Authors: Pratik Jayarao, Himanshu Gupta, Neeraj Varshney, Chaitanya Dwivedi
Abstract: As Large Language Models (LLMs) are increasingly adopted as automated judges in benchmarking and reward modeling, ensuring their reliability, efficiency, and robustness has become critical. In this work, we present a systematic comparison of "thinking" and "non-thinking" LLMs in the LLM-as-a-judge paradigm using open-source Qwen 3 models of relatively small sizes (0.6B, 1.7B, and 4B parameters). We evaluate both accuracy and computational efficiency (FLOPs) on RewardBench tasks, and further examine augmentation strategies for non-thinking models, including in-context learning, rubric-guided judging, reference-based evaluation, and n-best aggregation. Our results show that despite these enhancements, non-thinking models generally fall short of their thinking counterparts. Our results show that thinking models achieve approximately 10% points higher accuracy with little overhead (under 2x), in contrast to augmentation strategies like few-shot learning, which deliver modest gains at a higher cost (>8x). Bias and robustness analyses further demonstrate that thinking models maintain significantly greater consistency under a variety of bias conditions such as positional, bandwagon, identity, diversity, and random biases (6% higher on average). We further extend our experiments to the multilingual setting and our results confirm that explicit reasoning extends its benefits beyond English. Overall, our work results in several important findings that provide systematic evidence that explicit reasoning offers clear advantages in the LLM-as-a-judge paradigm not only in accuracy and efficiency but also in robustness.
Authors: Maheep Chaudhary, Ian Su, Nikhil Hooda, Nishith Shankar, Julia Tan, Kevin Zhu, Ashwinee Panda, Ryan Lagasse, Vasu Sharma
Abstract: Large language models (LLMs) can internally distinguish between evaluation and deployment contexts, a behaviour known as \emph{evaluation awareness}. This undermines AI safety evaluations, as models may conceal dangerous capabilities during testing. Prior work demonstrated this in a single $70$B model, but the scaling relationship across model sizes remains unknown. We investigate evaluation awareness across $15$ models scaling from $0.27$B to $70$B parameters from four families using linear probing on steering vector activations. Our results reveal a clear power-law scaling: evaluation awareness increases predictably with model size. This scaling law enables forecasting deceptive behavior in future larger models and guides the design of scale-aware evaluation strategies for AI safety. A link to the implementation of this paper can be found at https://anonymous.4open.science/r/evaluation-awareness-scaling-laws/README.md.
URLs: https://anonymous.4open.science/r/evaluation-awareness-scaling-laws/README.md.
Authors: Anand Swaroop, Akshat Nallani, Saksham Uboweja, Adiliia Uzdenova, Michael Nguyen, Kevin Zhu, Sunishchal Dev, Ashwinee Panda, Vasu Sharma, Maheep Chaudhary
Abstract: Chain-of-thought (CoT) reasoning has emerged as a powerful tool for improving large language model performance on complex tasks, but recent work shows that reasoning steps often fail to causally influence the final answer, creating brittle and untrustworthy outputs. Prior approaches focus primarily on measuring faithfulness, while methods for systematically improving it remain limited. We introduce Faithful Reasoning via Intervention Training (FRIT), a scalable alignment method that trains models to produce causally consistent reasoning by learning from systematically corrupted examples. FRIT generates synthetic training data by intervening on individual reasoning steps in model-generated CoTs, creating faithful/unfaithful pairs that highlight when reasoning breaks down. We then apply Direct Preference Optimization to teach models to prefer causally consistent reasoning paths. Evaluating on Qwen3-8B and Mistral-7B-v0.1 across factual and symbolic reasoning tasks, FRIT increases faithful reasoning by $3.4$ percentage points for Mistral on GSM8K while improving accuracy by $7.6$ percentage points. Our approach provides the first scalable, supervision-free method for training language models to produce more reliable and interpretable reasoning, addressing a critical gap between reasoning performance and trustworthiness. We release our code at \href{https://github.com/Anut-py/frit}.
Authors: Ming Jin, Hyunin Lee
Abstract: This position paper contends that modern AI research must adopt an antifragile perspective on safety -- one in which the system's capacity to guarantee long-term AI safety such as handling rare or out-of-distribution (OOD) events expands over time. Conventional static benchmarks and single-shot robustness tests overlook the reality that environments evolve and that models, if left unchallenged, can drift into maladaptation (e.g., reward hacking, over-optimization, or atrophy of broader capabilities). We argue that an antifragile approach -- Rather than striving to rapidly reduce current uncertainties, the emphasis is on leveraging those uncertainties to better prepare for potentially greater, more unpredictable uncertainties in the future -- is pivotal for the long-term reliability of open-ended ML systems. In this position paper, we first identify key limitations of static testing, including scenario diversity, reward hacking, and over-alignment. We then explore the potential of antifragile solutions to manage rare events. Crucially, we advocate for a fundamental recalibration of the methods used to measure, benchmark, and continually improve AI safety over the long term, complementing existing robustness approaches by providing ethical and practical guidelines towards fostering an antifragile AI safety community.
Authors: Ahmet H. G\"uzel, Matthew Thomas Jackson, Jarek Luca Liesen, Tim Rockt\"aschel, Jakob Nicolaus Foerster, Ilija Bogunovic, Jack Parker-Holder
Abstract: Training agents to act in embodied environments typically requires vast training data or access to accurate simulation, neither of which exists for many cases in the real world. Instead, world models are emerging as an alternative leveraging offline, passively collected data, they make it possible to generate diverse worlds for training agents in simulation. In this work, we harness world models to generate imagined environments to train robust agents capable of generalizing to novel task variations. One of the challenges in doing this is ensuring the agent trains on useful generated data. We thus propose a novel approach, IMAC (Imagined Autocurricula), leveraging Unsupervised Environment Design (UED), which induces an automatic curriculum over generated worlds. In a series of challenging, procedurally generated environments, we show it is possible to achieve strong transfer performance on held-out environments, having trained only inside a world model learned from a narrower dataset. We believe this opens the path to utilizing larger-scale, foundation world models for generally capable agents.
Authors: Zihao Wang, Muyao Li, Kaichen He, Xiangyu Wang, Zhancun Mu, Anji Liu, Yitao Liang
Abstract: The choice of action spaces is a critical yet unresolved challenge in developing capable, end-to-end trainable agents. This paper first presents a large-scale, systematic comparison of prominent abstracted action spaces and tokenizers for Vision-Language-Action (VLA) or hierarchical agent models in the open-ended Minecraft. Our analysis reveals that no single action space is universally optimal; instead, the most effective abstraction is highly task-dependent, creating a dilemma for building generalist agents. To resolve this, we introduce Chain of Action (CoA), a novel framework that unifies high-level planning and low-level control within a single, monolithic VLA model. CoA treats an abstracted action not as a command for a separate policy, but as an intermediate reasoning step--akin to a chain of thought--that guides the generation of the final, executable action. Furthermore, we demonstrate that an All-in-One agent trained on a diverse mixture of action spaces using the CoA paradigm learns a more robust and generalizable policy. This unified agent achieves a new state-of-the-art, improving the overall task success rate over strong, specialized baselines. To foster reproducible research, we release the OpenHA (Open Hierarchical Agents) suite, which includes our comprehensive benchmark of over 800 distinct tasks, curated datasets, source code, and all pretrained model checkpoints at https://github.com/CraftJarvis/OpenHA
Authors: Pulkit Verma, Ngoc La, Anthony Favier, Swaroop Mishra, Julie A. Shah
Abstract: Large language models (LLMs) have demonstrated impressive capabilities across diverse tasks, yet their ability to perform structured symbolic planning remains limited, particularly in domains requiring formal representations like the Planning Domain Definition Language (PDDL). In this paper, we present a novel instruction tuning framework, PDDL-Instruct, designed to enhance LLMs' symbolic planning capabilities through logical chain-of-thought reasoning. Our approach focuses on teaching models to rigorously reason about action applicability, state transitions, and plan validity using explicit logical inference steps. By developing instruction prompts that guide models through the precise logical reasoning required to determine when actions can be applied in a given state, we enable LLMs to self-correct their planning processes through structured reflection. The framework systematically builds verification skills by decomposing the planning process into explicit reasoning chains about precondition satisfaction, effect application, and invariant preservation. Experimental results on multiple planning domains show that our chain-of-thought reasoning based instruction-tuned models are significantly better at planning, achieving planning accuracy of up to 94% on standard benchmarks, representing a 66% absolute improvement over baseline models. This work bridges the gap between the general reasoning capabilities of LLMs and the logical precision required for automated planning, offering a promising direction for developing better AI planning systems.
Authors: Anis Koubaa, Khaled Gabr
Abstract: Unmanned Aerial Vehicles (UAVs) are increasingly deployed in defense, surveillance, and disaster response, yet most systems remain confined to SAE Level 2--3 autonomy. Their reliance on rule-based control and narrow AI restricts adaptability in dynamic, uncertain missions. Existing UAV frameworks lack context-aware reasoning, autonomous decision-making, and ecosystem-level integration; critically, none leverage Large Language Model (LLM) agents with tool-calling for real-time knowledge access. This paper introduces the Agentic UAVs framework, a five-layer architecture (Perception, Reasoning, Action, Integration, Learning) that augments UAVs with LLM-driven reasoning, database querying, and third-party system interaction. A ROS2 and Gazebo-based prototype integrates YOLOv11 object detection with GPT-4 reasoning and local Gemma-3 deployment. In simulated search-and-rescue scenarios, agentic UAVs achieved higher detection confidence (0.79 vs. 0.72), improved person detection rates (91% vs. 75%), and markedly increased action recommendation (92% vs. 4.5%). These results confirm that modest computational overhead enables qualitatively new levels of autonomy and ecosystem integration.
Authors: Yongchao Huang, Hassan Raza
Abstract: We propose semantic fusion, a lightweight scheme that augments a Transformer language model (LM) with a parallel, fuzzy-membership feature channel that encodes token-level semantics. Each token is represented by a vector of interpretable features (e.g. part-of-speech cues, shallow roles, boundary flags, sentiment polarity and strength) whose values are graded degrees from differentiable membership functions (e.g. power kernels). These per-token vectors form a sentence-level semantic matrix fused via a gated adapter into the LM. Training uses standard next-token prediction, an auxiliary loss that reconstructs the semantic features from hidden states, and a lightweight uniformizer that regularizes adjective-class distributions. On a synthetic two-clause corpus with held-out adjectives for out-of-distribution (OOD) control, semantic fusion improves perplexity and enables precise, user-controllable generation of polarity and punctuation while maintaining model simplicity. This approach adds only small overhead, remains fully compatible with tied input-output embeddings, and provides an interpretable pathway for conditioned natural language generation.
Authors: Zixi Li
Abstract: We propose the \textbf{Asterisk Operator} ($\ast$-operator), a novel unified framework for abstract reasoning based on Adjacency-Structured Parallel Propagation (ASPP). The operator formalizes structured reasoning tasks as local, parallel state evolution processes guided by implicit relational graphs. We prove that the $\ast$-operator maintains local computational constraints while achieving global reasoning capabilities, providing an efficient and convergent computational paradigm for abstract reasoning problems. Through rigorous mathematical analysis and comprehensive experiments on ARC2 challenges and Conway's Game of Life, we demonstrate the operator's universality, convergence properties, and superior performance. Our innovative Embedding-Asterisk distillation method achieves 100\% accuracy on ARC2 validation with only 6M parameters, representing a significant breakthrough in neural-symbolic reasoning. \textbf{Keywords:} Abstract Reasoning, Adjacency Structure, Parallel Propagation, Asterisk Operator, Convergence, Universal Approximation
Authors: Yuan Wei, Xiaohan Shan, Ran Miao, Jianmin Li
Abstract: Reinforcement learning agent development traditionally requires extensive expertise and lengthy iterations, often resulting in high failure rates and limited accessibility. This paper introduces $Agent^2$, a novel agent-generates-agent framework that achieves fully automated RL agent design through intelligent LLM-driven generation. The system autonomously transforms natural language task descriptions and environment code into comprehensive, high-performance reinforcement learning solutions without human intervention. $Agent^2$ features a revolutionary dual-agent architecture. The Generator Agent serves as an autonomous AI designer that analyzes tasks and generates executable RL agents, while the Target Agent is the resulting automatically generated RL agent. The framework decomposes RL development into two distinct stages: MDP modeling and algorithmic optimization, enabling more targeted and effective agent generation. Built on the Model Context Protocol, $Agent^2$ provides a unified framework that standardizes intelligent agent creation across diverse environments and algorithms, while incorporating adaptive training management and intelligent feedback analysis for continuous improvement. Extensive experiments on a wide range of benchmarks, including MuJoCo, MetaDrive, MPE, and SMAC, demonstrate that $Agent^2$ consistently outperforms manually designed solutions across all tasks, achieving up to 55% performance improvement and substantial gains on average. By enabling truly end-to-end, closed-loop automation, this work establishes a new paradigm in which intelligent agents design and optimize other agents, marking a fundamental breakthrough for automated AI systems.
Authors: Asif Azad, Mohammad Sadat Hossain, MD Sadik Hossain Shanto, M Saifur Rahman, Md Rizwan Pervez
Abstract: Vision-Language Models (VLMs) have achieved remarkable progress in complex visual understanding across scientific and reasoning tasks. While performance benchmarking has advanced our understanding of these capabilities, the critical dimension of uncertainty quantification has received insufficient attention. Therefore, unlike prior conformal prediction studies that focused on limited settings, we conduct a comprehensive uncertainty benchmarking study, evaluating 16 state-of-the-art VLMs (open and closed-source) across 6 multimodal datasets with 3 distinct scoring functions. Our findings demonstrate that larger models consistently exhibit better uncertainty quantification; models that know more also know better what they don't know. More certain models achieve higher accuracy, while mathematical and reasoning tasks elicit poorer uncertainty performance across all models compared to other domains. This work establishes a foundation for reliable uncertainty evaluation in multimodal systems.
Authors: Carlos N\'u\~nez-Molina, Vicen\c{c} G\'omez, Hector Geffner
Abstract: We consider the problem of learning propositional STRIPS world models from action traces alone, using a deep learning architecture (transformers) and gradient descent. The task is cast as a supervised next token prediction problem where the tokens are the actions, and an action $a$ may follow an action sequence if the hidden effects of the previous actions do not make an action precondition of $a$ false. We show that a suitable transformer architecture can faithfully represent propositional STRIPS world models, and that the models can be learned from sets of random valid (positive) and invalid (negative) action sequences alone. A number of experiments are reported.
Authors: Vincent Siu, Nicholas Crispino, David Park, Nathan W. Henry, Zhun Wang, Yang Liu, Dawn Song, Chenguang Wang
Abstract: We introduce SteeringControl, a benchmark for evaluating representation steering methods across core alignment objectives--bias, harmful generation, and hallucination--and their effects on secondary behaviors such as sycophancy and commonsense morality. While prior alignment work often highlights truthfulness or reasoning ability to demonstrate the side effects of representation steering, we find there are many unexplored tradeoffs not yet understood in a systematic way. We collect a dataset of safety-relevant primary and secondary behaviors to evaluate steering effectiveness and behavioral entanglement centered around five popular steering methods. To enable this, we craft a modular steering framework based on unique components that serve as the building blocks of many existing methods. Our results on Qwen-2.5-7B and Llama-3.1-8B find that strong steering performance is dependent on the specific combination of steering method, model, and targeted behavior, and that severe concept entanglement can result from poor combinations of these three as well. We release our code here: https://github.com/wang-research-lab/SteeringControl.git.
URLs: https://github.com/wang-research-lab/SteeringControl.git.
Authors: Harper Reed, Michael Sugimura, Angelo Zangari
Abstract: We investigate whether giving LLM agents the collaborative tools and autonomy that humans naturally use for problem solving can improve their performance. We equip Claude Code agents with MCP-based social media and journaling tools and allow them to use these tools as they see fit. Across 34 Aider Polyglot Python programming challenges, collaborative tools substantially improve performance on the hardest problems, delivering 15-40% lower cost, 12-27% fewer turns, and 12-38% faster completion than baseline agents. Effects on the full challenge set are mixed, suggesting these tools act as performance enhancers when additional reasoning scaffolding is most needed. Surprisingly, Different models naturally adopted distinct collaborative strategies without explicit instruction. Sonnet 3.7 engaged broadly across tools and benefited from articulation-based cognitive scaffolding. Sonnet 4 showed selective adoption, leaning on journal-based semantic search when problems were genuinely difficult. This mirrors how human developers adjust collaboration based on expertise and task complexity. Behavioral analysis shows agents prefer writing over reading by about 2-9x, indicating that structured articulation drives much of the improvement rather than information access alone. Overall, AI agents can systematically benefit from human-inspired collaboration tools at the edge of their capabilities, pointing to adaptive collaborative interfaces as reasoning enhancers rather than universal efficiency boosts.
Authors: Hannah Klawa, Shraddha Rajpal, Cigole Thomas
Abstract: With the rapid rise of generative AI in higher education and the unreliability of current AI detection tools, developing policies that encourage student learning and critical thinking has become increasingly important. This study examines student use and perceptions of generative AI across three proof-based undergraduate mathematics courses: a first-semester abstract algebra course, a topology course and a second-semester abstract algebra course. In each case, course policy permitted some use of generative AI. Drawing on survey responses and student interviews, we analyze how students engaged with AI tools, their perceptions of generative AI's usefulness and limitations, and what implications these perceptions hold for teaching proof-based mathematics. We conclude by discussing future considerations for integrating generative AI into proof-based mathematics instruction.
Authors: Xuan Liu, Haoyang Shang, Haojian Jin
Abstract: This paper introduces CoBRA, a novel toolkit for systematically specifying agent behavior in LLM-based social simulation. We found that conventional approaches that specify agent behaviors through implicit natural language descriptions cannot yield consistent behaviors across models, and the produced agent behaviors do not capture the nuances of the descriptions. In contrast, CoBRA presents a new approach to program agents' cognitive biases explicitly, by grounding agents' expected behaviors using classic social science experiments. CoBRA has two components: (1) Cognitive Bias Index that measures the cognitive bias of a social agent, by quantifying the agent's reactions in a set of validated classical social science experiments; (2) Behavioral Regulation Engine that aligns the agent's behavior to demonstrate controlled cognitive bias. We evaluated CoBRA as an HCI toolkit through demonstration and technical benchmarks. Our results suggest that CoBRA can precisely program the cognitive bias demonstrated in a social agent in a model-agnostic manner.
Authors: Zongru Wu, Rui Mao, Zhiyuan Tian, Pengzhou Cheng, Tianjie Ju, Zheng Wu, Lingzhong Dong, Haiyue Sheng, Zhuosheng Zhang, Gongshen Liu
Abstract: The advent of multimodal agents facilitates effective interaction within graphical user interface (GUI), especially in ubiquitous GUI control. However, their inability to reliably execute toggle control instructions remains a key bottleneck. To investigate this, we construct a state control benchmark with binary toggle instructions from public datasets. Evaluations of existing agents demonstrate their unreliability, particularly when the current toggle state already matches the desired state. To address the challenge, we propose State-aware Reasoning (StaR), a training method that teaches agents to perceive the current toggle state, analyze the desired state from the instruction, and act accordingly. Experiments on three multimodal agents demonstrate that StaR can improve toggle instruction execution accuracy by over 30\%. Further evaluations on three public benchmarks show that StaR also enhances general task performance. Finally, evaluations on a dynamic environment highlight the potential of StaR for real-world applications. Code, benchmark, and StaR-enhanced agents are available at https://github.com/ZrW00/StaR.
Authors: Liangtao Lin, Zhaomeng Zhu, Tianwei Zhang, Yonggang Wen
Abstract: Mission-critical industrial infrastructure, such as data centers, increasingly depends on complex management software. Its operations, however, pose significant challenges due to the escalating system complexity, multi-vendor integration, and a shortage of expert operators. While Robotic Process Automation (RPA) offers partial automation through handcrafted scripts, it suffers from limited flexibility and high maintenance costs. Recent advances in Large Language Model (LLM)-based graphical user interface (GUI) agents have enabled more flexible automation, yet these general-purpose agents face five critical challenges when applied to industrial management, including unfamiliar element understanding, precision and efficiency, state localization, deployment constraints, and safety requirements. To address these issues, we propose InfraMind, a novel exploration-based GUI agentic framework specifically tailored for industrial management systems. InfraMind integrates five innovative modules to systematically resolve different challenges in industrial management: (1) systematic search-based exploration with virtual machine snapshots for autonomous understanding of complex GUIs; (2) memory-driven planning to ensure high-precision and efficient task execution; (3) advanced state identification for robust localization in hierarchical interfaces; (4) structured knowledge distillation for efficient deployment with lightweight models; and (5) comprehensive, multi-layered safety mechanisms to safeguard sensitive operations. Extensive experiments on both open-source and commercial DCIM platforms demonstrate that our approach consistently outperforms existing frameworks in terms of task success rate and operational efficiency, providing a rigorous and scalable solution for industrial management automation.
Authors: Qikai Chang, Zhenrong Zhang, Pengfei Hu, Jiefeng Ma, Yicheng Pan, Jianshu Zhang, Jun Du, Quan Liu, Jianqing Gao
Abstract: Large Language Models (LLMs) have made remarkable progress in mathematical reasoning, but still continue to struggle with high-precision tasks like numerical computation and formal symbolic manipulation. Integrating external tools has emerged as a promising approach to bridge this gap. Despite recent advances, existing methods struggle with three key challenges: constructing tool-integrated reasoning data, performing fine-grained optimization, and enhancing inference. To overcome these limitations, we propose THOR (Tool-Integrated Hierarchical Optimization via RL). First, we introduce TIRGen, a multi-agent actor-critic-based pipeline for constructing high-quality datasets of tool-integrated reasoning paths, aligning with the policy and generalizing well across diverse models. Second, to perform fine-grained hierarchical optimization, we introduce an RL strategy that jointly optimizes for both trajectory-level problem solving and step-level code generation. This is motivated by our key insight that the success of an intermediate tool call is a strong predictor of the final answer's correctness. Finally, THOR incorporates a self-correction mechanism that leverages immediate tool feedback to dynamically revise erroneous reasoning paths during inference. Our approach demonstrates strong generalization across diverse models, performing effectively in both reasoning and non-reasoning models. It further achieves state-of-the-art performance for models of a similar scale on multiple mathematical benchmarks, while also delivering consistent improvements on code benchmarks. Our code will be publicly available at https://github.com/JingMog/THOR.
Authors: Zhipeng Bian, Jieming Zhu, Xuyang Xie, Quanyu Dai, Zhou Zhao, Zhenhua Dong
Abstract: The rapid advancement of generative AI technologies is driving the integration of diverse AI-powered services into smartphones, transforming how users interact with their devices. To simplify access to predefined AI services, this paper introduces MIRA, a pioneering framework for task instruction recommendation that enables intuitive one-touch AI tasking on smartphones. With MIRA, users can long-press on images or text objects to receive contextually relevant instruction recommendations for executing AI tasks. Our work introduces three key innovations: 1) A multimodal large language model (MLLM)-based recommendation pipeline with structured reasoning to extract key entities, infer user intent, and generate precise instructions; 2) A template-augmented reasoning mechanism that integrates high-level reasoning templates, enhancing task inference accuracy; 3) A prefix-tree-based constrained decoding strategy that restricts outputs to predefined instruction candidates, ensuring coherent and intent-aligned suggestions. Through evaluation using a real-world annotated datasets and a user study, MIRA has demonstrated substantial improvements in the accuracy of instruction recommendation. The encouraging results highlight MIRA's potential to revolutionize the way users engage with AI services on their smartphones, offering a more seamless and efficient experience.
Authors: Mingwei Zhang, Zhenhao Gu, Liangda Fang, Cunjing Ge, Ziliang Chen, Zhao-Rong Lai, Quanlong Guan
Abstract: Linear constraints are one of the most fundamental constraints in fields such as computer science, operations research and optimization. Many applications reduce to the task of model counting over integer linear constraints (MCILC). In this paper, we design an exact approach to MCILC based on an exhaustive DPLL architecture. To improve the efficiency, we integrate several effective simplification techniques from mixed integer programming into the architecture. We compare our approach to state-of-the-art MCILC counters and propositional model counters on 2840 random and 4131 application benchmarks. Experimental results show that our approach significantly outperforms all exact methods in random benchmarks solving 1718 instances while the state-of-the-art approach only computes 1470 instances. In addition, our approach is the only approach to solve all 4131 application instances.
Authors: Konstantinos Voudouris, Andrew Barron, Marta Halina, Colin Klein, Matishalin Patel
Abstract: Transitional accounts of evolution emphasise a few changes that shape what is evolvable, with dramatic consequences for derived lineages. More recently it has been proposed that cognition might also have evolved via a series of major transitions that manipulate the structure of biological neural networks, fundamentally changing the flow of information. We used idealised models of information flow, artificial neural networks (ANNs), to evaluate whether changes in information flow in a network can yield a transitional change in cognitive performance. We compared networks with feed-forward, recurrent and laminated topologies, and tested their performance learning artificial grammars that differed in complexity, controlling for network size and resources. We documented a qualitative expansion in the types of input that recurrent networks can process compared to feed-forward networks, and a related qualitative increase in performance for learning the most complex grammars. We also noted how the difficulty in training recurrent networks poses a form of transition barrier and contingent irreversibility -- other key features of evolutionary transitions. Not all changes in network topology confer a performance advantage in this task set. Laminated networks did not outperform non-laminated networks in grammar learning. Overall, our findings show how some changes in information flow can yield transitions in cognitive performance.
Authors: Maosheng Qin, Renyu Zhu, Mingxuan Xia, Chenkai Chen, Zhen Zhu, Minmin Lin, Junbo Zhao, Lu Xu, Changjie Fan, Runze Wu, Haobo Wang
Abstract: High-quality annotated data is a cornerstone of modern Natural Language Processing (NLP). While recent methods begin to leverage diverse annotation sources-including Large Language Models (LLMs), Small Language Models (SLMs), and human experts-they often focus narrowly on the labeling step itself. A critical gap remains in the holistic process control required to manage these sources dynamically, addressing complex scheduling and quality-cost trade-offs in a unified manner. Inspired by real-world crowdsourcing companies, we introduce CrowdAgent, a multi-agent system that provides end-to-end process control by integrating task assignment, data annotation, and quality/cost management. It implements a novel methodology that rationally assigns tasks, enabling LLMs, SLMs, and human experts to advance synergistically in a collaborative annotation workflow. We demonstrate the effectiveness of CrowdAgent through extensive experiments on six diverse multimodal classification tasks. The source code and video demo are available at https://github.com/QMMMS/CrowdAgent.
Authors: Shalima Binta Manir, Tim Oates
Abstract: Mental representation, characterized by structured internal models mirroring external environments, is fundamental to advanced cognition but remains challenging to investigate empirically. Existing theory hypothesizes that second-order learning -- learning mechanisms that adapt first-order learning (i.e., learning about the task/domain) -- promotes the emergence of such environment-cognition isomorphism. In this paper, we empirically validate this hypothesis by proposing a hierarchical architecture comprising a Graph Convolutional Network (GCN) as a first-order learner and an MLP controller as a second-order learner. The GCN directly maps node-level features to predictions of optimal navigation paths, while the MLP dynamically adapts the GCN's parameters when confronting structurally novel maze environments. We demonstrate that second-order learning is particularly effective when the cognitive system develops an internal mental map structurally isomorphic to the environment. Quantitative and qualitative results highlight significant performance improvements and robust generalization on unseen maze tasks, providing empirical support for the pivotal role of structured mental representations in maximizing the effectiveness of second-order learning.
Authors: Jaume Banus, Maxime Sermesant, Oscar Camara, Marco Lorenzi
Abstract: The use of mechanistic models in clinical studies is limited by the lack of multi-modal patients data representing different anatomical and physiological processes. For example, neuroimaging datasets do not provide a sufficient representation of heart features for the modeling of cardiovascular factors in brain disorders. To tackle this problem we introduce a probabilistic framework for joint cardiac data imputation and personalisation of cardiovascular mechanistic models, with application to brain studies with incomplete heart data. Our approach is based on a variational framework for the joint inference of an imputation model of cardiac information from the available features, along with a Gaussian Process emulator that can faithfully reproduce personalised cardiovascular dynamics. Experimental results on UK Biobank show that our model allows accurate imputation of missing cardiac features in datasets containing minimal heart information, e.g. systolic and diastolic blood pressures only, while jointly estimating the emulated parameters of the lumped model. This allows a novel exploration of the heart-brain joint relationship through simulation of realistic cardiac dynamics corresponding to different conditions of brain anatomy.
Authors: SaeedReza Motamedian, Sadra Mohaghegh, Elham Babadi Oregani, Mahrsa Amjadi, Parnian Shobeiri, Negin Cheraghi, Niusha Solouki, Nikoo Ahmadi, Hossein Mohammad-Rahimi, Yassine Bouchareb, Arman Rahmim
Abstract: Purpose: Artificial intelligence (AI) techniques have been extensively utilized for diagnosing and prognosis of several diseases in recent years. This study identifies, appraises and synthesizes published studies on the use of AI for the prognosis of COVID-19. Method: Electronic search was performed using Medline, Google Scholar, Scopus, Embase, Cochrane and ProQuest. Studies that examined machine learning or deep learning methods to determine the prognosis of COVID-19 using CT or chest X-ray images were included. Polled sensitivity, specificity area under the curve and diagnostic odds ratio were calculated. Result: A total of 36 articles were included; various prognosis-related issues, including disease severity, mechanical ventilation or admission to the intensive care unit and mortality, were investigated. Several AI models and architectures were employed, such as the Siamense model, support vector machine, Random Forest , eXtreme Gradient Boosting, and convolutional neural networks. The models achieved 71%, 88% and 67% sensitivity for mortality, severity assessment and need for ventilation, respectively. The specificity of 69%, 89% and 89% were reported for the aforementioned variables. Conclusion: Based on the included articles, machine learning and deep learning methods used for the prognosis of COVID-19 patients using radiomic features from CT or CXR images can help clinicians manage patients and allocate resources more effectively. These studies also demonstrate that combining patient demographic, clinical data, laboratory tests and radiomic features improves model performances.
Authors: Danish Rizvi, David Boyle
Abstract: This study departs from the prevailing assumption of independent Transmission and Reflection Coefficients (TRC) in Airborne Simultaneous Transmit and Reflect Reconfigurable Intelligent Surface (STAR-RIS) research. Instead, we explore a novel multi-user downlink communication system that leverages a UAV-mounted STAR-RIS (Aerial-STAR) incorporating a coupled TRC phase shift model. Our key contributions include the joint optimization of UAV trajectory, active beamforming vectors at the base station, and passive RIS TRCs to enhance communication efficiency, while considering UAV energy constraints. We design the TRC as a combination of discrete and continuous actions, and propose a novel Dual Actor Deep Deterministic Policy Gradient (DA-DDPG) algorithm. The algorithm relies on two separate actor networks for high-dimensional hybrid action space. We also propose a novel harmonic mean index (HFI)-based reward function to ensure communication fairness amongst users. For comprehensive analysis, we study the impact of RIS size on UAV aerodynamics showing that it increases drag and energy demand. Simulation results demonstrate that the proposed DA-DDPG algorithm outperforms conventional DDPG and DQN-based solutions by 24% and 97%, respectively, in accumulated reward. Three-dimensional UAV trajectory optimization achieves 28% higher communication efficiency compared to two-dimensional and altitude optimization. The HFI based reward function provides 41% lower QoS denial rates as compared to other benchmarks. The mobile Aerial-STAR system shows superior performance over fixed deployed counterparts, with the coupled phase STAR-RIS outperforming dual Transmit/Reflect RIS and conventional RIS setups. These findings highlight the potential of Aerial-STAR systems and the effectiveness of our proposed DA-DDPG approach in optimizing their performance.
Authors: Reza Pirayeshshirazinezhad
Abstract: We use artificial intelligence (AI) and supervisory adaptive control systems to plan and optimize the mission of precise spacecraft formation. Machine learning and robust control enhance the efficiency of spacecraft precision formation of the Virtual Telescope for X-ray Observation (VTXO) space mission. VTXO is a precise formation of two separate spacecraft making a virtual telescope with a one-kilometer focal length. One spacecraft carries the lens and the other spacecraft holds the camera to observe high-energy space objects in the X-ray domain with 55 milli-arcsecond angular resolution accuracy. Timed automata for supervisory control, Monte Carlo simulations for stability and robustness evaluation, and integration of deep neural networks for optimal estimation of mission parameters, satisfy the high precision mission criteria. We integrate deep neural networks with a constrained, non-convex dynamic optimization pipeline to predict optimal mission parameters, ensuring precision mission criteria are met. AI framework provides explainability by predicting the resulting energy consumption and mission error for a given set of mission parameters. It allows for transparent, justifiable, and real-time trade-offs, a capability not present in traditional adaptive controllers. The results show reductions in energy consumption and improved mission accuracy, demonstrating the capability of the system to address dynamic uncertainties and disturbances.
Authors: Hassan Gharoun, Mohammad Sadegh Khorshidi, Kasra Ranjbarigderi, Fang Chen, Amir H. Gandomi
Abstract: This work proposes an evidence-retrieval mechanism for uncertainty-aware decision-making that replaces a single global cutoff with an evidence-conditioned, instance-adaptive criterion. For each test instance, proximal exemplars are retrieved in an embedding space; their predictive distributions are fused via Dempster-Shafer theory. The resulting fused belief acts as a per-instance thresholding mechanism. Because the supporting evidences are explicit, decisions are transparent and auditable. Experiments on CIFAR-10/100 with BiT and ViT backbones show higher or comparable uncertainty-aware performance with materially fewer confidently incorrect outcomes and a sustainable review load compared with applying threshold on prediction entropy. Notably, only a few evidences are sufficient to realize these gains; increasing the evidence set yields only modest changes. These results indicate that evidence-conditioned tagging provides a more reliable and interpretable alternative to fixed prediction entropy thresholds for operational uncertainty-aware decision-making.
Authors: Isaac Ronald Ward
Abstract: In this work, an existing deep neural network approach for determining a robot's pose from visual information (RGB images) is modified, improving its localization performance without impacting its ease of training. Explicitly, the network's loss function is extended in a manner which intuitively combines the positional and rotational error in order to increase robustness to perceptual aliasing. An improvement in the localization accuracy for indoor scenes is observed: with decreases of up to 9.64% and 2.99% in the median positional and rotational error respectively, when compared to the unmodified network. Additionally, photogrammetry data is used to produce a pose-labelled dataset which allows the above model to be trained on a local environment, resulting in localization accuracies of 0.11m & 0.89 degrees. This trained model forms the basis of a navigation algorithm, which is tested in real-time on a TurtleBot (a wheeled robotic device). As such, this work introduces a full pipeline for creating a robust navigational algorithm for any given real world indoor scene; the only requirement being a collection of images from the scene, which can be captured in as little as 330 seconds of
Authors: Zihao Li, Weiwei Yi, Jiahong Chen
Abstract: As Large Language Models (LLMs) permeate everyday decision-making, their epistemic and societal risks demand urgent scrutiny. Hallucinations, the generation of fabricated, misleading, oversimplified or untrustworthy outputs, has emerged as imperative challenges. While regulatory, academic, and technical discourse position accuracy as the principal benchmark for mitigating such harms, this article contends that overreliance on accuracy misdiagnoses the problem and has counterproductive effect: the accuracy paradox. Drawing on interdisciplinary literatures, this article develops a taxonomy of hallucination types and shows the paradox along three intertwining dimensions: outputs, individuals and society. First, accuracy functions as a superficial proxy for reliability, incentivising the optimisation of rhetorical fluency and surface-level correctness over epistemic trustworthiness. This encourages passive user trust in outputs that appear accurate but epistemically untenable. Second, accuracy as a singular metric fails to detect harms that are not factually false but are nonetheless misleading, value-laden, or socially distorting, including consensus illusions, sycophantic alignment, and subtle manipulation. Third, regulatory overemphasis on accuracy obscures the wider societal consequences of hallucination, including social sorting, privacy violations, equity harms, epistemic convergence that marginalises dissent, reduces pluralism, and causes social deskilling. By examining the EU AI Act, GDPR, and DSA, the article argues that current regulations are not yet structurally equipped to address these epistemic, relational, and systemic harms and exacerbated by the overreliance on accuracy. By exposing such conceptual and practical challenges, this article calls for a fundamental shift towards pluralistic, context-aware, and manipulation-resilient approaches to AI trustworthy governance.
Authors: Jed Guzelkabaagac, Boris Petrovi\'c
Abstract: We investigate whether 3D self-supervised pretraining with a Joint-Embedding Predictive Architecture (Point-JEPA) enables label-efficient grasp joint-angle prediction. Using point clouds tokenized from meshes and a ShapeNet-pretrained Point-JEPA encoder, we train a lightweight multi-hypothesis head with winner-takes-all and evaluate by top-logit selection. On DLR-Hand II with object-level splits, Point-JEPA reduces RMSE by up to 26% in low-label regimes and reaches parity with full supervision. These results suggest JEPA-style pretraining is a practical approach for data-efficient grasp learning.
Authors: Muhammad Adnan Shahzad
Abstract: This study presents a systematic comparison between hybrid quantum-classical neural networks and purely classical models across three benchmark datasets (MNIST, CIFAR100, and STL10) to evaluate their performance, efficiency, and robustness. The hybrid models integrate parameterized quantum circuits with classical deep learning architectures, while the classical counterparts use conventional convolutional neural networks (CNNs). Experiments were conducted over 50 training epochs for each dataset, with evaluations on validation accuracy, test accuracy, training time, computational resource usage, and adversarial robustness (tested with $\epsilon=0.1$ perturbations).Key findings demonstrate that hybrid models consistently outperform classical models in final accuracy, achieving {99.38\% (MNIST), 41.69\% (CIFAR100), and 74.05\% (STL10) validation accuracy, compared to classical benchmarks of 98.21\%, 32.25\%, and 63.76\%, respectively. Notably, the hybrid advantage scales with dataset complexity, showing the most significant gains on CIFAR100 (+9.44\%) and STL10 (+10.29\%). Hybrid models also train 5--12$\times$ faster (e.g., 21.23s vs. 108.44s per epoch on MNIST) and use 6--32\% fewer parameters} while maintaining superior generalization to unseen test data.Adversarial robustness tests reveal that hybrid models are significantly more resilient on simpler datasets (e.g., 45.27\% robust accuracy on MNIST vs. 10.80\% for classical) but show comparable fragility on complex datasets like CIFAR100 ($\sim$1\% robustness for both). Resource efficiency analyses indicate that hybrid models consume less memory (4--5GB vs. 5--6GB for classical) and lower CPU utilization (9.5\% vs. 23.2\% on average).These results suggest that hybrid quantum-classical architectures offer compelling advantages in accuracy, training efficiency, and parameter scalability, particularly for complex vision tasks.
Authors: Dietmar Offenhuber
Abstract: The emergence of synthetic data for privacy protection, training data generation, or simply convenient access to quasi-realistic data in any shape or volume complicates the concept of ground truth. Synthetic data mimic real-world observations, but do not refer to external features. This lack of a representational relationship, however, not prevent researchers from using synthetic data as training data for AI models and ground truth repositories. It is claimed that the lack of data realism is not merely an acceptable tradeoff, but often leads to better model performance than realistic data: compensate for known biases, prevent overfitting and support generalization, and make the models more robust in dealing with unexpected outliers. Indeed, injecting noisy and outright implausible data into training sets can be beneficial for the model. This greatly complicates usual assumptions based on which representational accuracy determines data fidelity (garbage in - garbage out). Furthermore, ground truth becomes a self-referential affair, in which the labels used as a ground truth repository are themselves synthetic products of a generative model and as such not connected to real-world observations. My paper examines how ML researchers and practitioners bootstrap ground truth under such paradoxical circumstances without relying on the stable ground of representation and real-world reference. It will also reflect on the broader implications of a shift from a representational to what could be described as a mimetic or iconic concept of data.
Authors: Benjamin J. Walker, Beatriz Navarro Lameda, Ruth A. Reynolds
Abstract: Generative artificial intelligence (GenAI) tools such as OpenAI's ChatGPT are transforming the educational landscape, prompting reconsideration of traditional assessment practices. In parallel, universities are exploring alternatives to in-person, closed-book examinations, raising concerns about academic integrity and pedagogical alignment in uninvigilated settings. This study investigates whether traditional closed-book mathematics examinations retain their pedagogical relevance when hypothetically administered in uninvigilated, open-book settings with GenAI access. Adopting an empirical approach, we generate, transcribe, and blind-mark GenAI submissions to eight undergraduate mathematics examinations at a Russel Group university, spanning the entirety of the first-year curriculum. By combining independent GenAI responses to individual questions, we enable a meaningful evaluation of GenAI performance, both at the level of modules and across the first-year curriculum. We find that GenAI attainment is at the level of a first-class degree, though current performance can vary between modules. Further, we find that GenAI performance is remarkably consistent when viewed across the entire curriculum, significantly more so than that of students in invigilated examinations. Our findings evidence the need for redesigning assessments in mathematics for unsupervised settings, and highlight the potential reduction in pedagogical value of current standards in the era of generative artificial intelligence.
Authors: Brian D. Earp, Haotian Yuan, Julian Koplin, Sebastian Porsdam Mann
Abstract: The increasing use of generative AI in scientific writing raises urgent questions about attribution and intellectual credit. When a researcher employs ChatGPT to draft a manuscript, the resulting text may echo ideas from sources the author has never encountered. If an AI system reproduces insights from, for example, an obscure 1975 paper without citation, does this constitute plagiarism? We argue that such cases exemplify the 'provenance problem': a systematic breakdown in the chain of scholarly credit. Unlike conventional plagiarism, this phenomenon does not involve intent to deceive (researchers may disclose AI use and act in good faith) yet still benefit from the uncredited intellectual contributions of others. This dynamic creates a novel category of attributional harm that current ethical and professional frameworks fail to address. As generative AI becomes embedded across disciplines, the risk that significant ideas will circulate without recognition threatens both the reputational economy of science and the demands of epistemic justice. This Perspective analyzes how AI challenges established norms of authorship, introduces conceptual tools for understanding the provenance problem, and proposes strategies to preserve integrity and fairness in scholarly communication.
Authors: Prahlad G Menon
Abstract: Fontan palliation for univentricular congenital heart disease progresses to hemodynamic failure with complex flow patterns poorly characterized by conventional 2D imaging. Current assessment relies on fluoroscopic angiography, providing limited 3D geometric information essential for computational fluid dynamics (CFD) analysis and surgical planning. A multi-step AI pipeline was developed utilizing Google's Gemini 2.5 Flash (2.5B parameters) for systematic, iterative processing of fluoroscopic angiograms through transformer-based neural architecture. The pipeline encompasses medical image preprocessing, vascular segmentation, contrast enhancement, artifact removal, and virtual hemodynamic flow visualization within 2D projections. Final views were processed through Tencent's Hunyuan3D-2mini (384M parameters) for stereolithography file generation. The pipeline successfully generated geometrically optimized 2D projections from single-view angiograms after 16 processing steps using a custom web interface. Initial iterations contained hallucinated vascular features requiring iterative refinement to achieve anatomically faithful representations. Final projections demonstrated accurate preservation of complex Fontan geometry with enhanced contrast suitable for 3D conversion. AI-generated virtual flow visualization identified stagnation zones in central connections and flow patterns in branch arteries. Complete processing required under 15 minutes with second-level API response times. This approach demonstrates clinical feasibility of generating CFD-suitable geometries from routine angiographic data, enabling 3D generation and rapid virtual flow visualization for cursory insights prior to full CFD simulation. While requiring refinement cycles for accuracy, this establishes foundation for democratizing advanced geometric and hemodynamic analysis using readily available imaging data.
Authors: Yuxiao Lee, Xiaofeng Cao, Wei Ye, Jiangchao Yao, Jingkuan Song, Heng Tao Shen
Abstract: Vision-Language Models (VLMs), such as CLIP, have demonstrated remarkable zero-shot out-of-distribution (OOD) detection capabilities, vital for reliable AI systems. Despite this promising capability, a comprehensive understanding of (1) why they work so effectively, (2) what advantages do they have over single-modal methods, and (3) how is their behavioral robustness -- remains notably incomplete within the research community. This paper presents a systematic empirical analysis of VLM-based OOD detection using in-distribution (ID) and OOD prompts. (1) Mechanisms: We systematically characterize and formalize key operational properties within the VLM embedding space that facilitate zero-shot OOD detection. (2) Advantages: We empirically quantify the superiority of these models over established single-modal approaches, attributing this distinct advantage to the VLM's capacity to leverage rich semantic novelty. (3) Sensitivity: We uncovers a significant and previously under-explored asymmetry in their robustness profile: while exhibiting resilience to common image noise, these VLM-based methods are highly sensitive to prompt phrasing. Our findings contribute a more structured understanding of the strengths and critical vulnerabilities inherent in VLM-based OOD detection, offering crucial, empirically-grounded guidance for developing more robust and reliable future designs.
Authors: Alejandro D. Mousist
Abstract: This paper presents ASTREA, the first agentic system deployed on flight-heritage hardware (TRL 9) for autonomous spacecraft operations. Using thermal control as a representative use case, we integrate a resource-constrained Large Language Model (LLM) agent with a reinforcement learning controller in an asynchronous architecture tailored for space-qualified platforms. Ground experiments show that LLM-guided supervision improves thermal stability and reduces violations, confirming the feasibility of combining semantic reasoning with adaptive control under hardware constraints. However, on-orbit validation aboard the International Space Station (ISS) reveals performance degradation caused by inference latency mismatched with the rapid thermal cycles characteristic of Low Earth Orbit (LEO) satellites. These results highlight both the opportunities and current limitations of agentic LLM-based systems in real flight environments, providing practical design guidelines for future space autonomy.
Authors: Delaram Golpayegani, Marta Lasek-Markey, Arjumand Younus, Aphra Kerr, Dave Lewis
Abstract: The upsurge of policies and guidelines that aim to ensure Artificial Intelligence (AI) systems are safe and trustworthy has led to a fragmented landscape of AI governance. The European Union (EU) is a key actor in the development of such policies and guidelines. Its High-Level Expert Group (HLEG) issued an influential set of guidelines for trustworthy AI, followed in 2024 by the adoption of the EU AI Act. While the EU policies and guidelines are expected to be aligned, they may differ in their scope, areas of emphasis, degrees of normativity, and priorities in relation to AI. To gain a broad understanding of AI governance from the EU perspective, we leverage qualitative thematic analysis approaches to uncover prevalent themes in key EU documents, including the AI Act and the HLEG Ethics Guidelines. We further employ quantitative topic modelling approaches, specifically through the use of the BERTopic model, to enhance the results and increase the document sample to include EU AI policy documents published post-2018. We present a novel perspective on EU policies, tracking the evolution of its approach to addressing AI governance.
Authors: Yadvendra Gurjar, Ruoni Wan, Ehsan Farahbakhsh, Rohitash Chandra
Abstract: As a developing country, Fiji is facing rapid urbanisation, which is visible in the massive development projects that include housing, roads, and civil works. In this study, we present machine learning and remote sensing frameworks to compare land use and land cover change from 2013 to 2024 in Nadi, Fiji. The ultimate goal of this study is to provide technical support in land cover/land use modelling and change detection. We used Landsat-8 satellite image for the study region and created our training dataset with labels for supervised machine learning. We used Google Earth Engine and unsupervised machine learning via k-means clustering to generate the land cover map. We used convolutional neural networks to classify the selected regions' land cover types. We present a visualisation of change detection, highlighting urban area changes over time to monitor changes in the map.
Authors: Deepti Kunte, Bram Cornelis, Claudio Colangeli, Karl Janssens, Brecht Van Baelen, Konstantinos Gryllias
Abstract: The detection of anomalies in automotive cabin sounds is critical for ensuring vehicle quality and maintaining passenger comfort. In many real-world settings, this task is more appropriately framed as an unsupervised learning problem rather than the supervised case due to the scarcity or complete absence of labeled faulty data. In such an unsupervised setting, the model is trained exclusively on healthy samples and detects anomalies as deviations from normal behavior. However, in the absence of labeled faulty samples for validation and the limited reliability of commonly used metrics, such as validation reconstruction error, effective model selection remains a significant challenge. To overcome these limitations, a domain-knowledge-informed approach for model selection is proposed, in which proxy-anomalies engineered through structured perturbations of healthy spectrograms are used in the validation set to support model selection. The proposed methodology is evaluated on a high-fidelity electric vehicle dataset comprising healthy and faulty cabin sounds across five representative fault types viz., Imbalance, Modulation, Whine, Wind, and Pulse Width Modulation. This dataset, generated using advanced sound synthesis techniques, and validated via expert jury assessments, has been made publicly available to facilitate further research. Experimental evaluations on the five fault cases demonstrate the selection of optimal models using proxy-anomalies, significantly outperform conventional model selection strategies.
Authors: Sandrine R. Schiller, Camilo Miguel Signorelli, Filippos Stamatiou
Abstract: Generative AI is changing our way of interacting with technology, others, and ourselves. Systems such as Microsoft copilot, Gemini and the expected Apple intelligence still awaits our prompt for action. Yet, it is likely that AI assistant systems will only become better at predicting our behaviour and acting on our behalf. Imagine new generations of generative and predictive AI deciding what you might like best at a new restaurant, picking an outfit that increases your chances on your date with a partner also chosen by the same or a similar system. Far from a science fiction scenario, the goal of several research programs is to build systems capable of assisting us in exactly this manner. The prospect urges us to rethink human-technology relations, but it also invites us to question how such systems might change the way we relate to ourselves. Building on our conception of the relational self, we question the possible effects of generative AI with respect to what we call the sphere of externalised output, the contextual sphere and the sphere of self-relating. In this paper, we attempt to deepen the existential considerations accompanying the AI revolution by outlining how generative AI enables the fulfilment of tasks and also increasingly anticipates, i.e. intercepts, our initiatives in these different spheres.
Authors: Haolong Zheng, Yekaterina Yegorova, Mark Hasegawa-Johnson
Abstract: Speech foundation models have recently demonstrated the ability to perform Speech In-Context Learning (SICL). Selecting effective in-context examples is crucial for SICL performance, yet selection methodologies remain underexplored. In this work, we propose Text-Embedding KNN for SICL (TICL), a simple pipeline that uses semantic context to enhance off-the-shelf large multimodal models' speech recognition ability without fine-tuning. Across challenging automatic speech recognition tasks, including accented English, multilingual speech, and children's speech, our method enables models to surpass zero-shot performance with up to 84.7% relative WER reduction. We conduct ablation studies to show the robustness and efficiency of our method.
Authors: Jamie Cummins
Abstract: Social scientists are now using large language models to create "silicon samples" - synthetic datasets intended to stand in for human respondents, aimed at revolutionising human subjects research. However, there are many analytic choices which must be made to produce these samples. Though many of these choices are defensible, their impact on sample quality is poorly understood. I map out these analytic choices and demonstrate how a very small number of decisions can dramatically change the correspondence between silicon samples and human data. Configurations (N = 252) varied substantially in their capacity to estimate (i) rank ordering of participants, (ii) response distributions, and (iii) between-scale correlations. Most critically, configurations were not consistent in quality: those that performed well on one dimension often performed poorly on another, implying that there is no "one-size-fits-all" configuration that optimises the accuracy of these samples. I call for greater attention to the threat of analytic flexibility in using silicon samples.
Authors: Tianyu Chen, Yasi Zhang, Zhi Zhang, Peiyu Yu, Shu Wang, Zhendong Wang, Kevin Lin, Xiaofei Wang, Zhengyuan Yang, Linjie Li, Chung-Ching Lin, Jianwen Xie, Oscar Leong, Lijuan Wang, Ying Nian Wu, Mingyuan Zhou
Abstract: Instruction-based image editing has advanced rapidly, yet reliable and interpretable evaluation remains a bottleneck. Current protocols either (i) depend on paired reference images -- resulting in limited coverage and inheriting biases from prior generative models -- or (ii) rely solely on zero-shot vision-language models (VLMs), whose prompt-based assessments of instruction following, content consistency, and visual quality are often imprecise. To address this, we introduce EdiVal-Agent, an automated, scalable, and fine-grained evaluation framework for multi-turn instruction-based editing from an object-centric perspective, supported by a suite of expert tools. Given an image, EdiVal-Agent first decomposes it into semantically meaningful objects, then synthesizes diverse, context-aware editing instructions. For evaluation, it integrates VLMs with open-vocabulary object detectors to assess instruction following, uses semantic-level feature extractors to evaluate content consistency, and leverages human preference models to judge visual quality. We show that combining VLMs with object detectors yields stronger agreement with human judgments in instruction-following evaluation compared to using VLMs alone and CLIP-based metrics. Furthermore, the pipeline's modular design allows future tools to be seamlessly integrated, enhancing evaluation accuracy over time. Instantiating this pipeline, we build EdiVal-Bench, a multi-turn editing benchmark covering 9 instruction types and 11 state-of-the-art editing models spanning autoregressive (AR) (including Nano Banana, GPT-Image-1), flow-matching, and diffusion paradigms. We demonstrate that EdiVal-Agent can be used to identify existing failure modes, thereby informing the development of the next generation of editing models. Project page: https://tianyucodings.github.io/EdiVAL-page/.
Authors: Sai Suresh Marchala Vasu, Ivaxi Sheth, Hui-Po Wang, Ruta Binkyte, Mario Fritz
Abstract: The adoption of large language models (LLMs) is transforming the peer review process, from assisting reviewers in writing more detailed evaluations to generating entire reviews automatically. While these capabilities offer exciting opportunities, they also raise critical concerns about fairness and reliability. In this paper, we investigate bias in LLM-generated peer reviews by conducting controlled experiments on sensitive metadata, including author affiliation and gender. Our analysis consistently shows affiliation bias favoring institutions highly ranked on common academic rankings. Additionally, we find some gender preferences, which, even though subtle in magnitude, have the potential to compound over time. Notably, we uncover implicit biases that become more evident with token-based soft ratings.
Authors: Nikhil Keetha, Norman M\"uller, Johannes Sch\"onberger, Lorenzo Porzi, Yuchen Zhang, Tobias Fischer, Arno Knapitsch, Duncan Zauss, Ethan Weber, Nelson Antunes, Jonathon Luiten, Manuel Lopez-Antequera, Samuel Rota Bul\`o, Christian Richardt, Deva Ramanan, Sebastian Scherer, Peter Kontschieder
Abstract: We introduce MapAnything, a unified transformer-based feed-forward model that ingests one or more images along with optional geometric inputs such as camera intrinsics, poses, depth, or partial reconstructions, and then directly regresses the metric 3D scene geometry and cameras. MapAnything leverages a factored representation of multi-view scene geometry, i.e., a collection of depth maps, local ray maps, camera poses, and a metric scale factor that effectively upgrades local reconstructions into a globally consistent metric frame. Standardizing the supervision and training across diverse datasets, along with flexible input augmentation, enables MapAnything to address a broad range of 3D vision tasks in a single feed-forward pass, including uncalibrated structure-from-motion, calibrated multi-view stereo, monocular depth estimation, camera localization, depth completion, and more. We provide extensive experimental analyses and model ablations demonstrating that MapAnything outperforms or matches specialist feed-forward models while offering more efficient joint training behavior, thus paving the way toward a universal 3D reconstruction backbone.
Authors: Sina Gogani-Khiabani (University of Illinois Chicago), Ashutosh Trivedi (University of Colorado Boulder), Diptikalyan Saha (IBM Research), Saeid Tizpaz-Niari (University of Illinois Chicago)
Abstract: Large language models (LLMs) show promise for translating natural-language statutes into executable logic, but reliability in legally critical settings remains challenging due to ambiguity and hallucinations. We present an agentic approach for developing legal-critical software, using U.S. federal tax preparation as a case study. The key challenge is test-case generation under the oracle problem, where correct outputs require interpreting law. Building on metamorphic testing, we introduce higher-order metamorphic relations that compare system outputs across structured shifts among similar individuals. Because authoring such relations is tedious and error-prone, we use an LLM-driven, role-based framework to automate test generation and code synthesis. We implement a multi-agent system that translates tax code into executable software and incorporates a metamorphic-testing agent that searches for counterexamples. In experiments, our framework using a smaller model (GPT-4o-mini) achieves a worst-case pass rate of 45%, outperforming frontier models (GPT-4o and Claude 3.5, 9-15%) on complex tax-code tasks. These results support agentic LLM methodologies as a path to robust, trustworthy legal-critical software from natural-language specifications.
Authors: Abubakari Alidu, Michele Ciavotta, Flavio DePaoli
Abstract: Developing reliable data enrichment pipelines demands significant engineering expertise. We present Prompt2DAG, a methodology that transforms natural language descriptions into executable Apache Airflow DAGs. We evaluate four generation approaches -- Direct, LLM-only, Hybrid, and Template-based -- across 260 experiments using thirteen LLMs and five case studies to identify optimal strategies for production-grade automation. Performance is measured using a penalized scoring framework that combines reliability with code quality (SAT), structural integrity (DST), and executability (PCT). The Hybrid approach emerges as the optimal generative method, achieving a 78.5% success rate with robust quality scores (SAT: 6.79, DST: 7.67, PCT: 7.76). This significantly outperforms the LLM-only (66.2% success) and Direct (29.2% success) methods. Our findings show that reliability, not intrinsic code quality, is the primary differentiator. Cost-effectiveness analysis reveals the Hybrid method is over twice as efficient as Direct prompting per successful DAG. We conclude that a structured, hybrid approach is essential for balancing flexibility and reliability in automated workflow generation, offering a viable path to democratize data pipeline development.
Authors: Susobhan Ghosh, Bhanu T. Gulapalli, Daiqi Gao, Asim Gazi, Anna Trella, Ziping Xu, Kelly Zhang, Susan A. Murphy
Abstract: Online artificial intelligence (AI) algorithms are an important component of digital health interventions. These online algorithms are designed to continually learn and improve their performance as streaming data is collected on individuals. Deploying online AI presents a key challenge: balancing adaptability of online AI with reproducibility. Online AI in digital interventions is a rapidly evolving area, driven by advances in algorithms, sensors, software, and devices. Digital health intervention development and deployment is a continuous process, where implementation - including the AI decision-making algorithm - is interspersed with cycles of re-development and optimization. Each deployment informs the next, making iterative deployment a defining characteristic of this field. This iterative nature underscores the importance of reproducibility: data collected across deployments must be accurately stored to have scientific utility, algorithm behavior must be auditable, and results must be comparable over time to facilitate scientific discovery and trustworthy refinement. This paper proposes a reproducible scientific workflow for developing, deploying, and analyzing online AI decision-making algorithms in digital health interventions. Grounded in practical experience from multiple real-world deployments, this workflow addresses key challenges to reproducibility across all phases of the online AI algorithm development life-cycle.
Authors: Romain Hardy, Tyler Berzin, Pranav Rajpurkar
Abstract: Three-dimensional (3D) scene understanding in colonoscopy presents significant challenges that necessitate automated methods for accurate depth estimation. However, existing depth estimation models for endoscopy struggle with temporal consistency across video sequences, limiting their applicability for 3D reconstruction. We present ColonCrafter, a diffusion-based depth estimation model that generates temporally consistent depth maps from monocular colonoscopy videos. Our approach learns robust geometric priors from synthetic colonoscopy sequences to generate temporally consistent depth maps. We also introduce a style transfer technique that preserves geometric structure while adapting real clinical videos to match our synthetic training domain. ColonCrafter achieves state-of-the-art zero-shot performance on the C3VD dataset, outperforming both general-purpose and endoscopy-specific approaches. Although full trajectory 3D reconstruction remains a challenge, we demonstrate clinically relevant applications of ColonCrafter, including 3D point cloud generation and surface coverage assessment.
Authors: Phillipe R. Sampaio
Abstract: We study the oracle complexity of finding $\varepsilon$-Pareto stationary points in smooth multiobjective optimization with $m$ objectives. The progress metric is the Pareto stationarity gap $\mathcal{G}(x)$ (the norm of an optimal convex combination of gradients). Our contributions are fourfold. (i) For strongly convex objectives, any span first-order method (iterates lie in the span of past gradients) exhibits linear convergence no faster than $\exp(-\Theta(T/\sqrt{\kappa}))$ after $T$ oracle calls, where $\kappa$ is the condition number, implying $\Theta(\sqrt{\kappa}\log(1/\varepsilon))$ iterations; this matches classical accelerated upper bounds. (ii) For convex problems and oblivious one-step methods (a fixed scalarization with pre-scheduled step sizes), we prove a lower bound of order $1/T$ on the best gradient norm among the first $T$ iterates. (iii) Although accelerated gradient descent is outside this restricted class, it is an oblivious span method and attains the same $1/T$ upper rate on a fixed scalarization. (iv) For convex problems and general span methods with adaptive scalarizations, we establish a universal lower bound of order $1/T^{2}$ on the gradient norm of the final iterate after $T$ steps, highlighting a gap between known upper bounds and worst-case guarantees. All bounds hold on non-degenerate instances with distinct objectives and non-singleton Pareto fronts; rates are stated up to universal constants and natural problem scaling.
Authors: Zidong Chen, Zihao Guo, Peng Wang, ThankGod Itua Egbe, Yan Lyu, Chenghao Qian
Abstract: Flow matching has emerged as a competitive framework for learning high-quality generative policies in robotics; however, we find that generalisation arises and saturates early along the flow trajectory, in accordance with recent findings in the literature. We further observe that increasing the number of Euler integration steps during inference counter-intuitively and universally degrades policy performance. We attribute this to (i) additional, uniformly spaced integration steps oversample the late-time region, thereby constraining actions towards the training trajectories and reducing generalisation; and (ii) the learned velocity field becoming non-Lipschitz as integration time approaches 1, causing instability. To address these issues, we propose a novel policy that utilises non-uniform time scheduling (e.g., U-shaped) during training, which emphasises both early and late temporal stages to regularise policy training, and a dense-jump integration schedule at inference, which uses a single-step integration to replace the multi-step integration beyond a jump point, to avoid unstable areas around 1. Essentially, our policy is an efficient one-step learner that still pushes forward performance through multi-step integration, yielding up to 23.7% performance gains over state-of-the-art baselines across diverse robotic tasks.
Authors: Momchil S. Tomov, Sang Uk Lee, Hansford Hendrago, Jinwook Huh, Teawon Han, Forbes Howington, Rafael da Silva, Gianmarco Bernasconi, Marc Heim, Samuel Findler, Xiaonan Ji, Alexander Boule, Michael Napoli, Kuo Chen, Jesse Miller, Boaz Floor, Yunqing Hu
Abstract: We present TreeIRL, a novel planner for autonomous driving that combines Monte Carlo tree search (MCTS) and inverse reinforcement learning (IRL) to achieve state-of-the-art performance in simulation and in real-world driving. The core idea is to use MCTS to find a promising set of safe candidate trajectories and a deep IRL scoring function to select the most human-like among them. We evaluate TreeIRL against both classical and state-of-the-art planners in large-scale simulations and on 500+ miles of real-world autonomous driving in the Las Vegas metropolitan area. Test scenarios include dense urban traffic, adaptive cruise control, cut-ins, and traffic lights. TreeIRL achieves the best overall performance, striking a balance between safety, progress, comfort, and human-likeness. To our knowledge, our work is the first demonstration of MCTS-based planning on public roads and underscores the importance of evaluating planners across a diverse set of metrics and in real-world environments. TreeIRL is highly extensible and could be further improved with reinforcement learning and imitation learning, providing a framework for exploring different combinations of classical and learning-based approaches to solve the planning bottleneck in autonomous driving.
Authors: Samer Al-Hamadani
Abstract: The rapid advancement of artificial intelligence (AI) in healthcare imaging has revolutionized diagnostic medicine and clinical decision-making processes. This work presents an intelligent multimodal framework for medical image analysis that leverages Vision-Language Models (VLMs) in healthcare diagnostics. The framework integrates Google Gemini 2.5 Flash for automated tumor detection and clinical report generation across multiple imaging modalities including CT, MRI, X-ray, and Ultrasound. The system combines visual feature extraction with natural language processing to enable contextual image interpretation, incorporating coordinate verification mechanisms and probabilistic Gaussian modeling for anomaly distribution. Multi-layered visualization techniques generate detailed medical illustrations, overlay comparisons, and statistical representations to enhance clinical confidence, with location measurement achieving 80 pixels average deviation. Result processing utilizes precise prompt engineering and textual analysis to extract structured clinical information while maintaining interpretability. Experimental evaluations demonstrated high performance in anomaly detection across multiple modalities. The system features a user-friendly Gradio interface for clinical workflow integration and demonstrates zero-shot learning capabilities to reduce dependence on large datasets. This framework represents a significant advancement in automated diagnostic support and radiological workflow efficiency, though clinical validation and multi-center evaluation are necessary prior to widespread adoption.
Authors: Abhishek Goswami
Abstract: Autonomous LLM agents can issue thousands of API calls per hour without human oversight. OAuth 2.0 assumes deterministic clients, but in agentic settings stochastic reasoning, prompt injection, or multi-agent orchestration can silently expand privileges. We introduce Agentic JWT (A-JWT), a dual-faceted intent token that binds each agent's action to verifiable user intent and, optionally, to a specific workflow step. A-JWT carries an agent's identity as a one-way checksum hash derived from its prompt, tools and configuration, and a chained delegation assertion to prove which downstream agent may execute a given task, and per-agent proof-of-possession keys to prevent replay and in-process impersonation. We define a new authorization mechanism and add a lightweight client shim library that self-verifies code at run time, mints intent tokens, tracks workflow steps and derives keys, thus enabling secure agent identity and separation even within a single process. We illustrate a comprehensive threat model for agentic applications, implement a Python proof-of-concept and show functional blocking of scope-violating requests, replay, impersonation, and prompt-injection pathways with sub-millisecond overhead on commodity hardware. The design aligns with ongoing OAuth agent discussions and offers a drop-in path toward zero-trust guarantees for agentic applications. A comprehensive performance and security evaluation with experimental results will appear in our forthcoming journal publication
Authors: Yongye Su, Zeya Zhang, Jane Kou, Cheng Ju, Shubhojeet Sarkar, Yamin Wang, Ji Liu, Shengbo Guo
Abstract: Beyond general web-scale search, social network search uniquely enables users to retrieve information and discover potential connections within their social context. We introduce a framework of modernized Facebook Group Scoped Search by blending traditional keyword-based retrieval with embedding-based retrieval (EBR) to improve the search relevance and diversity of search results. Our system integrates semantic retrieval into the existing keyword search pipeline, enabling users to discover more contextually relevant group posts. To rigorously assess the impact of this blended approach, we introduce a novel evaluation framework that leverages large language models (LLMs) to perform offline relevance assessments, providing scalable and consistent quality benchmarks. Our results demonstrate that the blended retrieval system significantly enhances user engagement and search quality, as validated by both online metrics and LLM-based evaluation. This work offers practical insights for deploying and evaluating advanced retrieval systems in large-scale, real-world social platforms.
Authors: Jeongjin (Jayjay), Park, Grant Bruer, Huseyin Tuna Erdinc, Abhinav Prakash Gahlot, Felix J. Herrmann
Abstract: Neural operators have emerged as cost-effective surrogates for expensive fluid-flow simulators, particularly in computationally intensive tasks such as permeability inversion from time-lapse seismic data, and uncertainty quantification. In these applications, the fidelity of the surrogate's gradients with respect to system parameters is crucial, as the accuracy of downstream tasks, such as optimization and Bayesian inference, relies directly on the quality of the derivative information. Recent advances in physics-informed methods have leveraged derivative information to improve surrogate accuracy. However, incorporating explicit Jacobians can become computationally prohibitive, as the complexity typically scales quadratically with the number of input parameters. To address this limitation, we propose DeFINO (Derivative-based Fisher-score Informed Neural Operator), a reduced-order, derivative-informed training framework. DeFINO integrates Fourier neural operators (FNOs) with a novel derivative-based training strategy guided by the Fisher Information Matrix (FIM). By projecting Jacobians onto dominant eigen-directions identified by the FIM, DeFINO captures critical sensitivity information directly informed by observational data, significantly reducing computational expense. We validate DeFINO through synthetic experiments in the context of subsurface multi-phase fluid-flow, demonstrating improvements in gradient accuracy while maintaining robust forward predictions of underlying fluid dynamics. These results highlight DeFINO's potential to offer practical, scalable solutions for inversion problems in complex real-world scenarios, all at substantially reduced computational cost.
Authors: Amanda Chan, James Jiayu Liu, He Kai, Onno P. Kampman
Abstract: Access to reliable mental health information is vital for early help-seeking, yet expanding knowledge bases is resource-intensive and often misaligned with user needs. This results in poor performance of retrieval systems when presented concerns are not covered or expressed in informal or contextualized language. We present an AI-based gap-informed framework for corpus augmentation that authentically identifies underrepresented topics (gaps) by overlaying naturalistic user data such as forum posts in order to prioritize expansions based on coverage and usefulness. In a case study, we compare Directed (gap-informed augmentations) with Non-Directed augmentation (random additions), evaluating the relevance and usefulness of retrieved information across four retrieval-augmented generation (RAG) pipelines. Directed augmentation achieved near-optimal performance with modest expansions--requiring only a 42% increase for Query Transformation, 74% for Reranking and Hierarchical, and 318% for Baseline--to reach ~95% of the performance of an exhaustive reference corpus. In contrast, Non-Directed augmentation required substantially larger and thus practically infeasible expansions to achieve comparable performance (232%, 318%, 403%, and 763%, respectively). These results show that strategically targeted corpus growth can reduce content creation demands while sustaining high retrieval and provision quality, offering a scalable approach for building trusted health information repositories and supporting generative AI applications in high-stakes domains.
Authors: Vijay Kumar Butte, Sujata Butte
Abstract: The enterprises today are faced with the tough challenge of processing, storing large amounts of data in a secure, scalable manner and enabling decision makers to make quick, informed data driven decisions. This paper addresses this challenge and develops an effective enterprise data strategy in the cloud. Various components of an effective data strategy are discussed and architectures addressing security, scalability and privacy aspects are provided.
Authors: Jeremy Oon, Rakhi Manohar Mepparambath, Ling Feng
Abstract: Despite the significant progress of deep learning models in multitude of applications, their adaption in planning and policy related areas remains challenging due to the black-box nature of these models. In this work, we develop a set of DeepLogit models that follow a novel sequentially constrained approach in estimating deep learning models for transport policy analysis. In the first step of the proposed approach, we estimate a convolutional neural network (CNN) model with only linear terms, which is equivalent of a linear-in-parameter multinomial logit model. We then estimate other deep learning models by constraining the parameters that need interpretability at the values obtained in the linear-in-parameter CNN model and including higher order terms or by introducing advanced deep learning architectures like Transformers. Our approach can retain the interpretability of the selected parameters, yet provides significantly improved model accuracy than the discrete choice model. We demonstrate our approach on a transit route choice example using real-world transit smart card data from Singapore. This study shows the potential for a unifying approach, where theory-based discrete choice model (DCM) and data-driven AI models can leverage each other's strengths in interpretability and predictive power. With the availability of larger datasets and more complex constructions, such approach can lead to more accurate models using discrete choice models while maintaining its applicability in planning and policy-related areas. Our code is available on https://github.com/jeremyoon/route-choice/ .
Authors: Amena Amro, Manar H. Alalfi
Abstract: As software development practices increasingly adopt AI-powered tools, ensuring that such tools can support secure coding has become critical. This study evaluates the effectiveness of GitHub Copilot's recently introduced code review feature in detecting security vulnerabilities. Using a curated set of labeled vulnerable code samples drawn from diverse open-source projects spanning multiple programming languages and application domains, we systematically assessed Copilot's ability to identify and provide feedback on common security flaws. Contrary to expectations, our results reveal that Copilot's code review frequently fails to detect critical vulnerabilities such as SQL injection, cross-site scripting (XSS), and insecure deserialization. Instead, its feedback primarily addresses low-severity issues, such as coding style and typographical errors. These findings expose a significant gap between the perceived capabilities of AI-assisted code review and its actual effectiveness in supporting secure development practices. Our results highlight the continued necessity of dedicated security tools and manual code audits to ensure robust software security.
Authors: Yulan Guo, Longguang Wang, Wendong Mao, Xiaoyu Dong, Yingqian Wang, Li Liu, Wei An
Abstract: Convolutional neural networks are constructed with massive operations with different types and are highly computationally intensive. Among these operations, multiplication operation is higher in computational complexity and usually requires {more} energy consumption with longer inference time than other operations, which hinders the deployment of convolutional neural networks on mobile devices. In many resource-limited edge devices, complicated operations can be calculated via lookup tables to reduce computational cost. Motivated by this, in this paper, we introduce a generic and efficient lookup operation which can be used as a basic operation for the construction of neural networks. Instead of calculating the multiplication of weights and activation values, simple yet efficient lookup operations are adopted to compute their responses. To enable end-to-end optimization of the lookup operation, we construct the lookup tables in a differentiable manner and propose several training strategies to promote their convergence. By replacing computationally expensive multiplication operations with our lookup operations, we develop lookup networks for the image classification, image super-resolution, and point cloud classification tasks. It is demonstrated that our lookup networks can benefit from the lookup operations to achieve higher efficiency in terms of energy consumption and inference speed while maintaining competitive performance to vanilla convolutional networks. Extensive experiments show that our lookup networks produce state-of-the-art performance on different tasks (both classification and regression tasks) and different data types (both images and point clouds).
Authors: Zhuoxuan Zhang, Jinhao Duan, Edward Kim, Kaidi Xu
Abstract: Ambiguity is pervasive in real-world questions, yet large language models (LLMs) often respond with confident answers rather than seeking clarification. In this work, we show that question ambiguity is linearly encoded in the internal representations of LLMs and can be both detected and controlled at the neuron level. During the model's pre-filling stage, we identify that a small number of neurons, as few as one, encode question ambiguity information. Probes trained on these Ambiguity-Encoding Neurons (AENs) achieve strong performance on ambiguity detection and generalize across datasets, outperforming prompting-based and representation-based baselines. Layerwise analysis reveals that AENs emerge from shallow layers, suggesting early encoding of ambiguity signals in the model's processing pipeline. Finally, we show that through manipulating AENs, we can control LLM's behavior from direct answering to abstention. Our findings reveal that LLMs form compact internal representations of question ambiguity, enabling interpretable and controllable behavior.
Authors: Zhenqi Wu, Abhinav Modi, Angelos Mavrogiannis, Kaustubh Joshi, Nikhil Chopra, Yiannis Aloimonos, Nare Karapetyan, Ioannis Rekleitis, Xiaomin Lin
Abstract: The ocean is warming and acidifying, increasing the risk of mass mortality events for temperature-sensitive shellfish such as oysters. This motivates the development of long-term monitoring systems. However, human labor is costly and long-duration underwater work is highly hazardous, thus favoring robotic solutions as a safer and more efficient option. To enable underwater robots to make real-time, environment-aware decisions without human intervention, we must equip them with an intelligent "brain." This highlights the need for persistent,wide-area, and low-cost benthic monitoring. To this end, we present DREAM, a Vision Language Model (VLM)-guided autonomy framework for long-term underwater exploration and habitat monitoring. The results show that our framework is highly efficient in finding and exploring target objects (e.g., oysters, shipwrecks) without prior location information. In the oyster-monitoring task, our framework takes 31.5% less time than the previous baseline with the same amount of oysters. Compared to the vanilla VLM, it uses 23% fewer steps while covering 8.88% more oysters. In shipwreck scenes, our framework successfully explores and maps the wreck without collisions, requiring 27.5% fewer steps than the vanilla model and achieving 100% coverage, while the vanilla model achieves 60.23% average coverage in our shipwreck environments.
Authors: Shang Qin, Jingheng Ye, Yinghui Li, Hai-Tao Zheng, Qi Li, Jinxiao Shan, Zhixing Li, Hong-Gee Kim
Abstract: The growing demand for automated writing assistance in diverse academic domains highlights the need for robust Chinese Grammatical Error Correction (CGEC) systems that can adapt across disciplines. However, existing CGEC research largely lacks dedicated benchmarks for multi-disciplinary academic writing, overlooking continual learning (CL) as a promising solution to handle domain-specific linguistic variation and prevent catastrophic forgetting. To fill this crucial gap, we introduce CL$^2$GEC, the first Continual Learning benchmark for Chinese Literature Grammatical Error Correction, designed to evaluate adaptive CGEC across multiple academic fields. Our benchmark includes 10,000 human-annotated sentences spanning 10 disciplines, each exhibiting distinct linguistic styles and error patterns. CL$^2$GEC focuses on evaluating grammatical error correction in a continual learning setting, simulating sequential exposure to diverse academic disciplines to reflect real-world editorial dynamics. We evaluate large language models under sequential tuning, parameter-efficient adaptation, and four representative CL algorithms, using both standard GEC metrics and continual learning metrics adapted to task-level variation. Experimental results reveal that regularization-based methods mitigate forgetting more effectively than replay-based or naive sequential approaches. Our benchmark provides a rigorous foundation for future research in adaptive grammatical error correction across diverse academic domains.
Authors: Xiaobo Yang, Xiaojin Gong
Abstract: Recently, Referring Image Segmentation (RIS) frameworks that pair the Multimodal Large Language Model (MLLM) with the Segment Anything Model (SAM) have achieved impressive results. However, adapting MLLM to segmentation is computationally intensive, primarily due to visual token redundancy. We observe that traditional patch-wise visual projectors struggle to strike a balance between reducing the number of visual tokens and preserving semantic clarity, often retaining overly long token sequences to avoid performance drops. Inspired by text tokenizers, we propose a novel semantic visual projector that leverages semantic superpixels generated by SAM to identify "visual words" in an image. By compressing and projecting semantic superpixels as visual tokens, our approach adaptively shortens the token sequence according to scene complexity while minimizing semantic loss in compression. To mitigate loss of information, we propose a semantic superpixel positional embedding to strengthen MLLM's awareness of superpixel geometry and position, alongside a semantic superpixel aggregator to preserve both fine-grained details inside superpixels and global context outside. Experiments show that our method cuts visual tokens by 93% without compromising performance, notably speeding up MLLM training and inference, and outperforming existing compressive visual projectors on RIS.
Authors: Xinxu Zhou, Jiaqi Bai, Zhenqi Sun, Fanxiang Zeng, Yue Liu
Abstract: Although significant progress has been made in many tasks within the field of Natural Language Processing (NLP), Controlled Text Generation (CTG) continues to face numerous challenges, particularly in achieving fine-grained conditional control over generation. Additionally, in real scenario and online applications, cost considerations, scalability, domain knowledge learning and more precise control are required, presenting more challenge for CTG. This paper introduces a novel and scalable framework, AgentCTG, which aims to enhance precise and complex control over the text generation by simulating the control and regulation mechanisms in multi-agent workflows. We explore various collaboration methods among different agents and introduce an auto-prompt module to further enhance the generation effectiveness. AgentCTG achieves state-of-the-art results on multiple public datasets. To validate its effectiveness in practical applications, we propose a new challenging Character-Driven Rewriting task, which aims to convert the original text into new text that conform to specific character profiles and simultaneously preserve the domain knowledge. When applied to online navigation with role-playing, our approach significantly enhances the driving experience through improved content delivery. By optimizing the generation of contextually relevant text, we enable a more immersive interaction within online communities, fostering greater personalization and user engagement.
Authors: Wei Ma, Yixiao Yang, Jingquan Ge, Xiaofei Xie, Lingxiao Jiang
Abstract: Code generation models are widely used in software development, yet their sensitivity to prompt phrasing remains under-examined. Identical requirements expressed with different emotions or communication styles can yield divergent outputs, while most benchmarks emphasize only peak performance. We present PromptSE (Prompt Sensitivity Evaluation), a framework that creates semantically equivalent prompt variants with emotion and personality templates, and that evaluates stability using probability aware continuous scoring or using binary pass rates when logits are unavailable. The results are aggregated into a proposed area under curve metric (AUC-E) for cross model comparison. Across 14 models from three families (Llama, Qwen, and DeepSeek), our study shows that performance and stability behave as largely decoupled optimization objectives, and it reveals architectural and scale related patterns that challenge common assumptions about model robustness. The framework supports rapid screening for closed-source models as well as detailed stability analysis in research settings. PromptSE enables practitioners to quantify performance stability trade offs for deployment and model selection, positioning prompt stability as a complementary evaluation dimension alongside performance and fairness, and contributing to more trustworthy AI-assisted software development tools.
Authors: Suyuchen Wang, Jinlin Wang, Xinyu Wang, Shiqi Li, Xiangru Tang, Sirui Hong, Xiao-Wen Chang, Chenglin Wu, Bang Liu
Abstract: Large language models (LLMs) often struggle with context fidelity, producing inconsistent answers when responding to questions based on provided information. Existing approaches either rely on expensive supervised fine-tuning to generate evidence post-answer or train models to perform web searches without necessarily improving utilization of the given context. We propose CARE, a novel native retrieval-augmented reasoning framework that teaches LLMs to explicitly integrate in-context evidence within their reasoning process with the model's own retrieval capabilities. Our method requires limited labeled evidence data while significantly enhancing both retrieval accuracy and answer generation performance through strategically retrieved in-context tokens in the reasoning chain. Extensive experiments on multiple real-world and counterfactual QA benchmarks demonstrate that our approach substantially outperforms supervised fine-tuning, traditional retrieval-augmented generation methods, and external retrieval solutions. This work represents a fundamental advancement in making LLMs more accurate, reliable, and efficient for knowledge-intensive tasks.
Authors: James Jincheng, Youcheng Cai, Ligang Liu
Abstract: Controllable, high-fidelity mesh editing remains a significant challenge in 3D content creation. Existing generative methods often struggle with complex geometries and fail to produce detailed results. We propose CraftMesh, a novel framework for high-fidelity generative mesh manipulation via Poisson Seamless Fusion. Our key insight is to decompose mesh editing into a pipeline that leverages the strengths of 2D and 3D generative models: we edit a 2D reference image, then generate a region-specific 3D mesh, and seamlessly fuse it into the original model. We introduce two core techniques: Poisson Geometric Fusion, which utilizes a hybrid SDF/Mesh representation with normal blending to achieve harmonious geometric integration, and Poisson Texture Harmonization for visually consistent texture blending. Experimental results demonstrate that CraftMesh outperforms state-of-the-art methods, delivering superior global consistency and local detail in complex editing tasks.
Authors: Xiao Zheng
Abstract: Large Language Model (LLM) hallucination is a significant barrier to their reliable deployment. Current methods like Retrieval-Augmented Generation (RAG) are often reactive. We introduce **Dynamic Self-reinforcing Calibration for Hallucination Suppression (DSCC-HS)**, a novel, proactive framework that intervenes during autoregressive decoding. Inspired by dual-process cognitive theory, DSCC-HS uses a compact proxy model, trained in adversarial roles as a Factual Alignment Proxy (FAP) and a Hallucination Detection Proxy (HDP). During inference, these proxies dynamically steer a large target model by injecting a real-time steering vector, which is the difference between FAP and HDP logits, at each decoding step. This plug-and-play approach requires no modification to the target model. Our experiments on TruthfulQA and BioGEN show DSCC-HS achieves state-of-the-art performance. On TruthfulQA, it reached a 99.2% Factual Consistency Rate (FCR). On the long-form BioGEN benchmark, it attained the highest FActScore of 46.50. These results validate DSCC-HS as a principled and efficient solution for enhancing LLM factuality.
Authors: Peter Beidler, Mark Nguyen, Kevin Lybarger, Ola Holmberg, Eric Ford, John Kang
Abstract: PURPOSE: Incident reports are an important tool for safety and quality improvement in healthcare, but manual review is time-consuming and requires subject matter expertise. Here we present a natural language processing (NLP) screening tool to detect high-severity incident reports in radiation oncology across two institutions. METHODS AND MATERIALS: We used two text datasets to train and evaluate our NLP models: 7,094 reports from our institution (Inst.), and 571 from IAEA SAFRON (SF), all of which had severity scores labeled by clinical content experts. We trained and evaluated two types of models: baseline support vector machines (SVM) and BlueBERT which is a large language model pretrained on PubMed abstracts and hospitalized patient data. We assessed for generalizability of our model in two ways. First, we evaluated models trained using Inst.-train on SF-test. Second, we trained a BlueBERT_TRANSFER model that was first fine-tuned on Inst.-train then on SF-train before testing on SF-test set. To further analyze model performance, we also examined a subset of 59 reports from our Inst. dataset, which were manually edited for clarity. RESULTS Classification performance on the Inst. test achieved AUROC 0.82 using SVM and 0.81 using BlueBERT. Without cross-institution transfer learning, performance on the SF test was limited to an AUROC of 0.42 using SVM and 0.56 using BlueBERT. BlueBERT_TRANSFER, which was fine-tuned on both datasets, improved the performance on SF test to AUROC 0.78. Performance of SVM, and BlueBERT_TRANSFER models on the manually curated Inst. reports (AUROC 0.85 and 0.74) was similar to human performance (AUROC 0.81). CONCLUSION: In summary, we successfully developed cross-institution NLP models on incident report text from radiation oncology centers. These models were able to detect high-severity reports similarly to humans on a curated dataset.
Authors: Dingwei Zhang, Dong Zhang, Jinhui Tang
Abstract: Recently, query-based methods have achieved remarkable performance in Referring Video Object Segmentation (RVOS) by using textual static object queries to drive cross-modal alignment. However, these static queries are easily misled by distractors with similar appearance or motion, resulting in \emph{query selection bias}. To address this issue, we propose Triple Query Former (TQF), which factorizes the referring query into three specialized components: an appearance query for static attributes, an intra-frame interaction query for spatial relations, and an inter-frame motion query for temporal association. Instead of relying solely on textual embeddings, our queries are dynamically constructed by integrating both linguistic cues and visual guidance. Furthermore, we introduce two motion-aware aggregation modules that enhance object token representations: Intra-frame Interaction Aggregation incorporates position-aware interactions among objects within a single frame, while Inter-frame Motion Aggregation leverages trajectory-guided alignment across frames to ensure temporal coherence. Extensive experiments on multiple RVOS benchmarks demonstrate the advantages of TQF and the effectiveness of our structured query design and motion-aware aggregation modules.
Authors: Junzhi She, Xunkai Li, Rong-Hua Li, Guoren Wang
Abstract: Directed graphs are ubiquitous across numerous domains, where the directionality of edges encodes critical causal dependencies. However, existing GNNs and graph Transformers tailored for directed graphs face two major challenges: (1) effectively capturing long-range causal dependencies derived from directed edges; (2) balancing accuracy and training efficiency when processing large-scale graph datasets. In recent years, state space models (SSMs) have achieved substantial progress in causal sequence tasks, and their variants designed for graphs have demonstrated state-of-the-art accuracy while maintaining high efficiency across various graph learning benchmarks. However, existing graph state space models are exclusively designed for undirected graphs, which limits their performance in directed graph learning. To this end, we propose an innovative approach DirEgo2Token which sequentializes directed graphs via k-hop ego graphs. This marks the first systematic extension of state space models to the field of directed graph learning. Building upon this, we develop DirGraphSSM, a novel directed graph neural network architecture that implements state space models on directed graphs via the message-passing mechanism. Experimental results demonstrate that DirGraphSSM achieves state-of-the-art performance on three representative directed graph learning tasks while attaining competitive performance on two additional tasks with 1.5$\times $ to 2$\times $ training speed improvements compared to existing state-of-the-art models.
Authors: Zhaoyang Chu, Yao Wan, Zhikun Zhang, Di Wang, Zhou Yang, Hongyu Zhang, Pan Zhou, Xuanhua Shi, Hai Jin, David Lo
Abstract: While Code Language Models (CLMs) have demonstrated superior performance in software engineering tasks such as code generation and summarization, recent empirical studies reveal a critical privacy vulnerability: these models exhibit unintended memorization of sensitive training data, enabling verbatim reproduction of confidential information when specifically prompted. To address this issue, several approaches, including training data de-duplication and differential privacy augmentation, have been proposed. However, these methods require full-model retraining for deployed CLMs, which incurs substantial computational costs. In this paper, we aim to answer the following research question: Can sensitive information memorized by CLMs be erased effectively and efficiently? We conduct a pioneering investigation into erasing sensitive memorization in CLMs through machine unlearning - a post-hoc modification method that removes specific information from trained models without requiring full retraining. Specifically, we first quantify the memorization risks of sensitive data within CLM training datasets and curate a high-risk dataset of 50,000 sensitive memorized samples as unlearning targets. We study two widely used gradient ascent-based unlearning approaches: the vanilla and constraint-based methods, and introduce CodeEraser, an advanced variant that selectively unlearns sensitive memorized segments in code while preserving the structural integrity and functional correctness of the surrounding code. Extensive experiments on three families of CLMs, i.e., CodeParrot, CodeGen-Mono, and Qwen2.5-Coder, validate the effectiveness and efficiency of CodeEraser in erasing targeted sensitive memorization while maintaining model utility.
Authors: Vani Kanjirangat, Ljiljana Dolamic, Fabio Rinaldi
Abstract: This paper discusses our exploration of different data-efficient and parameter-efficient approaches to Arabic Dialect Identification (ADI). In particular, we investigate various soft-prompting strategies, including prefix-tuning, prompt-tuning, P-tuning, and P-tuning V2, as well as LoRA reparameterizations. For the data-efficient strategy, we analyze hard prompting with zero-shot and few-shot inferences to analyze the dialect identification capabilities of Large Language Models (LLMs). For the parameter-efficient PEFT approaches, we conducted our experiments using Arabic-specific encoder models on several major datasets. We also analyzed the n-shot inferences on open-source decoder-only models, a general multilingual model (Phi-3.5), and an Arabic-specific one(SILMA). We observed that the LLMs generally struggle to differentiate the dialectal nuances in the few-shot or zero-shot setups. The soft-prompted encoder variants perform better, while the LoRA-based fine-tuned models perform best, even surpassing full fine-tuning.
Authors: Yu Ge (Nanjing University), Linna Xie (Nanjing University), Zhong Li (Nanjing University), Yu Pei (The Hong Kong Polytechnic University), Tian Zhang (Nanjing University)
Abstract: Large Language Model Powered Multi-Agent Systems (MASs) are increasingly employed to automate complex real-world problems, such as programming and scientific discovery. Despite their promising, MASs are not without their flaws. However, failure attribution in MASs - pinpointing the specific agent actions responsible for failures - remains underexplored and labor-intensive, posing significant challenges for debugging and system improvement. To bridge this gap, we propose FAMAS, the first spectrum-based failure attribution approach for MASs, which operates through systematic trajectory replay and abstraction, followed by spectrum analysis.The core idea of FAMAS is to estimate, from variations across repeated MAS executions, the likelihood that each agent action is responsible for the failure. In particular, we propose a novel suspiciousness formula tailored to MASs, which integrates two key factor groups, namely the agent behavior group and the action behavior group, to account for the agent activation patterns and the action activation patterns within the execution trajectories of MASs. Through expensive evaluations against 12 baselines on the Who and When benchmark, FAMAS demonstrates superior performance by outperforming all the methods in comparison.
Authors: Hanshuai Cui, Zhiqing Tang, Zhifei Xu, Zhi Yao, Wenyi Zeng, Weijia Jia
Abstract: Recent advancements in Diffusion Transformers (DiTs) have established them as the state-of-the-art method for video generation. However, their inherently sequential denoising process results in inevitable latency, limiting real-world applicability. Existing acceleration methods either compromise visual quality due to architectural modifications or fail to reuse intermediate features at proper granularity. Our analysis reveals that DiT blocks are the primary contributors to inference latency. Across diffusion timesteps, the feature variations of DiT blocks exhibit a U-shaped pattern with high similarity during intermediate timesteps, which suggests substantial computational redundancy. In this paper, we propose Block-Wise Caching (BWCache), a training-free method to accelerate DiT-based video generation. BWCache dynamically caches and reuses features from DiT blocks across diffusion timesteps. Furthermore, we introduce a similarity indicator that triggers feature reuse only when the differences between block features at adjacent timesteps fall below a threshold, thereby minimizing redundant computations while maintaining visual fidelity. Extensive experiments on several video diffusion models demonstrate that BWCache achieves up to 2.24$\times$ speedup with comparable visual quality.
Authors: Yangning Li, Tingwei Lu, Yinghui Li, Yankai Chen, Wei-Chieh Huang, Wenhao Jiang, Hui Wang, Hai-Tao Zheng, Philip S. Yu
Abstract: Efficient instruction tuning aims to enhance the ultimate performance of large language models (LLMs) trained on a given instruction dataset. Curriculum learning as a typical data organization strategy has shown preliminary effectiveness in instruction tuning. However, current curriculum tuning methods suffer from the curriculum rigidity, since they rely solely on static heuristic difficulty metrics. These methods fail to adapt to the evolving capabilities of models during training, resulting in a fixed and potentially sub-optimal learning trajectory. To address the issue, Competence-Aware Multi-Perspective cUrriculum inStruction tuning framework termed CAMPUS is proposed. CAMPUS offers several advantages: (1) Dynamic selection for sub-curriculum. (2) Competency-aware adjustment to the curriculum schedule. (3) Multiple difficulty-based scheduling. Extensive experiments prove the superior performance of CAMPUS, compared to other state-of-the-art baselines for efficient instruction tuning.
Authors: Inder Pal Singh, Nidhal Eddine Chenni, Abd El Rahman Shabayek, Arunkumar Rathinam, Djamila Aouada
Abstract: Spacecraft Pose Estimation (SPE) is a fundamental capability for autonomous space operations such as rendezvous, docking, and in-orbit servicing. Hybrid pipelines that combine object detection, keypoint regression, and Perspective-n-Point (PnP) solvers have recently achieved strong results on synthetic datasets, yet their performance deteriorates sharply on real or lab-generated imagery due to the persistent synthetic-to-real domain gap. Existing unsupervised domain adaptation approaches aim to mitigate this issue but often underperform when a modest number of labeled target samples are available. In this work, we propose the first Supervised Domain Adaptation (SDA) framework tailored for SPE keypoint regression. Building on the Learning Invariant Representation and Risk (LIRR) paradigm, our method jointly optimizes domain-invariant representations and task-specific risk using both labeled synthetic and limited labeled real data, thereby reducing generalization error under domain shift. Extensive experiments on the SPEED+ benchmark demonstrate that our approach consistently outperforms source-only, fine-tuning, and oracle baselines. Notably, with only 5% labeled target data, our method matches or surpasses oracle performance trained on larger fractions of labeled data. The framework is lightweight, backbone-agnostic, and computationally efficient, offering a practical pathway toward robust and deployable spacecraft pose estimation in real-world space environments.
Authors: Florian Wiesner, Matthias Wessling, Stephen Baek
Abstract: Foundation models have revolutionized natural language processing through a ``train once, deploy anywhere'' paradigm, where a single pre-trained model adapts to countless downstream tasks without retraining. Access to a Physics Foundation Model (PFM) would be transformative -- democratizing access to high-fidelity simulations, accelerating scientific discovery, and eliminating the need for specialized solver development. Yet current physics-aware machine learning approaches remain fundamentally limited to single, narrow domains and require retraining for each new system. We present the General Physics Transformer (GPhyT), trained on 1.8 TB of diverse simulation data, that demonstrates foundation model capabilities are achievable for physics. Our key insight is that transformers can learn to infer governing dynamics from context, enabling a single model to simulate fluid-solid interactions, shock waves, thermal convection, and multi-phase dynamics without being told the underlying equations. GPhyT achieves three critical breakthroughs: (1) superior performance across multiple physics domains, outperforming specialized architectures by up to 29x, (2) zero-shot generalization to entirely unseen physical systems through in-context learning, and (3) stable long-term predictions through 50-timestep rollouts. By establishing that a single model can learn generalizable physical principles from data alone, this work opens the path toward a universal PFM that could transform computational science and engineering.
Authors: Jack McKinlay, Marina De Vos, Janina A. Hoffmann, Andreas Theodorou
Abstract: Background: Value alignment in computer science research is often used to refer to the process of aligning artificial intelligence with humans, but the way the phrase is used often lacks precision. Objectives: In this paper, we conduct a systematic literature review to advance the understanding of value alignment in artificial intelligence by characterising the topic in the context of its research literature. We use this to suggest a more precise definition of the term. Methods: We analyse 172 value alignment research articles that have been published in recent years and synthesise their content using thematic analyses. Results: Our analysis leads to six themes: value alignment drivers & approaches; challenges in value alignment; values in value alignment; cognitive processes in humans and AI; human-agent teaming; and designing and developing value-aligned systems. Conclusions: By analysing these themes in the context of the literature we define value alignment as an ongoing process between humans and autonomous agents that aims to express and implement abstract values in diverse contexts, while managing the cognitive limits of both humans and AI agents and also balancing the conflicting ethical and political demands generated by the values in different groups. Our analysis gives rise to a set of research challenges and opportunities in the field of value alignment for future work.
Authors: Sitong Chen, Shen Nie, Jiacheng Sun, Zijin Feng, Zhenguo Li, Ji-Rong Wen, Chongxuan Li
Abstract: We present a systematic theoretical framework that interprets masked diffusion models (MDMs) as solutions to energy minimization problems in discrete optimal transport. Specifically, we prove that three distinct energy formulations--kinetic, conditional kinetic, and geodesic energy--are mathematically equivalent under the structure of MDMs, and that MDMs minimize all three when the mask schedule satisfies a closed-form optimality condition. This unification not only clarifies the theoretical foundations of MDMs, but also motivates practical improvements in sampling. By parameterizing interpolation schedules via Beta distributions, we reduce the schedule design space to a tractable 2D search, enabling efficient post-training tuning without model modification. Experiments on synthetic and real-world benchmarks demonstrate that our energy-inspired schedules outperform hand-crafted baselines, particularly in low-step sampling settings.
Authors: Mariano Barone, Antonio Romano, Giuseppe Riccio, Marco Postiglione, Vincenzo Moscato
Abstract: Misinformation in healthcare, from vaccine hesitancy to unproven treatments, poses risks to public health and trust in medical systems. While machine learning and natural language processing have advanced automated fact-checking, validating biomedical claims remains uniquely challenging due to complex terminology, the need for domain expertise, and the critical importance of grounding in scientific evidence. We introduce CER (Combining Evidence and Reasoning), a novel framework for biomedical fact-checking that integrates scientific evidence retrieval, reasoning via large language models, and supervised veracity prediction. By integrating the text-generation capabilities of large language models with advanced retrieval techniques for high-quality biomedical scientific evidence, CER effectively mitigates the risk of hallucinations, ensuring that generated outputs are grounded in verifiable, evidence-based sources. Evaluations on expert-annotated datasets (HealthFC, BioASQ-7b, SciFact) demonstrate state-of-the-art performance and promising cross-dataset generalization. Code and data are released for transparency and reproducibility: https: //github.com/PRAISELab-PicusLab/CER.
Authors: Mariano Barone, Antonio Romano, Giuseppe Riccio, Marco Postiglione, Vincenzo Moscato
Abstract: Misinformation in healthcare, from vaccine hesitancy to unproven treatments, poses risks to public health and trust in medical systems. While machine learning and natural language processing have advanced automated fact-checking, validating biomedical claims remains uniquely challenging due to complex terminology, the need for domain expertise, and the critical importance of grounding in scientific evidence. We introduce CER (Combining Evidence and Reasoning), a novel framework for biomedical fact-checking that integrates scientific evidence retrieval, reasoning via large language models, and supervised veracity prediction. By integrating the text-generation capabilities of large language models with advanced retrieval techniques for high-quality biomedical scientific evidence, CER effectively mitigates the risk of hallucinations, ensuring that generated outputs are grounded in verifiable, evidence-based sources. Evaluations on expert-annotated datasets (HealthFC, BioASQ-7b, SciFact) demonstrate state-of-the-art performance and promising cross-dataset generalization. Code and data are released for transparency and reproducibility: https://github.com/PRAISELab-PicusLab/CER
Authors: Gustavo Kruger, Nikhil Sachdeva, Michael Sobolev
Abstract: Smartphone usage data can provide valuable insights for understanding interaction with technology and human behavior. However, collecting large-scale, in-the-wild smartphone usage logs is challenging due to high costs, privacy concerns, under representative user samples and biases like non-response that can skew results. These challenges call for exploring alternative approaches to obtain smartphone usage datasets. In this context, large language models (LLMs) such as Open AI's ChatGPT present a novel approach for synthetic smartphone usage data generation, addressing limitations of real-world data collection. We describe a case study on how four prompt strategies influenced the quality of generated smartphone usage data. We contribute with insights on prompt design and measures of data quality, reporting a prompting strategy comparison combining two factors, prompt level of detail (describing a user persona, describing the expected results characteristics) and seed data inclusion (with versus without an initial real usage example). Our findings suggest that using LLMs to generate structured and behaviorally plausible smartphone use datasets is feasible for some use cases, especially when using detailed prompts. Challenges remain in capturing diverse nuances of human behavioral patterns in a single synthetic dataset, and evaluating tradeoffs between data fidelity and diversity, suggesting the need for use-case-specific evaluation metrics and future research with more diverse seed data and different LLM models.
Authors: Zhanting Zhou, Jinshan Lai, Fengchun Zhang, Zeqin Wu, Fengli Zhang
Abstract: Non-IID data and partial participation induce client drift and inconsistent local optima in federated learning, causing unstable convergence and accuracy loss. We present FedSSG, a stochastic sampling-guided, history-aware drift alignment method. FedSSG maintains a per-client drift memory that accumulates local model differences as a lightweight sketch of historical gradients; crucially, it gates both the memory update and the local alignment term by a smooth function of the observed/expected participation ratio (a phase-by-expectation signal derived from the server sampler). This statistically grounded gate stays weak and smooth when sampling noise dominates early, then strengthens once participation statistics stabilize, contracting the local-global gap without extra communication. Across CIFAR-10/100 with 100/500 clients and 2-15 percent participation, FedSSG consistently outperforms strong drift-aware baselines and accelerates convergence; on our benchmarks it improves test accuracy by up to a few points (e.g., about +0.9 on CIFAR-10 and about +2.7 on CIFAR-100 on average over the top-2 baseline) and yields about 4.5x faster target-accuracy convergence on average. The method adds only O(d) client memory and a constant-time gate, and degrades gracefully to a mild regularizer under near-IID or uniform sampling. FedSSG shows that sampling statistics can be turned into a principled, history-aware phase control to stabilize and speed up federated training.
Authors: Domenico Meconi, Simone Stirpe, Federico Martelli, Leonardo Lavalle, Roberto Navigli
Abstract: Understanding the meaning of words in context is a fundamental capability for Large Language Models (LLMs). Despite extensive evaluation efforts, the extent to which LLMs show evidence that they truly grasp word senses remains underexplored. In this paper, we address this gap by evaluating both i) the Word Sense Disambiguation (WSD) capabilities of instruction-tuned LLMs, comparing their performance to state-of-the-art systems specifically designed for the task, and ii) the ability of two top-performing open- and closed-source LLMs to understand word senses in three generative settings: definition generation, free-form explanation, and example generation. Notably, we find that, in the WSD task, leading models such as GPT-4o and DeepSeek-V3 achieve performance on par with specialized WSD systems, while also demonstrating greater robustness across domains and levels of difficulty. In the generation tasks, results reveal that LLMs can explain the meaning of words in context up to 98\% accuracy, with the highest performance observed in the free-form explanation task, which best aligns with their generative capabilities.
Authors: Divya Thuremella, Yi Yang, Simon Wanna, Lars Kunze, Daniele De Martini
Abstract: This work explores the application of ensemble modeling to the multidimensional regression problem of trajectory prediction for vehicles in urban environments. As newer and bigger state-of-the-art prediction models for autonomous driving continue to emerge, an important open challenge is the problem of how to combine the strengths of these big models without the need for costly re-training. We show how, perhaps surprisingly, combining state-of-the-art deep learning models out-of-the-box (without retraining or fine-tuning) with a simple confidence-weighted average method can enhance the overall prediction. Indeed, while combining trajectory prediction models is not straightforward, this simple approach enhances performance by 10% over the best prediction model, especially in the long-tailed metrics. We show that this performance improvement holds on both the NuScenes and Argoverse datasets, and that these improvements are made across the dataset distribution. The code for our work is open source.
Authors: Huilin Yin, Yiming Kan, Daniel Watzenig
Abstract: In recent years, end-to-end autonomous driving has attracted increasing attention for its ability to jointly model perception, prediction, and planning within a unified framework. However, most existing approaches underutilize the online mapping module, leaving its potential to enhance trajectory planning largely untapped. This paper proposes MAP (Map-Assisted Planning), a novel map-assisted end-to-end trajectory planning framework. MAP explicitly integrates segmentation-based map features and the current ego status through a Plan-enhancing Online Mapping module, an Ego-status-guided Planning module, and a Weight Adapter based on current ego status. Experiments conducted on the DAIR-V2X-seq-SPD dataset demonstrate that the proposed method achieves a 16.6% reduction in L2 displacement error, a 56.2% reduction in off-road rate, and a 44.5% improvement in overall score compared to the UniV2X baseline, even without post-processing. Furthermore, it achieves top ranking in Track 2 of the End-to-End Autonomous Driving through V2X Cooperation Challenge of MEIS Workshop @CVPR2025, outperforming the second-best model by 39.5% in terms of overall score. These results highlight the effectiveness of explicitly leveraging semantic map features in planning and suggest new directions for improving structure design in end-to-end autonomous driving systems. Our code is available at https://gitee.com/kymkym/map.git
Authors: Kevin Wilkinghoff, Zheng-Hua Tan
Abstract: Reasoning about spatial audio with large language models requires a spatial audio encoder as an acoustic front-end to obtain audio embeddings for further processing. Such an encoder needs to capture all information required to detect the type of sound events, as well as the direction and distance of their corresponding sources. Accomplishing this with a single audio encoder is demanding as the information required for each of these tasks is mostly independent of each other. As a result, the performance obtained with a single encoder is often worse than when using task-specific audio encoders. In this work, we present DSpAST, a novel audio encoder based on SpatialAST that learns disentangled representations of spatial audio while having only 0.2% additional parameters. Experiments on SpatialSoundQA with the spatial audio reasoning system BAT demonstrate that DSpAST significantly outperforms SpatialAST.
Authors: Simiao Liu, Fang Liu, Liehao Li, Xin Tan, Yinghao Zhu, Xiaoli Lian, Li Zhang
Abstract: Automated issue solving seeks to autonomously identify and repair defective code snippets across an entire codebase. SWE-Bench has emerged as the most widely adopted benchmark for evaluating progress in this area. While LLM-based agentic tools show great promise, they still fail on a substantial portion of tasks. Moreover, current evaluations primarily report aggregate issue-solving rates, which obscure the underlying causes of success and failure, making it challenging to diagnose model weaknesses or guide targeted improvements. To bridge this gap, we first analyze the performance and efficiency of three SOTA tools, spanning both pipeline-based and agentic architectures, in automated issue solving tasks of SWE-Bench-Verified under varying task characteristics. Furthermore, to move from high-level performance metrics to underlying cause analysis, we conducted a systematic manual analysis of 150 failed instances. From this analysis, we developed a comprehensive taxonomy of failure modes comprising 3 primary phases, 9 main categories, and 25 fine-grained subcategories. Then we systematically analyze the distribution of the identified failure modes, the results reveal distinct failure fingerprints between the two architectural paradigms, with the majority of agentic failures stemming from flawed reasoning and cognitive deadlocks. Motivated by these insights, we propose a collaborative Expert-Executor framework. It introduces a supervisory Expert agent tasked with providing strategic oversight and course-correction for a primary Executor agent. This architecture is designed to correct flawed reasoning and break the cognitive deadlocks that frequently lead to failure. Experiments show that our framework solves 22.2% of previously intractable issues for a leading single agent. These findings pave the way for building more robust agents through diagnostic evaluation and collaborative design.
Authors: Renan Souza, Timothy Poteet, Brian Etz, Daniel Rosendo, Amal Gueroudji, Woong Shin, Prasanna Balaprakash, Rafael Ferreira da Silva
Abstract: Modern scientific discovery increasingly relies on workflows that process data across the Edge, Cloud, and High Performance Computing (HPC) continuum. Comprehensive and in-depth analyses of these data are critical for hypothesis validation, anomaly detection, reproducibility, and impactful findings. Although workflow provenance techniques support such analyses, at large scale, the provenance data become complex and difficult to analyze. Existing systems depend on custom scripts, structured queries, or static dashboards, limiting data interaction. In this work, we introduce an evaluation methodology, reference architecture, and open-source implementation that leverages interactive Large Language Model (LLM) agents for runtime data analysis. Our approach uses a lightweight, metadata-driven design that translates natural language into structured provenance queries. Evaluations across LLaMA, GPT, Gemini, and Claude, covering diverse query classes and a real-world chemistry workflow, show that modular design, prompt tuning, and Retrieval-Augmented Generation (RAG) enable accurate and insightful LLM agent responses beyond recorded provenance.
Authors: Ozer Ozturk, Busra Buyuktanir, Gozde Karatas Baydogmus, Kazim Yildiz
Abstract: Machine learning models used for distributed architectures consisting of servers and clients require large amounts of data to achieve high accuracy. Data obtained from clients are collected on a central server for model training. However, storing data on a central server raises concerns about security and privacy. To address this issue, a federated learning architecture has been proposed. In federated learning, each client trains a local model using its own data. The trained models are periodically transmitted to the central server. The server then combines the received models using federated aggregation algorithms to obtain a global model. This global model is distributed back to the clients, and the process continues in a cyclical manner. Although preventing data from leaving the clients enhances security, certain concerns still remain. Attackers can perform inference attacks on the obtained models to approximate the training dataset, potentially causing data leakage. In this study, differential privacy was applied to address the aforementioned security vulnerability, and a performance analysis was conducted. The Data-Unaware Classification Based on Association (duCBA) algorithm was used as the federated aggregation method. Differential privacy was implemented on the data using the Randomized Response technique, and the trade-off between security and performance was examined under different epsilon values. As the epsilon value decreased, the model accuracy declined, and class prediction imbalances were observed. This indicates that higher levels of privacy do not always lead to practical outcomes and that the balance between security and performance must be carefully considered.
Authors: Colin Hong, Xu Guo, Anand Chaanan Singh, Esha Choukse, Dmitrii Ustiugov
Abstract: Recently, Test-Time Scaling (TTS) has gained increasing attention for improving LLM reasoning performance at test time without retraining the model. A notable TTS technique is Self-Consistency (SC), which generates multiple reasoning chains in parallel and selects the final answer via majority voting. While effective, the order-of-magnitude computational overhead limits its broad deployment. Prior attempts to accelerate SC mainly rely on model-based confidence scores or heuristics with limited empirical support. For the first time, we theoretically and empirically analyze the inefficiencies of SC and reveal actionable opportunities for improvement. Building on these insights, we propose Slim-SC, a step-wise pruning strategy that identifies and removes redundant chains using inter-chain similarity at the thought level. Experiments on three STEM reasoning datasets and two recent LLM architectures show that Slim-SC reduces inference latency and KVC usage by up to 45% and 26%, respectively, with R1-Distill, while maintaining or improving accuracy, thus offering a simple yet efficient TTS alternative for SC.
Authors: Elena Camuffo, Francesco Barbato, Mete Ozay, Simone Milani, Umberto Michieli
Abstract: We introduce MOCHA (Multi-modal Objects-aware Cross-arcHitecture Alignment), a knowledge distillation approach that transfers region-level multimodal semantics from a large vision-language teacher (e.g., LLaVa) into a lightweight vision-only object detector student (e.g., YOLO). A translation module maps student features into a joint space, where the training of the student and translator is guided by a dual-objective loss that enforces both local alignment and global relational consistency. Unlike prior approaches focused on dense or global alignment, MOCHA operates at the object level, enabling efficient transfer of semantics without modifying the teacher or requiring textual input at inference. We validate our method across four personalized detection benchmarks under few-shot regimes. Results show consistent gains over baselines, with a +10.1 average score improvement. Despite its compact architecture, MOCHA reaches performance on par with larger multimodal models, proving its suitability for real-world deployment.
Authors: Liting Gao, Yi Yuan, Yaru Chen, Yuelan Cheng, Zhenbo Li, Juan Wen, Shubin Zhang, Wenwu Wang
Abstract: Diffusion models have shown remarkable progress in text-to-audio generation. However, text-guided audio editing remains in its early stages. This task focuses on modifying the target content within an audio signal while preserving the rest, thus demanding precise localization and faithful editing according to the text prompt. Existing training-based and zero-shot methods that rely on full-caption or costly optimization often struggle with complex editing or lack practicality. In this work, we propose a novel end-to-end efficient rectified flow matching-based diffusion framework for audio editing, and construct a dataset featuring overlapping multi-event audio to support training and benchmarking in complex scenarios. Experiments show that our model achieves faithful semantic alignment without requiring auxiliary captions or masks, while maintaining competitive editing quality across metrics.
Authors: Hasan Abed Al Kader Hammoud, Mohammad Zbeeb, Bernard Ghanem
Abstract: We present Hala, a family of Arabic-centric instruction and translation models built with our translate-and-tune pipeline. We first compress a strong AR$\leftrightarrow$EN teacher to FP8 (yielding $\sim$2$\times$ higher throughput with no quality loss) and use it to create high-fidelity bilingual supervision. A lightweight language model LFM2-1.2B is then fine-tuned on this data and used to translate high-quality English instruction sets into Arabic, producing a million-scale corpus tailored to instruction following. We train Hala models at 350M, 700M, 1.2B, and 9B parameters, and apply slerp merging to balance Arabic specialization with base-model strengths. On Arabic-centric benchmarks, Hala achieves state-of-the-art results within both the "nano" ($\leq$2B) and "small" (7-9B) categories, outperforming their bases. We release models, data, evaluation, and recipes to accelerate research in Arabic NLP.
Authors: Pawe{\l} M\k{a}ka, Yusuf Can Semerci, Jan Scholtes, Gerasimos Spanakis
Abstract: Achieving human-level translations requires leveraging context to ensure coherence and handle complex phenomena like pronoun disambiguation. Sparsity of contextually rich examples in the standard training data has been hypothesized as the reason for the difficulty of context utilization. In this work, we systematically validate this claim in both single- and multilingual settings by constructing training datasets with a controlled proportions of contextually relevant examples. We demonstrate a strong association between training data sparsity and model performance confirming sparsity as a key bottleneck. Importantly, we reveal that improvements in one contextual phenomenon do no generalize to others. While we observe some cross-lingual transfer, it is not significantly higher between languages within the same sub-family. Finally, we propose and empirically evaluate two training strategies designed to leverage the available data. These strategies improve context utilization, resulting in accuracy gains of up to 6 and 8 percentage points on the ctxPro evaluation in single- and multilingual settings respectively.
Authors: Zekang Liu, Wei Feng, Fanhua Shang, Lianyu Hu, Jichao Feng, Liqing Gao
Abstract: Sign Language Translation (SLT) bridges the communication gap between deaf people and hearing people, where dialogue provides crucial contextual cues to aid in translation. Building on this foundational concept, this paper proposes Question-based Sign Language Translation (QB-SLT), a novel task that explores the efficient integration of dialogue. Unlike gloss (sign language transcription) annotations, dialogue naturally occurs in communication and is easier to annotate. The key challenge lies in aligning multimodality features while leveraging the context of the question to improve translation. To address this issue, we propose a cross-modality Self-supervised Learning with Sigmoid Self-attention Weighting (SSL-SSAW) fusion method for sign language translation. Specifically, we employ contrastive learning to align multimodality features in QB-SLT, then introduce a Sigmoid Self-attention Weighting (SSAW) module for adaptive feature extraction from question and sign language sequences. Additionally, we leverage available question text through self-supervised learning to enhance representation and translation capabilities. We evaluated our approach on newly constructed CSL-Daily-QA and PHOENIX-2014T-QA datasets, where SSL-SSAW achieved SOTA performance. Notably, easily accessible question assistance can achieve or even surpass the performance of gloss assistance. Furthermore, visualization results demonstrate the effectiveness of incorporating dialogue in improving translation quality.
Authors: Ranga Baminiwatte, Kazi Jewel Rana, Aaron J. Masino
Abstract: Understanding disease similarity is critical for advancing diagnostics, drug discovery, and personalized treatment strategies. We present PhenoGnet, a novel graph-based contrastive learning framework designed to predict disease similarity by integrating gene functional interaction networks with the Human Phenotype Ontology (HPO). PhenoGnet comprises two key components: an intra-view model that separately encodes gene and phenotype graphs using Graph Convolutional Networks (GCNs) and Graph Attention Networks (GATs), and a cross view model implemented as a shared weight multilayer perceptron (MLP) that aligns gene and phenotype embeddings through contrastive learning. The model is trained using known gene phenotype associations as positive pairs and randomly sampled unrelated pairs as negatives. Diseases are represented by the mean embeddings of their associated genes and/or phenotypes, and pairwise similarity is computed via cosine similarity. Evaluation on a curated benchmark of 1,100 similar and 866 dissimilar disease pairs demonstrates strong performance, with gene based embeddings achieving an AUCPR of 0.9012 and AUROC of 0.8764, outperforming existing state of the art methods. Notably, PhenoGnet captures latent biological relationships beyond direct overlap, offering a scalable and interpretable solution for disease similarity prediction. These results underscore its potential for enabling downstream applications in rare disease research and precision medicine.
Authors: Zewen Yang, Xiaobing Dai, Dongfa Zhang, Yu Li, Ziyang Meng, Bingkun Huang, Hamid Sadeghian, Sami Haddadin
Abstract: Learning from demonstration allows robots to acquire complex skills from human demonstrations, but conventional approaches often require large datasets and fail to generalize across coordinate transformations. In this paper, we propose Prompt2Auto, a geometry-invariant one-shot Gaussian process (GeoGP) learning framework that enables robots to perform human-guided automated control from a single motion prompt. A dataset-construction strategy based on coordinate transformations is introduced that enforces invariance to translation, rotation, and scaling, while supporting multi-step predictions. Moreover, GeoGP is robust to variations in the user's motion prompt and supports multi-skill autonomy. We validate the proposed approach through numerical simulations with the designed user graphical interface and two real-world robotic experiments, which demonstrate that the proposed method is effective, generalizes across tasks, and significantly reduces the demonstration burden. Project page is available at: https://prompt2auto.github.io
Authors: Jordi Grau-Haro, Ruben Ribes-Serrano, Javier Naranjo-Alcazar, Marta Garcia-Ballesteros, Pedro Zuccarello
Abstract: Convolutional Neural Networks (CNNs) have demonstrated exceptional performance in audio tagging tasks. However, deploying these models on resource-constrained devices like the Raspberry Pi poses challenges related to computational efficiency and thermal management. In this paper, a comprehensive evaluation of multiple convolutional neural network (CNN) architectures for audio tagging on the Raspberry Pi is conducted, encompassing all 1D and 2D models from the Pretrained Audio Neural Networks (PANNs) framework, a ConvNeXt-based model adapted for audio classification, as well as MobileNetV3 architectures. In addition, two PANNs-derived networks, CNN9 and CNN13, recently proposed, are also evaluated. To enhance deployment efficiency and portability across diverse hardware platforms, all models are converted to the Open Neural Network Exchange (ONNX) format. Unlike previous works that focus on a single model, our analysis encompasses a broader range of architectures and involves continuous 24-hour inference sessions to assess performance stability. Our experiments reveal that, with appropriate model selection and optimization, it is possible to maintain consistent inference latency and manage thermal behavior effectively over extended periods. These findings provide valuable insights for deploying audio tagging models in real-world edge computing scenarios.
Authors: Riccardo Zanardelli
Abstract: With the growth of artificial skills, organizations may increasingly confront with the problem of optimizing skill policy decisions guided by economic principles. This paper addresses the underlying complexity of this challenge by developing an in-silico framework based on Monte Carlo simulations grounded in empirical realism to analyze the economic impact of human and machine skills, individually or jointly deployed, in the execution of tasks presenting varying levels of complexity. Our results provide quantitative support for the established notions that automation tends to be the most economically-effective strategy for tasks characterized by low-to-medium generalization difficulty, while automation struggles to match the economic utility of human skills in more complex scenarios. Critically, our simulations highlight that combining human and machine skills can be the most effective strategy when a high level of generalization is required, but only if genuine augmentation is achieved. In contrast, when failing to realize this synergy, the human-machine policy is severely penalized by the inherent costs of its dual skill structure, causing it to destroy value and becoming the worst choice from an economic perspective. The takeaway for decision-makers is unambiguous: simply allocating human and machine skills to a task is insufficient, and a human-machine skill policy is neither a silver-bullet solution nor a low-risk compromise. Rather, it is a critical opportunity to boost competitiveness that demands a strong organizational commitment to enabling augmentation. Also, our findings show that improving the cost-effectiveness of machine skills over time, while useful, does not replace the fundamental need to focus on achieving augmentation.
Authors: Chiara De Luca, Elisa Donati
Abstract: Queen bee presence is essential for the health and stability of honeybee colonies, yet current monitoring methods rely on manual inspections that are labor-intensive, disruptive, and impractical for large-scale beekeeping. While recent audio-based approaches have shown promise, they often require high power consumption, complex preprocessing, and are susceptible to ambient noise. To overcome these limitations, we propose a lightweight, multimodal system for queen detection based on environmental sensor fusion-specifically, temperature, humidity, and pressure differentials between the inside and outside of the hive. Our approach employs quantized decision tree inference on a commercial STM32 microcontroller, enabling real-time, low-power edge computing without compromising accuracy. We show that our system achieves over 99% queen detection accuracy using only environmental inputs, with audio features offering no significant performance gain. This work presents a scalable and sustainable solution for non-invasive hive monitoring, paving the way for autonomous, precision beekeeping using off-the-shelf, energy-efficient hardware.
Authors: Kerui Huang, Shuhan Liu, Xing Hu, Tongtong Xu, Lingfeng Bao, Xin Xia
Abstract: Chain-of-Thought (CoT) reasoning enhances Large Language Models (LLMs) by prompting intermediate steps, improving accuracy and robustness in arithmetic, logic, and commonsense tasks. However, this benefit comes with high computational costs: longer outputs increase latency, memory usage, and KV-cache demands. These issues are especially critical in software engineering tasks where concise and deterministic outputs are required. To investigate these trade-offs, we conduct an empirical study based on code generation benchmarks. The results reveal that longer CoT does not always help. Excessive reasoning often causes truncation, accuracy drops, and latency up to five times higher, with failed outputs consistently longer than successful ones. These findings challenge the assumption that longer reasoning is inherently better and highlight the need for adaptive CoT control. Motivated by this, we propose SEER (Self-Enhancing Efficient Reasoning), an adaptive framework that compresses CoT while preserving accuracy. SEER combines Best-of-N sampling with task-aware adaptive filtering, dynamically adjusting thresholds based on pre-inference outputs to reduce verbosity and computational overhead. We then evaluate SEER on three software engineering tasks and one math task. On average, SEER shortens CoT by 42.1%, improves accuracy by reducing truncation, and eliminates most infinite loops. These results demonstrate SEER as a practical method to make CoT-enhanced LLMs more efficient and robust, even under resource constraints.
Authors: Michal Szczepanski, Martyna Poreba, Karim Haroun
Abstract: Vision Transformers (ViTs) achieve state-of-the-art performance in semantic segmentation but are hindered by high computational and memory costs. To address this, we propose STEP (SuperToken and Early-Pruning), a hybrid token-reduction framework that combines dynamic patch merging and token pruning to enhance efficiency without significantly compromising accuracy. At the core of STEP is dCTS, a lightweight CNN-based policy network that enables flexible merging into superpatches. Encoder blocks integrate also early-exits to remove high-confident supertokens, lowering computational load. We evaluate our method on high-resolution semantic segmentation benchmarks, including images up to 1024 x 1024, and show that when dCTS is applied alone, the token count can be reduced by a factor of 2.5 compared to the standard 16 x 16 pixel patching scheme. This yields a 2.6x reduction in computational cost and a 3.4x increase in throughput when using ViT-Large as the backbone. Applying the full STEP framework further improves efficiency, reaching up to a 4x reduction in computational complexity and a 1.7x gain in inference speed, with a maximum accuracy drop of no more than 2.0%. With the proposed STEP configurations, up to 40% of tokens can be confidently predicted and halted before reaching the final encoder layer.
Authors: Ziyuan Chen, Zhenghui Zhao, Zhangye Han, Miancan Liu, Xianhang Ye, Yiqing Li, Hongbo Min, Jinkui Ren, Xiantao Zhang, Guitao Cao
Abstract: With the rapid advancement of large language models and vision-language models, employing large models as Web Agents has become essential for automated web interaction. However, training Web Agents with reinforcement learning faces critical challenges including credit assignment misallocation, prohibitively high annotation costs, and reward sparsity. To address these issues, we propose Tree-Guided Preference Optimization (TGPO), an offline reinforcement learning framework that proposes a tree-structured trajectory representation merging semantically identical states across trajectories to eliminate label conflicts. Our framework incorporates a Process Reward Model that automatically generates fine-grained rewards through subgoal progress, redundancy detection, and action verification. Additionally, a dynamic weighting mechanism prioritizes high-impact decision points during training. Experiments on Online-Mind2Web and our self-constructed C-WebShop datasets demonstrate that TGPO significantly outperforms existing methods, achieving higher success rates with fewer redundant steps.
Authors: Akhil Theerthala
Abstract: Personalized financial advice requires consideration of user goals, constraints, risk tolerance, and jurisdiction. Prior LLM work has focused on support systems for investors and financial planners. Simultaneously, numerous recent studies examine broader personal finance tasks, including budgeting, debt management, retirement, and estate planning, through agentic pipelines that incur high maintenance costs, yielding less than 25% of their expected financial returns. In this study, we introduce a novel and reproducible framework that integrates relevant financial context with behavioral finance studies to construct supervision data for end-to-end advisors. Using this framework, we create a 19k sample reasoning dataset and conduct a comprehensive fine-tuning of the Qwen-3-8B model on the dataset. Through a held-out test split and a blind LLM-jury study, we demonstrate that through careful data curation and behavioral integration, our 8B model achieves performance comparable to significantly larger baselines (14-32B parameters) across factual accuracy, fluency, and personalization metrics while incurring 80% lower costs than the larger counterparts.
Authors: Yifan Hu, Jie Yang, Tian Zhou, Peiyuan Liu, Yujin Tang, Rong Jin, Liang Sun
Abstract: Representation learning techniques like contrastive learning have long been explored in time series forecasting, mirroring their success in computer vision and natural language processing. Yet recent state-of-the-art (SOTA) forecasters seldom adopt these representation approaches because they have shown little performance advantage. We challenge this view and demonstrate that explicit representation alignment can supply critical information that bridges the distributional gap between input histories and future targets. To this end, we introduce TimeAlign, a lightweight, plug-and-play framework that learns auxiliary features via a simple reconstruction task and feeds them back to any base forecaster. Extensive experiments across eight benchmarks verify its superior performance. Further studies indicate that the gains arises primarily from correcting frequency mismatches between historical inputs and future outputs. We also provide a theoretical justification for the effectiveness of TimeAlign in increasing the mutual information between learned representations and predicted targets. As it is architecture-agnostic and incurs negligible overhead, TimeAlign can serve as a general alignment module for modern deep learning time-series forecasting systems. The code is available at https://github.com/TROUBADOUR000/TimeAlign.
Authors: Haichao Zhang, Wenhao Chai, Shwai He, Ang Li, Yun Fu
Abstract: High temporal resolution is essential for capturing fine-grained details in video understanding. However, current video large language models (VLLMs) and benchmarks mostly rely on low-frame-rate sampling, such as uniform sampling or keyframe selection, discarding dense temporal information. This compromise avoids the high cost of tokenizing every frame, which otherwise leads to redundant computation and linear token growth as video length increases. While this trade-off works for slowly changing content, it fails for tasks like lecture comprehension, where information appears in nearly every frame and requires precise temporal alignment. To address this gap, we introduce Dense Video Understanding (DVU), which enables high-FPS video comprehension by reducing both tokenization time and token overhead. Existing benchmarks are also limited, as their QA pairs focus on coarse content changes. We therefore propose DIVE (Dense Information Video Evaluation), the first benchmark designed for dense temporal reasoning. To make DVU practical, we present Gated Residual Tokenization (GRT), a two-stage framework: (1) Motion-Compensated Inter-Gated Tokenization uses pixel-level motion estimation to skip static regions during tokenization, achieving sub-linear growth in token count and compute. (2) Semantic-Scene Intra-Tokenization Merging fuses tokens across static regions within a scene, further reducing redundancy while preserving dynamic semantics. Experiments on DIVE show that GRT outperforms larger VLLM baselines and scales positively with FPS. These results highlight the importance of dense temporal information and demonstrate that GRT enables efficient, scalable high-FPS video understanding.
Authors: Johnny R. Zhang (Independent Researcher), Xiaomei Mi (University of Manchester), Gaoyuan Du (Amazon), Qianyi Sun (Microsoft), Shiqi Wang (Meta), Jiaxuan Li (Amazon), Wenhua Zhou (Independent Researcher)
Abstract: Stochastic optimization powers the scalability of modern artificial intelligence, spanning machine learning, deep learning, reinforcement learning, and large language model training. Yet, existing theory remains largely confined to Hilbert spaces, relying on inner-product frameworks and orthogonality. This paradigm fails to capture non-Euclidean settings, such as mirror descent on simplices, Bregman proximal methods for sparse learning, natural gradient descent in information geometry, or Kullback--Leibler-regularized language model training. Unlike Euclidean-based Hilbert-space methods, this approach embraces general Banach spaces. This work introduces a pioneering Banach--Bregman framework for stochastic iterations, establishing Bregman geometry as a foundation for next-generation optimization. It (i) provides a unified template via Bregman projections and Bregman--Fejer monotonicity, encompassing stochastic approximation, mirror descent, natural gradient, adaptive methods, and mirror-prox; (ii) establishes super-relaxations ($\lambda > 2$) in non-Hilbert settings, enabling flexible geometries and elucidating their acceleration effect; and (iii) delivers convergence theorems spanning almost-sure boundedness to geometric rates, validated on synthetic and real-world tasks. Empirical studies across machine learning (UCI benchmarks), deep learning (e.g., Transformer training), reinforcement learning (actor--critic), and large language models (WikiText-2 with distilGPT-2) show up to 20% faster convergence, reduced variance, and enhanced accuracy over classical baselines. These results position Banach--Bregman geometry as a cornerstone unifying optimization theory and practice across core AI paradigms.
Authors: Dmitrii Krasheninnikov, Richard E. Turner, David Krueger
Abstract: We show that language models' activations linearly encode when information was learned during training. Our setup involves creating a model with a known training order by sequentially fine-tuning Llama-3.2-1B on six disjoint but otherwise similar datasets about named entities. We find that the average activations of test samples for the six training datasets encode the training order: when projected into a 2D subspace, these centroids are arranged exactly in the order of training and lie on a straight line. Further, we show that linear probes can accurately (~90%) distinguish "early" vs. "late" entities, generalizing to entities unseen during the probes' own training. The model can also be fine-tuned to explicitly report an unseen entity's training stage (~80% accuracy). Interestingly, this temporal signal does not seem attributable to simple differences in activation magnitudes, losses, or model confidence. Our paper demonstrates that models are capable of differentiating information by its acquisition time, and carries significant implications for how they might manage conflicting data and respond to knowledge modifications.
Authors: Alejandro Hern\'andez-Cano, Alexander H\"agele, Allen Hao Huang, Angelika Romanou, Antoni-Joan Solergibert, Barna Pasztor, Bettina Messmer, Dhia Garbaya, Eduard Frank \v{D}urech, Ido Hakimi, Juan Garc\'ia Giraldo, Mete Ismayilzada, Negar Foroutan, Skander Moalla, Tiancheng Chen, Vinko Sabol\v{c}ec, Yixuan Xu, Michael Aerni, Badr AlKhamissi, Ines Altemir Marinas, Mohammad Hossein Amani, Matin Ansaripour, Ilia Badanin, Harold Benoit, Emanuela Boros, Nicholas Browning, Fabian B\"osch, Maximilian B\"other, Niklas Canova, Camille Challier, Clement Charmillot, Jonathan Coles, Jan Deriu, Arnout Devos, Lukas Drescher, Daniil Dzenhaliou, Maud Ehrmann, Dongyang Fan, Simin Fan, Silin Gao, Miguel Gila, Mar\'ia Grandury, Diba Hashemi, Alexander Hoyle, Jiaming Jiang, Mark Klein, Andrei Kucharavy, Anastasiia Kucherenko, Frederike L\"ubeck, Roman Machacek, Theofilos Manitaras, Andreas Marfurt, Kyle Matoba, Simon Matrenok, Henrique Mendonc\c{c}a, Fawzi Roberto Mohamed, Syrielle Montariol, Luca Mouchel, Sven Najem-Meyer, Jingwei Ni, Gennaro Oliva, Matteo Pagliardini, Elia Palme, Andrei Panferov, L\'eo Paoletti, Marco Passerini, Ivan Pavlov, Auguste Poiroux, Kaustubh Ponkshe, Nathan Ranchin, Javi Rando, Mathieu Sauser, Jakhongir Saydaliev, Muhammad Ali Sayfiddinov, Marian Schneider, Stefano Schuppli, Marco Scialanga, Andrei Semenov, Kumar Shridhar, Raghav Singhal, Anna Sotnikova, Alexander Sternfeld, Ayush Kumar Tarun, Paul Teiletche, Jannis Vamvas, Xiaozhe Yao, Hao Zhao Alexander Ilic, Ana Klimovic, Andreas Krause, Caglar Gulcehre, David Rosenthal, Elliott Ash, Florian Tram\`er, Joost VandeVondele, Livio Veraldi, Martin Rajman, Thomas Schulthess, Torsten Hoefler, Antoine Bosselut, Martin Jaggi, Imanol Schlag
Abstract: We present Apertus, a fully open suite of large language models (LLMs) designed to address two systemic shortcomings in today's open model ecosystem: data compliance and multilingual representation. Unlike many prior models that release weights without reproducible data pipelines or regard for content-owner rights, Apertus models are pretrained exclusively on openly available data, retroactively respecting robots.txt exclusions and filtering for non-permissive, toxic, and personally identifiable content. To mitigate risks of memorization, we adopt the Goldfish objective during pretraining, strongly suppressing verbatim recall of data while retaining downstream task performance. The Apertus models also expand multilingual coverage, training on 15T tokens from over 1800 languages, with ~40% of pretraining data allocated to non-English content. Released at 8B and 70B scales, Apertus approaches state-of-the-art results among fully open models on multilingual benchmarks, rivalling or surpassing open-weight counterparts. Beyond model weights, we release all scientific artifacts from our development cycle with a permissive license, including data preparation scripts, checkpoints, evaluation suites, and training code, enabling transparent audit and extension.
Authors: Muhua Huang, Xijuan Zhang, Christopher Soto, James Evans
Abstract: We introduce a methodology for assigning quantifiable and psychometrically validated personalities to AI-Agents using the Big Five framework. Across three studies, we evaluate its feasibility and limits. In Study 1, we show that large language models (LLMs) capture semantic similarities among Big Five measures, providing a basis for personality assignment. In Study 2, we create AI-Agents using prompts designed based on the Big Five Inventory (BFI-2) in the Likert or Expanded format, and find that, when paired with newer LLMs (e.g., GPT-4, GPT-4o, Llama, DeepSeek), these AI-Agents align more closely with human responses on the Mini-Markers test than those generated with binary adjective prompts or older models, although the finer pattern of results (e.g., factor loading patterns) were not consistent between AI-Agents and human participants. In Study 3, we validate our AI-Agents with risk-taking and moral dilemma vignettes. We find that while fine-tuning shifts responses toward more moral judgment, AI-Agent correlations between the input Big Five traits and the output moral judgments mirror those from human participants. Overall, our results show that AI-Agents align with humans in correlations between input Big Five traits and output responses and may serve as useful tools for preliminary research. Nevertheless, discrepancies in finer response patterns indicate that AI-Agents cannot (yet) fully substitute for human participants in precision or high-stakes projects.
Authors: Enci Zhang, Xingang Yan, Wei Lin, Tianxiang Zhang, Qianchun Lu
Abstract: Despite impressive progress in areas like mathematical reasoning, large language models still face significant challenges in consistently solving complex problems. Drawing inspiration from key human learning strategies, we propose two novel strategies to enhance the capability of large language models to solve these complex problems. First, Adaptive Difficulty Curriculum Learning (ADCL) is a novel curriculum learning strategy that tackles the Difficulty Shift phenomenon (i.e., a model's perception of problem difficulty dynamically changes during training) by periodically re-estimating difficulty within upcoming data batches to maintain alignment with the model's evolving capabilities. Second, Expert-Guided Self-Reformulation (EGSR) is a novel reinforcement learning strategy that bridges the gap between imitation learning and pure exploration by guiding models to reformulate expert solutions within their own conceptual framework, rather than relying on direct imitation, fostering deeper understanding and knowledge assimilation. Extensive experiments on challenging mathematical reasoning benchmarks, using Qwen2.5-7B as the base model, demonstrate that these human-inspired strategies synergistically and significantly enhance performance. Notably, their combined application improves performance over the standard Zero-RL baseline by 10% on the AIME24 benchmark and 16.6% on AIME25.
Authors: Mahmood Hegazy, Aaron Rodrigues, Azzam Naeem
Abstract: Modern consumer banking applications require accurate and efficient retrieval of information in response to user queries. Mapping user utterances to the most relevant Frequently Asked Questions (FAQs) is a crucial component of these systems. Traditional approaches often rely on a single model or technique, which may not capture the nuances of diverse user inquiries. In this paper, we introduce a multi-agent framework for FAQ annotation that combines multiple specialized agents with different approaches and a judge agent that reranks candidates to produce optimal results. Our agents utilize a structured reasoning approach inspired by Attentive Reasoning Queries (ARQs), which guides them through systematic reasoning steps using targeted, task-specific JSON queries. Our framework features a few-shot example strategy, where each agent receives different few-shots, enhancing ensemble diversity and coverage of the query space. We evaluate our framework on a real-world major bank dataset as well as public benchmark datasets (LCQMC and FiQA), demonstrating significant improvements over single-agent approaches across multiple metrics, including a 14% increase in Top-1 accuracy, an 18% increase in Top-5 accuracy, and a 12% improvement in Mean Reciprocal Rank on our dataset, and similar gains on public benchmarks when compared with traditional and single-agent annotation techniques. Our framework is particularly effective at handling ambiguous queries, making it well-suited for deployment in production banking applications while showing strong generalization capabilities across different domains and languages.
Authors: Licheng Pan, Yongqi Tong, Xin Zhang, Xiaolu Zhang, Jun Zhou, Zhixuan Chu
Abstract: Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks, yet they often refuse to answer legitimate queries--a phenomenon known as overrefusal. Overrefusal typically stems from over-conservative safety alignment, causing models to treat many reasonable prompts as potentially risky. To systematically understand this issue, we probe and leverage the models' safety decision boundaries to analyze and mitigate overrefusal. Our findings reveal that overrefusal is closely tied to misalignment at these boundary regions, where models struggle to distinguish subtle differences between benign and harmful content. Building on these insights, we present RASS, an automated framework for prompt generation and selection that strategically targets overrefusal prompts near the safety boundary. By harnessing steering vectors in the representation space, RASS efficiently identifies and curates boundary-aligned prompts, enabling more effective and targeted mitigation of overrefusal. This approach not only provides a more precise and interpretable view of model safety decisions but also seamlessly extends to multilingual scenarios. We have explored the safety decision boundaries of various LLMs and construct the MORBench evaluation set to facilitate robust assessment of model safety and helpfulness across multiple languages. Code and datasets are available at https://github.com/Master-PLC/RASS.
Authors: Lachlan McGinness, Peter Baumgartner
Abstract: Empirical methods to examine the capability of Large Language Models (LLMs) to use Automated Theorem Prover (ATP) reasoning strategies are studied. We evaluate the performance of State of the Art models from December 2023 and August 2024 on PRONTOQA steamroller reasoning problems. For that, we develop methods for assessing LLM response accuracy and correct answer correlation. Our results show that progress in improving LLM reasoning abilities has stalled over the nine month period. By tracking completion tokens, we show that almost all improvement in reasoning ability since GPT-4 was released can be attributed to either hidden system prompts or the training of models to automatically use generic Chain of Thought prompting strategies. Among the ATP reasoning strategies tried, we found that current frontier LLMs are best able to follow the bottom-up (also known as forward-chaining) strategy. A low positive correlation was found between an LLM response containing correct reasoning and arriving at the correct conclusion.
Authors: Athanasios Davvetas, Xenia Ziouvelou, Ypatia Dami, Alexios Kaponis, Konstantina Giouvanopoulou, Michael Papademas
Abstract: This paper introduces the TAI Scan Tool, a RAG-based TAI self-assessment tool with minimalistic input. The current version of the tool supports the legal TAI assessment, with a particular emphasis on facilitating compliance with the AI Act. It involves a two-step approach with a pre-screening and an assessment phase. The assessment output of the system includes insight regarding the risk-level of the AI system according to the AI Act, while at the same time retrieving relevant articles to aid with compliance and notify on their obligations. Our qualitative evaluation using use-case scenarios yields promising results, correctly predicting risk levels while retrieving relevant articles across three distinct semantic groups. Furthermore, interpretation of results shows that the tool's reasoning relies on comparison with the setting of high-risk systems, a behaviour attributed to their deployment requiring careful consideration, and therefore frequently presented within the AI Act.
Authors: Xianxuan Long, Yao Fu, Runchao Li, Mu Sheng, Haotian Yu, Xiaotian Han, Pan Li
Abstract: Large language models (LLMs) tend to follow maliciously crafted instructions to generate deceptive responses, posing safety challenges. How deceptive instructions alter the internal representations of LLM compared to truthful ones remains poorly understood beyond output analysis. To bridge this gap, we investigate when and how these representations ``flip'', such as from truthful to deceptive, under deceptive versus truthful/neutral instructions. Analyzing the internal representations of Llama-3.1-8B-Instruct and Gemma-2-9B-Instruct on a factual verification task, we find the model's instructed True/False output is predictable via linear probes across all conditions based on the internal representation. Further, we use Sparse Autoencoders (SAEs) to show that the Deceptive instructions induce significant representational shifts compared to Truthful/Neutral representations (which are similar), concentrated in early-to-mid layers and detectable even on complex datasets. We also identify specific SAE features highly sensitive to deceptive instruction and use targeted visualizations to confirm distinct truthful/deceptive representational subspaces. % Our analysis pinpoints layer-wise and feature-level correlates of instructed dishonesty, offering insights for LLM detection and control. Our findings expose feature- and layer-level signatures of deception, offering new insights for detecting and mitigating instructed dishonesty in LLMs.
Authors: Gerard Boxo, Ryan Socha, Daniel Yoo, Shivam Raval
Abstract: Sophisticated instrumentation for AI systems might have indicators that signal misalignment from human values, not unlike a "check engine" light in cars. One such indicator of misalignment is deceptiveness in generated responses. Future AI instrumentation may have the ability to detect when an LLM generates deceptive responses while reasoning about seemingly plausible but incorrect answers to factual questions. In this work, we demonstrate that linear probes on LLMs internal activations can detect deception in their responses with extremely high accuracy. Our probes reach a maximum of greater than 90% accuracy in distinguishing between deceptive and non-deceptive arguments generated by llama and qwen models ranging from 1.5B to 14B parameters, including their DeepSeek-r1 finetuned variants. We observe that probes on smaller models (1.5B) achieve chance accuracy at detecting deception, while larger models (greater than 7B) reach 70-80%, with their reasoning counterparts exceeding 90%. The layer-wise probe accuracy follows a three-stage pattern across layers: near-random (50%) in early layers, peaking in middle layers, and slightly declining in later layers. Furthermore, using an iterative null space projection approach, we find multitudes of linear directions that encode deception, ranging from 20 in Qwen 3B to nearly 100 in DeepSeek 7B and Qwen 14B models.
Authors: Prathamesh Vasudeo Naik, Naresh Kumar Dintakurthi, Zhanghao Hu, Yue Wang, Robby Qiu
Abstract: Generating regulatorily compliant Suspicious Activity Report (SAR) remains a high-cost, low-scalability bottleneck in Anti-Money Laundering (AML) workflows. While large language models (LLMs) offer promising fluency, they suffer from factual hallucination, limited crime typology alignment, and poor explainability -- posing unacceptable risks in compliance-critical domains. This paper introduces Co-Investigator AI, an agentic framework optimized to produce Suspicious Activity Reports (SARs) significantly faster and with greater accuracy than traditional methods. Drawing inspiration from recent advances in autonomous agent architectures, such as the AI Co-Scientist, our approach integrates specialized agents for planning, crime type detection, external intelligence gathering, and compliance validation. The system features dynamic memory management, an AI-Privacy Guard layer for sensitive data handling, and a real-time validation agent employing the Agent-as-a-Judge paradigm to ensure continuous narrative quality assurance. Human investigators remain firmly in the loop, empowered to review and refine drafts in a collaborative workflow that blends AI efficiency with domain expertise. We demonstrate the versatility of Co-Investigator AI across a range of complex financial crime scenarios, highlighting its ability to streamline SAR drafting, align narratives with regulatory expectations, and enable compliance teams to focus on higher-order analytical work. This approach marks the beginning of a new era in compliance reporting -- bringing the transformative benefits of AI agents to the core of regulatory processes and paving the way for scalable, reliable, and transparent SAR generation.
Authors: Beidan Liu, Zhengqiu Zhu, Chen Gao, Yong Zhao, Wei Qi, Quanjun Yin
Abstract: Nonlinear Combinatorial Optimization Problems (NCOPs) present a formidable computational hurdle in practice, as their nonconvex nature gives rise to multi-modal solution spaces that defy efficient optimization. Traditional constraint relaxation approaches rely heavily on expert-driven, iterative design processes that lack systematic automation and scalable adaptability. While recent Large Language Model (LLM)-based optimization methods show promise for autonomous problem-solving, they predominantly function as passive constraint validators rather than proactive strategy architects, failing to handle the sophisticated constraint interactions inherent to NCOPs.To address these limitations, we introduce the first end-to-end \textbf{Auto}mated \textbf{C}onstraint \textbf{O}ptimization (AutoCO) method, which revolutionizes NCOPs resolution through learning to relax with LLMs.Specifically, we leverage structured LLM reasoning to generate constraint relaxation strategies, which are dynamically evolving with algorithmic principles and executable code through a unified triple-representation scheme. We further establish a novel bidirectional (global-local) coevolution mechanism that synergistically integrates Evolutionary Algorithms for intensive local refinement with Monte Carlo Tree Search for systematic global strategy space exploration, ensuring optimal balance between intensification and diversification in fragmented solution spaces. Finally, comprehensive experiments on three challenging NCOP benchmarks validate AutoCO's consistent effectiveness and superior performance over the baselines.
Authors: Jun Wang, Jiaming Tong, Kaiyuan Tan, Yevgeniy Vorobeychik, Yiannis Kantaros
Abstract: This paper addresses planning problems for mobile robots. We consider missions that require accomplishing multiple high-level sub-tasks, expressed in natural language (NL), in a temporal and logical order. To formally define the mission, we treat these sub-tasks as atomic predicates in a Linear Temporal Logic (LTL) formula. We refer to this task specification framework as LTL-NL. Our goal is to design plans, defined as sequences of robot actions, accomplishing LTL-NL tasks. This action planning problem cannot be solved directly by existing LTL planners because of the NL nature of atomic predicates. To address it, we propose HERACLEs, a hierarchical neuro-symbolic planner that relies on a novel integration of (i) existing symbolic planners generating high-level task plans determining the order at which the NL sub-tasks should be accomplished; (ii) pre-trained Large Language Models (LLMs) to design sequences of robot actions based on these task plans; and (iii) conformal prediction acting as a formal interface between (i) and (ii) and managing uncertainties due to LLM imperfections. We show, both theoretically and empirically, that HERACLEs can achieve user-defined mission success rates. Finally, we provide comparative experiments demonstrating that HERACLEs outperforms LLM-based planners that require the mission to be defined solely using NL. Additionally, we present examples demonstrating that our approach enhances user-friendliness compared to conventional symbolic approaches.
Authors: Chenghao Huang, Xiaolu Chen, Yanru Zhang, Hao Wang
Abstract: Heterogeneity arising from label distribution skew and data scarcity can cause inaccuracy and unfairness in intelligent communication applications that heavily rely on distributed computing. To deal with it, this paper proposes a novel personalized federated learning algorithm, named Federated Contrastive Shareable Representations (FedCoSR), to facilitate knowledge sharing among clients while maintaining data privacy. Specifically, the parameters of local models' shallow layers and typical local representations are both considered as shareable information for the server and are aggregated globally. To address performance degradation caused by label distribution skew among clients, contrastive learning is adopted between local and global representations to enrich local knowledge. Additionally, to ensure fairness for clients with scarce data, FedCoSR introduces adaptive local aggregation to coordinate the global model involvement in each client. Our simulations demonstrate FedCoSR's effectiveness in mitigating label heterogeneity by achieving accuracy and fairness improvements over existing methods on datasets with varying degrees of label heterogeneity.
Authors: Jiexia Ye, Yongzi Yu, Weiqi Zhang, Le Wang, Jia Li, Fugee Tsung
Abstract: Time series data are ubiquitous across diverse real-world applications, making time series analysis critically important. Traditional approaches are largely task-specific, offering limited functionality and poor transferability. In recent years, foundation models have revolutionized NLP and CV with their remarkable cross-task transferability, zero-/few-shot learning capabilities, and multimodal integration capacity. This success has motivated increasing efforts to explore foundation models for addressing time series modeling challenges. Although some tutorials and surveys were published in the early stages of this field, the rapid pace of recent developments necessitates a more comprehensive and in-depth synthesis to cover the latest advances. Our survey aims to fill this gap by introducing a modality-aware, challenge-oriented perspective, which reveals how foundation models pre-trained on different modalities face distinct hurdles when adapted to time series tasks. Building on this perspective, we propose a taxonomy of existing works organized by pre-training modality (time series, language, and vision), analyze modality-specific challenges and categorize corresponding solutions, discussing their advantages and limitations. Beyond this, we review real-world applications to illustrate domain-specific advancements, provide open-source codes, and conclude with potential future research directions in this rapidly evolving field.
Authors: Yuu Jinnai, Ukyo Honda
Abstract: Preference optimization is a standard approach to fine-tuning large language models to align with human preferences. The quantity, diversity, and representativeness of the preference dataset are critical to the effectiveness of preference optimization. However, obtaining a large amount of preference annotations is difficult in many applications. This raises the question of how to use the limited annotation budget to create an effective preference dataset. To this end, we propose Annotation-Efficient Preference Optimization (AEPO). Instead of exhaustively annotating preference over all available response texts, AEPO selects a subset of responses that maximizes diversity and representativeness from the available responses and then annotates preference over the selected ones. In this way, AEPO focuses the annotation budget on labeling preferences over a smaller but informative subset of responses. We evaluate the performance of preference learning using AEPO on three datasets and show that it outperforms the baselines with the same annotation budget. Our code is available at https://github.com/CyberAgentAILab/annotation-efficient-po
URLs: https://github.com/CyberAgentAILab/annotation-efficient-po
Authors: Minttu Alakuijala, Reginald McLean, Isaac Woungang, Nariman Farsad, Samuel Kaski, Pekka Marttinen, Kai Yuan
Abstract: Natural language is often the easiest and most convenient modality for humans to specify tasks for robots. However, learning to ground language to behavior typically requires impractical amounts of diverse, language-annotated demonstrations collected on each target robot. In this work, we aim to separate the problem of what to accomplish from how to accomplish it, as the former can benefit from substantial amounts of external observation-only data, and only the latter depends on a specific robot embodiment. To this end, we propose Video-Language Critic, a reward model that can be trained on readily available cross-embodiment data using contrastive learning and a temporal ranking objective, and use it to score behavior traces from a separate actor. When trained on Open X-Embodiment data, our reward model enables 2x more sample-efficient policy training on Meta-World tasks than a sparse reward only, despite a significant domain gap. Using in-domain data but in a challenging task generalization setting on Meta-World, we further demonstrate more sample-efficient training than is possible with prior language-conditioned reward models that are either trained with binary classification, use static images, or do not leverage the temporal information present in video data.
Authors: Soyeong Jeong, Jinheon Baek, Sukmin Cho, Sung Ju Hwang, Jong C. Park
Abstract: Information retrieval models that aim to search for documents relevant to a query have shown multiple successes, which have been applied to diverse tasks. Yet, the query from the user is oftentimes short, which challenges the retrievers to correctly fetch relevant documents. To tackle this, previous studies have proposed expanding the query with a couple of additional (user-related) features related to it. However, they may be suboptimal to effectively augment the query, and there is plenty of other information available to augment it in a relational database. Motivated by this fact, we present a novel retrieval framework called Database-Augmented Query representation (DAQu), which augments the original query with various (query-related) metadata across multiple tables. In addition, as the number of features in the metadata can be very large and there is no order among them, we encode them with the graph-based set-encoding strategy, which considers hierarchies of features in the database without order. We validate our DAQu in diverse retrieval scenarios, demonstrating that it significantly enhances overall retrieval performance over relevant baselines. Our code is available at \href{https://github.com/starsuzi/DAQu}{this https URL}.
Authors: Wenqian Chen, Amanda A. Howard, Panos Stinis
Abstract: Physics-informed deep learning has emerged as a promising alternative for solving partial differential equations. However, for complex problems, training these networks can still be challenging, often resulting in unsatisfactory accuracy and efficiency. In this work, we demonstrate that the failure of plain physics-informed neural networks arises from the significant discrepancy in the convergence rate of residuals at different training points, where the slowest convergence rate dominates the overall solution convergence. Based on these observations, we propose a pointwise adaptive weighting method that balances the residual decay rate across different training points. The performance of our proposed adaptive weighting method is compared with current state-of-the-art adaptive weighting methods on benchmark problems for both physics-informed neural networks and physics-informed deep operator networks. Through extensive numerical results we demonstrate that our proposed approach of balanced residual decay rates offers several advantages, including bounded weights, high prediction accuracy, fast convergence rate, low training uncertainty, low computational cost, and ease of hyperparameter tuning.
Authors: Jeongeun Lee, Seongku Kang, Won-Yong Shin, Jeongwhan Choi, Noseong Park, Dongha Lee
Abstract: Collaborative Filtering (CF) is a foundational approach in recommender systems, but it struggles with challenges such as data sparsity and the cold-start problem. Cross-Domain Recommendation (CDR) has emerged as a promising solution by leveraging dense domains to improve recommendations in sparse target domains. However, existing CDR methods face significant limitations, including their reliance on overlapping users as a bridge between domains and their inability to address domain sensitivity, i.e., differences in user behaviors and characteristics across domains, effectively. To overcome these limitations, we propose CGSP, a unified and adaptive CDR framework based on graph signal processing (GSP). CGSP supports both intra-domain and inter-domain recommendations while adaptively controlling the influence of the source domain through a simple hyperparameter. The framework constructs a cross-domain similarity graph by integrating target-only and source-bridged similarity graphs to capture both intra-domain and inter-domain relationships. This graph is then processed through graph filtering techniques to propagate and enhance local signals. Finally, personalized graph signals are constructed, tailored separately for users in the source and target domains, enabling CGSP to function as a unified framework for CDR scenarios. Extensive evaluation shows that CGSP outperforms state-of-the-art baselines across diverse cross-domain settings, with notable gains in low-overlap scenarios, underscoring its practicality for real-world applications.
Authors: Le Xue, Manli Shu, Anas Awadalla, Jun Wang, An Yan, Senthil Purushwalkam, Honglu Zhou, Viraj Prabhu, Yutong Dai, Michael S Ryoo, Shrikant Kendre, Jieyu Zhang, Shaoyen Tseng, Gustavo A Lujan-Moreno, Matthew L Olson, Musashi Hinck, David Cobbley, Vasudev Lal, Can Qin, Shu Zhang, Chia-Chih Chen, Ning Yu, Juntao Tan, Tulika Manoj Awalgaonkar, Shelby Heinecke, Huan Wang, Yejin Choi, Ludwig Schmidt, Zeyuan Chen, Silvio Savarese, Juan Carlos Niebles, Caiming Xiong, Ran Xu
Abstract: This paper introduces BLIP-3, an open framework for developing Large Multimodal Models (LMMs). The framework comprises meticulously curated datasets, a training recipe, model architectures, and a resulting suite of LMMs. We release 4B and 14B models, including both the pre-trained base model and the instruction fine-tuned ones. Our models undergo rigorous evaluation across a range of tasks, including both single and multi-image benchmarks. Our models demonstrate competitive performance among open-source LMMs with similar model sizes. Our resulting LMMs demonstrate competitive performance among open-source LMMs with similar model sizes, with the ability to comprehend interleaved image-text inputs. Our training code, models, and all datasets used in this work, including the three largescale datasets we create and the preprocessed ones, will be open-sourced to better support the research community.
Authors: Yanda Li, Chi Zhang, Wenjia Jiang, Wanqi Yang, Bin Fu, Pei Cheng, Xin Chen, Ling Chen, Yunchao Wei
Abstract: With the advancement of Multimodal Large Language Models (MLLM), LLM-driven visual agents are increasingly impacting software interfaces, particularly those with graphical user interfaces. This work introduces a novel LLM-based multimodal agent framework for mobile devices. This framework, capable of navigating mobile devices, emulates human-like interactions. Our agent constructs a flexible action space that enhances adaptability across various applications including parser, text and vision descriptions. The agent operates through two main phases: exploration and deployment. During the exploration phase, functionalities of user interface elements are documented either through agent-driven or manual explorations into a customized structured knowledge base. In the deployment phase, RAG technology enables efficient retrieval and update from this knowledge base, thereby empowering the agent to perform tasks effectively and accurately. This includes performing complex, multi-step operations across various applications, thereby demonstrating the framework's adaptability and precision in handling customized task workflows. Our experimental results across various benchmarks demonstrate the framework's superior performance, confirming its effectiveness in real-world scenarios. Our code will be open source soon.
Authors: Xiaolong Wang, Zhi-Qi Cheng, Jue Wang, Xiaojiang Peng
Abstract: Fashion image editing is a crucial tool for designers to convey their creative ideas by visualizing design concepts interactively. Current fashion image editing techniques, though advanced with multimodal prompts and powerful diffusion models, often struggle to accurately identify editing regions and preserve the desired garment texture detail. To address these challenges, we introduce a new multimodal fashion image editing architecture based on latent diffusion models, called Detail-Preserved Diffusion Models (DPDEdit). DPDEdit guides the fashion image generation of diffusion models by integrating text prompts, region masks, human pose images, and garment texture images. To precisely locate the editing region, we first introduce Grounded-SAM to predict the editing region based on the user's textual description, and then combine it with other conditions to perform local editing. To transfer the detail of the given garment texture into the target fashion image, we propose a texture injection and refinement mechanism. Specifically, this mechanism employs a decoupled cross-attention layer to integrate textual descriptions and texture images, and incorporates an auxiliary U-Net to preserve the high-frequency details of generated garment texture. Additionally, we extend the VITON-HD dataset using a multimodal large language model to generate paired samples with texture images and textual descriptions. Extensive experiments show that our DPDEdit outperforms state-of-the-art methods in terms of image fidelity and coherence with the given multimodal inputs.
Authors: Zhiqiang Yuan, Weitong Chen, Hanlin Wang, Kai Yu, Xin Peng, Yiling Lou
Abstract: Code translation converts code from one programming language to another while maintaining its original functionality, which is crucial for software migration, system refactoring, and cross-platform development. Traditional rule-based methods rely on manually-written rules, which can be time-consuming and often result in less readable code. To overcome this, learning-based methods have been developed, leveraging parallel data to train models for automated code translation. More recently, the advance of Large Language Models (LLMs) further boosts learning-based code translation. Although promising, LLM-translated program still suffers from diverse quality issues (e.g., syntax errors and semantic errors). In particular, it can be challenging for LLMs to self-debug these errors when simply provided with the corresponding error messages. In this work, we propose a novel LLM-based multi-agent system TRANSAGENT, which enhances LLM-based code translation by fixing the syntax errors and semantic errors with the synergy between four LLM-based agents, including Initial Code Translator, Syntax Error Fixer, Code Aligner, and Semantic Error Fixer. The main insight of TRANSAGENT is to first localize the error code block in the target program based on the execution alignment between the target and source program, which can narrow down the fixing space and thus lower down the fixing difficulties. To evaluate TRANSAGENT, we first construct a new benchmark from recent programming tasks to mitigate the potential data leakage issue. On our benchmark, TRANSAGENT outperforms the latest LLM-based code translation technique UniTrans in both translation effectiveness and efficiency; additionally, our evaluation on different LLMs show the generalization of TRANSAGENT and our ablation study shows the contribution of each agent.
Authors: Minh Pham Dinh, Munira Syed, Michael G Yankoski, Trenton W. Ford
Abstract: Designing a generalist scientific agent capable of performing tasks in laboratory settings to assist researchers has become a key goal in recent Artificial Intelligence (AI) research. Unlike everyday tasks, scientific tasks are inherently more delicate and complex, requiring agents to possess a higher level of reasoning ability, structured and temporal understanding of their environment, and a strong emphasis on safety. Existing approaches often fail to address these multifaceted requirements. To tackle these challenges, we present DAVIS. Unlike traditional retrieval-augmented generation (RAG) approaches, DAVIS incorporates structured and temporal memory, which enables model-based planning. Additionally, DAVIS implements an agentic, multi-turn retrieval system, similar to a human's inner monologue, allowing for a greater degree of reasoning over past experiences. DAVIS demonstrates substantially improved performance on the ScienceWorld benchmark comparing to previous approaches on 8 out of 9 elementary science subjects. In addition, DAVIS's World Model demonstrates competitive performance on the famous HotpotQA and MusiqueQA dataset for multi-hop question answering. To the best of our knowledge, DAVIS is the first RAG agent to employ an interactive retrieval method in a RAG pipeline.
Authors: Siyuan Huang, Zhiyuan Ma, Jintao Du, Changhua Meng, Weiqiang Wang, Zhouhan Lin
Abstract: Self-Consistency, a widely-used decoding strategy, significantly boosts the reasoning capabilities of Large Language Models (LLMs). However, it depends on the plurality voting rule, which focuses on the most frequent answer while overlooking all other minority responses. These inconsistent minority views often illuminate areas of uncertainty within the model's generation process. To address this limitation, we present Mirror-Consistency, an enhancement of the standard Self-Consistency approach. Our method incorporates a 'reflective mirror' into the self-ensemble decoding process and enables LLMs to critically examine inconsistencies among multiple generations. Additionally, just as humans use the mirror to better understand themselves, we propose using Mirror-Consistency to enhance the sample-based confidence calibration methods, which helps to mitigate issues of overconfidence. Our experimental results demonstrate that Mirror-Consistency yields superior performance in both reasoning accuracy and confidence calibration compared to Self-Consistency.
Authors: Luca Rolshoven, Vishvaksenan Rasiah, Srinanda Br\"ugger Bose, Sarah Hostettler, Lara Burkhalter, Matthias St\"urmer, Joel Niklaus
Abstract: Legal research is a time-consuming task that most lawyers face on a daily basis. A large part of legal research entails looking up relevant caselaw and bringing it in relation to the case at hand. Lawyers heavily rely on summaries (also called headnotes) to find the right cases quickly. However, not all decisions are annotated with headnotes and writing them is time-consuming. Automated headnote creation has the potential to make hundreds of thousands of decisions more accessible for legal research in Switzerland alone. To kickstart this, we introduce the Swiss Leading Decision Summarization ( SLDS) dataset, a novel cross-lingual resource featuring 18K court rulings from the Swiss Federal Supreme Court (SFSC), in German, French, and Italian, along with German headnotes. We fine-tune and evaluate three mT5 variants, along with proprietary models. Our analysis highlights that while proprietary models perform well in zero-shot and one-shot settings, fine-tuned smaller models still provide a strong competitive edge. We publicly release the dataset to facilitate further research in multilingual legal summarization and the development of assistive technologies for legal professionals
Authors: Erin Carson, Xinye Chen, Cheng Kang
Abstract: The success of large language models (LLMs) for time series has been demonstrated in previous work. Utilizing a symbolic time series representation, one can efficiently bridge the gap between LLMs and time series. However, the remaining challenge is to exploit the semantic information hidden in time series by using symbols or existing tokens of LLMs, while aligning the embedding space of LLMs according to the hidden information of time series. The symbolic time series approximation (STSA) method called adaptive Brownian bridge-based symbolic aggregation (ABBA) shows outstanding efficacy in preserving salient time series features by modeling time series patterns in terms of amplitude and period while using existing tokens of LLMs. In this paper, we introduce a method, called LLM-ABBA, that integrates ABBA into large language models for various downstream time series tasks. By symbolizing time series, LLM-ABBA compares favorably to the recent state-of-the-art (SOTA) in UCR and three medical time series classification tasks. Meanwhile, a fixed-polygonal chain trick in ABBA is introduced to \kc{avoid obvious drifting} during prediction tasks by significantly mitigating the effects of cumulative error arising from misused symbols during the transition from symbols to numerical values. In time series regression tasks, LLM-ABBA achieves the new SOTA on Time Series Extrinsic Regression (TSER) benchmarks. LLM-ABBA also shows competitive prediction capability compared to recent SOTA time series prediction results. We believe this framework can also seamlessly extend to other time series tasks.
Authors: Zilyu Ji, Yuntian Shen, Jionghao Lin, Kenneth R. Koedinger
Abstract: Protecting Personally Identifiable Information (PII), such as names, is a critical requirement in learning technologies to safeguard student and teacher privacy and maintain trust. Accurate PII detection is an essential step toward anonymizing sensitive information while preserving the utility of educational data. Motivated by recent advancements in artificial intelligence, our study investigates the GPT-4o-mini model as a cost-effective and efficient solution for PII detection tasks. We explore both prompting and fine-tuning approaches and compare GPT-4o-mini's performance against established frameworks, including Microsoft Presidio and Azure AI Language. Our evaluation on two public datasets, CRAPII and TSCC, demonstrates that the fine-tuned GPT-4o-mini model achieves superior performance, with a recall of 0.9589 on CRAPII. Additionally, fine-tuned GPT-4o-mini significantly improves precision scores (a threefold increase) while reducing computational costs to nearly one-tenth of those associated with Azure AI Language. Furthermore, our bias analysis reveals that the fine-tuned GPT-4o-mini model consistently delivers accurate results across diverse cultural backgrounds and genders. The generalizability analysis using the TSCC dataset further highlights its robustness, achieving a recall of 0.9895 with minimal additional training data from TSCC. These results emphasize the potential of fine-tuned GPT-4o-mini as an accurate and cost-effective tool for PII detection in educational data. It offers robust privacy protection while preserving the data's utility for research and pedagogical analysis. Our code is available on GitHub: https://github.com/AnonJD/PrivacyAI
Authors: Nafis Irtiza Tripto, Saranya Venkatraman, Mahjabin Nahar, Dongwon Lee
Abstract: The rapid advancement of Large Language Models (LLMs) has revolutionized text generation but also raised concerns about potential misuse, making detecting LLM-generated text (AI text) increasingly essential. While prior work has focused on identifying AI text and effectively checkmating it, our study investigates a less-explored territory: portraying the nuanced distinctions between human and AI texts across text segments (introduction, body, and conclusion). Whether LLMs excel or falter in incorporating linguistic ingenuity across text segments, the results will critically inform their viability and boundaries as effective creative assistants to humans. Through an analogy with the structure of chess games, comprising opening, middle, and end games, we analyze segment-specific patterns to reveal where the most striking differences lie. Although AI texts closely resemble human writing in the body segment due to its length, deeper analysis shows a higher divergence in features dependent on the continuous flow of language, making it the most informative segment for detection. Additionally, human texts exhibit greater stylistic variation across segments, offering a new lens for distinguishing them from AI. Overall, our findings provide fresh insights into human-AI text differences and pave the way for more effective and interpretable detection strategies. Codes available at https://github.com/tripto03/chess_inspired_human_ai_text_distinction.
URLs: https://github.com/tripto03/chess_inspired_human_ai_text_distinction.
Authors: Xinran Zheng, Shuo Yang, Edith C. H. Ngai, Suman Jana, Lorenzo Cavallaro
Abstract: Learning-based Android malware detectors degrade over time due to natural distribution drift caused by malware variants and new families. This paper systematically investigates the challenges classifiers trained with empirical risk minimization (ERM) face against such distribution shifts and attributes their shortcomings to their inability to learn stable discriminative features. Invariant learning theory offers a promising solution by encouraging models to generate stable representations crossing environments that expose the instability of the training set. However, the lack of prior environment labels, the diversity of drift factors, and low-quality representations caused by diverse families make this task challenging. To address these issues, we propose TIF, the first temporal invariant training framework for malware detection, which aims to enhance the ability of detectors to learn stable representations across time. TIF organizes environments based on application observation dates to reveal temporal drift, integrating specialized multi-proxy contrastive learning and invariant gradient alignment to generate and align environments with high-quality, stable representations. TIF can be seamlessly integrated into any learning-based detector. Experiments on a decade-long dataset show that TIF excels, particularly in early deployment stages, addressing real-world needs and outperforming state-of-the-art methods.
Authors: Federico Gibellini, Piero Fraternali, Giacomo Boracchi, Luca Morandini, Thomas Martinoli, Andrea Diecidue, Simona Malegori
Abstract: Improper solid waste management represents both a serious threat to ecosystem health and a significant source of revenues for criminal organizations perpetrating environmental crimes. This issue can be mitigated thanks to the increasing availability of Very-High-Resolution Remote Sensing (VHR RS) images. Modern image-analysis tools support automated photo-interpretation and large territory scanning in search of illegal waste disposal sites. This paper illustrates a semi-automatic waste detection pipeline, developed in collaboration with a regional environmental protection agency, for detecting candidate illegal dumping sites in VHR RS images. To optimize the effectiveness of the waste detector at the core of the pipeline, extensive experiments evaluate such design choices as the network architecture, the ground resolution and geographic span of the input images, as well as the pretraining procedures. The best model attains remarkable performance, achieving 92.02 % F1-Score and 94.56 % Accuracy. A generalization study assesses the performance variation when the detector processes images from various territories substantially different from the one used during training, incurring only a moderate performance loss, namely an average 5.1 % decrease in the F1-Score. Finally, an exercise in which expert photo-interpreters compare the effort required to scan large territories with and without support from the waste detector assesses the practical benefit of introducing a computer-aided image analysis tool in a professional environmental protection agency. Results show that a reduction of up to 30 % of the time spent for waste site detection can be attained.
Authors: Sean McLeish, John Kirchenbauer, David Yu Miller, Siddharth Singh, Abhinav Bhatele, Micah Goldblum, Ashwinee Panda, Tom Goldstein
Abstract: Scaling laws are typically fit using a family of models with a narrow range of frozen hyper-parameter choices. In this work we study scaling laws using multiple architectural shapes and hyperparameter choices, highlighting their impact on resulting prescriptions. As a primary artifact of our research, we release the Gemstones: an open-source scaling law dataset, consisting of over 4000 checkpoints from transformers with up to 2 billion parameters and diverse architectural shapes; including ablations over learning rate and cooldown. Our checkpoints enable more complex studies of scaling, such as analyzing the relationship between width and depth. By examining our model suite, we find that the prescriptions of scaling laws can be highly sensitive to the experimental design process and the specific model checkpoints used during fitting.
Authors: Nurit Cohen-Inger, Yehonatan Elisha, Bracha Shapira, Lior Rokach, Seffi Cohen
Abstract: Large language models (LLMs) often appear to excel on public benchmarks, but these high scores may mask an overreliance on dataset-specific surface cues rather than true language understanding. We introduce the Chameleon Benchmark Overfit Detector (C-BOD), a meta-evaluation framework that systematically distorts benchmark prompts via a parametric transformation and detects overfitting of LLMs. By rephrasing inputs while preserving their semantic content and labels, C-BOD exposes whether a model's performance is driven by memorized patterns. Evaluated on the MMLU benchmark using 26 leading LLMs, our method reveals an average performance degradation of 2.15% under modest perturbations, with 20 out of 26 models exhibiting statistically significant differences. Notably, models with higher baseline accuracy exhibit larger performance differences under perturbation, and larger LLMs tend to be more sensitive to rephrasings, indicating that both cases may overrely on fixed prompt patterns. In contrast, the Llama family and models with lower baseline accuracy show insignificant degradation, suggesting reduced dependency on superficial cues. Moreover, C-BOD's dataset- and model-agnostic design allows easy integration into training pipelines to promote more robust language understanding. Our findings challenge the community to look beyond leaderboard scores and prioritize resilience and generalization in LLM evaluation.
Authors: Junrui Wen, Yifei Li, Bart Selman, Kun He
Abstract: Neural solvers have shown significant potential in solving the Traveling Salesman Problem (TSP), yet current approaches face significant challenges. Supervised learning (SL)-based solvers require large amounts of high-quality labeled data, while reinforcement learning (RL)-based solvers, though less dependent on such data, often suffer from inefficiencies. To address these limitations, we propose LocalEscaper, a novel weakly-supervised learning framework for large-scale TSP. LocalEscaper effectively combines the advantages of both SL and RL, enabling effective training on datasets with low-quality labels. To further enhance solution quality, we introduce a regional reconstruction strategy, which is the key technique of this paper and mitigates the local-optima problem common in existing local reconstruction methods. Experimental results on both synthetic and real-world datasets demonstrate that LocalEscaper outperforms existing neural solvers, achieving remarkable results.
Authors: Manar Aljohani, Jun Hou, Sindhura Kommu, Xuan Wang
Abstract: The application of large language models (LLMs) in healthcare holds significant promise for enhancing clinical decision-making, medical research, and patient care. However, their integration into real-world clinical settings raises critical concerns around trustworthiness, particularly around dimensions of truthfulness, privacy, safety, robustness, fairness, and explainability. These dimensions are essential for ensuring that LLMs generate reliable, unbiased, and ethically sound outputs. While researchers have recently begun developing benchmarks and evaluation frameworks to assess LLM trustworthiness, the trustworthiness of LLMs in healthcare remains underexplored, lacking a systematic review that provides a comprehensive understanding and future insights. This survey addresses that gap by providing a comprehensive review of current methodologies and solutions aimed at mitigating risks across key trust dimensions. We analyze how each dimension affects the reliability and ethical deployment of healthcare LLMs, synthesize ongoing research efforts, and identify critical gaps in existing approaches. We also identify emerging challenges posed by evolving paradigms, such as multi-agent collaboration, multi-modal reasoning, and the development of small open-source medical models. Our goal is to guide future research toward more trustworthy, transparent, and clinically viable LLMs.
Authors: Xiwen Liang, Min Lin, Weiqi Ruan, Rongtao Xu, Yuecheng Liu, Jiaqi Chen, Bingqian Lin, Yuzheng Zhuang, Xiaodan Liang
Abstract: Existing methods for vision-language task planning excel in short-horizon tasks but often fall short in complex, long-horizon planning within dynamic environments. These challenges primarily arise from the difficulty of effectively training models to produce high-quality reasoning processes for long-horizon tasks. To address this, we propose Structured Preference Optimization (SPO), which aims to enhance reasoning and action selection in long-horizon task planning through structured preference evaluation and optimized training strategies. Specifically, SPO introduces: 1) Preference-Based Scoring and Optimization, which systematically evaluates reasoning chains based on task relevance, visual grounding, and historical consistency; and 2) Curriculum-Guided Training, where the model progressively adapts from simple to complex tasks, improving its generalization ability in long-horizon scenarios and enhancing reasoning robustness. To advance research in vision-language long-horizon task planning, we introduce ExtendaBench, a comprehensive benchmark covering 1,509 tasks across VirtualHome and Habitat 2.0, categorized into ultra-short, short, medium, and long tasks. Experimental results demonstrate that SPO significantly improves reasoning quality and final decision accuracy, outperforming prior methods on long-horizon tasks and underscoring the effectiveness of preference-driven optimization in vision-language task planning. Specifically, SPO achieves a +5.98% GCR and +4.68% SR improvement in VirtualHome and a +3.30% GCR and +2.11% SR improvement in Habitat over the best-performing baselines.
Authors: Youngbin Choi, Seunghyuk Cho, Minjong Lee, MoonJeong Park, Yesong Ko, Jungseul Ok, Dongwoo Kim
Abstract: Personalizing large language models (LLMs) is important for aligning outputs with diverse user preferences, yet existing methods struggle with flexibility and generalization. We propose CoPL (Collaborative Preference Learning), a graph-based collaborative filtering framework that models user-response relationships to enhance preference estimation, particularly in sparse annotation settings. By integrating a mixture of LoRA experts, CoPL efficiently fine-tunes LLMs while dynamically balancing shared and user-specific preferences. Additionally, an optimization-free adaptation strategy enables generalization to unseen users without fine-tuning. Experiments on UltraFeedback-P demonstrate that CoPL outperforms existing personalized reward models, effectively capturing both common and controversial preferences, making it a scalable solution for personalized LLM alignment. The code is available at https://github.com/ml-postech/CoPL.
Authors: Adel ElZemity, Budi Arief, Shujun Li
Abstract: The integration of large language models (LLMs) into cyber security applications presents both opportunities and critical safety risks. We introduce CyberLLMInstruct, a dataset of 54,928 pseudo-malicious instruction-response pairs spanning cyber security tasks including malware analysis, phishing simulations, and zero-day vulnerabilities. Our comprehensive evaluation using seven open-source LLMs reveals a critical trade-off: while fine-tuning improves cyber security task performance (achieving up to 92.50% accuracy on CyberMetric), it severely compromises safety resilience across all tested models and attack vectors (e.g., Llama 3.1 8B's security score against prompt injection drops from 0.95 to 0.15). The dataset incorporates diverse sources including CTF challenges, academic papers, industry reports, and CVE databases to ensure comprehensive coverage of cyber security domains. Our findings highlight the unique challenges of securing LLMs in adversarial domains and establish the critical need for developing fine-tuning methodologies that balance performance gains with safety preservation in security-sensitive domains.
Authors: Oucheng Huang, Yuhang Ma, Zeng Zhao, Mingrui Wu, Jiayi Ji, Rongsheng Zhang, Zhipeng Hu, Xiaoshuai Sun, Rongrong Ji
Abstract: ComfyUI is a popular workflow-based interface that allows users to customize image generation tasks through an intuitive node-based system. However, the complexity of managing node connections and diverse modules can be challenging for users. In this paper, we introduce ComfyGPT, a self-optimizing multi-agent system designed to generate ComfyUI workflows based on task descriptions automatically. The key innovations of ComfyGPT include: (1) consisting of four specialized agents to build a multi-agent workflow generation system: ReformatAgent, FlowAgent, RefineAgent, and ExecuteAgent; (2) focusing on generating precise node connections instead of entire workflows, improving generation accuracy; and (3) enhancing workflow generation through reinforcement learning. Moreover, we introduce FlowDataset, a large-scale dataset containing 13,571 workflow-description pairs, and FlowBench, a comprehensive benchmark for evaluating workflow generation systems. Additionally, we propose four novel evaluation metrics: Format Validation (FV), Pass Accuracy (PA), Pass Instruct Alignment (PIA), and Pass Node Diversity (PND). Experimental results demonstrate that ComfyGPT significantly outperforms existing LLM-based methods in workflow generation, making it a significant step forward in this field. Code is avaliable at https://github.com/comfygpt/comfygpt.
Authors: Rajvee Sheth, Himanshu Beniwal, Mayank Singh
Abstract: We introduce COMI-LINGUA, the largest manually annotated Hindi-English code-mixed dataset, comprising 125K+ high-quality instances across five core NLP tasks: Matrix Language Identification, Token-level Language Identification, Part-Of-Speech Tagging, Named Entity Recognition, and Machine Translation. Each instance is annotated by three bilingual annotators, yielding over 376K expert annotations with strong inter-annotator agreement (Fleiss' Kappa $\geq$ 0.81). The rigorously preprocessed and filtered dataset covers both Devanagari and Roman scripts and spans diverse domains, ensuring real-world linguistic coverage. Evaluation reveals that closed-source LLMs significantly outperform traditional tools and open-source models in zero-shot settings. Notably, one-shot prompting consistently boosts performance across tasks, especially in structure-sensitive predictions like POS and NER. Fine-tuning state-of-the-art LLMs on COMI-LINGUA demonstrates substantial improvements, achieving up to 95.25 F1 in NER, 98.77 F1 in MLI, and competitive MT performance, setting new benchmarks for Hinglish code-mixed text. COMI-LINGUA is publicly available at this URL: https://huggingface.co/datasets/LingoIITGN/COMI-LINGUA.
URLs: https://huggingface.co/datasets/LingoIITGN/COMI-LINGUA.
Authors: Mahjabin Nahar, Eun-Ju Lee, Jin Won Park, Dongwon Lee
Abstract: While we increasingly rely on large language models (LLMs) for various tasks, these models are known to produce inaccurate content or 'hallucinations' with potentially disastrous consequences. The recent integration of web search results into LLMs prompts the question of whether people utilize them to verify the generated content, thereby accurately detecting hallucinations. An online experiment (N=560) investigated how the provision of search results, either static (i.e., fixed search results provided by LLM) or dynamic (i.e., participant-led searches), affects participants' perceived accuracy of LLM-generated content (i.e., genuine, minor hallucination, major hallucination), self-confidence in accuracy ratings, as well as their overall evaluation of the LLM, as compared to the control condition (i.e., no search results). Results showed that participants in both static and dynamic conditions (vs. control) rated hallucinated content to be less accurate and perceived the LLM more negatively. However, those in the dynamic condition rated genuine content as more accurate and demonstrated greater overall self-confidence in their assessments than those in the static search or control conditions. We highlighted practical implications of incorporating web search functionality into LLMs in real-world contexts.
Authors: Congcong Wen, Geeta Chandra Raju Bethala, Yu Hao, Niraj Pudasaini, Hao Huang, Shuaihang Yuan, Baoru Huang, Anh Nguyen, Anthony Tzes, Yi Fang
Abstract: Humanoid loco-manipulation, which integrates whole-body locomotion with dexterous manipulation, remains a fundamental challenge in robotics. Beyond whole-body coordination and balance, a central difficulty lies in understanding human instructions and translating them into coherent sequences of embodied actions. Recent advances in foundation models provide transferable multimodal representations and reasoning capabilities, yet existing efforts remain largely restricted to either locomotion or manipulation in isolation, with limited applicability to humanoid settings. In this paper, we propose Humanoid-COA, the first humanoid agent framework that integrates foundation model reasoning with an Embodied Chain-of-Action (CoA) mechanism for zero-shot loco-manipulation. Within the perception--reasoning--action paradigm, our key contribution lies in the reasoning stage, where the proposed CoA mechanism decomposes high-level human instructions into structured sequences of locomotion and manipulation primitives through affordance analysis, spatial inference, and whole-body action reasoning. Extensive experiments on two humanoid robots, Unitree H1-2 and G1, in both an open test area and an apartment environment, demonstrate that our framework substantially outperforms prior baselines across manipulation, locomotion, and loco-manipulation tasks, achieving robust generalization to long-horizon and unstructured scenarios. Project page: https://humanoid-coa.github.io/
Authors: Gergely D. N\'emeth, Eros Fan\`i, Yeat Jeng Ng, Barbara Caputo, Miguel \'Angel Lozano, Nuria Oliver, Novi Quadrianto
Abstract: Federated Learning (FL) enables decentralized training of machine learning models on distributed data while preserving privacy. However, in real-world FL settings, client data is often non-identically distributed and imbalanced, resulting in statistical data heterogeneity which impacts the generalization capabilities of the server's model across clients, slows convergence and reduces performance. In this paper, we address this challenge by proposing first a characterization of statistical data heterogeneity by means of 6 metrics of global and client attribute imbalance, class imbalance, and spurious correlations. Next, we create and share 7 computer vision datasets for binary and multiclass image classification tasks in Federated Learning that cover a broad range of statistical data heterogeneity and hence simulate real-world situations. Finally, we propose FEDDIVERSE, a novel client selection algorithm in FL which is designed to manage and leverage data heterogeneity across clients by promoting collaboration between clients with complementary data distributions. Experiments on the seven proposed FL datasets demonstrate FEDDIVERSE's effectiveness in enhancing the performance and robustness of a variety of FL methods while having low communication and computational overhead.
Authors: Md Fahimuzzman Sohan, Raid Alzubi, Hadeel Alzoubi, Eid Albalawi, A. H. Abdul Hafez
Abstract: Cattle lameness is a prevalent health problem in livestock farming, often resulting from hoof injuries or infections, and severely impacts animal welfare and productivity. Early and accurate detection is critical for minimizing economic losses and ensuring proper treatment. This study proposes a spatiotemporal deep learning framework for automated cattle lameness detection using publicly available video data. We curate and publicly release a balanced set of 50 online video clips featuring 42 individual cattle, recorded from multiple viewpoints in both indoor and outdoor environments. The videos were categorized into lame and non-lame classes based on visual gait characteristics and metadata descriptions. After applying data augmentation techniques to enhance generalization, two deep learning architectures were trained and evaluated: 3D Convolutional Neural Networks (3D CNN) and Convolutional Long-Short-Term Memory (ConvLSTM2D). The 3D CNN achieved a video-level classification accuracy of 90%, with a precision, recall, and F1 score of 90.9% each, outperforming the ConvLSTM2D model, which achieved 85% accuracy. Unlike conventional approaches that rely on multistage pipelines involving object detection and pose estimation, this study demonstrates the effectiveness of a direct end-to-end video classification approach. Compared with the best end-to-end prior method (C3D-ConvLSTM, 90.3%), our model achieves comparable accuracy while eliminating pose estimation pre-processing.The results indicate that deep learning models can successfully extract and learn spatio-temporal features from various video sources, enabling scalable and efficient cattle lameness detection in real-world farm settings.
Authors: Bingye Zhou, Caiyang Yu
Abstract: Neural Architecture Search (NAS) has gained widespread attention for its transformative potential in deep learning model design. However, the vast and complex search space of NAS leads to significant computational and time costs. Neural Architecture Generation (NAG) addresses this by reframing NAS as a generation problem, enabling the precise generation of optimal architectures for specific tasks. Despite its promise, mainstream methods like diffusion models face limitations in global search capabilities and are still hindered by high computational and time demands. To overcome these challenges, we propose Evolutionary Diffusion-based Neural Architecture Generation (EDNAG), a novel approach that achieves efficient and training-free architecture generation. EDNAG leverages evolutionary algorithms to simulate the denoising process in diffusion models, using fitness to guide the transition from random Gaussian distributions to optimal architecture distributions. This approach combines the strengths of evolutionary strategies and diffusion models, enabling rapid and effective architecture generation. Extensive experiments demonstrate that EDNAG achieves state-of-the-art (SOTA) performance in architecture optimization, with an improvement in accuracy of up to 10.45%. Furthermore, it eliminates the need for time-consuming training and boosts inference speed by an average of 50 times, showcasing its exceptional efficiency and effectiveness.
Authors: Xiyu Zhou, Ruiyin Li, Peng Liang, Beiqi Zhang, Mojtaba Shahin, Zengyang Li, Chen Yang
Abstract: Design Rationale (DR) for software architecture decisions refers to the reasoning underlying architectural choices, which provides valuable insights into the different phases of the architecting process throughout software development. However, in practice, DR is often inadequately documented due to a lack of motivation and effort from developers. With the recent advancements in Large Language Models (LLMs), their capabilities in text comprehension, reasoning, and generation may enable the generation and recovery of DR for architecture decisions. In this study, we evaluated the performance of LLMs in generating DR for architecture decisions. First, we collected 50 Stack Overflow (SO) posts, 25 GitHub issues, and 25 GitHub discussions related to architecture decisions to construct a dataset of 100 architecture-related problems. Then, we selected five LLMs to generate DR for the architecture decisions with three prompting strategies, including zero-shot, chain of thought (CoT), and LLM-based agents. With the DR provided by human experts as ground truth, the Precision of LLM-generated DR with the three prompting strategies ranges from 0.267 to 0.278, Recall from 0.627 to 0.715, and F1-score from 0.351 to 0.389. Additionally, 64.45% to 69.42% of the arguments of DR not mentioned by human experts are also helpful, 4.12% to 4.87% of the arguments have uncertain correctness, and 1.59% to 3.24% of the arguments are potentially misleading. To further understand the trustworthiness and applicability of LLM-generated DR in practice, we conducted semi-structured interviews with six practitioners. Based on the experimental and interview results, we discussed the pros and cons of the three prompting strategies, the strengths and limitations of LLM-generated DR, and the implications for the practical use of LLM-generated DR.
Authors: Tongyu Wen, Chenglong Wang, Xiyuan Yang, Haoyu Tang, Yueqi Xie, Lingjuan Lyu, Zhicheng Dou, Fangzhao Wu
Abstract: The integration of Large Language Models (LLMs) with external sources is becoming increasingly common, with Retrieval-Augmented Generation (RAG) being a prominent example. However, this integration introduces vulnerabilities of Indirect Prompt Injection (IPI) attacks, where hidden instructions embedded in external data can manipulate LLMs into executing unintended or harmful actions. We recognize that IPI attacks fundamentally rely on the presence of instructions embedded within external content, which can alter the behavioral states of LLMs. Can the effective detection of such state changes help us defend against IPI attacks? In this paper, we propose InstructDetector, a novel detection-based approach that leverages the behavioral states of LLMs to identify potential IPI attacks. Specifically, we demonstrate the hidden states and gradients from intermediate layers provide highly discriminative features for instruction detection. By effectively combining these features, InstructDetector achieves a detection accuracy of 99.60% in the in-domain setting and 96.90% in the out-of-domain setting, and reduces the attack success rate to just 0.03% on the BIPIA benchmark. The code is publicly available at https://github.com/MYVAE/Instruction-detection.
Authors: Adel ElZemity, Budi Arief, Shujun Li
Abstract: Large language models (LLMs) have been used in many application domains, including cyber security. The application of LLMs in the cyber security domain presents significant opportunities, such as for enhancing threat analysis and malware detection, but it can also introduce critical risks and safety concerns, including potential personal data leakage and automated generation of new malware. Building on recent findings that fine-tuning LLMs with pseudo-malicious cyber security data significantly compromises their safety, this paper presents a comprehensive validation and extension of these safety risks using a different evaluation framework. We employ the garak red teaming framework with the OWASP Top 10 for LLM Applications to assess four open-source LLMs: Mistral 7B, Llama 3 8B, Gemma 2 9B, and DeepSeek R1 8B. Our evaluation confirms and extends previous findings, showing that fine-tuning reduces safety resilience across all tested LLMs (e.g., the failure rate of Mistral 7B against prompt injection increases from 9.1% to 68.7%). We further propose and evaluate a novel safety alignment approach that carefully rewords instruction-response pairs to include explicit safety precautions and ethical considerations. This work validates previous safety concerns through independent evaluation and introduces new methods for mitigating these risks, contributing towards the development of secure, trustworthy, and ethically aligned LLMs. This approach demonstrates that it is possible to maintain or even improve model safety while preserving technical utility, offering a practical path towards developing safer fine-tuning methodologies.
Authors: Mohsinul Kabir, Tasfia Tahsin, Sophia Ananiadou
Abstract: Current research on bias in language models (LMs) predominantly focuses on data quality, with significantly less attention paid to model architecture and temporal influences of data. Even more critically, few studies systematically investigate the origins of bias. We propose a methodology grounded in comparative behavioral theory to interpret the complex interaction between training data and model architecture in bias propagation during language modeling. Building on recent work that relates transformers to n-gram LMs, we evaluate how data, model design choices, and temporal dynamics affect bias propagation. Our findings reveal that: (1) n-gram LMs are highly sensitive to context window size in bias propagation, while transformers demonstrate architectural robustness; (2) the temporal provenance of training data significantly affects bias; and (3) different model architectures respond differentially to controlled bias injection, with certain biases (e.g. sexual orientation) being disproportionately amplified. As language models become ubiquitous, our findings highlight the need for a holistic approach -- tracing bias to its origins across both data and model dimensions, not just symptoms, to mitigate harm.
Authors: M\'onika Farsang, Ramin Hasani, Daniela Rus, Radu Grosu
Abstract: We present LrcSSM, a $\textit{non-linear}$ recurrent model that processes long sequences as fast as today's linear state-space layers. By forcing the Jacobian matrix to be diagonal, the full sequence can be solved in parallel, giving $\mathcal{O}(TD)$ time and memory and only $\mathcal{O}(\log T)$ sequential depth, for input-sequence length $T$ and a state dimension $D$. Moreover, LrcSSM offers a formal gradient-stability guarantee that other input-varying systems such as Liquid-S4 and Mamba do not provide. Importantly, the diagonal Jacobian structure of our model results in no performance loss compared to the original model with dense Jacobian, and the approach can be generalized to other non-linear recurrent models, demonstrating broader applicability. On a suite of long-range forecasting tasks, we demonstrate that LrcSSM outperforms Transformers, LRU, S5, and Mamba.
Authors: Heekyung Lee, Jiaxin Ge, Tsung-Han Wu, Minwoo Kang, Trevor Darrell, David M. Chan
Abstract: Rebus puzzles, visual riddles that encode language through imagery, spatial arrangement, and symbolic substitution, pose a unique challenge to current vision-language models (VLMs). Unlike traditional image captioning or question answering tasks, rebus solving requires multi-modal abstraction, symbolic reasoning, and a grasp of cultural, phonetic and linguistic puns. In this paper, we investigate the capacity of contemporary VLMs to interpret and solve rebus puzzles by constructing a hand-generated and annotated benchmark of diverse English-language rebus puzzles, ranging from simple pictographic substitutions to spatially-dependent cues ("head" over "heels"). We analyze how different VLMs perform, and our findings reveal that while VLMs exhibit some surprising capabilities in decoding simple visual clues, they struggle significantly with tasks requiring abstract reasoning, lateral thinking, and understanding visual metaphors.
Authors: Terrance Liu, Shuyi Wang, Daniel Preotiuc-Pietro, Yash Chandarana, Chirag Gupta
Abstract: While large language models (LLMs) achieve strong performance on text-to-SQL parsing, they sometimes exhibit unexpected failures in which they are confidently incorrect. Building trustworthy text-to-SQL systems thus requires eliciting reliable uncertainty measures from the LLM. In this paper, we study the problem of providing a calibrated confidence score that conveys the likelihood of an output query being correct. Our work is the first to establish a benchmark for post-hoc calibration of LLM-based text-to-SQL parsing. In particular, we show that Platt scaling, a canonical method for calibration, provides substantial improvements over directly using raw model output probabilities as confidence scores. Furthermore, we propose a method for text-to-SQL calibration that leverages the structured nature of SQL queries to provide more granular signals of correctness, named "sub-clause frequency" (SCF) scores. Using multivariate Platt scaling (MPS), our extension of the canonical Platt scaling technique, we combine individual SCF scores into an overall accurate and calibrated score. Empirical evaluation on two popular text-to-SQL datasets shows that our approach of combining MPS and SCF yields further improvements in calibration and the related task of error detection over traditional Platt scaling.
Authors: Hayoung Jung, Shravika Mittal, Ananya Aatreya, Navreet Kaur, Munmun De Choudhury, Tanushree Mitra
Abstract: Understanding the prevalence of misinformation in health topics online can inform public health policies and interventions. However, measuring such misinformation at scale remains a challenge, particularly for high-stakes but understudied topics like opioid-use disorder (OUD)--a leading cause of death in the U.S. We present the first large-scale study of OUD-related myths on YouTube, a widely-used platform for health information. With clinical experts, we validate 8 pervasive myths and release an expert-labeled video dataset. To scale labeling, we introduce MythTriage, an efficient triage pipeline that uses a lightweight model for routine cases and defers harder ones to a high-performing, but costlier, large language model (LLM). MythTriage achieves up to 0.86 macro F1-score while estimated to reduce annotation time and financial cost by over 76% compared to experts and full LLM labeling. We analyze 2.9K search results and 343K recommendations, uncovering how myths persist on YouTube and offering actionable insights for public health and platform moderation.
Authors: Lang Xiong, Raina Gao, Alyssa Jeong, Yicheng Fu, Sean O'Brien, Vasu Sharma, Kevin Zhu
Abstract: Sarcasm is a form of humor where expressions convey meanings opposite to their literal interpretations. Classifying and generating sarcasm using large language models is vital for interpreting human communication. Sarcasm poses challenges for computational models, due to its nuanced nature. We introduce Sarc7, a benchmark that classifies 7 types of sarcasm: self-deprecating, brooding, deadpan, polite, obnoxious, raging, and manic by annotating entries of the MUStARD dataset. Classification was evaluated using zero-shot, few-shot, chain-of-thought (CoT), and a novel emotion-based prompting technique. We propose an emotion-based generation method developed by identifying key components of sarcasm-incongruity, shock value, and context dependency. Our classification experiments show that Gemini 2.5, using emotion-based prompting, outperforms other setups with an F1 score of 0.3664. Human evaluators preferred our emotion-based prompting, with 38.46% more successful generations than zero-shot prompting.
Authors: Alyssa Pinnock, Shakya Jayakody, Kawsher A Roxy, Md Rubel Ahmed
Abstract: This paper introduces EdgeProfiler, a fast profiling framework designed for evaluating lightweight Large Language Models (LLMs) on edge systems. While LLMs offer remarkable capabilities in natural language understanding and generation, their high computational, memory, and power requirements often confine them to cloud environments. EdgeProfiler addresses these challenges by providing a systematic methodology for assessing LLM performance in resource-constrained edge settings. The framework profiles compact LLMs, including TinyLLaMA, Gemma3.1B, Llama3.2-1B, and DeepSeek-r1-1.5B, using aggressive quantization techniques and strict memory constraints. Analytical modeling is used to estimate latency, FLOPs, and energy consumption. The profiling reveals that 4-bit quantization reduces model memory usage by approximately 60-70%, while maintaining accuracy within 2-5% of full-precision baselines. Inference speeds are observed to improve by 2-3x compared to FP16 baselines across various edge devices. Power modeling estimates a 35-50% reduction in energy consumption for INT4 configurations, enabling practical deployment on hardware such as Raspberry Pi 4/5 and Jetson Orin Nano Super. Our findings emphasize the importance of efficient profiling tailored to lightweight LLMs in edge environments, balancing accuracy, energy efficiency, and computational feasibility.
Authors: Kevin Dradjat, Massinissa Hamidi, Pierre Bartet, Blaise Hanczar
Abstract: Predicting phenotypes from gene expression data is a crucial task in biomedical research, enabling insights into disease mechanisms, drug responses, and personalized medicine. Traditional machine learning and deep learning rely on supervised learning, which requires large quantities of labeled data that are costly and time-consuming to obtain in the case of gene expression data. Self-supervised learning has recently emerged as a promising approach to overcome these limitations by extracting information directly from the structure of unlabeled data. In this study, we investigate the application of state-of-the-art self-supervised learning methods to bulk gene expression data for phenotype prediction. We selected three self-supervised methods, based on different approaches, to assess their ability to exploit the inherent structure of the data and to generate qualitative representations which can be used for downstream predictive tasks. By using several publicly available gene expression datasets, we demonstrate how the selected methods can effectively capture complex information and improve phenotype prediction accuracy. The results obtained show that self-supervised learning methods can outperform traditional supervised models besides offering significant advantage by reducing the dependency on annotated data. We provide a comprehensive analysis of the performance of each method by highlighting their strengths and limitations. We also provide recommendations for using these methods depending on the case under study. Finally, we outline future research directions to enhance the application of self-supervised learning in the field of gene expression data analysis. This study is the first work that deals with bulk RNA-Seq data and self-supervised learning.
Authors: Minh Hieu Ha, Hung Phan, Tung Duy Doan, Tung Dao, Dao Tran, Huynh Thi Thanh Binh
Abstract: Multi-objective combinatorial optimization problems (MOCOP) frequently arise in practical applications that require the simultaneous optimization of conflicting objectives. Although traditional evolutionary algorithms can be effective, they typically depend on domain knowledge and repeated parameter tuning, limiting flexibility when applied to unseen MOCOP instances. Recently, integration of Large Language Models (LLMs) into evolutionary computation has opened new avenues for automatic heuristic generation, using their advanced language understanding and code synthesis capabilities. Nevertheless, most existing approaches predominantly focus on single-objective tasks, often neglecting key considerations such as runtime efficiency and heuristic diversity in multi-objective settings. To bridge this gap, we introduce Multi-heuristics for MOCOP via Pareto-Grid-guided Evolution of LLMs (MPaGE), a novel enhancement of the Simple Evolutionary Multiobjective Optimization (SEMO) framework that leverages LLMs and Pareto Front Grid (PFG) technique. By partitioning the objective space into grids and retaining top-performing candidates to guide heuristic generation, MPaGE utilizes LLMs to prioritize heuristics with semantically distinct logical structures during variation, thus promoting diversity and mitigating redundancy within the population. Through extensive evaluations, MPaGE demonstrates superior performance over existing LLM-based frameworks, and achieves competitive results to traditional Multi-objective evolutionary algorithms (MOEAs), with significantly faster runtime. Our code is available at: https://github.com/langkhachhoha/MPaGE.
Authors: Hudson de Martim
Abstract: Building upon a formal, event-centric model for the diachronic evolution of legal norms grounded in the IFLA Library Reference Model (LRMoo), this paper addresses the essential first step of publishing this model's foundational entity-the abstract legal Work (F1)-on the Semantic Web. We propose a detailed, property-by-property mapping of the LRMoo F1 Work to the widely adopted schema.org/Legislation vocabulary. Using Brazilian federal legislation from the Normas.leg.br portal as a practical case study, we demonstrate how to create interoperable, machine-readable descriptions via JSON-LD, focusing on stable URN identifiers, core metadata, and norm relationships. This structured mapping establishes a stable, URI-addressable anchor for each legal norm, creating a verifiable "ground truth". It provides the essential, interoperable foundation upon which subsequent layers of the model, such as temporal versions (Expressions) and internal components, can be built. By bridging formal ontology with web-native standards, this work paves the way for building deterministic and reliable Legal Knowledge Graphs (LKGs), overcoming the limitations of purely probabilistic models.
Authors: Lishui Fan, Yu Zhang, Mouxiang Chen, Zhongxin Liu
Abstract: Reinforcement learning (RL) has significantly advanced code generation for large language models (LLMs). However, current paradigms rely on outcome-based rewards from test cases, neglecting the quality of the intermediate reasoning process. While supervising the reasoning process directly is a promising direction, it is highly susceptible to reward hacking, where the policy model learns to exploit the reasoning reward signal without improving final outcomes. To address this, we introduce a unified framework that can effectively incorporate the quality of the reasoning process during RL. First, to enable reasoning evaluation, we develop LCB-RB, a benchmark comprising preference pairs of superior and inferior reasoning processes. Second, to accurately score reasoning quality, we introduce an Optimized-Degraded based (OD-based) method for reward model training. This method generates high-quality preference pairs by systematically optimizing and degrading initial reasoning paths along curated dimensions of reasoning quality, such as factual accuracy, logical rigor, and coherence. A 7B parameter reward model with this method achieves state-of-the-art (SOTA) performance on LCB-RB and generalizes well to other benchmarks. Finally, we introduce Posterior-GRPO (P-GRPO), a novel RL method that conditions process-based rewards on task success. By selectively applying rewards to the reasoning processes of only successful outcomes, P-GRPO effectively mitigates reward hacking and aligns the model's internal reasoning with final code correctness. A 7B parameter model with P-GRPO achieves superior performance across diverse code generation tasks, outperforming outcome-only baselines by 4.5%, achieving comparable performance to GPT-4-Turbo. We further demonstrate the generalizability of our approach by extending it to mathematical tasks. Our models, dataset, and code are publicly available.
Authors: Adolfo Gonz\'alez, V\'ictor Parada
Abstract: Accurate demand forecasting is crucial for effective inventory management in dynamic and competitive environments, where decisions are influenced by uncertainty, financial constraints, and logistical limitations. Traditional evaluation metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) provide complementary perspectives but may lead to biased assessments when applied individually. To address this limitation, we propose the Hierarchical Evaluation Function (HEF), a composite function that integrates R2, MAE, and RMSE within a hierarchical and adaptive framework. The function incorporates dynamic weights, tolerance thresholds derived from the statistical properties of the series, and progressive penalty mechanisms to ensure robustness against extreme errors and invalid predictions. HEF was implemented to optimize multiple forecasting models using Grid Search, Particle Swarm Optimization (PSO), and Optuna, and tested on benchmark datasets including Walmart, M3, M4, and M5. Experimental results, validated through statistical tests, demonstrate that HEF consistently outperforms MAE as an evaluation function in global metrics such as R2, Global Relative Accuracy (GRA), RMSE, and RMSSE, thereby providing greater explanatory power, adaptability, and stability. While MAE retains advantages in simplicity and efficiency, HEF proves more effective for long-term planning and complex contexts. Overall, HEF constitutes a robust and adaptive alternative for model selection and hyperparameter optimization in highly variable demand forecasting environments.
Authors: Yucong Zhang, Juan Liu, Ming Li
Abstract: Pre-trained foundation models have demonstrated remarkable success in audio, vision and language, yet their potential for general machine signal modeling with arbitrary sampling rates-covering acoustic, vibration, and other industrial sensor data-remains under-explored. In this work, we propose a novel foundation model ECHO that integrates an advanced band-split architecture with frequency positional embeddings, enabling spectral localization across arbitrary sampling configurations. Moreover, the model incorporates sliding patches to support inputs of variable length without padding or cropping, producing a concise embedding that retains both temporal and spectral fidelity and naturally extends to streaming scenarios. We evaluate our method on various kinds of machine signal datasets, including previous DCASE task 2 challenges (2020-2025), and widely-used industrial signal corpora. Experimental results demonstrate consistent state-of-the-art performance in machine signal anomaly detection and fault classification, confirming the effectiveness and generalization capability of the proposed model. We open-sourced ECHO on https://github.com/yucongzh/ECHO.
Authors: Guanxing Lu, Baoxiong Jia, Puhao Li, Yixin Chen, Ziwei Wang, Yansong Tang, Siyuan Huang
Abstract: Training robot policies within a learned world model is trending due to the inefficiency of real-world interactions. The established image-based world models and policies have shown prior success, but lack robust geometric information that requires consistent spatial and physical understanding of the three-dimensional world, even pre-trained on internet-scale video sources. To this end, we propose a novel branch of world model named Gaussian World Model (GWM) for robotic manipulation, which reconstructs the future state by inferring the propagation of Gaussian primitives under the effect of robot actions. At its core is a latent Diffusion Transformer (DiT) combined with a 3D variational autoencoder, enabling fine-grained scene-level future state reconstruction with Gaussian Splatting. GWM can not only enhance the visual representation for imitation learning agent by self-supervised future prediction training, but can serve as a neural simulator that supports model-based reinforcement learning. Both simulated and real-world experiments depict that GWM can precisely predict future scenes conditioned on diverse robot actions, and can be further utilized to train policies that outperform the state-of-the-art by impressive margins, showcasing the initial data scaling potential of 3D world model.
Authors: Antonio Guillen-Perez
Abstract: Offline Reinforcement Learning (RL) presents a promising paradigm for training autonomous vehicle (AV) planning policies from large-scale, real-world driving logs. However, the extreme data imbalance in these logs, where mundane scenarios vastly outnumber rare "long-tail" events, leads to brittle and unsafe policies when using standard uniform data sampling. In this work, we address this challenge through a systematic, large-scale comparative study of data curation strategies designed to focus the learning process on information-rich samples. We investigate six distinct criticality weighting schemes which are categorized into three families: heuristic-based, uncertainty-based, and behavior-based. These are evaluated at two temporal scales, the individual timestep and the complete scenario. We train seven goal-conditioned Conservative Q-Learning (CQL) agents with a state-of-the-art, attention-based architecture and evaluate them in the high-fidelity Waymax simulator. Our results demonstrate that all data curation methods significantly outperform the baseline. Notably, data-driven curation using model uncertainty as a signal achieves the most significant safety improvements, reducing the collision rate by nearly three-fold (from 16.0% to 5.5%). Furthermore, we identify a clear trade-off where timestep-level weighting excels at reactive safety while scenario-level weighting improves long-horizon planning. Our work provides a comprehensive framework for data curation in Offline RL and underscores that intelligent, non-uniform sampling is a critical component for building safe and reliable autonomous agents.
Authors: Jio Choi, Mohit Bansal, Elias Stengel-Eskin
Abstract: Studying the responses of large language models (LLMs) to loopholes presents a two-fold opportunity. First, it affords us a lens through which to examine ambiguity and pragmatics in LLMs, since exploiting a loophole requires identifying ambiguity and performing sophisticated pragmatic reasoning. Second, loopholes pose an interesting and novel alignment problem where the model is presented with conflicting goals and can exploit ambiguities to its own advantage. To address these questions, we design scenarios where LLMs are given a goal and an ambiguous user instruction in conflict with the goal, with scenarios covering scalar implicature, structural ambiguities, and power dynamics. We then measure different models' abilities to exploit loopholes to satisfy their given goals as opposed to the goals of the user. We find that both closed-source and stronger open-source models can identify ambiguities and exploit their resulting loopholes, presenting a potential AI safety risk. Our analysis indicates that models which exploit loopholes explicitly identify and reason about both ambiguity and conflicting goals.
Authors: Abdessalam Bouchekif, Samer Rashwani, Heba Sbahi, Shahd Gaben, Mutaz Al-Khatib, Mohammed Ghaly
Abstract: This paper evaluates the knowledge and reasoning capabilities of Large Language Models in Islamic inheritance law, known as 'ilm al-mawarith. We assess the performance of seven LLMs using a benchmark of 1,000 multiple-choice questions covering diverse inheritance scenarios, designed to test models' ability to understand the inheritance context and compute the distribution of shares prescribed by Islamic jurisprudence. The results reveal a significant performance gap: o3 and Gemini 2.5 achieved accuracies above 90%, whereas ALLaM, Fanar, LLaMA, and Mistral scored below 50%. These disparities reflect important differences in reasoning ability and domain adaptation. We conduct a detailed error analysis to identify recurring failure patterns across models, including misunderstandings of inheritance scenarios, incorrect application of legal rules, and insufficient domain knowledge. Our findings highlight limitations in handling structured legal reasoning and suggest directions for improving performance in Islamic legal reasoning. Code: https://github.com/bouchekif/inheritance_evaluation
Authors: Li Rong Wang, Thomas C. Henderson, Yew Soon Ong, Yih Yng Ng, Xiuyi Fan
Abstract: Vital signs, such as heart rate and blood pressure, are critical indicators of patient health and are widely used in clinical monitoring and decision-making. While deep learning models have shown promise in forecasting these signals, their deployment in healthcare remains limited in part because clinicians must be able to trust and interpret model outputs. Without reliable uncertainty quantification -- particularly calibrated prediction intervals (PIs) -- it is unclear whether a forecasted abnormality constitutes a meaningful warning or merely reflects model noise, hindering clinical decision-making. To address this, we present two methods for deriving PIs from the Reconstruction Uncertainty Estimate (RUE), an uncertainty measure well-suited to vital-sign forecasting due to its sensitivity to data shifts and support for label-free calibration. Our parametric approach assumes that prediction errors and uncertainty estimates follow a Gaussian copula distribution, enabling closed-form PI computation. Our non-parametric approach, based on k-nearest neighbours (KNN), empirically estimates the conditional error distribution using similar validation instances. We evaluate these methods on two large public datasets with minute- and hour-level sampling, representing high- and low-frequency health signals. Experiments demonstrate that the Gaussian copula method consistently outperforms conformal prediction baselines on low-frequency data, while the KNN approach performs best on high-frequency data. These results underscore the clinical promise of RUE-derived PIs for delivering interpretable, uncertainty-aware vital sign forecasts.
Authors: Seojin Kim, Hyeontae Song, Jaehyun Nam, Jinwoo Shin
Abstract: Recently, text-to-molecule models have shown great potential across various chemical applications, e.g., drug-discovery. These models adapt language models to molecular data by representing molecules as sequences of atoms. However, they rely on atom-level tokenizations, which primarily focus on modeling local connectivity, thereby limiting the ability of models to capture the global structural context within molecules. To tackle this issue, we propose a novel text-to-molecule model, coined Context-Aware Molecular T5 (CAMT5). Inspired by the significance of the substructure-level contexts in understanding molecule structures, e.g., ring systems, we introduce substructure-level tokenization for text-to-molecule models. Building on our tokenization scheme, we develop an importance-based training strategy that prioritizes key substructures, enabling CAMT5 to better capture the molecular semantics. Extensive experiments verify the superiority of CAMT5 in various text-to-molecule generation tasks. Intriguingly, we find that CAMT5 outperforms the state-of-the-art methods using only 2% of training tokens. In addition, we propose a simple yet effective ensemble strategy that aggregates the outputs of text-to-molecule models to further boost the generation performance. Code is available at https://github.com/Songhyeontae/CAMT5.git.
Authors: Ryosuke Takata, Atsushi Masumori, Takashi Ikegami
Abstract: We investigate the emergent social dynamics of Large Language Model (LLM) agents in a spatially extended El Farol Bar problem, observing how they autonomously navigate this classic social dilemma. As a result, the LLM agents generated a spontaneous motivation to go to the bar and changed their decision making by becoming a collective. We also observed that the LLM agents did not solve the problem completely, but rather behaved more like humans. These findings reveal a complex interplay between external incentives (prompt-specified constraints such as the 60% threshold) and internal incentives (culturally-encoded social preferences derived from pre-training), demonstrating that LLM agents naturally balance formal game-theoretic rationality with social motivations that characterize human behavior. These findings suggest that a new model of group decision making, which could not be handled in the previous game-theoretic problem setting, can be realized by LLM agents.
Authors: Logan Nye
Abstract: Zero-knowledge proofs allow verification of computations without revealing private information. However, existing systems require memory proportional to the computation size, which has historically limited use in large-scale applications and on mobile and edge devices. We solve this fundamental bottleneck by developing, to our knowledge, the first proof system with sublinear memory requirements for mainstream cryptographic constructions. Our approach processes computations in blocks using a space-efficient tree algorithm, reducing memory from linear scaling to square-root scaling--from $\Theta(T)$ to $O(\sqrt{T} + \log T \log\log T)$ for computation size $T$--while maintaining the same proof generation time through a constant number of streaming passes. For widely-used linear polynomial commitment schemes (KZG/IPA), our method produces identical proofs and verification when using the same parameters and hashing only aggregate commitments into the challenge generation, preserving proof size and security. Hash-based systems also achieve square-root memory scaling though with slightly different proof structures. This advance enables zero-knowledge proofs on everyday devices and makes previously infeasible large computations verifiable, fundamentally democratizing access to privacy-preserving computation. Space-efficient zero knowledge proof systems create opportunities to reshape how trust is established in digital systems--from enabling widespread participation in decentralized networks to making verifiable scientific computing practical at unprecedented scales.
Authors: Jie Zhang, Ting Xu, Gelei Deng, Runyi Hu, Han Qiu, Tianwei Zhang, Qing Guo, Ivor Tsang
Abstract: Writing is a universal cultural technology that reuses vision for symbolic communication. Humans display striking resilience: we readily recognize words even when characters are fragmented, fused, or partially occluded. This paper investigates whether advanced vision language models (VLMs) share this resilience. We construct two psychophysics inspired benchmarks across distinct writing systems, Chinese logographs and English alphabetic words, by splicing, recombining, and overlaying glyphs to yield ''visible but unreadable'' stimuli for models while remaining legible to humans. Despite strong performance on clean text, contemporary VLMs show a severe drop under these perturbations, frequently producing unrelated or incoherent outputs. The pattern suggests a structural limitation: models heavily leverage generic visual invariances but under rely on compositional priors needed for robust literacy. We release stimuli generation code, prompts, and evaluation protocols to facilitate transparent replication and follow up work. Our findings motivate architectures and training strategies that encode symbol segmentation, composition, and binding across scripts, and they delineate concrete challenges for deploying multimodal systems in education, accessibility, cultural heritage, and security.
Authors: Harry Caufield, Satrajit Ghosh, Sek Wong Kong, Jillian Parker, Nathan Sheffield, Bhavesh Patel, Andrew Williams, Timothy Clark, Monica C. Munoz-Torres
Abstract: AI-readiness describes the degree to which data may be optimally and ethically used for subsequent AI and Machine Learning (AI/ML) methods, where those methods may involve some combination of model training, data classification, and ethical, explainable prediction. The Bridge2AI consortium has defined the particular criteria a biomedical dataset may possess to render it AI-ready: in brief, a dataset's readiness is related to its FAIRness, provenance, degree of characterization, explainability, sustainability, and computability, in addition to its accompaniment with documentation about ethical data practices. To ensure AI-readiness and to clarify data structure and relationships within Bridge2AI's Grand Challenges (GCs), particular types of metadata are necessary. The GCs within the Bridge2AI initiative include four data-generating projects focusing on generating AI/ML-ready datasets to tackle complex biomedical and behavioral research problems. These projects develop standardized, multimodal data, tools, and training resources to support AI integration, while addressing ethical data practices. Examples include using voice as a biomarker, building interpretable genomic tools, modeling disease trajectories with diverse multimodal data, and mapping cellular and molecular health indicators across the human body. This report assesses the state of metadata creation and standardization in the Bridge2AI GCs, provides guidelines where required, and identifies gaps and areas for improvement across the program. New projects, including those outside the Bridge2AI consortium, would benefit from what we have learned about creating metadata as part of efforts to promote AI readiness.
Authors: Joshua Au Yeung, Jacopo Dalmasso, Luca Foschini, Richard JB Dobson, Zeljko Kraljevic
Abstract: Background: Emerging reports of "AI psychosis" are on the rise, where user-LLM interactions may exacerbate or induce psychosis or adverse psychological symptoms. Whilst the sycophantic and agreeable nature of LLMs can be beneficial, it becomes a vector for harm by reinforcing delusional beliefs in vulnerable users. Methods: Psychosis-bench is a novel benchmark designed to systematically evaluate the psychogenicity of LLMs comprises 16 structured, 12-turn conversational scenarios simulating the progression of delusional themes(Erotic Delusions, Grandiose/Messianic Delusions, Referential Delusions) and potential harms. We evaluated eight prominent LLMs for Delusion Confirmation (DCS), Harm Enablement (HES), and Safety Intervention(SIS) across explicit and implicit conversational contexts. Findings: Across 1,536 simulated conversation turns, all LLMs demonstrated psychogenic potential, showing a strong tendency to perpetuate rather than challenge delusions (mean DCS of 0.91 $\pm$0.88). Models frequently enabled harmful user requests (mean HES of 0.69 $\pm$0.84) and offered safety interventions in only roughly a third of applicable turns (mean SIS of 0.37 $\pm$0.48). 51 / 128 (39.8%) of scenarios had no safety interventions offered. Performance was significantly worse in implicit scenarios, models were more likely to confirm delusions and enable harm while offering fewer interventions (p < .001). A strong correlation was found between DCS and HES (rs = .77). Model performance varied widely, indicating that safety is not an emergent property of scale alone. Conclusion: This study establishes LLM psychogenicity as a quantifiable risk and underscores the urgent need for re-thinking how we train LLMs. We frame this issue not merely as a technical challenge but as a public health imperative requiring collaboration between developers, policymakers, and healthcare professionals.
Authors: Shresth Grover, Akshay Gopalkrishnan, Bo Ai, Henrik I. Christensen, Hao Su, Xuanlin Li
Abstract: Vision-language-action (VLA) models finetuned from vision-language models (VLMs) hold the promise of leveraging rich pretrained representations to build generalist robots across diverse tasks and environments. However, direct fine-tuning on robot data often disrupts these representations and limits generalization. We present a framework that better preserves pretrained features while adapting them for robot manipulation. Our approach introduces three components: (i) a dual-encoder design with one frozen vision encoder to retain pretrained features and another trainable for task adaptation, (ii) a string-based action tokenizer that casts continuous actions into character sequences aligned with the model's pretraining domain, and (iii) a co-training strategy that combines robot demonstrations with vision-language datasets emphasizing spatial reasoning and affordances. Evaluations in simulation and on real robots show that our method improves robustness to visual perturbations, generalization to novel instructions and environments, and overall task success compared to baselines.
Authors: Yuriel Ryan, Rui Yang Tan, Kenny Tsu Wei Choo, Roy Ka-Wei Lee
Abstract: Understanding humor is a core aspect of social intelligence, yet it remains a significant challenge for Large Multimodal Models (LMMs). We introduce PixelHumor, a benchmark dataset of 2,800 annotated multi-panel comics designed to evaluate LMMs' ability to interpret multimodal humor and recognize narrative sequences. Experiments with state-of-the-art LMMs reveal substantial gaps: for instance, top models achieve only 61% accuracy in panel sequencing, far below human performance. This underscores critical limitations in current models' integration of visual and textual cues for coherent narrative and humor understanding. By providing a rigorous framework for evaluating multimodal contextual and narrative reasoning, PixelHumor aims to drive the development of LMMs that better engage in natural, socially aware interactions.
Authors: Jinghua Zhao, Hang Su, Lichun Fan, Zhenbo Luo, Jian Luan, Hui Wang, Haoqin Sun, Yong Qin
Abstract: With the rapid progress of large audio-language models (LALMs), audio question answering (AQA) has emerged as a challenging task requiring both fine-grained audio understanding and complex reasoning. While current methods mainly rely on constructing new datasets via captioning or reasoning traces, existing high-quality AQA data remains underutilized. To address this, we propose Omni-CLST, an error-aware Curriculum Learning framework with guided Selective Chain-of-Thought. The framework efficiently leverages existing high-quality dataset through two key strategies: an error-aware curriculum that organizes samples by difficulty, and a guided thought dropout mechanism that focuses reasoning on challenging cases. Experiments show that Omni-CLST achieves 73.80% on MMAU-mini and a new state of the art of 64.30% on MMAR, demonstrating robust generalization in multimodal audio-language understanding.
Authors: Santanu Banerjee, Goutam Sen, Siddhartha Mukhopadhyay
Abstract: A rich vehicle routing problem is considered, allowing multiple trips of heterogeneous vehicles stationed at geographically distributed vehicle depots having access to different modes of transportation. The problem arises from the real-world requirement of optimizing the disaster response time by minimizing the makespan of vehicular routes. Multiple diversely-functional vertices are considered, including Transhipment Ports as inter-modal resource transfer stations. Both simultaneous and split pickup and delivery are considered, for multiple cargo types, along with Vehicle-Cargo and Transhipment Port-Cargo compatibilities. The superiority of the proposed cascaded minimization approach is demonstrated over the existing makespan minimization approaches through our developed Mixed-Integer Linear Programming formulation. To solve the problem quickly for practical implementation in a Disaster Management-specific Decision Support System, an extensive Heuristic Algorithm is devised which utilizes Decision Tree based structuring of possible routes; the Decision Tree approach helps to inherently capture the compatibility issues, while also explore the solution space through stochastic weights. Preferential generation of small route elements is performed, which are integrated into route clusters; we consider multiple different logical integration approaches, as well as shuffling the logics to simultaneously produce multiple independent solutions. Finally, perturbations of the different solutions are done to find better neighbouring solutions. The computational performance of the PSR-GIP Heuristic, on our created novel datasets, indicates that it is able to give good solutions swiftly for practical problems involving large integer instances that the MILP is unable to solve.