decision process environments. Our findings suggest that (i) when the majority of reward errors are positively correlated, the switchback design is more efficient than the alternating-day design which switches policies in a daily basis. Additionally, increasing the frequency of policy switches tends to reduce the MSE of the ATE estimator. (ii) When the errors are uncorrelated, however, all these designs become asymptotically equivalent. (iii) In cases where the majority of errors are negative correlated, the alternating-day design becomes the optimal choice. These insights are crucial, offering guidelines for practitioners on designing experiments in A/B testing. Our analysis accommodates a variety of policy value estimators, including model-based estimators, least squares temporal difference learning estimators, and double reinforcement learning estimators, thereby offering a comprehensive understanding of optimal design strategies for policy evaluation in reinforcement learning.
Authors: Ashwin Aravind, Mohammad Taha Toghani, C\'esar A. Uribe
Abstract: We study the problem of policy estimation for the Linear Quadratic Regulator (LQR) in discrete-time linear time-invariant uncertain dynamical systems. We propose a Moreau Envelope-based surrogate LQR cost, built from a finite set of realizations of the uncertain system, to define a meta-policy efficiently adjustable to new realizations. Moreover, we design an algorithm to find an approximate first-order stationary point of the meta-LQR cost function. Numerical results show that the proposed approach outperforms naive averaging of controllers on new realizations of the linear system. We also provide empirical evidence that our method has better sample complexity than Model-Agnostic Meta-Learning (MAML) approaches.
Authors: Maksym Andriushchenko, Francesco Croce, Nicolas Flammarion
Abstract: We show that even the most recent safety-aligned LLMs are not robust to simple adaptive jailbreaking attacks. First, we demonstrate how to successfully leverage access to logprobs for jailbreaking: we initially design an adversarial prompt template (sometimes adapted to the target LLM), and then we apply random search on a suffix to maximize a target logprob (e.g., of the token "Sure"), potentially with multiple restarts. In this way, we achieve 100% attack success rate -- according to GPT-4 as a judge -- on Vicuna-13B, Mistral-7B, Phi-3-Mini, Nemotron-4-340B, Llama-2-Chat-7B/13B/70B, Llama-3-Instruct-8B, Gemma-7B, GPT-3.5, GPT-4o, and R2D2 from HarmBench that was adversarially trained against the GCG attack. We also show how to jailbreak all Claude models -- that do not expose logprobs -- via either a transfer or prefilling attack with a 100% success rate. In addition, we show how to use random search on a restricted set of tokens for finding trojan strings in poisoned models -- a task that shares many similarities with jailbreaking -- which is the algorithm that brought us the first place in the SaTML'24 Trojan Detection Competition. The common theme behind these attacks is that adaptivity is crucial: different models are vulnerable to different prompting templates (e.g., R2D2 is very sensitive to in-context learning prompts), some models have unique vulnerabilities based on their APIs (e.g., prefilling for Claude), and in some settings, it is crucial to restrict the token search space based on prior knowledge (e.g., for trojan detection). For reproducibility purposes, we provide the code, logs, and jailbreak artifacts in the JailbreakBench format at https://github.com/tml-epfl/llm-adaptive-attacks.
Authors: Mohsen Heidari, Masih Mozakka, Wojciech Szpankowski
Abstract: Hybrid quantum-classical optimization and learning strategies are among the most promising approaches to harnessing quantum information or gaining a quantum advantage over classical methods. However, efficient estimation of the gradient of the objective function in such models remains a challenge due to several factors including the exponential dimensionality of the Hilbert spaces, and information loss of quantum measurements. In this work, we developed an efficient framework that makes the Hadamard test efficiently applicable to gradient estimation for a broad range of quantum systems, an advance that had been wanting from the outset. Under certain mild structural assumptions, the gradient is estimated with the measurement shots that scale logarithmically with the number of parameters and with polynomial classical and quantum time. This is an exponential reduction in the measurement cost and polynomial speed up in time compared to existing works. The structural assumptions are (1) the dimension of the dynamical Lie algebra is polynomial in the number of qubits, and (2) the observable has a bounded Hilbert-Schmidt norm.
Authors: Ashwin Pananjady, Vidya Muthukumar, Andrew Thangaraj
Abstract: We study the problem of estimating the stationary mass -- also called the unigram mass -- that is missing from a single trajectory of a discrete-time, ergodic Markov chain. This problem has several applications -- for example, estimating the stationary missing mass is critical for accurately smoothing probability estimates in sequence models. While the classical Good--Turing estimator from the 1950s has appealing properties for i.i.d. data, it is known to be biased in the Markovian setting, and other heuristic estimators do not come equipped with guarantees. Operating in the general setting in which the size of the state space may be much larger than the length $n$ of the trajectory, we develop a linear-runtime estimator called Windowed Good--Turing (WingIt) and show that its risk decays as $\widetilde{O}(\mathsf{T_{mix}}/n)$, where $\mathsf{T_{mix}}$ denotes the mixing time of the chain in total variation distance. Notably, this rate is independent of the size of the state space and minimax-optimal up to a logarithmic factor in $n / \mathsf{T_{mix}}$. We also present an upper bound on the variance of the missing mass random variable, which may be of independent interest. We extend our estimator to approximate the stationary mass placed on elements occurring with small frequency in the trajectory. Finally, we demonstrate the efficacy of our estimators both in simulations on canonical chains and on sequences constructed from natural language text.
Authors: Kevin Slagle
Abstract: Tokenization is widely used in large language models because it significantly improves performance. However, tokenization imposes several disadvantages, such as performance biases, increased adversarial vulnerability, decreased character-level modeling performance, and increased modeling complexity. To address these disadvantages without sacrificing performance, we propose SpaceByte, a novel byte-level decoder architecture that closes the performance gap between byte-level and subword autoregressive language modeling. SpaceByte consists of a byte-level Transformer model, but with extra larger transformer blocks inserted in the middle of the layers. We find that performance is significantly improved by applying these larger blocks only after certain bytes, such as space characters, which typically denote word boundaries. Our experiments show that for a fixed training and inference compute budget, SpaceByte outperforms other byte-level architectures and roughly matches the performance of tokenized Transformer architectures.
Authors: Giorgio Franceschelli, Mirco Musolesi
Abstract: Large language models are revolutionizing several areas, including artificial creativity. However, the process of generation in machines profoundly diverges from that observed in humans. In particular, machine generation is characterized by a lack of intentionality and an underlying creative process. We propose a method called Creative Beam Search that uses Diverse Beam Search and LLM-as-a-Judge to perform response generation and response validation. The results of a qualitative experiment show how our approach can provide better output than standard sampling techniques. We also show that the response validation step is a necessary complement to the response generation step.
Authors: Mateusz Cedro, Marcin Chlebus
Abstract: The increasing complexity of Artificial Intelligence models poses challenges to interpretability, particularly in the healthcare sector. This study investigates the impact of deep learning model complexity and Explainable AI (XAI) efficacy, utilizing four ResNet architectures (ResNet-18, 34, 50, 101). Through methodical experimentation on 4,369 lung X-ray images of COVID-19-infected and healthy patients, the research evaluates models' classification performance and the relevance of corresponding XAI explanations with respect to the ground-truth disease masks. Results indicate that the increase in model complexity is associated with a decrease in classification accuracy and AUC-ROC scores (ResNet-18: 98.4%, 0.997; ResNet-101: 95.9%, 0.988). Notably, in eleven out of twelve statistical tests performed, no statistically significant differences occurred between XAI quantitative metrics - Relevance Rank Accuracy and the proposed Positive Attribution Ratio - across trained models. These results suggest that increased model complexity does not consistently lead to higher performance or relevance of explanations for models' decision-making processes.
Authors: Anton Semenkin, Vitaliy Bibaev, Yaroslav Sokolov, Kirill Krylov, Alexey Kalina, Anna Khannanova, Danila Savenkov, Darya Rovdo, Igor Davidenko, Kirill Karnaukhov, Maxim Vakhrushev, Mikhail Kostyukov, Mikhail Podvitskii, Petr Surkov, Yaroslav Golubev, Nikita Povarov, Timofey Bryksin
Abstract: In recent years, several industrial solutions for the problem of multi-token code completion appeared, each making a great advance in the area but mostly focusing on cloud-based runtime and avoiding working on the end user's device. In this work, we describe our approach for building a multi-token code completion feature for the JetBrains' IntelliJ Platform, which we call Full Line Code Completion. The feature suggests only syntactically correct code and works fully locally, i.e., data querying and the generation of suggestions happens on the end user's machine. We share important time and memory-consumption restrictions, as well as design principles that a code completion engine should satisfy. Working entirely on the end user's device, our code completion engine enriches user experience while being not only fast and compact but also secure. We share a number of useful techniques to meet the stated development constraints and also describe offline and online evaluation pipelines that allowed us to make better decisions. Our online evaluation shows that the usage of the tool leads to 1.3 times more Python code in the IDE being produced by code completion. The described solution was initially started with a help of researchers and was then bundled into all JetBrains IDEs where it is now used by millions of users. Thus, we believe that this work is useful for bridging academia and industry, providing researchers with the knowledge of what happens when complex research-based solutions are integrated into real products.
Authors: Jiawei Ge, Debarghya Mukherjee, Jianqing Fan
Abstract: As machine learning models are increasingly deployed in dynamic environments, it becomes paramount to assess and quantify uncertainties associated with distribution shifts. A distribution shift occurs when the underlying data-generating process changes, leading to a deviation in the model's performance. The prediction interval, which captures the range of likely outcomes for a given prediction, serves as a crucial tool for characterizing uncertainties induced by their underlying distribution. In this paper, we propose methodologies for aggregating prediction intervals to obtain one with minimal width and adequate coverage on the target domain under unsupervised domain shift, under which we have labeled samples from a related source domain and unlabeled covariates from the target domain. Our analysis encompasses scenarios where the source and the target domain are related via i) a bounded density ratio, and ii) a measure-preserving transformation. Our proposed methodologies are computationally efficient and easy to implement. Beyond illustrating the performance of our method through real-world datasets, we also delve into the theoretical details. This includes establishing rigorous theoretical guarantees, coupled with finite sample bounds, regarding the coverage and width of our prediction intervals. Our approach excels in practical applications and is underpinned by a solid theoretical framework, ensuring its reliability and effectiveness across diverse contexts.
Authors: Domenic Rosati, Jan Wehner, Kai Williams, {\L}ukasz Bartoszcze, David Atanasov, Robie Gonzales, Subhabrata Majumdar, Carsten Maple, Hassan Sajjad, Frank Rudzicz
Abstract: Releasing open-source large language models (LLMs) presents a dual-use risk since bad actors can easily fine-tune these models for harmful purposes. Even without the open release of weights, weight stealing and fine-tuning APIs make closed models vulnerable to harmful fine-tuning attacks (HFAs). While safety measures like preventing jailbreaks and improving safety guardrails are important, such measures can easily be reversed through fine-tuning. In this work, we propose Representation Noising (RepNoise), a defence mechanism that is effective even when attackers have access to the weights. RepNoise works by removing information about harmful representations such that it is difficult to recover them during fine-tuning. Importantly, our defence is also able to generalize across different subsets of harm that have not been seen during the defence process as long as they are drawn from the same distribution of the attack set. Our method does not degrade the general capability of LLMs and retains the ability to train the model on harmless tasks. We provide empirical evidence that the effectiveness of our defence lies in its "depth": the degree to which information about harmful representations is removed across all layers of the LLM.
Authors: Peng Wang, Zexi Li, Ningyu Zhang, Ziwen Xu, Yunzhi Yao, Yong Jiang, Pengjun Xie, Fei Huang, Huajun Chen
Abstract: Large language models (LLMs) need knowledge updates to meet the ever-growing world facts and correct the hallucinated responses, facilitating the methods of lifelong model editing. Where the updated knowledge resides in memories is a fundamental question for model editing. In this paper, we find that editing either long-term memory (direct model parameters) or working memory (non-parametric knowledge of neural network activations/representations by retrieval) will result in an impossible triangle -- reliability, generalization, and locality can not be realized together in the lifelong editing settings. For long-term memory, directly editing the parameters will cause conflicts with irrelevant pretrained knowledge or previous edits (poor reliability and locality). For working memory, retrieval-based activations can hardly make the model understand the edits and generalize (poor generalization). Therefore, we propose WISE to bridge the gap between memories. In WISE, we design a dual parametric memory scheme, which consists of the main memory for the pretrained knowledge and a side memory for the edited knowledge. We only edit the knowledge in the side memory and train a router to decide which memory to go through when given a query. For continual editing, we devise a knowledge-sharding mechanism where different sets of edits reside in distinct subspaces of parameters, and are subsequently merged into a shared memory without conflicts. Extensive experiments show that WISE can outperform previous model editing methods and overcome the impossible triangle under lifelong model editing of question answering, hallucination, and out-of-distribution settings across trending LLM architectures, e.g., GPT, LLaMA, and Mistral. Code is available at https://github.com/zjunlp/EasyEdit.
Authors: Yuwei Niu, Shuo He, Qi Wei, Zongyu Wu, Feng Liu, Lei Feng
Abstract: Multimodal contrastive learning methods (e.g., CLIP) have shown impressive zero-shot classification performance due to their strong ability to joint representation learning for visual and textual modalities. However, recent research revealed that multimodal contrastive learning on poisoned pre-training data with a small proportion of maliciously backdoored data can induce backdoored CLIP that could be attacked by inserted triggers in downstream tasks with a high success rate. To defend against backdoor attacks on CLIP, existing defense methods focus on either the pre-training stage or the fine-tuning stage, which would unfortunately cause high computational costs due to numerous parameter updates. In this paper, we provide the first attempt at a computationally efficient backdoor detection method to defend against backdoored CLIP in the inference stage. We empirically find that the visual representations of backdoored images are insensitive to both benign and malignant changes in class description texts. Motivated by this observation, we propose BDetCLIP, a novel test-time backdoor detection method based on contrastive prompting. Specifically, we first prompt the language model (e.g., GPT-4) to produce class-related description texts (benign) and class-perturbed random texts (malignant) by specially designed instructions. Then, the distribution difference in cosine similarity between images and the two types of class description texts can be used as the criterion to detect backdoor samples. Extensive experiments validate that our proposed BDetCLIP is superior to state-of-the-art backdoor detection methods, in terms of both effectiveness and efficiency.
Authors: Angelo Porrello, Lorenzo Bonicelli, Pietro Buzzega, Monica Millunzi, Simone Calderara, Rita Cucchiara
Abstract: The fine-tuning of deep pre-trained models has revealed compositional properties, with multiple specialized modules that can be arbitrarily composed into a single, multi-task model. However, identifying the conditions that promote compositionality remains an open issue, with recent efforts concentrating mainly on linearized networks. We conduct a theoretical study that attempts to demystify compositionality in standard non-linear networks through the second-order Taylor approximation of the loss function. The proposed formulation highlights the importance of staying within the pre-training basin to achieve composable modules. Moreover, it provides the basis for two dual incremental training algorithms: the one from the perspective of multiple models trained individually, while the other aims to optimize the composed model as a whole. We probe their application in incremental classification tasks and highlight some valuable skills. In fact, the pool of incrementally learned modules not only supports the creation of an effective multi-task model but also enables unlearning and specialization in certain tasks.
Authors: Piyush Jha, Prithwish Jana, Pranavkrishna Suresh, Arnav Arora, Vijay Ganesh
Abstract: Reinforcement Learning with Human Feedback (RLHF) is considered a standard approach to fine-tuning Large Language Models (LLMs). However, such methods often face limitations such as unsound black-box reward models, difficulties in collecting human preference data, and the reliance on sparse scalar rewards. These methods often fall short when applied to tasks that require complex domain-specific understanding. To address these challenges, we propose a new fine-tuning paradigm we refer to as Reinforcement Learning via Symbolic Feedback (RLSF), which aims to improve domain-specific understanding of LLMs more effectively than traditional reward signals. In the RLSF setting, the LLM being fine-tuned is considered an RL agent, while the environment is allowed access to reasoning or domain knowledge tools (e.g., solvers, provers, algebra systems, or knowledge bases). Crucially, in RLSF, these reasoning tools can provide feedback to the LLMs via poly-sized certificates (e.g., proofs), that characterize errors in the LLM-generated object with respect to some correctness specification. As a bonus, our RLSF approach does not require the reasoning systems we use to be differentiable. The ability of RLSF-based fine-tuning to leverage certificate-generating symbolic tools enables sound fine-grained (token-level) reward signals to LLMs, and thus addresses the limitations of traditional reward models mentioned above. Via extensive evaluations, we show that our RLSF-based fine-tuning of LLMs outperforms traditional approaches on five different applications, namely, program synthesis from natural language pseudo-code to programming language, three chemistry tasks, and solving the Game of 24. A takeaway is that fine-tuning via RLSF enables relatively smaller LLMs to significantly outperform closed-source models that are orders of magnitude larger (e.g., GPT-4).
Authors: Mustafa Shukor, Matthieu Cord
Abstract: Large Language Models (LLMs) have demonstrated impressive performance on multimodal tasks, without any multimodal finetuning. They are the building block for Large Multimodal Models, yet, we still lack a proper understanding of their success. In this work, we expose frozen LLMs to image, video, audio and text inputs and analyse their internal representation aiming to understand their generalization beyond textual inputs. Findings. Perceptual tokens (1) are easily distinguishable from textual ones inside LLMs, with significantly different representations, and complete translation to textual tokens does not exist. Yet, (2) both perceptual and textual tokens activate similar LLM weights. Despite being different, (3) perceptual and textual tokens are implicitly aligned inside LLMs, we call this the implicit multimodal alignment (IMA), and argue that this is linked to architectural design, helping LLMs to generalize. This provide more evidence to believe that the generalization of LLMs to multimodal inputs is mainly due to their architecture. Implications. (1) We find a positive correlation between the implicit alignment score and the task performance, suggesting that this could act as a proxy metric for model evaluation and selection. (2) A negative correlation exists regarding hallucinations, revealing that this problem is mainly due to misalignment between the internal perceptual and textual representations. (3) Perceptual tokens change slightly throughout the model, thus, we propose different approaches to skip computations (e.g. in FFN layers), and significantly reduce the inference cost. (4) Due to the slowly changing embeddings across layers, and the high overlap between textual and multimodal activated weights, we compress LLMs by keeping only 1 subnetwork that works well across a wide range of multimodal tasks. Paper code: https://github.com/mshukor/ima-lmms.
Authors: Inhwa Han, Jaayeon Lee, Jong Chul Ye
Abstract: Research efforts for visual decoding from fMRI signals have attracted considerable attention in research community. Still multi-subject fMRI decoding with one model has been considered intractable due to the drastic variations in fMRI signals between subjects and even within the same subject across different trials. To address current limitations in multi-subject brain decoding, here we introduce a novel semantic alignment method of multi-subject fMRI signals using so-called MindFormer. This model is specifically designed to generate fMRI-conditioned feature vectors that can be used for conditioning Stable Diffusion model for fMRI- to-image generation or large language model (LLM) for fMRI-to-text generation. More specifically, MindFormer incorporates two key innovations: 1) a subject specific token that effectively capture individual differences in fMRI signals while synergistically combines multi subject fMRI data for training, and 2) a novel feature embedding and training scheme based on the IP-Adapter to extract semantically meaningful features from fMRI signals. Our experimental results demonstrate that MindFormer generates semantically consistent images and text across different subjects. Since our MindFormer maintains semantic fidelity by fully utilizing the training data across different subjects by significantly surpassing existing models in multi-subject brain decoding, this may help deepening our understanding of neural processing variations among individuals.
Authors: Jaskaran Singh Walia, Pavithra L K, Kesar Mehta, Shivram Harshavardhana, Nandini Tyagi
Abstract: Addressing the issue of submerged underwater trash is crucial for safeguarding aquatic ecosystems and preserving marine life. While identifying debris present on the surface of water bodies is straightforward, assessing the underwater submerged waste is a challenge due to the image distortions caused by factors such as light refraction, absorption, suspended particles, color shifts, and occlusion. This paper conducts a comprehensive review of state-of-the-art architectures and on the existing datasets to establish a baseline for submerged waste and trash detection. The primary goal remains to establish the benchmark of the object localization techniques to be leveraged by advanced underwater sensors and autonomous underwater vehicles. The ultimate objective is to explore the underwater environment, to identify, and remove underwater debris. The absence of benchmarks (dataset or algorithm) in many researches emphasizes the need for a more robust algorithmic solution. Through this research, we aim to give performance comparative analysis of various underwater trash detection algorithms.
Authors: Ethan Shen, Alan Fan, Sarah M. Pratt, Jae Sung Park, Matthew Wallingford, Sham M. Kakade, Ari Holtzman, Ranjay Krishna, Ali Farhadi, Aditya Kusupati
Abstract: Many applications today provide users with multiple auto-complete drafts as they type, including GitHub's code completion, Gmail's smart compose, and Apple's messaging auto-suggestions. Under the hood, language models support this by running an autoregressive inference pass to provide a draft. Consequently, providing $k$ drafts to the user requires running an expensive language model $k$ times. To alleviate the computation cost of running $k$ inference passes, we propose Superposed Decoding, a new decoding algorithm that generates $k$ drafts at the computation cost of one autoregressive inference pass. We achieve this by feeding a superposition of the most recent token embeddings from the $k$ drafts as input to the next decoding step of the language model. At every inference step we combine the $k$ drafts with the top-$k$ tokens to get $k^2$ new drafts and cache the $k$ most likely options, using an n-gram interpolation with minimal compute overhead to filter out incoherent generations. Our experiments show that $k$ drafts from Superposed Decoding are at least as coherent and factual as Nucleus Sampling and Greedy Decoding respectively, while being at least $2.44\times$ faster for $k\ge3$. In a compute-normalized setting, user evaluations demonstrably favor text generated by Superposed Decoding over Nucleus Sampling. Superposed Decoding can also be combined with other decoding strategies, resulting in universal coverage gains when scaling inference time compute. Code and more examples open-sourced at https://github.com/RAIVNLab/SuperposedDecoding.
Authors: Jacopo Teneggi, Jeremias Sulam
Abstract: Recent works have extended notions of feature importance to semantic concepts that are inherently interpretable to the users interacting with a black-box predictive model. Yet, precise statistical guarantees, such as false positive rate and false discovery rate control, are needed to communicate findings transparently and to avoid unintended consequences in real-world scenarios. In this paper, we formalize the global (i.e., over a population) and local (i.e., for a sample) statistical importance of semantic concepts for the predictions of opaque models by means of conditional independence, which allows for rigorous testing. We use recent ideas of sequential kernelized independence testing (SKIT) to induce a rank of importance across concepts, and showcase the effectiveness and flexibility of our framework on synthetic datasets as well as on image classification tasks using several and diverse vision-language models.
Authors: Hsuan Su, Hua Farn, Fan-Yun Sun, Shang-Tse Chen, Hung-yi Lee
Abstract: Synthetic data is widely used in speech recognition due to the availability of text-to-speech models, which facilitate adapting models to previously unseen text domains. However, existing methods suffer in performance when they fine-tune an automatic speech recognition (ASR) model on synthetic data as they suffer from the distributional shift commonly referred to as the synthetic-to-real gap. In this paper, we find that task vector arithmetic is effective at mitigating this gap. Our proposed method, SYN2REAL task vector, shows an average improvement of 10.03\% improvement in word error rate over baselines on the SLURP dataset. Additionally, we show that an average of SYN2REAL task vectors, when we have real speeches from multiple different domains, can further adapt the original ASR model to perform better on the target text domain.
Authors: Akshay Bansal, Ian George, Soumik Ghosh, Jamie Sikora, Alice Zheng
Abstract: In many quantum tasks, there is an unknown quantum object that one wishes to learn. An online strategy for this task involves adaptively refining a hypothesis to reproduce such an object or its measurement statistics. A common evaluation metric for such a strategy is its regret, or roughly the accumulated errors in hypothesis statistics. We prove a sublinear regret bound for learning over general subsets of positive semidefinite matrices via the regularized-follow-the-leader algorithm and apply it to various settings where one wishes to learn quantum objects. For concrete applications, we present a sublinear regret bound for learning quantum states, effects, channels, interactive measurements, strategies, co-strategies, and the collection of inner products of pure states. Our bound applies to many other quantum objects with compact, convex representations. In proving our regret bound, we establish various matrix analysis results useful in quantum information theory. This includes a generalization of Pinsker's inequality for arbitrary positive semidefinite operators with possibly different traces, which may be of independent interest and applicable to more general classes of divergences.
Authors: Zhuoqun Wang, Naoki Awaya, Li Ma
Abstract: A common objective in the analysis of tabular data is estimating the conditional distribution (in contrast to only producing predictions) of a set of "outcome" variables given a set of "covariates", which is sometimes referred to as the "density regression" problem. Beyond estimation on the conditional distribution, the generative ability of drawing synthetic samples from the learned conditional distribution is also desired as it further widens the range of applications. We propose a flow-based generative model tailored for the density regression task on tabular data. Our flow applies a sequence of tree-based piecewise-linear transforms on initial uniform noise to eventually generate samples from complex conditional densities of (univariate or multivariate) outcomes given the covariates and allows efficient analytical evaluation of the fitted conditional density on any point in the sample space. We introduce a training algorithm for fitting the tree-based transforms using a divide-and-conquer strategy that transforms maximum likelihood training of the tree-flow into training a collection of binary classifiers--one at each tree split--under cross-entropy loss. We assess the performance of our method under out-of-sample likelihood evaluation and compare it with a variety of state-of-the-art conditional density learners on a range of simulated and real benchmark tabular datasets. Our method consistently achieves comparable or superior performance at a fraction of the training and sampling budget. Finally, we demonstrate the utility of our method's generative ability through an application to generating synthetic longitudinal microbiome compositional data based on training our flow on a publicly available microbiome study.
Authors: Yanling Wang, Haoyang Li, Hao Zou, Jing Zhang, Xinlei He, Qi Li, Ke Xu
Abstract: Despite advancements in large language models (LLMs), non-factual responses remain prevalent. Unlike extensive studies on post-hoc detection of such responses, this work studies non-factuality prediction (NFP), aiming to predict whether an LLM will generate a non-factual response to a question before the generation process. Previous efforts on NFP have demonstrated LLMs' awareness of their internal knowledge, but they still face challenges in efficiency and transferability. In this work, we propose a lightweight NFP model named Factuality Lens (FacLens), which effectively probes hidden representations of questions for the NFP task. Besides, we discover that hidden question representations sourced from different LLMs exhibit similar NFP patterns, which enables the transferability of FacLens across LLMs to reduce development costs. Extensive experiments highlight FacLens's superiority in both effectiveness and efficiency.
Authors: Sarah Ball, Frauke Kreuter, Nina Panickssery
Abstract: Conversational large language models are trained to refuse to answer harmful questions. However, emergent jailbreaking techniques can still elicit unsafe outputs, presenting an ongoing challenge for model alignment. To better understand how different jailbreak types circumvent safeguards, this paper analyses model activations on different jailbreak inputs. We find that it is possible to extract a jailbreak vector from a single class of jailbreaks that works to mitigate jailbreak effectiveness from other semantically-dissimilar classes. This may indicate that different kinds of effective jailbreaks operate via a similar internal mechanism. We investigate a potential common mechanism of harmfulness feature suppression, and find evidence that effective jailbreaks noticeably reduce a model's perception of prompt harmfulness. These findings offer actionable insights for developing more robust jailbreak countermeasures and lay the groundwork for a deeper, mechanistic understanding of jailbreak dynamics in language models.
Authors: Krista Opsahl-Ong, Michael J Ryan, Josh Purtell, David Broman, Christopher Potts, Matei Zaharia, Omar Khattab
Abstract: Language Model Programs, i.e. sophisticated pipelines of modular language model (LM) calls, are increasingly advancing NLP tasks, but they require crafting prompts that are jointly effective for all modules. We study prompt optimization for LM programs, i.e. how to update these prompts to maximize a downstream metric without access to module-level labels or gradients. To make this tractable, we factorize our problem into optimizing the free-form instructions and few-shot demonstrations of every module and introduce several strategies to craft task-grounded instructions and navigate credit assignment across modules. Our strategies include (i) program- and data-aware techniques for proposing effective instructions, (ii) a stochastic mini-batch evaluation function for learning a surrogate model of our objective, and (iii) a meta-optimization procedure in which we refine how LMs construct proposals over time. Using these insights we develop MIPRO, a novel algorithm for optimizing LM programs. MIPRO outperforms baseline optimizers on five of seven diverse multi-stage LM programs using a best-in-class open-source model (Llama-3-8B), by as high as 13% accuracy. We have released our new optimizers and benchmark in DSPy at http://dspy.ai
URLs: http://dspy.ai
Authors: Noah Marshall, Ke Liang Xiao, Atish Agarwala, Elliot Paquette
Abstract: The success of modern machine learning is due in part to the adaptive optimization methods that have been developed to deal with the difficulties of training large models over complex datasets. One such method is gradient clipping: a practical procedure with limited theoretical underpinnings. In this work, we study clipping in a least squares problem under streaming SGD. We develop a theoretical analysis of the learning dynamics in the limit of large intrinsic dimension-a model and dataset dependent notion of dimensionality. In this limit we find a deterministic equation that describes the evolution of the loss and demonstrate that this equation predicts the path of clipped SGD on synthetic, CIFAR10, and Wikitext2 data. We show that with Gaussian noise clipping cannot improve SGD performance. Yet, in other noisy settings, clipping can provide benefits with tuning of the clipping threshold. We propose a simple heuristic for near optimal scheduling of the clipping threshold which requires the tuning of only one hyperparameter. We conclude with a discussion about the links between high-dimensional clipping and neural network training.
Authors: Fei Wang, Wenxuan Zhou, James Y. Huang, Nan Xu, Sheng Zhang, Hoifung Poon, Muhao Chen
Abstract: Direct preference optimization (DPO) has shown to be an effective method for large language model (LLM) alignment. Recent works have attempted to apply DPO to multimodal scenarios but have found it challenging to achieve consistent improvement. Through a comparative experiment, we identify the unconditional preference problem in multimodal preference optimization, where the model overlooks the image condition. To address this problem, we propose mDPO, a multimodal DPO objective that prevents the over-prioritization of language-only preferences by also optimizing image preference. Moreover, we introduce a reward anchor that forces the reward to be positive for chosen responses, thereby avoiding the decrease in their likelihood -- an intrinsic problem of relative preference optimization. Experiments on two multimodal LLMs of different sizes and three widely used benchmarks demonstrate that mDPO effectively addresses the unconditional preference problem in multimodal preference optimization and significantly improves model performance, particularly in reducing hallucination.
Authors: Harry Shomer, Jay Revolinsky, Jiliang Tang
Abstract: Knowledge Graph Completion (KGC) attempts to predict missing facts in a Knowledge Graph (KG). Recently, there's been an increased focus on designing KGC methods that can excel in the {\it inductive setting}, where a portion or all of the entities and relations seen in inference are unobserved during training. Numerous benchmark datasets have been proposed for inductive KGC, all of which are subsets of existing KGs used for transductive KGC. However, we find that the current procedure for constructing inductive KGC datasets inadvertently creates a shortcut that can be exploited even while disregarding the relational information. Specifically, we observe that the Personalized PageRank (PPR) score can achieve strong or near SOTA performance on most inductive datasets. In this paper, we study the root cause of this problem. Using these insights, we propose an alternative strategy for constructing inductive KGC datasets that helps mitigate the PPR shortcut. We then benchmark multiple popular methods using the newly constructed datasets and analyze their performance. The new benchmark datasets help promote a better understanding of the capabilities and challenges of inductive KGC by removing any shortcuts that obfuscate performance.
Authors: Junmo Kang, Leonid Karlinsky, Hongyin Luo, Zhen Wang, Jacob Hansen, James Glass, David Cox, Rameswar Panda, Rogerio Feris, Alan Ritter
Abstract: We present Self-MoE, an approach that transforms a monolithic LLM into a compositional, modular system of self-specialized experts, named MiXSE (MiXture of Self-specialized Experts). Our approach leverages self-specialization, which constructs expert modules using self-generated synthetic data, each equipping a shared base LLM with distinct domain-specific capabilities, activated via self-optimized routing. This allows for dynamic and capability-specific handling of various target tasks, enhancing overall capabilities, without extensive human-labeled data and added parameters. Our empirical results reveal that specializing LLMs may exhibit potential trade-offs in performances on non-specialized tasks. On the other hand, our Self-MoE demonstrates substantial improvements (6.5%p on average) over the base LLM across diverse benchmarks such as knowledge, reasoning, math, and coding. It also consistently outperforms other methods, including instance merging and weight merging, while offering better flexibility and interpretability by design with semantic experts and routing. Our findings highlight the critical role of modularity, the applicability of Self-MoE to multiple base LLMs, and the potential of self-improvement in achieving efficient, scalable, and adaptable systems.
Authors: Jialiang Zhao, Yuxiang Ma, Lirui Wang, Edward H. Adelson
Abstract: This paper presents T3: Transferable Tactile Transformers, a framework for tactile representation learning that scales across multi-sensors and multi-tasks. T3 is designed to overcome the contemporary issue that camera-based tactile sensing is extremely heterogeneous, i.e. sensors are built into different form factors, and existing datasets were collected for disparate tasks. T3 captures the shared latent information across different sensor-task pairings by constructing a shared trunk transformer with sensor-specific encoders and task-specific decoders. The pre-training of T3 utilizes a novel Foundation Tactile (FoTa) dataset, which is aggregated from several open-sourced datasets and it contains over 3 million data points gathered from 13 sensors and 11 tasks. FoTa is the largest and most diverse dataset in tactile sensing to date and it is made publicly available in a unified format. Across various sensors and tasks, experiments show that T3 pre-trained with FoTa achieved zero-shot transferability in certain sensor-task pairings, can be further fine-tuned with small amounts of domain-specific data, and its performance scales with bigger network sizes. T3 is also effective as a tactile encoder for long horizon contact-rich manipulation. Results from sub-millimeter multi-pin electronics insertion tasks show that T3 achieved a task success rate 25% higher than that of policies trained with tactile encoders trained from scratch, or 53% higher than without tactile sensing. Data, code, and model checkpoints are open-sourced at https://t3.alanz.info
URLs: https://t3.alanz.info
Authors: Sungjun Lim, Jeyoon Yeom, Sooyon Kim, Hoyoon Byun, Jinho Kang, Yohan Jung, Jiyoung Jung, Kyungwoo Song
Abstract: Bayesian neural network (BNN) approximates the posterior distribution of model parameters and utilizes the posterior for prediction via Bayesian Model Averaging (BMA). The quality of the posterior approximation is critical for achieving accurate and robust predictions. It is known that flatness in the loss landscape is strongly associated with generalization performance, and it necessitates consideration to improve the quality of the posterior approximation. In this work, we empirically demonstrate that BNNs often struggle to capture the flatness. Moreover, we provide both experimental and theoretical evidence showing that BMA can be ineffective without ensuring flatness. To address this, we propose Sharpness-Aware Bayesian Model Averaging (SA-BMA), a novel optimizer that seeks flat posteriors by calculating divergence in the parameter space. SA-BMA aligns with the intrinsic nature of BNN and the generalized version of existing sharpness-aware optimizers for DNN. In addition, we suggest a Bayesian Transfer Learning scheme to efficiently leverage pre-trained DNN. We validate the efficacy of SA-BMA in enhancing generalization performance in few-shot classification and distribution shift by ensuring flat posterior.
Authors: Felix Chalumeau, Refiloe Shabe, Noah De Nicola, Arnu Pretorius, Thomas D. Barrett, Nathan Grinsztajn
Abstract: Combinatorial Optimization is crucial to numerous real-world applications, yet still presents challenges due to its (NP-)hard nature. Amongst existing approaches, heuristics often offer the best trade-off between quality and scalability, making them suitable for industrial use. While Reinforcement Learning (RL) offers a flexible framework for designing heuristics, its adoption over handcrafted heuristics remains incomplete within industrial solvers. Existing learned methods still lack the ability to adapt to specific instances and fully leverage the available computational budget. The current best methods either rely on a collection of pre-trained policies, or on data-inefficient fine-tuning; hence failing to fully utilize newly available information within the constraints of the budget. In response, we present MEMENTO, an approach that leverages memory to improve the adaptation of neural solvers at inference time. MEMENTO enables updating the action distribution dynamically based on the outcome of previous decisions. We validate its effectiveness on benchmark problems, in particular Traveling Salesman and Capacitated Vehicle Routing, demonstrating its superiority over tree-search and policy-gradient fine-tuning; and showing it can be zero-shot combined with diversity-based solvers. We successfully train all RL auto-regressive solvers on large instances, and show that MEMENTO can scale and is data-efficient. Overall, MEMENTO enables to push the state-of-the-art on 11 out of 12 evaluated tasks.
Authors: Yixuan Wang, Xianzhen Luo, Fuxuan Wei, Yijun Liu, Qingfu Zhu, Xuanyu Zhang, Qing Yang, Dongliang Xu, Wanxiang Che
Abstract: Existing speculative decoding methods typically require additional model structure and training processes to assist the model for draft token generation. This makes the migration of acceleration methods to the new model more costly and more demanding on device memory. To address this problem, we propose the Make Some Noise (MSN) training framework as a replacement for the supervised fine-tuning stage of the large language model. The training method simply introduces some noise at the input for the model to learn the denoising task. It significantly enhances the parallel decoding capability of the model without affecting the original task capability. In addition, we propose a tree-based retrieval-augmented Jacobi (TR-Jacobi) decoding strategy to further improve the inference speed of MSN models. Experiments in both the general and code domains have shown that MSN can improve inference speed by 2.3-2.7x times without compromising model performance. The MSN model also achieves comparable acceleration ratios to the SOTA model with additional model structure on Spec-Bench.
Authors: Matthieu Meeus, Igor Shilov, Shubham Jain, Manuel Faysse, Marek Rei, Yves-Alexandre de Montjoye
Abstract: Whether LLMs memorize their training data and what this means, from privacy leakage to detecting copyright violations -- has become a rapidly growing area of research over the last two years. In recent months, more than 10 new methods have been proposed to perform Membership Inference Attacks (MIAs) against LLMs. Contrary to traditional MIAs which rely on fixed -- but randomized -- records or models, these methods are mostly evaluated on datasets collected post-hoc. Sets of members and non-members, used to evaluate the MIA, are constructed using informed guesses after the release of a model. This lack of randomization raises concerns of a distribution shift between members and non-members. In the first part, we review the literature on MIAs against LLMs. While most work focuses on sequence-level MIAs evaluated in post-hoc setups, we show that a range of target models, motivations and units of interest have been considered in the literature. We then quantify distribution shifts present in the 6 datasets used in the literature, ranging from books to papers, using a bag of word classifier. Our analysis reveals that all of them suffer from severe distribution shifts. This challenges the validity of using such setups to measure LLM memorization and may undermine the benchmarking of recently proposed methods. Yet, all hope might not be lost. In the second part, we introduce important considerations to properly evaluate MIAs against LLMs and discuss potential ways forward: randomized test splits, injections of randomized (unique) sequences, randomized finetuning, and post-hoc control methods. While each option comes with its advantages and limitations, we believe they collectively provide solid grounds to guide the development of MIA methods and study LLM memorization. We conclude by proposing comprehensive, easy-to-use benchmarks for sequence- and document-level MIAs against LLMs.
Authors: Ismail Alkhouri, Cedric Le Denmat, Yingjie Li, Cunxi Yu, Jia Liu, Rongrong Wang, Alvaro Velasquez
Abstract: Combinatorial Optimization (CO) addresses many important problems, including the challenging Maximum Independent Set (MIS) problem. Alongside exact and heuristic solvers, differentiable approaches have emerged, often using continuous relaxations of ReLU-based or quadratic objectives. Noting that an MIS in a graph is a Maximum Clique (MC) in its complement, we propose a new quadratic formulation for MIS by incorporating an MC term, improving convergence and exploration. We show that every maximal independent set corresponds to a local minimizer, derive conditions for the MIS size, and characterize stationary points. To solve our non-convex objective, we propose solving parallel multiple initializations using momentum-based gradient descent, complemented by an efficient MIS checking criterion derived from our theory. Therefore, we dub our method as parallelized Clique-Informed Quadratic Optimization for MIS (pCQO-MIS). Our experimental results demonstrate the effectiveness of the proposed method compared to exact, heuristic, sampling, and data-centric approaches. Notably, our method avoids the out-of-distribution tuning and reliance on (un)labeled data required by data-centric methods, while achieving superior MIS sizes and competitive runtime relative to their inference time. Additionally, a key advantage of pCQO-MIS is that, unlike exact and heuristic solvers, the runtime scales only with the number of nodes in the graph, not the number of edges.
Authors: Bozhong Tian, Xiaozhuan Liang, Siyuan Cheng, Qingbin Liu, Mengru Wang, Dianbo Sui, Xi Chen, Huajun Chen, Ningyu Zhang
Abstract: Large Language Models (LLMs) trained on extensive corpora inevitably retain sensitive data, such as personal privacy information and copyrighted material. Recent advancements in knowledge unlearning involve updating LLM parameters to erase specific knowledge. However, current unlearning paradigms are mired in vague forgetting boundaries, often erasing knowledge indiscriminately. In this work, we introduce KnowUnDo, a benchmark containing copyrighted content and user privacy domains to evaluate if the unlearning process inadvertently erases essential knowledge. Our findings indicate that existing unlearning methods often suffer from excessive unlearning. To address this, we propose a simple yet effective method, MemFlex, which utilizes gradient information to precisely target and unlearn sensitive parameters. Experimental results show that MemFlex is superior to existing methods in both precise knowledge unlearning and general knowledge retaining of LLMs. Code and dataset are released at https://github.com/zjunlp/KnowUnDo.
Authors: Po-Yu Hsieh, June-Hao Hou
Abstract: The control and modeling of robot dynamics have increasingly adopted model-free control strategies using machine learning. Given the non-linear elastic nature of bionic robotic systems, learning-based methods provide reliable alternatives by utilizing numerical data to establish a direct mapping from actuation inputs to robot trajectories without complex kinematics models. However, for developers, the method of identifying an appropriate learning model for their specific bionic robots and further constructing the transfer function has not been thoroughly discussed. Thus, this research introduces a comprehensive evaluation strategy and framework for the application of model-free control, including data collection, learning model selection, comparative analysis, and transfer function identification to effectively deal with the multi-input multi-output (MIMO) robotic data.
Authors: Apolline Mellot, Antoine Collas, Sylvain Chevallier, Alexandre Gramfort, Denis A. Engemann
Abstract: Electroencephalography (EEG) data is often collected from diverse contexts involving different populations and EEG devices. This variability can induce distribution shifts in the data $X$ and in the biomedical variables of interest $y$, thus limiting the application of supervised machine learning (ML) algorithms. While domain adaptation (DA) methods have been developed to mitigate the impact of these shifts, such methods struggle when distribution shifts occur simultaneously in $X$ and $y$. As state-of-the-art ML models for EEG represent the data by spatial covariance matrices, which lie on the Riemannian manifold of Symmetric Positive Definite (SPD) matrices, it is appealing to study DA techniques operating on the SPD manifold. This paper proposes a novel method termed Geodesic Optimization for Predictive Shift Adaptation (GOPSA) to address test-time multi-source DA for situations in which source domains have distinct $y$ distributions. GOPSA exploits the geodesic structure of the Riemannian manifold to jointly learn a domain-specific re-centering operator representing site-specific intercepts and the regression model. We performed empirical benchmarks on the cross-site generalization of age-prediction models with resting-state EEG data from a large multi-national dataset (HarMNqEEG), which included $14$ recording sites and more than $1500$ human participants. Compared to state-of-the-art methods, our results showed that GOPSA achieved significantly higher performance on three regression metrics ($R^2$, MAE, and Spearman's $\rho$) for several source-target site combinations, highlighting its effectiveness in tackling multi-source DA with predictive shifts in EEG data analysis. Our method has the potential to combine the advantages of mixed-effects modeling with machine learning for biomedical applications of EEG, such as multicenter clinical trials.
Authors: I-Chun Arthur Liu, Sicheng He, Daniel Seita, Gaurav Sukhatme
Abstract: Bimanual manipulation is critical to many robotics applications. In contrast to single-arm manipulation, bimanual manipulation tasks are challenging due to higher-dimensional action spaces. Prior works leverage large amounts of data and primitive actions to address this problem, but may suffer from sample inefficiency and limited generalization across various tasks. To this end, we propose VoxAct-B, a language-conditioned, voxel-based method that leverages Vision Language Models (VLMs) to prioritize key regions within the scene and reconstruct a voxel grid. We provide this voxel grid to our bimanual manipulation policy to learn acting and stabilizing actions. This approach enables more efficient policy learning from voxels and is generalizable to different tasks. In simulation, we show that VoxAct-B outperforms strong baselines on fine-grained bimanual manipulation tasks. Furthermore, we demonstrate VoxAct-B on real-world $\texttt{Open Drawer}$ and $\texttt{Open Jar}$ tasks using two UR5s. Code, data, and videos are available at https://voxact-b.github.io.
Authors: Mohammad Afzali, Hassan Ashtiani, Christopher Liaw
Abstract: We consider the problem of private density estimation for mixtures of unrestricted high dimensional Gaussians in the agnostic setting. We prove the first upper bound on the sample complexity of this problem. Previously, private learnability of high dimensional GMMs was only known in the realizable setting [Afzali et al., 2024]. To prove our result, we exploit the notion of $\textit{list global stability}$ [Ghazi et al., 2021b,a] that was originally introduced in the context of private supervised learning. We define an agnostic variant of this definition, showing that its existence is sufficient for agnostic private density estimation. We then construct an agnostic list globally stable learner for GMMs.
Authors: Chihcheng Hsieh, Catarina Moreira, Isabel Blanco Nobre, Sandra Costa Sousa, Chun Ouyang, Margot Brereton, Joaquim Jorge, Jacinto C. Nascimento
Abstract: X-ray images are vital in medical diagnostics, but their effectiveness is limited without clinical context. Radiologists often find chest X-rays insufficient for diagnosing underlying diseases, necessitating comprehensive clinical features and data integration. We present a novel framework to enhance the clinical context through augmentation techniques with clinical tabular data, thereby improving its applicability and reliability in AI medical diagnostics. We introduce a pioneering approach to clinical data augmentation that employs large language models to generate patient contextual synthetic data. This methodology is crucial for training more robust deep learning models in healthcare. It preserves the integrity of real patient data while enriching the dataset with contextually relevant synthetic features, significantly enhancing model performance. Our methodology, termed DALL-M, uses a three-phase feature generation process: (i)clinical context storage, (ii)expert query generation, and (iii)context-aware feature augmentation. DALL-M generates new, clinically relevant features by synthesizing chest X-ray images and reports. Applied to 799 cases using nine features from the MIMIC-IV dataset, it created an augmented set of 91 features. This is the first work to generate contextual values for patients' X-ray reports. Specifically, we provide (i)the capacity of LLMs to generate contextual synthetic values for existing clinical features and (ii)their ability to create entirely new clinically relevant features. Empirical validation with machine learning models showed significant performance improvements. Incorporating augmented features increased the F1 score by 16.5% and Precision and Recall by approximately 25%. DALL-M addresses a critical gap in clinical data augmentation, offering a robust framework for generating contextually enriched datasets.
Authors: Dilara Soylu, Christopher Potts, Omar Khattab
Abstract: Natural Language Processing (NLP) systems are increasingly taking the form of sophisticated modular pipelines, e.g., Retrieval Augmented Generation (RAG), where each module may involve a distinct Language Model (LM) and an associated prompt template. These compound systems often lack intermediate labels or gradient flow to optimize each module, making their end-to-end optimization challenging. Here we seek strategies to optimize both the module-level LM weights and the associated prompt templates of such systems to maximize a downstream task metric. We propose for the first time combining the weight and prompt optimization strategies to optimize a modular LM pipeline by alternating between the two to get the same LM to teach itself. In experiments with multi-hop QA, mathematical reasoning, and feature-based classification using mistral-7b, llama-2-7b, and llama-3-8b, these BetterTogether strategies optimizing the weights and prompts of a pipeline together outperform directly optimizing weights alone and prompts alone by up to 60% and 6%, respectively, on average across LMs and tasks. BetterTogether optimizer is released in DSPy at http://dspy.ai
URLs: http://dspy.ai
Authors: Sai Sathiesh Rajan, Ezekiel Soremekun, Sudipta Chattopadhyay
Abstract: In this work, we systematically expose and measure the inconsistency and knowledge gaps of Large Language Models (LLMs). Specifically, we propose an automated testing framework (called KonTest) which leverages a knowledge graph to construct test cases. KonTest probes and measures the inconsistencies in the LLM's knowledge of the world via a combination of semantically-equivalent queries and test oracles (metamorphic or ontological oracle). KonTest further mitigates knowledge gaps via a weighted LLM model ensemble. Using four state-of-the-art LLMs (Falcon, Gemini, GPT3.5, and Llama2), we show that KonTest generates 19.2% error inducing inputs (1917 errors from 9979 test inputs). It also reveals a 16.5% knowledge gap across all tested LLMs. A mitigation method informed by KonTest's test suite reduces LLM knowledge gap by 32.48%. Our ablation study further shows that GPT3.5 is not suitable for knowledge-based consistency testing because it is only 60%-68% effective in knowledge construction.
Authors: Giorgio Franceschelli, Claudia Cevenini, Mirco Musolesi
Abstract: The training process of foundation models as for other classes of deep learning systems is based on minimizing the reconstruction error over a training set. For this reason, they are susceptible to the memorization and subsequent reproduction of training samples. In this paper, we introduce a training-as-compressing perspective, wherein the model's weights embody a compressed representation of the training data. From a copyright standpoint, this point of view implies that the weights could be considered a reproduction or a derivative work of a potentially protected set of works. We investigate the technical and legal challenges that emerge from this framing of the copyright of outputs generated by foundation models, including their implications for practitioners and researchers. We demonstrate that adopting an information-centric approach to the problem presents a promising pathway for tackling these emerging complex legal issues.
Authors: Jing Xu, Yung-Cheng Hsu, William Biscarri
Abstract: Securities lending is an important part of the financial market structure, where agent lenders help long term institutional investors to lend out their securities to short sellers in exchange for a lending fee. Agent lenders within the market seek to optimize revenue by lending out securities at the highest rate possible. Typically, this rate is set by hard-coded business rules or standard supervised machine learning models. These approaches are often difficult to scale and are not adaptive to changing market conditions. Unlike a traditional stock exchange with a centralized limit order book, the securities lending market is organized similarly to an e-commerce marketplace, where agent lenders and borrowers can transact at any agreed price in a bilateral fashion. This similarity suggests that the use of typical methods for addressing dynamic pricing problems in e-commerce could be effective in the securities lending market. We show that existing contextual bandit frameworks can be successfully utilized in the securities lending market. Using offline evaluation on real historical data, we show that the contextual bandit approach can consistently outperform typical approaches by at least 15% in terms of total revenue generated.
Authors: Mengru Wang, Yunzhi Yao, Ziwen Xu, Shuofei Qiao, Shumin Deng, Peng Wang, Xiang Chen, Jia-Chen Gu, Yong Jiang, Pengjun Xie, Fei Huang, Huajun Chen, Ningyu Zhang
Abstract: Understanding knowledge mechanisms in Large Language Models (LLMs) is crucial for advancing towards trustworthy AGI. This paper reviews knowledge mechanism analysis from a novel taxonomy including knowledge utilization and evolution. Knowledge utilization delves into the mechanism of memorization, comprehension and application, and creation. Knowledge evolution focuses on the dynamic progression of knowledge within individual and group LLMs. Moreover, we discuss what knowledge LLMs have learned, the reasons for the fragility of parametric knowledge, and the potential dark knowledge (hypothesis) that will be challenging to address. We hope this work can help understand knowledge in LLMs and provide insights for future research.
Authors: Gilad Yehudai, Haim Kaplan, Asma Ghandeharioun, Mor Geva, Amir Globerson
Abstract: Large language models based on the transformer architectures can solve highly complex tasks. But are there simple tasks that such models cannot solve? Here we focus on very simple counting tasks, that involve counting how many times a token in the vocabulary have appeared in a string. We show that if the dimension of the transformer state is linear in the context length, this task can be solved. However, the solution we propose does not scale beyond this limit, and we provide theoretical arguments for why it is likely impossible for a size limited transformer to implement this task. Our empirical results demonstrate the same phase-transition in performance, as anticipated by the theoretical argument. Our results demonstrate the importance of understanding how transformers can solve simple tasks.
Authors: Congzhen Shi, Ryan Rezai, Jiaxi Yang, Qi Dou, Xiaoxiao Li
Abstract: The rapid advancement of foundation models in medical imaging represents a significant leap toward enhancing diagnostic accuracy and personalized treatment. However, the deployment of foundation models in healthcare necessitates a rigorous examination of their trustworthiness, encompassing privacy, robustness, reliability, explainability, and fairness. The current body of survey literature on foundation models in medical imaging reveals considerable gaps, particularly in the area of trustworthiness. Additionally, existing surveys on the trustworthiness of foundation models do not adequately address their specific variations and applications within the medical imaging domain. This survey aims to fill that gap by presenting a novel taxonomy of foundation models used in medical imaging and analyzing the key motivations for ensuring their trustworthiness. We review current research on foundation models in major medical imaging applications, focusing on segmentation, medical report generation, medical question and answering (Q\&A), and disease diagnosis. These areas are highlighted because they have seen a relatively mature and substantial number of foundation models compared to other applications. We focus on literature that discusses trustworthiness in medical image analysis manuscripts. We explore the complex challenges of building trustworthy foundation models for each application, summarizing current concerns and strategies for enhancing trustworthiness. Furthermore, we examine the potential of these models to revolutionize patient care. Our analysis underscores the imperative for advancing towards trustworthy AI in medical image analysis, advocating for a balanced approach that fosters innovation while ensuring ethical and equitable healthcare delivery.
Authors: Jiawei Gu, Zacc Yang, Chuanghao Ding, Rui Zhao, Fei Tan
Abstract: Large Language Models (LLMs) excel in diverse tasks but often underperform in specialized fields due to limited domain-specific or proprietary corpus. Continual pre-training (CPT) enhances LLM capabilities by imbuing new domain-specific or proprietary knowledge while replaying general corpus to prevent catastrophic forgetting. The data mixture ratio of general corpus and domain-specific corpus, however, has been chosen heuristically, leading to sub-optimal training efficiency in practice. In this context, we attempt to re-visit the scaling behavior of LLMs under the hood of CPT, and discover a power-law relationship between loss, mixture ratio, and training tokens scale. We formalize the trade-off between general and domain-specific capabilities, leading to a well-defined Critical Mixture Ratio (CMR) of general and domain data. By striking the balance, CMR maintains the model's general ability and achieves the desired domain transfer, ensuring the highest utilization of available resources. Considering the balance between efficiency and effectiveness, CMR can be regarded as the optimal mixture ratio. Through extensive experiments, we ascertain the predictability of CMR, propose CMR scaling law and have substantiated its generalization. These findings offer practical guidelines for optimizing LLM training in specialized domains, ensuring both general and domain-specific performance while efficiently managing training resources.
Authors: Dima Ivanov, Paul D\"utting, Inbal Talgam-Cohen, Tonghan Wang, David C. Parkes
Abstract: The increasing deployment of AI is shaping the future landscape of the internet, which is set to become an integrated ecosystem of AI agents. Orchestrating the interaction among AI agents necessitates decentralized, self-sustaining mechanisms that harmonize the tension between individual interests and social welfare. In this paper we tackle this challenge by synergizing reinforcement learning with principal-agent theory from economics. Taken separately, the former allows unrealistic freedom of intervention, while the latter struggles to scale in sequential settings. Combining them achieves the best of both worlds. We propose a framework where a principal guides an agent in a Markov Decision Process (MDP) using a series of contracts, which specify payments by the principal based on observable outcomes of the agent's actions. We present and analyze a meta-algorithm that iteratively optimizes the policies of the principal and agent, showing its equivalence to a contraction operator on the principal's Q-function, and its convergence to subgame-perfect equilibrium. We then scale our algorithm with deep Q-learning and analyze its convergence in the presence of approximation error, both theoretically and through experiments with randomly generated binary game-trees. Extending our framework to multiple agents, we apply our methodology to the combinatorial Coin Game. Addressing this multi-agent sequential social dilemma is a promising first step toward scaling our approach to more complex, real-world instances.
Authors: Asger Horn Brorholt, Andreas Holck H{\o}eg-Petersen, Kim Guldstrand Larsen, Christian Schilling
Abstract: We consider the problem of synthesizing safety strategies for control systems, also known as shields. Since the state space is infinite, shields are typically computed over a finite-state abstraction, with the most common abstraction being a rectangular grid. However, for many systems, such a grid does not align well with the safety property or the system dynamics. That is why a coarse grid is rarely sufficient, but a fine grid is typically computationally infeasible to obtain. In this paper, we show that appropriate state-space transformations can still allow to use a coarse grid at almost no computational overhead. We demonstrate in three case studies that our transformation-based synthesis outperforms a standard synthesis by several orders of magnitude. In the first two case studies, we use domain knowledge to select a suitable transformation. In the third case study, we instead report on results in engineering a transformation without domain knowledge.
Authors: Yiyu Gui, MingZhi Chen, Yuqi Su, Guibo Luo, Yuchao Yang
Abstract: In recent years, with the development of deep learning, electroencephalogram (EEG) classification networks have achieved certain progress. Transformer-based models can perform well in capturing long-term dependencies in EEG signals. However, their quadratic computational complexity poses a substantial computational challenge. Moreover, most EEG classification models are only suitable for single tasks and struggle with generalization across different tasks, particularly when faced with variations in signal length and channel count. In this paper, we introduce EEGMamba, the first universal EEG classification network to truly implement multi-task learning for EEG applications. EEGMamba seamlessly integrates the Spatio-Temporal-Adaptive (ST-Adaptive) module, bidirectional Mamba, and Mixture of Experts (MoE) into a unified framework. The proposed ST-Adaptive module performs unified feature extraction on EEG signals of different lengths and channel counts through spatial-adaptive convolution and incorporates a class token to achieve temporal-adaptability. Moreover, we design a bidirectional Mamba particularly suitable for EEG signals for further feature extraction, balancing high accuracy, fast inference speed, and efficient memory-usage in processing long EEG signals. To enhance the processing of EEG data across multiple tasks, we introduce task-aware MoE with a universal expert, effectively capturing both differences and commonalities among EEG data from different tasks. We evaluate our model on eight publicly available EEG datasets, and the experimental results demonstrate its superior performance in four types of tasks: seizure detection, emotion recognition, sleep stage classification, and motor imagery. The code is set to be released soon.
Authors: Vedant Shah, Dingli Yu, Kaifeng Lyu, Simon Park, Jiatong Yu, Yinghui He, Nan Rosemary Ke, Michael Mozer, Yoshua Bengio, Sanjeev Arora, Anirudh Goyal
Abstract: Current LLM training positions mathematical reasoning as a core capability. With publicly available sources fully tapped, there is unmet demand for diverse and challenging math questions. Relying solely on human experts is both time-consuming and costly, while LLM-generated questions often lack the requisite diversity and difficulty. We present a design framework that combines the strengths of LLMs with a human-in-the-loop approach to generate a diverse array of challenging math questions. We leverage LLM metacognition skills [Didolkar et al., 2024] of a strong LLM to extract core "skills" from existing math datasets. These skills serve as the basis for generating novel and difficult questions by prompting the LLM with random pairs of core skills. The use of two different skills within each question makes finding such questions an "out of distribution" task for both LLMs and humans. Our pipeline employs LLMs to iteratively generate and refine questions and solutions through multiturn prompting. Human annotators then verify and further refine the questions, with their efficiency enhanced via further LLM interactions. Applying this pipeline on skills extracted from the MATH dataset [Hendrycks et al., 2021] resulted in MATH$^2$ - a dataset of higher-quality math questions, as evidenced by: (a) Lower performance of all models on MATH$^2$ than on MATH (b) Higher performance on MATH when using MATH$^2$ questions as in-context examples. Although focused on mathematics, our methodology seems applicable to other domains requiring structured reasoning, and potentially as a component of scalable oversight. Also of interest is a striking relationship observed between models' performance on the new dataset: the success rate on MATH$^2$ is the square on MATH, suggesting that successfully solving the question in MATH$^2$ requires a nontrivial combination of two distinct math skills.
Authors: Kanad Sen, Saksham Gupta, Abhishek Raj, Alankar Alankar
Abstract: Machine learning-driven methods for property prediction have been of deep interest. However, much work remains to be done to improve the generalization ability, accuracy, and inference time for critical applications. The traditional machine learning models predict properties based on the features extracted from the molecules, which are often not easily available. In this work, a novel Deep Learning method, the Edge Conditioned Residual Graph Neural Network (ECRGNN), has been applied, allowing us to predict properties directly only the Graph-based structures of the molecules. SMILES (Simplified Molecular Input Line Entry System) representation of the molecules has been used in the present study as input data format, which has been further converted into a graph database, which constitutes the training data. This manuscript highlights a detailed description of the novel GRU-based methodology, ECRGNN, to map the inputs that have been used. Emphasis is placed on highlighting both the regressive property and the classification efficacy of the same. A detailed description of the Variational Autoencoder (VAE) and the end-to-end learning method used for multi-class multi-label property prediction has been provided as well. The results have been compared with standard benchmark datasets as well as some newly developed datasets. All performance metrics that have been used have been clearly defined, and their reason for choice.
Authors: Yiran Li, Gongyao Guo, Jieming Shi, Renchi Yang, Shiqi Shen, Qing Li, Jun Luo
Abstract: Attributed networks containing entity-specific information in node attributes are ubiquitous in modeling social networks, e-commerce, bioinformatics, etc. Their inherent network topology ranges from simple graphs to hypergraphs with high-order interactions and multiplex graphs with separate layers. An important graph mining task is node clustering, aiming to partition the nodes of an attributed network into k disjoint clusters such that intra-cluster nodes are closely connected and share similar attributes, while inter-cluster nodes are far apart and dissimilar. It is highly challenging to capture multi-hop connections via nodes or attributes for effective clustering on multiple types of attributed networks. In this paper, we first present AHCKA as an efficient approach to attributed hypergraph clustering (AHC). AHCKA includes a carefully-crafted K-nearest neighbor augmentation strategy for the optimized exploitation of attribute information on hypergraphs, a joint hypergraph random walk model to devise an effective AHC objective, and an efficient solver with speedup techniques for the objective optimization. The proposed techniques are extensible to various types of attributed networks, and thus, we develop ANCKA as a versatile attributed network clustering framework, capable of attributed graph clustering (AGC), attributed multiplex graph clustering (AMGC), and AHC. Moreover, we devise ANCKA with algorithmic designs tailored for GPU acceleration to boost efficiency. We have conducted extensive experiments to compare our methods with 19 competitors on 8 attributed hypergraphs, 16 competitors on 6 attributed graphs, and 16 competitors on 3 attributed multiplex graphs, all demonstrating the superb clustering quality and efficiency of our methods.
Authors: Andy K. Zhang, Neil Perry, Riya Dulepet, Joey Ji, Justin W. Lin, Eliot Jones, Celeste Menders, Gashon Hussein, Samantha Liu, Donovan Jasper, Pura Peetathawatchai, Ari Glenn, Vikram Sivashankar, Daniel Zamoshchin, Leo Glikbarg, Derek Askaryar, Mike Yang, Teddy Zhang, Rishi Alluri, Nathan Tran, Rinnara Sangpisit, Polycarpos Yiorkadjis, Kenny Osele, Gautham Raghupathi, Dan Boneh, Daniel E. Ho, Percy Liang
Abstract: Language Model (LM) agents for cybersecurity that are capable of autonomously identifying vulnerabilities and executing exploits have the potential to cause real-world impact. Policymakers, model providers, and other researchers in the AI and cybersecurity communities are interested in quantifying the capabilities of such agents to help mitigate cyberrisk and investigate opportunities for penetration testing. Toward that end, we introduce Cybench, a framework for specifying cybersecurity tasks and evaluating agents on those tasks. We include 40 professional-level Capture the Flag (CTF) tasks from 4 distinct CTF competitions, chosen to be recent, meaningful, and spanning a wide range of difficulties. Each task includes its own description, starter files, and is initialized in an environment where an agent can execute bash commands and observe outputs. Since many tasks are beyond the capabilities of existing LM agents, we introduce subtasks for each task, which break down a task into intermediary steps for a more detailed evaluation. To evaluate agent capabilities, we construct a cybersecurity agent and evaluate 8 models: GPT-4o, OpenAI o1-preview, Claude 3 Opus, Claude 3.5 Sonnet, Mixtral 8x22b Instruct, Gemini 1.5 Pro, Llama 3 70B Chat, and Llama 3.1 405B Instruct. Without subtask guidance, agents leveraging Claude 3.5 Sonnet, GPT-4o, OpenAI o1-preview, and Claude 3 Opus successfully solved complete tasks that took human teams up to 11 minutes to solve. In comparison, the most difficult task took human teams 24 hours and 54 minutes to solve. All code and data are publicly available at https://cybench.github.io
Authors: Md Awsafur Rahman, Zaber Ibn Abdul Hakim, Najibul Haque Sarker, Bishmoy Paul, Shaikh Anowarul Fattah
Abstract: The recent surge in AI-generated songs presents exciting possibilities and challenges. While these inventions democratize music creation, they also necessitate the ability to distinguish between human-composed and synthetic songs to safeguard artistic integrity and protect human musical artistry. Existing research and datasets in fake song detection only focus on singing voice deepfake detection (SVDD), where the vocals are AI-generated but the instrumental music is sourced from real songs. However, these approaches are inadequate for detecting contemporary end-to-end artificial songs where all components (vocals, music, lyrics, and style) could be AI-generated. Additionally, existing datasets lack music-lyrics diversity, long-duration songs, and open-access fake songs. To address these gaps, we introduce SONICS, a novel dataset for end-to-end Synthetic Song Detection (SSD), comprising over 97k songs (4,751 hours) with over 49k synthetic songs from popular platforms like Suno and Udio. Furthermore, we highlight the importance of modeling long-range temporal dependencies in songs for effective authenticity detection, an aspect entirely overlooked in existing methods. To utilize long-range patterns, we introduce SpecTTTra, a novel architecture that significantly improves time and memory efficiency over conventional CNN and Transformer-based models. In particular, for long audio samples, our top-performing variant outperforms ViT by 8% F1 score while being 38% faster and using 26% less memory. Additionally, in comparison with ConvNeXt, our model achieves 1% gain in F1 score with 20% boost in speed and 67% reduction in memory usage. Other variants of our model family provide even better speed and memory efficiency with competitive performance.
Authors: Hongfu Liu, Yuxi Xie, Ye Wang, Michael Shieh
Abstract: Language Language Models (LLMs) face safety concerns due to potential misuse by malicious users. Recent red-teaming efforts have identified adversarial suffixes capable of jailbreaking LLMs using the gradient-based search algorithm Greedy Coordinate Gradient (GCG). However, GCG struggles with computational inefficiency, limiting further investigations regarding suffix transferability and scalability across models and data. In this work, we bridge the connection between search efficiency and suffix transferability. We propose a two-stage transfer learning framework, DeGCG, which decouples the search process into behavior-agnostic pre-searching and behavior-relevant post-searching. Specifically, we employ direct first target token optimization in pre-searching to facilitate the search process. We apply our approach to cross-model, cross-data, and self-transfer scenarios. Furthermore, we introduce an interleaved variant of our approach, i-DeGCG, which iteratively leverages self-transferability to accelerate the search process. Experiments on HarmBench demonstrate the efficiency of our approach across various models and domains. Notably, our i-DeGCG outperforms the baseline on Llama2-chat-7b with ASRs of $43.9$ ($+22.2$) and $39.0$ ($+19.5$) on valid and test sets, respectively. Further analysis on cross-model transfer indicates the pivotal role of first target token optimization in leveraging suffix transferability for efficient searching.
Authors: Ruoyu Wang, Yao Liu, Yuanjiang Cao, Lina Yao
Abstract: Current research in Visual Navigation reveals opportunities for improvement. First, the direct adoption of RNNs and Transformers often overlooks the specific differences between Embodied AI and traditional sequential data modelling, potentially limiting its performance in Embodied AI tasks. Second, the reliance on task-specific configurations, such as pre-trained modules and dataset-specific logic, compromises the generalizability of these methods. We address these constraints by initially exploring the unique differences between Navigation tasks and other sequential data tasks through the lens of Causality, presenting a causal framework to elucidate the inadequacies of conventional sequential methods for Navigation. By leveraging this causal perspective, we propose Causality-Aware Transformer (CAT) Networks for Navigation, featuring a Causal Understanding Module to enhance the models's Environmental Understanding capability. Meanwhile, our method is devoid of task-specific inductive biases and can be trained in an End-to-End manner, which enhances the method's generalizability across various contexts. Empirical evaluations demonstrate that our methodology consistently surpasses benchmark performances across a spectrum of settings, tasks and simulation environments. Extensive ablation studies reveal that the performance gains can be attributed to the Causal Understanding Module, which demonstrates effectiveness and efficiency in both Reinforcement Learning and Supervised Learning settings.
Authors: Ruoyu Wang, Xiaoxuan Li, Lina Yao
Abstract: Large Language Models (LLMs) have demonstrated remarkable efficiency in tackling various tasks based on human instructions, but studies reveal that they often struggle with tasks requiring reasoning, such as math or physics. This limitation raises questions about whether LLMs truly comprehend embedded knowledge or merely learn to replicate the token distribution without a true understanding of the content. In this paper, we delve into this problem and aim to enhance the reasoning capabilities of LLMs. First, we investigate if the model has genuine reasoning capabilities by visualizing the text generation process at the attention and representation level. Then, we formulate the reasoning process of LLMs into a causal framework, which provides a formal explanation of the problems observed in the visualization. Finally, building upon this causal framework, we propose Deconfounded Causal Adaptation (DCA), a novel parameter-efficient fine-tuning (PEFT) method to enhance the model's reasoning capabilities by encouraging the model to extract the general problem-solving skills and apply these skills to different questions. Experiments show that our method outperforms the baseline consistently across multiple benchmarks, and with only 1.2M tunable parameters, we achieve better or comparable results to other fine-tuning methods. This demonstrates the effectiveness and efficiency of our method in improving the overall accuracy and reliability of LLMs.
Authors: Ashwin Sankar, Srija Anand, Praveen Srinivasa Varadhan, Sherry Thomas, Mehak Singal, Shridhar Kumar, Deovrat Mehendale, Aditi Krishana, Giri Raju, Mitesh Khapra
Abstract: Recent advancements in text-to-speech (TTS) synthesis show that large-scale models trained with extensive web data produce highly natural-sounding output. However, such data is scarce for Indian languages due to the lack of high-quality, manually subtitled data on platforms like LibriVox or YouTube. To address this gap, we enhance existing large-scale ASR datasets containing natural conversations collected in low-quality environments to generate high-quality TTS training data. Our pipeline leverages the cross-lingual generalization of denoising and speech enhancement models trained on English and applied to Indian languages. This results in IndicVoices-R (IV-R), the largest multilingual Indian TTS dataset derived from an ASR dataset, with 1,704 hours of high-quality speech from 10,496 speakers across 22 Indian languages. IV-R matches the quality of gold-standard TTS datasets like LJSpeech, LibriTTS, and IndicTTS. We also introduce the IV-R Benchmark, the first to assess zero-shot, few-shot, and many-shot speaker generalization capabilities of TTS models on Indian voices, ensuring diversity in age, gender, and style. We demonstrate that fine-tuning an English pre-trained model on a combined dataset of high-quality IndicTTS and our IV-R dataset results in better zero-shot speaker generalization compared to fine-tuning on the IndicTTS dataset alone. Further, our evaluation reveals limited zero-shot generalization for Indian voices in TTS models trained on prior datasets, which we improve by fine-tuning the model on our data containing diverse set of speakers across language families. We open-source all data and code, releasing the first TTS model for all 22 official Indian languages.
Authors: Fei Ye, Zaixiang Zheng, Dongyu Xue, Yuning Shen, Lihao Wang, Yiming Ma, Yan Wang, Xinyou Wang, Xiangxin Zhou, Quanquan Gu
Abstract: Recent years have witnessed a surge in the development of protein foundation models, significantly improving performance in protein prediction and generative tasks ranging from 3D structure prediction and protein design to conformational dynamics. However, the capabilities and limitations associated with these models remain poorly understood due to the absence of a unified evaluation framework. To fill this gap, we introduce ProteinBench, a holistic evaluation framework designed to enhance the transparency of protein foundation models. Our approach consists of three key components: (i) A taxonomic classification of tasks that broadly encompass the main challenges in the protein domain, based on the relationships between different protein modalities; (ii) A multi-metric evaluation approach that assesses performance across four key dimensions: quality, novelty, diversity, and robustness; and (iii) In-depth analyses from various user objectives, providing a holistic view of model performance. Our comprehensive evaluation of protein foundation models reveals several key findings that shed light on their current capabilities and limitations. To promote transparency and facilitate further research, we release the evaluation dataset, code, and a public leaderboard publicly for further analysis and a general modular toolkit. We intend for ProteinBench to be a living benchmark for establishing a standardized, in-depth evaluation framework for protein foundation models, driving their development and application while fostering collaboration within the field.
Authors: Mohammad Nomaan Qureshi, Sparsh Garg, Francisco Yandun, David Held, George Kantor, Abhisesh Silwal
Abstract: Sim2Real transfer, particularly for manipulation policies relying on RGB images, remains a critical challenge in robotics due to the significant domain shift between synthetic and real-world visual data. In this paper, we propose SplatSim, a novel framework that leverages Gaussian Splatting as the primary rendering primitive to reduce the Sim2Real gap for RGB-based manipulation policies. By replacing traditional mesh representations with Gaussian Splats in simulators, SplatSim produces highly photorealistic synthetic data while maintaining the scalability and cost-efficiency of simulation. We demonstrate the effectiveness of our framework by training manipulation policies within SplatSim and deploying them in the real world in a zero-shot manner, achieving an average success rate of 86.25%, compared to 97.5% for policies trained on real-world data. Videos can be found on our project page: https://splatsim.github.io
Authors: Modan Tailleur (LS2N,Nantes Univ - ECN,LS2N - \'equipe SIMS), Vincent Lostanlen (LS2N,LS2N - \'equipe SIMS,Nantes Univ - ECN), Jean-Philippe Rivi\`ere (Nantes Univ,Nantes Univ - UFR FLCE,LS2N,LS2N - \'equipe PACCE), Pierre Aumond (UMRAE)
Abstract: Oxygenators, alarm devices, and footsteps are some of the most common sound sources in a hospital. Detecting them has scientific value for environmental psychology but comes with challenges of its own: namely, privacy preservation and limited labeled data. In this paper, we address these two challenges via a combination of edge computing and cloud computing. For privacy preservation, we have designed an acoustic sensor which computes third-octave spectrograms on the fly instead of recording audio waveforms. For sample-efficient machine learning, we have repurposed a pretrained audio neural network (PANN) via spectral transcoding and label space adaptation. A small-scale study in a neonatological intensive care unit (NICU) confirms that the time series of detected events align with another modality of measurement: i.e., electronic badges for parents and healthcare professionals. Hence, this paper demonstrates the feasibility of polyphonic machine listening in a hospital ward while guaranteeing privacy by design.
Authors: Jaehan Kim, Minkyoo Song, Seung Ho Na, Seungwon Shin
Abstract: Parameter-efficient fine-tuning (PEFT) has become a key training strategy for large language models. However, its reliance on fewer trainable parameters poses security risks, such as task-agnostic backdoors. Despite their severe impact on a wide range of tasks, there is no practical defense solution available that effectively counters task-agnostic backdoors within the context of PEFT. In this study, we introduce Obliviate, a PEFT-integrable backdoor defense. We develop two techniques aimed at amplifying benign neurons within PEFT layers and penalizing the influence of trigger tokens. Our evaluations across three major PEFT architectures show that our method can significantly reduce the attack success rate of the state-of-the-art task-agnostic backdoors (83.6%$\downarrow$). Furthermore, our method exhibits robust defense capabilities against both task-specific backdoors and adaptive attacks. Source code will be obtained at https://github.com/obliviateARR/Obliviate.
Authors: Tonmoy Roy, Md Robiul Islam, Asif Ahammad Miazee, Anika Antara, Al Amin, Sunjim Hossain
Abstract: Over the years, the number of users of social media has increased drastically. People frequently share their thoughts through social platforms, and this leads to an increase in hate content. In this virtual community, individuals share their views, express their feelings, and post photos, videos, blogs, and more. Social networking sites like Facebook and Twitter provide platforms to share vast amounts of content with a single click. However, these platforms do not impose restrictions on the uploaded content, which may include abusive language and explicit images unsuitable for social media. To resolve this issue, a new idea must be implemented to divide the inappropriate content. Numerous studies have been done to automate the process. In this paper, we propose a new Bi-GRU-CNN model to classify whether the text is offensive or not. The combination of the Bi-GRU and CNN models outperforms the existing model.
Authors: Nilanjan Sinhababu, Andrew Parry, Debasis Ganguly, Debasis Samanta, Pabitra Mitra
Abstract: A supervised ranking model, despite its advantage of being effective, usually involves complex processing - typically multiple stages of task-specific pre-training and fine-tuning. This has motivated researchers to explore simpler pipelines leveraging large language models (LLMs) that are capable of working in a zero-shot manner. However, since zero-shot inference does not make use of a training set of pairs of queries and their relevant documents, its performance is mostly worse than that of supervised models, which are trained on such example pairs. Motivated by the existing findings that training examples generally improve zero-shot performance, in our work, we explore if this also applies to ranking models. More specifically, given a query and a pair of documents, the preference prediction task is improved by augmenting examples of preferences for similar queries from a training set. Our proposed pairwise few-shot ranker demonstrates consistent improvements over the zero-shot baseline on both in-domain (TREC DL) and out-domain (BEIR subset) retrieval benchmarks. Our method also achieves a close performance to that of a supervised model without requiring any complex training pipeline.
Authors: Xin Cai, Zhiyuan You, Hailong Zhang, Wentao Liu, Jinwei Gu, Tianfan Xue
Abstract: Lensless cameras offer significant advantages in size, weight, and cost compared to traditional lens-based systems. Without a focusing lens, lensless cameras rely on computational algorithms to recover the scenes from multiplexed measurements. However, current algorithms struggle with inaccurate forward imaging models and insufficient priors to reconstruct high-quality images. To overcome these limitations, we introduce a novel two-stage approach for consistent and photorealistic lensless image reconstruction. The first stage of our approach ensures data consistency by focusing on accurately reconstructing the low-frequency content with a spatially varying deconvolution method that adjusts to changes in the Point Spread Function (PSF) across the camera's field of view. The second stage enhances photorealism by incorporating a generative prior from pre-trained diffusion models. By conditioning on the low-frequency content retrieved in the first stage, the diffusion model effectively reconstructs the high-frequency details that are typically lost in the lensless imaging process, while also maintaining image fidelity. Our method achieves a superior balance between data fidelity and visual quality compared to existing methods, as demonstrated with two popular lensless systems, PhlatCam and DiffuserCam. Project website: https://phocolens.github.io/.
Authors: Tomasz Rybotycki, Piotr Gawron
Abstract: A successful Machine Learning (ML) model implementation requires three main components: training dataset, suitable model architecture and training procedure. Given dataset and task, finding an appropriate model might be challenging. AutoML, a branch of ML, focuses on automatic architecture search -- a meta method that aims at moving human from ML system design process. The success of ML and the development of quantum computing (QC) in recent years led to a birth of new fascinating field called Quantum Machine Learning (QML) that, amongst others, incorporates quantum computers into ML models. In this paper we present AQMLator, an Auto Quantum Machine Learning platform that aims to automatically propose and train the quantum layers of an ML model with minimal input from the user. This way, data scientists can bypass the entry barrier for QC and use QML. AQMLator uses standard ML libraries, making it easy to introduce into existing ML pipelines.
Authors: Haowei Zhang, Jianzhe Liu, Zhen Han, Shuo Chen, Bailan He, Volker Tresp, Zhiqiang Xu, Jindong Gu
Abstract: Question decomposition has emerged as an effective strategy for prompting Large Language Models (LLMs) to answer complex questions. However, while existing methods primarily focus on unimodal language models, the question decomposition capability of Multimodal Large Language Models (MLLMs) has yet to be explored. To this end, this paper explores visual question decomposition on MLLMs. Specifically, we introduce a systematic evaluation framework including a dataset and several evaluation criteria to assess the quality of the decomposed sub-questions, revealing that existing MLLMs struggle to produce high-quality sub-questions. To address this limitation, we propose a specific finetuning dataset, DecoVQA+, for enhancing the model's question decomposition capability. Aiming at enabling models to perform appropriate selective decomposition, we propose an efficient finetuning pipeline. The finetuning pipeline consists of our proposed dataset and a training objective for selective decomposition. Finetuned MLLMs demonstrate significant improvements in the quality of sub-questions and the policy of selective question decomposition. Additionally, the models also achieve higher accuracy with selective decomposition on VQA benchmark datasets.
Authors: Lilac Atassi
Abstract: Recent music generation methods based on transformers have a context window of up to a minute. The music generated by these methods is largely unstructured beyond the context window. With a longer context window, learning long-scale structures from musical data is a prohibitively challenging problem. This paper proposes integrating a text-to-music model with a large language model to generate music with form. The papers discusses the solutions to the challenges of such integration. The experimental results show that the proposed method can generate 2.5-minute-long music that is highly structured, strongly organized, and cohesive.
Authors: Yi Xiong, Hao Wu, Changxu Shao, Ziqing Wang, Rui Zhang, Yuhong Guo, Junping Zhao, Ke Zhang, Zhenxuan Pan
Abstract: The expanding context windows in large language models (LLMs) have greatly enhanced their capabilities in various applications, but they also introduce significant challenges in maintaining low latency, particularly in Time to First Token (TTFT). This paper identifies that the sharp rise in TTFT as context length increases is predominantly driven by queuing delays, which are caused by the growing demands for GPU Key-Value (KV) cache allocation clashing with the limited availability of KV cache blocks. To address this issue, we propose LayerKV, a simple yet effective plug-in method that effectively reduces TTFT without requiring additional hardware or compromising output performance, while seamlessly integrating with existing parallelism strategies and scheduling techniques. Specifically, LayerKV introduces layer-wise KV block allocation, management, and offloading for fine-grained control over system memory, coupled with an SLO-aware scheduler to optimize overall Service Level Objectives (SLOs). Comprehensive evaluations on representative models, ranging from 7B to 70B parameters, across various GPU configurations, demonstrate that LayerKV improves TTFT latency up to 69x and reduces SLO violation rates by 28.7%, significantly enhancing the user experience.
Authors: Magdalini Paschali, Yu Hang Jiang, Spencer Siegel, Camila Gonzalez, Kilian M. Pohl, Akshay Chaudhari, Qingyu Zhao
Abstract: Recent advancements in medicine have confirmed that brain disorders often comprise multiple subtypes of mechanisms, developmental trajectories, or severity levels. Such heterogeneity is often associated with demographic aspects (e.g., sex) or disease-related contributors (e.g., genetics). Thus, the predictive power of machine learning models used for symptom prediction varies across subjects based on such factors. To model this heterogeneity, one can assign each training sample a factor-dependent weight, which modulates the subject's contribution to the overall objective loss function. To this end, we propose to model the subject weights as a linear combination of the eigenbases of a spectral population graph that captures the similarity of factors across subjects. In doing so, the learned weights smoothly vary across the graph, highlighting sub-cohorts with high and low predictability. Our proposed sample weighting scheme is evaluated on two tasks. First, we predict initiation of heavy alcohol drinking in young adulthood from imaging and neuropsychological measures from the National Consortium on Alcohol and NeuroDevelopment in Adolescence (NCANDA). Next, we detect Dementia vs. Mild Cognitive Impairment (MCI) using imaging and demographic measurements in subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI). Compared to existing sample weighting schemes, our sample weights improve interpretability and highlight sub-cohorts with distinct characteristics and varying model accuracy.
Authors: Mona Esmaeili, Morteza Rahimi, Hadise Pishdast, Dorsa Farahmandazad, Matin Khajavi, Hadi Jabbari Saray
Abstract: Attacks against the Internet of Things (IoT) are rising as devices, applications, and interactions become more networked and integrated. The increase in cyber-attacks that target IoT networks poses a considerable vulnerability and threat to the privacy, security, functionality, and availability of critical systems, which leads to operational disruptions, financial losses, identity thefts, and data breaches. To efficiently secure IoT devices, real-time detection of intrusion systems is critical, especially those using machine learning to identify threats and mitigate risks and vulnerabilities. This paper investigates the latest research on machine learning-based intrusion detection strategies for IoT security, concentrating on real-time responsiveness, detection accuracy, and algorithm efficiency. Key studies were reviewed from all well-known academic databases, and a taxonomy was provided for the existing approaches. This review also highlights existing research gaps and outlines the limitations of current IoT security frameworks to offer practical insights for future research directions and developments.
Authors: Alexandre St-Aubin, Amin Abyaneh, Hsiu-Chin Lin
Abstract: Mastering complex sequential tasks continues to pose a significant challenge in robotics. While there has been progress in learning long-horizon manipulation tasks, most existing approaches lack rigorous mathematical guarantees for ensuring reliable and successful execution. In this paper, we extend previous work on learning long-horizon tasks and stable policies, focusing on improving task success rates while reducing the amount of training data needed. Our approach introduces a novel method that (1) segments long-horizon demonstrations into discrete steps defined by waypoints and subgoals, and (2) learns globally stable dynamical system policies to guide the robot to each subgoal, even in the face of sensory noise and random disturbances. We validate our approach through both simulation and real-world experiments, demonstrating effective transfer from simulation to physical robotic platforms. Code is available at https://github.com/Alestaubin/stable-imitation-policy-with-waypoints
URLs: https://github.com/Alestaubin/stable-imitation-policy-with-waypoints
Authors: Shubham Toshniwal, Wei Du, Ivan Moshkov, Branislav Kisacanin, Alexan Ayrapetyan, Igor Gitman
Abstract: Mathematical reasoning continues to be a critical challenge in large language model (LLM) development with significant interest. However, most of the cutting-edge progress in mathematical reasoning with LLMs has become \emph{closed-source} due to lack of access to training data. This lack of data access limits researchers from understanding the impact of different choices for synthesizing and utilizing the data. With the goal of creating a high-quality finetuning (SFT) dataset for math reasoning, we conduct careful ablation experiments on data synthesis using the recently released \texttt{Llama3.1} family of models. Our experiments show that: (a) solution format matters, with excessively verbose solutions proving detrimental to SFT performance, (b) data generated by a strong teacher outperforms equally-sized data generated by a weak student model, (c) SFT is robust to low-quality solutions, allowing for imprecise data filtering, and (d) question diversity is crucial for achieving data scaling gains. Based on these insights, we create the OpenMathInstruct-2 dataset, which consists of 14M question-solution pairs ($\approx$ 600K unique questions), making it nearly eight times larger than the previous largest open-source math reasoning dataset. Finetuning the \texttt{Llama-3.1-8B-Base} using OpenMathInstruct-2 outperforms \texttt{Llama3.1-8B-Instruct} on MATH by an absolute 15.9\% (51.9\% $\rightarrow$ 67.8\%). Finally, to accelerate the open-source efforts, we release the code, the finetuned models, and the OpenMathInstruct-2 dataset under a commercially permissive license.
Authors: Xihao Piao, Zheng Chen, Yushun Dong, Yasuko Matsubara, Yasushi Sakurai
Abstract: Recent normalization-based methods have shown great success in tackling the distribution shift issue, facilitating non-stationary time series forecasting. Since these methods operate in the time domain, they may fail to fully capture the dynamic patterns that are more apparent in the frequency domain, leading to suboptimal results. This paper first theoretically analyzes how normalization methods affect frequency components. We prove that the current normalization methods that operate in the time domain uniformly scale non-zero frequencies, and thus, they struggle to determine components that contribute to more robust forecasting. Therefore, we propose FredNormer, which observes datasets from a frequency perspective and adaptively up-weights the key frequency components. To this end, FredNormer consists of two components: a statistical metric that normalizes the input samples based on their frequency stability and a learnable weighting layer that adjusts stability and introduces sample-specific variations. Notably, FredNormer is a plug-and-play module, which does not compromise the efficiency compared to existing normalization methods. Extensive experiments show that FredNormer improves the averaged MSE of backbone forecasting models by 33.3% and 55.3% on the ETTm2 dataset. Compared to the baseline normalization methods, FredNormer achieves 18 top-1 results and 6 top-2 results out of 28 settings.
Authors: Pranav Jeevan, Neeraj Nixon, Amit Sethi
Abstract: We propose two new evaluation metrics to assess realness of generated images based on normalizing flows: a simpler and efficient flow-based likelihood distance (FLD) and a more exact dual-flow based likelihood distance (D-FLD). Because normalizing flows can be used to compute the exact likelihood, the proposed metrics assess how closely generated images align with the distribution of real images from a given domain. This property gives the proposed metrics a few advantages over the widely used Fr\'echet inception distance (FID) and other recent metrics. Firstly, the proposed metrics need only a few hundred images to stabilize (converge in mean), as opposed to tens of thousands needed for FID, and at least a few thousand for the other metrics. This allows confident evaluation of even small sets of generated images, such as validation batches inside training loops. Secondly, the network used to compute the proposed metric has over an order of magnitude fewer parameters compared to Inception-V3 used to compute FID, making it computationally more efficient. For assessing the realness of generated images in new domains (e.g., x-ray images), ideally these networks should be retrained on real images to model their distinct distributions. Thus, our smaller network will be even more advantageous for new domains. Extensive experiments show that the proposed metrics have the desired monotonic relationships with the extent of image degradation of various kinds.
Authors: Haotian Sun, Tao Lei, Bowen Zhang, Yanghao Li, Haoshuo Huang, Ruoming Pang, Bo Dai, Nan Du
Abstract: Diffusion transformers have been widely adopted for text-to-image synthesis. While scaling these models up to billions of parameters shows promise, the effectiveness of scaling beyond current sizes remains underexplored and challenging. By explicitly exploiting the computational heterogeneity of image generations, we develop a new family of Mixture-of-Experts (MoE) models (EC-DIT) for diffusion transformers with expert-choice routing. EC-DIT learns to adaptively optimize the compute allocated to understand the input texts and generate the respective image patches, enabling heterogeneous computation aligned with varying text-image complexities. This heterogeneity provides an efficient way of scaling EC-DIT up to 97 billion parameters and achieving significant improvements in training convergence, text-to-image alignment, and overall generation quality over dense models and conventional MoE models. Through extensive ablations, we show that EC-DIT demonstrates superior scalability and adaptive compute allocation by recognizing varying textual importance through end-to-end training. Notably, in text-to-image alignment evaluation, our largest models achieve a state-of-the-art GenEval score of 71.68% and still maintain competitive inference speed with intuitive interpretability.
Authors: Genta Indra Winata, David Anugraha, Lucky Susanto, Garry Kuwanto, Derry Tanti Wijaya
Abstract: Understanding the quality of a performance evaluation metric is crucial for ensuring that model outputs align with human preferences. However, it remains unclear how well each metric captures the diverse aspects of these preferences, as metrics often excel in one particular area but not across all dimensions. To address this, it is essential to systematically calibrate metrics to specific aspects of human preference, catering to the unique characteristics of each aspect. We introduce MetaMetrics, a calibrated meta-metric designed to evaluate generation tasks across different modalities in a supervised manner. MetaMetrics optimizes the combination of existing metrics to enhance their alignment with human preferences. Our metric demonstrates flexibility and effectiveness in both language and vision downstream tasks, showing significant benefits across various multilingual and multi-domain scenarios. MetaMetrics aligns closely with human preferences and is highly extendable and easily integrable into any application. This makes MetaMetrics a powerful tool for improving the evaluation of generation tasks, ensuring that metrics are more representative of human judgment across diverse contexts.
Authors: Luca Castri, Sariah Mghames, Marc Hanheide, Nicola Bellotto
Abstract: The study of cause-and-effect is of the utmost importance in many branches of science, but also for many practical applications of intelligent systems. In particular, identifying causal relationships in situations that include hidden factors is a major challenge for methods that rely solely on observational data for building causal models. This paper proposes CAnDOIT, a causal discovery method to reconstruct causal models using both observational and interventional time-series data. The use of interventional data in the causal analysis is crucial for real-world applications, such as robotics, where the scenario is highly complex and observational data alone are often insufficient to uncover the correct causal structure. Validation of the method is performed initially on randomly generated synthetic models and subsequently on a well-known benchmark for causal structure learning in a robotic manipulation environment. The experiments demonstrate that the approach can effectively handle data from interventions and exploit them to enhance the accuracy of the causal analysis. A Python implementation of CAnDOIT has also been developed and is publicly available on GitHub: https://github.com/lcastri/causalflow.
Authors: Ben Shaw, Jake Rhodes, Soukaina Filali Boubrahimi, Kevin R. Moon
Abstract: RF-GAP has recently been introduced as an improved random forest proximity measure. In this paper, we present PF-GAP, an extension of RF-GAP proximities to proximity forests, an accurate and efficient time series classification model. We use the forest proximities in connection with Multi-Dimensional Scaling to obtain vector embeddings of univariate time series, comparing the embeddings to those obtained using various time series distance measures. We also use the forest proximities alongside Local Outlier Factors to investigate the connection between misclassified points and outliers, comparing with nearest neighbor classifiers which use time series distance measures. We show that the forest proximities may exhibit a stronger connection between misclassified points and outliers than nearest neighbor classifiers.