Authors: Abhishek Sharma
Abstract: Recent progress in Learning by Reading and Machine Reading systems has significantly increased the capacity of knowledge-based systems to learn new facts. In this work, we discuss the problem of selecting a set of learning requests for these knowledge-based systems which would lead to maximum Q/A performance. To understand the dynamics of this problem, we simulate the properties of a learning strategy, which sends learning requests to an external knowledge source. We show that choosing an optimal set of facts for these learning systems is similar to a coordination game, and use reinforcement learning to solve this problem. Experiments show that such an approach can significantly improve Q/A performance.
Authors: Elija Perrier, Michael Timothy Bennett
Abstract: Language Model Agents (LMAs) are increasingly treated as capable of autonomously navigating interactions with humans and tools. Their design and deployment tends to presume they are normal agents capable of sustaining coherent goals, adapting across contexts and acting with a measure of intentionality. These assumptions are critical to prospective use cases in industrial, social and governmental settings. But LMAs are not normal agents. They inherit the structural problems of the large language models (LLMs) around which they are built: hallucinations, jailbreaking, misalignment and unpredictability. In this Position paper we argue LMAs should not be treated as normal agents, because doing so leads to problems that undermine their utility and trustworthiness. We enumerate pathologies of agency intrinsic to LMAs. Despite scaffolding such as external memory and tools, they remain ontologically stateless, stochastic, semantically sensitive, and linguistically intermediated. These pathologies destabilise the ontological properties of LMAs including identifiability, continuity, persistence and and consistency, problematising their claim to agency. In response, we argue LMA ontological properties should be measured before, during and after deployment so that the negative effects of pathologies can be mitigated.
Authors: Libo Wang
Abstract: To reduce the cost and consumption of computing resources caused by computational redundancy and delayed reward assignment in long CoT, this research proposes the dynamic chain-of-thought with adaptive reasoning time and steps. The researcher used simulation experiment to simulate the integration of D-CoT through Python 3.13 IDLE combined with a Python simulator based on GPTs. At the same time, the researcher used DeepSeek R1 as a control group to test and compare the performance of the D-CoT simulator in processing MIT OpenCourseWare's linear algebra exam questions. Experimental results show that D-CoT is better than DeepSeek R1 based on long CoT in three indicators: reasoning time, CoT length (reasoning steps) and token count, which achieves a significant reduction in computing resource consumption. In addition, this research has potential value in deep reasoning optimization and can be used as a reference for future dynamic deep reasoning frameworks.
Authors: Parashar Das
Abstract: There is a general concern that present developments in artificial intelligence (AI) research will lead to sentient AI systems, and these may pose an existential threat to humanity. But why cannot sentient AI systems benefit humanity instead? This paper endeavours to put this question in a tractable manner. I ask whether a putative AI system will develop an altruistic or a malicious disposition towards our society, or what would be the nature of its agency? Given that AI systems are being developed into formidable problem solvers, we can reasonably expect these systems to preferentially take on conscious aspects of human problem solving. I identify the relevant phenomenal aspects of agency in human problem solving. The functional aspects of conscious agency can be monitored using tools provided by functionalist theories of consciousness. A recent expert report (Butlin et al. 2023) has identified functionalist indicators of agency based on these theories. I show how to use the Integrated Information Theory (IIT) of consciousness, to monitor the phenomenal nature of this agency. If we are able to monitor the agency of AI systems as they develop, then we can dissuade them from becoming a menace to society while encouraging them to be an aid.
Authors: Maarten Buyl, Hadi Khalaf, Claudio Mayrink Verdun, Lucas Monteiro Paes, Caio C. Vieira Machado, Flavio du Pin Calmon
Abstract: In AI alignment, extensive latitude must be granted to annotators, either human or algorithmic, to judge which model outputs are `better' or `safer.' We refer to this latitude as alignment discretion. Such discretion remains largely unexamined, posing two risks: (i) annotators may use their power of discretion arbitrarily, and (ii) models may fail to mimic this discretion. To study this phenomenon, we draw on legal concepts of discretion that structure how decision-making authority is conferred and exercised, particularly in cases where principles conflict or their application is unclear or irrelevant. Extended to AI alignment, discretion is required when alignment principles and rules are (inevitably) conflicting or indecisive. We present a set of metrics to systematically analyze when and how discretion in AI alignment is exercised, such that both risks (i) and (ii) can be observed. Moreover, we distinguish between human and algorithmic discretion and analyze the discrepancy between them. By measuring both human and algorithmic discretion over safety alignment datasets, we reveal layers of discretion in the alignment process that were previously unaccounted for. Furthermore, we demonstrate how algorithms trained on these datasets develop their own forms of discretion in interpreting and applying these principles, which challenges the purpose of having any principles at all. Our paper presents the first step towards formalizing this core gap in current alignment processes, and we call on the community to further scrutinize and control alignment discretion.
Authors: Dan Elbaz, Oren Salzman
Abstract: Offline Reinforcement Learning (RL) algorithms learn a policy using a fixed training dataset, which is then deployed online to interact with the environment and make decisions. Transformers, a standard choice for modeling time-series data, are gaining popularity in offline RL. In this context, Beam Search (BS), an approximate inference algorithm, is the go-to decoding method. Offline RL eliminates the need for costly or risky online data collection. However, the restricted dataset induces uncertainty as the agent may encounter unfamiliar sequences of states and actions during execution that were not covered in the training data. In this context, BS lacks two important properties essential for offline RL: It does not account for the aforementioned uncertainty, and its greedy left-right search approach often results in sequences with minimal variations, failing to explore potentially better alternatives. To address these limitations, we propose Portfolio Beam Search (PBS), a simple-yet-effective alternative to BS that balances exploration and exploitation within a Transformer model during decoding. We draw inspiration from financial economics and apply these principles to develop an uncertainty-aware diversification mechanism, which we integrate into a sequential decoding algorithm at inference time. We empirically demonstrate the effectiveness of PBS on the D4RL locomotion benchmark, where it achieves higher returns and significantly reduces outcome variability.
Authors: Pulkit Rustagi, Yashwanthi Anand, Sandhya Saisubramanian
Abstract: Autonomous agents are often required to plan under multiple objectives whose preference ordering varies based on context. The agent may encounter multiple contexts during its course of operation, each imposing a distinct lexicographic ordering over the objectives, with potentially different reward functions associated with each context. Existing approaches to multi-objective planning typically consider a single preference ordering over the objectives, across the state space, and do not support planning under multiple objective orderings within an environment. We present Contextual Lexicographic Markov Decision Process (CLMDP), a framework that enables planning under varying lexicographic objective orderings, depending on the context. In a CLMDP, both the objective ordering at a state and the associated reward functions are determined by the context. We employ a Bayesian approach to infer a state-context mapping from expert trajectories. Our algorithm to solve a CLMDP first computes a policy for each objective ordering and then combines them into a single context-aware policy that is valid and cycle-free. The effectiveness of the proposed approach is evaluated in simulation and using a mobile robot.
Authors: Kumar Manas, Adrian Paschke
Abstract: This comprehensive survey examines the integration of knowledge-based approaches into autonomous driving systems, with a focus on trajectory prediction and planning. We systematically review methodologies for incorporating domain knowledge, traffic rules, and commonsense reasoning into these systems, spanning purely symbolic representations to hybrid neuro-symbolic architectures. In particular, we analyze recent advancements in formal logic and differential logic programming, reinforcement learning frameworks, and emerging techniques that leverage large foundation models and diffusion models for knowledge representation. Organized under a unified literature survey section, our discussion synthesizes the state-of-the-art into a high-level overview, supported by a detailed comparative table that maps key works to their respective methodological categories. This survey not only highlights current trends -- including the growing emphasis on interpretable AI, formal verification in safety-critical systems, and the increased use of generative models in prediction and planning -- but also outlines the challenges and opportunities for developing robust, knowledge-enhanced autonomous driving systems.
Authors: Andrew Kiruluta, Andreas Lemos, Priscilla Burity
Abstract: We propose a novel reinforcement learning framework for post training large language models that does not rely on human in the loop feedback. Instead, our approach uses cross attention signals within the model itself to derive a self supervised reward, thereby guiding iterative fine tuning of the model policy. By analyzing how the model attends to the input prompt during generation, we construct measures of prompt coverage, focus, and coherence. We then use these measures to rank or score candidate responses, providing a reward signal that encourages the model to produce well aligned, on topic text. In empirical comparisons against standard policy gradient methods and RL fine tuning with synthetic preference models, our method shows significant gains in prompt relevance and consistency over a non RL baseline. While it does not yet match the performance of fully human supervised RLHF systems, it highlights an important direction for scaling alignment with minimal human labeling. We provide a detailed analysis, discuss potential limitations, and outline future work for combining cross-attention based signals with smaller amounts of human feedback.
Authors: Shima Khoshraftar, Niaz Abedini, Amir Hajian
Abstract: The application of large language models (LLMs) to graph data has attracted a lot of attention recently. LLMs allow us to use deep contextual embeddings from pretrained models in text-attributed graphs, where shallow embeddings are often used for the text at- tributes of nodes. However, it is still challenging to efficiently en- code the graph structure and features into a sequential form for use by LLMs. In addition, the performance of an LLM alone, is highly dependent on the structure of the input prompt, which limits their effectiveness as a reliable approach and often requires iterative man- ual adjustments that could be slow, tedious and difficult to replicate programmatically. In this paper, we propose GraphiT (Graphs in Text), a framework for encoding graphs into a textual format and optimizing LLM prompts for graph prediction tasks. Here we focus on node classification for text-attributed graphs. We encode the graph data for every node and its neighborhood into a concise text to enable LLMs to better utilize the information in the graph. We then further programmatically optimize the LLM prompts us- ing the DSPy framework to automate this step and make it more efficient and reproducible. GraphiT outperforms our LLM-based baselines on three datasets and we show how the optimization step in GraphiT leads to measurably better results without manual prompt tweaking. We also demonstrated that our graph encoding approach is competitive to other graph encoding methods while being less expensive because it uses significantly less tokens for the same task.
Authors: Kiwon Song, James M. Jennings III, Clintin P. Davis-Stober
Abstract: Fundamental choice axioms, such as transitivity of preference, provide testable conditions for determining whether human decision making is rational, i.e., consistent with a utility representation. Recent work has demonstrated that AI systems trained on human data can exhibit similar reasoning biases as humans and that AI can, in turn, bias human judgments through AI recommendation systems. We evaluate the rationality of AI responses via a series of choice experiments designed to evaluate transitivity of preference in humans. We considered ten versions of Meta's Llama 2 and 3 LLM models. We applied Bayesian model selection to evaluate whether these AI-generated choices violated two prominent models of transitivity. We found that the Llama 2 and 3 models generally satisfied transitivity, but when violations did occur, occurred only in the Chat/Instruct versions of the LLMs. We argue that rationality axioms, such as transitivity of preference, can be useful for evaluating and benchmarking the quality of AI-generated responses and provide a foundation for understanding computational rationality in AI systems more generally.
Authors: Salom\'e Lepers, Vincent Thomas, Olivier Buffet
Abstract: In this article, we are interested in planning problems where the agent is aware of the presence of an observer, and where this observer is in a partial observability situation. The agent has to choose its strategy so as to optimize the information transmitted by observations. Building on observer-aware Markov decision processes (OAMDPs), we propose a framework to handle this type of problems and thus formalize properties such as legibility, explicability and predictability. This extension of OAMDPs to partial observability can not only handle more realistic problems, but also permits considering dynamic hidden variables of interest. These dynamic target variables allow, for instance, working with predictability, or with legibility problems where the goal might change during execution. We discuss theoretical properties of PO-OAMDPs and, experimenting with benchmark problems, we analyze HSVI's convergence behavior with dedicated initializations and study the resulting strategies.
Authors: Xueshen Li, Xinlong Hou, Ziyi Huang, Yu Gan
Abstract: Recent advancements in large language models (LLMs) have demonstrated extraordinary comprehension capabilities with remarkable breakthroughs on various vision-language tasks. However, the application of LLMs in generating reliable medical diagnostic reports remains in the early stages. Currently, medical LLMs typically feature a passive interaction model where doctors respond to patient queries with little or no involvement in analyzing medical images. In contrast, some ChatBots simply respond to predefined queries based on visual inputs, lacking interactive dialogue or consideration of medical history. As such, there is a gap between LLM-generated patient-ChatBot interactions and those occurring in actual patient-doctor consultations. To bridge this gap, we develop an LLM-based dialogue system, namely proactive multi-round vision-language interactions for computer-aided diagnosis (ProMRVL-CAD), to generate patient-friendly disease diagnostic reports. The proposed ProMRVL-CAD system allows proactive dialogue to provide patients with constant and reliable medical access via an integration of knowledge graph into a recommendation system. Specifically, we devise two generators: a Proactive Question Generator (Pro-Q Gen) to generate proactive questions that guide the diagnostic procedure and a Multi-Vision Patient-Text Diagnostic Report Generator (MVP-DR Gen) to produce high-quality diagnostic reports. Evaluating two real-world publicly available datasets, MIMIC-CXR and IU-Xray, our model has better quality in generating medical reports. We further demonstrate the performance of ProMRVL achieves robust under the scenarios with low image quality. Moreover, we have created a synthetic medical dialogue dataset that simulates proactive diagnostic interactions between patients and doctors, serving as a valuable resource for training LLM.
Authors: Hamed Rahimi, Adil Bahaj, Mouad Abrini, Mahdi Khoramshahi, Mounir Ghogho, Mohamed Chetouani
Abstract: The integration of vision-language models into robotic systems constitutes a significant advancement in enabling machines to interact with their surroundings in a more intuitive manner. While VLMs offer rich multimodal reasoning, existing approaches lack user-specific adaptability, often relying on generic interaction paradigms that fail to account for individual behavioral, contextual, or socio-emotional nuances. When customization is attempted, ethical concerns arise from unmitigated biases in user data, risking exclusion or unfair treatment. To address these dual challenges, we propose User-VLM 360{\deg}, a holistic framework integrating multimodal user modeling with bias-aware optimization. Our approach features: (1) user-aware tuning that adapts interactions in real time using visual-linguistic signals; (2) bias mitigation via preference optimization; and (3) curated 360{\deg} socio-emotive interaction datasets annotated with demographic, emotion, and relational metadata. Evaluations across eight benchmarks demonstrate state-of-the-art results: +35.3% F1 in personalized VQA, +47.5% F1 in facial features understanding, 15% bias reduction, and 30X speedup over baselines. Ablation studies confirm component efficacy, and deployment on the Pepper robot validates real-time adaptability across diverse users. We open-source parameter-efficient 3B/10B models and an ethical verification framework for responsible adaptation.
Authors: Hamed Rahimi, Mouad Abrini, Mahdi Khoramshahi, Mohamed Chetouani
Abstract: This paper investigates the performance of multimodal pre-trained models in user profiling tasks based on visual-linguistic demographic data. These models are critical for adapting to the needs and preferences of human users in social robotics, thereby providing personalized responses and enhancing interaction quality. First, we introduce two datasets specifically curated to represent demographic characteristics derived from user facial images. Next, we evaluate the performance of a prominent contrastive multimodal pre-trained model, CLIP, on these datasets, both in its out-of-the-box state and after fine-tuning. Initial results indicate that CLIP performs suboptimal in matching images to demographic descriptions without fine-tuning. Although fine-tuning significantly enhances its predictive capacity, the model continues to exhibit limitations in effectively generalizing subtle demographic nuances. To address this, we propose adopting a masked image modeling strategy to improve generalization and better capture subtle demographic attributes. This approach offers a pathway for enhancing demographic sensitivity in multimodal user modeling tasks.
Authors: Quanmin Wei, Penglin Dai, Wei Li, Bingyi Liu, Xiao Wu
Abstract: Multi-agent collaborative perception is expected to significantly improve perception performance by overcoming the limitations of single-agent perception through exchanging complementary information. However, training a robust collaborative perception model requires collecting sufficient training data that covers all possible collaboration scenarios, which is impractical due to intolerable deployment costs. Hence, the trained model is not robust against new traffic scenarios with inconsistent data distribution and fundamentally restricts its real-world applicability. Further, existing methods, such as domain adaptation, have mitigated this issue by exposing the deployment data during the training stage but incur a high training cost, which is infeasible for resource-constrained agents. In this paper, we propose a Parameter-Efficient Fine-Tuning-based lightweight framework, CoPEFT, for fast adapting a trained collaborative perception model to new deployment environments under low-cost conditions. CoPEFT develops a Collaboration Adapter and Agent Prompt to perform macro-level and micro-level adaptations separately. Specifically, the Collaboration Adapter utilizes the inherent knowledge from training data and limited deployment data to adapt the feature map to new data distribution. The Agent Prompt further enhances the Collaboration Adapter by inserting fine-grained contextual information about the environment. Extensive experiments demonstrate that our CoPEFT surpasses existing methods with less than 1\% trainable parameters, proving the effectiveness and efficiency of our proposed method.
Authors: Dezhi Luo, Yijiang Li, Hokin Deng
Abstract: Despite excelling in high-level reasoning, current language models lack robustness in real-world scenarios and perform poorly on fundamental problem-solving tasks that are intuitive to humans. This paper argues that both challenges stem from a core discrepancy between human and machine cognitive development. While both systems rely on increasing representational power, the absence of core knowledge-foundational cognitive structures in humans-prevents language models from developing robust, generalizable abilities, where complex skills are grounded in simpler ones within their respective domains. It explores empirical evidence of core knowledge in humans, analyzes why language models fail to acquire it, and argues that this limitation is not an inherent architectural constraint. Finally, it outlines a workable proposal for systematically integrating core knowledge into future multi-modal language models through the large-scale generation of synthetic training data using a cognitive prototyping strategy.
Authors: Zongqian Wu, Tianyu Li, Jiaying Yang, Mengmeng Zhan, Xiaofeng Zhu, Lei Feng
Abstract: Deep iterative chain-of-thought (CoT) reasoning enables LLMs to tackle complex tasks by progressively activating relevant pre-trained knowledge. However, it faces challenges in ensuring continual improvement and determining a stopping criterion. In this paper, we investigate whether the relevant knowledge that contributes directly to solving the given question can be activated from the initial reasoning path, thus circumventing the need for iterative refinement. Our experiments reveal that increasing the diversity of initial reasoning paths can achieve comparable or superior performance, a concept we term \textit{breadth reasoning}. However, existing breadth reasoning approaches, such as self-consistency, offer limited diversity. To address this limitation, we propose a simple yet effective method that enhances reasoning breadth by integrating contextual exploration with reduced sampling randomness. Extensive experiments demonstrate that our approach significantly outperforms deep iterative reasoning. Our code is provided in https://github.com/zongqianwu/breadth.
Authors: Jun Wang
Abstract: OpenAI o1 has shown that applying reinforcement learning to integrate reasoning steps directly during inference can significantly improve a model's reasoning capabilities. This result is exciting as the field transitions from the conventional autoregressive method of generating answers to a more deliberate approach that models the slow-thinking process through step-by-step reasoning training. Reinforcement learning plays a key role in both the model's training and decoding processes. In this article, we present a comprehensive formulation of reasoning problems and investigate the use of both model-based and model-free approaches to better support this slow-thinking framework.
Authors: In-Chang Baek, Sung-Hyun Kim, Sam Earle, Zehua Jiang, Noh Jin-Ha, Julian Togelius, Kyung-Joong Kim
Abstract: Reward design plays a pivotal role in the training of game AIs, requiring substantial domain-specific knowledge and human effort. In recent years, several studies have explored reward generation for training game agents and controlling robots using large language models (LLMs). In the content generation literature, there has been early work on generating reward functions for reinforcement learning agent generators. This work introduces PCGRLLM, an extended architecture based on earlier work, which employs a feedback mechanism and several reasoning-based prompt engineering techniques. We evaluate the proposed method on a story-to-reward generation task in a two-dimensional environment using two state-of-the-art LLMs, demonstrating the generalizability of our approach. Our experiments provide insightful evaluations that demonstrate the capabilities of LLMs essential for content generation tasks. The results highlight significant performance improvements of 415% and 40% respectively, depending on the zero-shot capabilities of the language model. Our work demonstrates the potential to reduce human dependency in game AI development, while supporting and enhancing creative processes.
Authors: Meet Udeshi, Minghao Shao, Haoran Xi, Nanda Rani, Kimberly Milner, Venkata Sai Charan Putrevu, Brendan Dolan-Gavitt, Sandeep Kumar Shukla, Prashanth Krishnamurthy, Farshad Khorrami, Ramesh Karri, Muhammad Shafique
Abstract: Large Language Models (LLMs) have been used in cybersecurity in many ways, including their recent use as intelligent agent systems for autonomous security analysis. Capture the Flag (CTF) challenges serve as benchmarks for assessing the automated task-planning abilities of LLM agents across various cybersecurity skill sets. Early attempts to apply LLMs for solving CTF challenges relied on single-agent systems, where feedback was restricted to a single reasoning-action loop. This approach proved inadequate for handling complex CTF tasks. Drawing inspiration from real-world CTF competitions, where teams of experts collaborate, we introduce the D-CIPHER multi-agent LLM framework for collaborative CTF challenge solving. D-CIPHER integrates agents with distinct roles, enabling dynamic feedback loops to enhance reasoning on CTF challenges. It introduces the Planner-Executor agent system, consisting of a Planner agent for overall problem-solving along with multiple heterogeneous Executor agents for individual tasks, facilitating efficient allocation of responsibilities among the LLMs. Additionally, D-CIPHER incorporates an Auto-prompter agent, which improves problem-solving by exploring the challenge environment and generating a highly relevant initial prompt. We evaluate D-CIPHER on CTF benchmarks using multiple LLM models and conduct comprehensive studies to highlight the impact of our enhancements. Our results demonstrate that the multi-agent D-CIPHER system achieves a significant improvement in challenges solved, setting a state-of-the-art performance on three benchmarks: 22.0% on NYU CTF Bench, 22.5% on Cybench, and 44.0% on HackTheBox. D-CIPHER is available at https://github.com/NYU-LLM-CTF/nyuctf_agents as the nyuctf_multiagent package.
Authors: Chengshuai Zhao, Zhen Tan, Chau-Wai Wong, Xinyan Zhao, Tianlong Chen, Huan Liu
Abstract: Content analysis breaks down complex and unstructured texts into theory-informed numerical categories. Particularly, in social science, this process usually relies on multiple rounds of manual annotation, domain expert discussion, and rule-based refinement. In this paper, we introduce SCALE, a novel multi-agent framework that effectively $\underline{\textbf{S}}$imulates $\underline{\textbf{C}}$ontent $\underline{\textbf{A}}$nalysis via $\underline{\textbf{L}}$arge language model (LLM) ag$\underline{\textbf{E}}$nts. SCALE imitates key phases of content analysis, including text coding, collaborative discussion, and dynamic codebook evolution, capturing the reflective depth and adaptive discussions of human researchers. Furthermore, by integrating diverse modes of human intervention, SCALE is augmented with expert input to further enhance its performance. Extensive evaluations on real-world datasets demonstrate that SCALE achieves human-approximated performance across various complex content analysis tasks, offering an innovative potential for future social science research.
Authors: Zi Wang, Shiwei Weng, Mohannad Alhanahnah, Somesh Jha, Tom Reps
Abstract: Large Language Models (LLMs) have exhibited remarkable capabilities across diverse domains, prompting investigations into their potential as generic reasoning engines. While recent studies have explored inference-time computation to enhance model performance on complex problems, current research lacks a formal framework to characterize the complexity of reasoning tasks. This study introduces the Predicate-Enumeration-Aggregation (PEA) framework, a formal approach to describe and solve a class of important reasoning tasks termed computational reasoning problems. The PEA framework decomposes these problems into predicate and enumeration components, using LLMs to synthesize programs based on specified predicates, enumeration, and aggregation rules. These synthesized programs are then executed to obtain solutions to the computational tasks. We demonstrate the framework's efficacy on benchmark tasks including Boolean satisfiability problems, game of $24$, and planning problems. Empirical evaluation reveals that PEA substantially enhances the performance of underlying models on benchmark computational problems, yielding an average accuracy improvement of approximately $50\%$, coupled with increased efficiency.
Authors: Antoine Dolant, Praveen Kumar
Abstract: Effective decision-making in complex systems requires synthesizing diverse perspectives to address multifaceted challenges under uncertainty. This study introduces a real-world inspired agentic Large Language Models (LLMs) framework, to simulate and enhance decision discourse-the deliberative process through which actionable strategies are collaboratively developed. Unlike traditional decision-support tools, the framework emphasizes dialogue, trade-off exploration, and the emergent synergies generated by interactions among agents embodying distinct personas. These personas simulate diverse stakeholder roles, each bringing unique priorities, expertise, and value-driven reasoning to the table. The framework incorporates adaptive and self-governing mechanisms, enabling agents to dynamically summon additional expertise and refine their assembly to address evolving challenges. An illustrative hypothetical example focused on extreme flooding in a Midwestern township demonstrates the framework's ability to navigate uncertainty, balance competing priorities, and propose mitigation and adaptation strategies by considering social, economic, and environmental dimensions. Results reveal how the breadth-first exploration of alternatives fosters robust and equitable recommendation pathways. This framework transforms how decisions are approached in high-stakes scenarios and can be incorporated in digital environments. It not only augments decision-makers' capacity to tackle complexity but also sets a foundation for scalable and context-aware AI-driven recommendations. This research explores novel and alternate routes leveraging agentic LLMs for adaptive, collaborative, and equitable recommendation processes, with implications across domains where uncertainty and complexity converge.
Authors: Robert Dahlke, Henrik Klagges, Dan Zecha, Benjamin Merkel, Sven Rohr, Fabian Klemm
Abstract: We present the Mixture-of-Tunable-Experts (MoTE), a method that extends the Mixture-of-Experts architecture of Large Language Models (LLMs). Without additional training, MoTE enables meaningful and focused behavior changes in LLMs on-the-fly during inference time. By analyzing the digital LLM brain of DeepSeek-R1 using a technique we dub 'functional Token Resonance Imaging' (fTRI) - inspired by fMRI and using prompts designed to elicit specific behavior (e.g., 'What happened {time}{place}?') - we empirically identify distinctive experts associated with behaviors like refusal responses. Using MoTE we are able to intervene and control such specific behavior. We switched off the top 10 most refusal-relevant experts (0.07% of R1's 14,848 routed experts), achieving a 52% refusal reduction on sensitive reference prompts without performance degradation on MT-Bench. Random expert deactivation resulted in smaller behavioral shifts with increased noise, whereas forced expert activation led to significantly higher refusal rates. Our approach shares similarities with sparse autoencoders (SAEs) in terms of explainability and steerability. Unlike SAEs, MoTE does not require large training efforts, as within MoEs with a vast number of experts, specialization already emerged naturally during pretraining. Our findings suggest that significant functional mechanisms in Mixture-of-Experts architectures can at least partially be localized in a small number of specific experts, rather than being distributed throughout the model's weights. Expert subgroups can be tuned to trigger significant behavior variations, providing insights into the inner workings of LLMs.
Authors: Zhao Wang, Sota Moriyama, Wei-Yao Wang, Briti Gangopadhyay, Shingo Takamatsu
Abstract: Recent advancements in LLM-based multi-agent (LLM-MA) systems have shown promise, yet significant challenges remain in managing communication and refinement when agents collaborate on complex tasks. In this paper, we propose \textit{Talk Structurally, Act Hierarchically (TalkHier)}, a novel framework that introduces a structured communication protocol for context-rich exchanges and a hierarchical refinement system to address issues such as incorrect outputs, falsehoods, and biases. \textit{TalkHier} surpasses various types of SoTA, including inference scaling model (OpenAI-o1), open-source multi-agent models (e.g., AgentVerse), and majority voting strategies on current LLM and single-agent baselines (e.g., ReAct, GPT4o), across diverse tasks, including open-domain question answering, domain-specific selective questioning, and practical advertisement text generation. These results highlight its potential to set a new standard for LLM-MA systems, paving the way for more effective, adaptable, and collaborative multi-agent frameworks. The code is available https://github.com/sony/talkhier.
Authors: Hongliang Lu, Zhonglin Xie, Yaoyu Wu, Can Ren, Yuxuan Chen, Zaiwen Wen
Abstract: Despite the rapid development of large language models (LLMs), a fundamental challenge persists: the lack of high-quality optimization modeling datasets hampers LLMs' robust modeling of practical optimization problems from natural language descriptions (NL). This data scarcity also contributes to the generalization difficulties experienced by learning-based methods. To address these challenges, we propose a scalable framework for synthesizing a high-quality dataset, named OptMATH. Starting from curated seed data with mathematical formulations (MF), this framework automatically generates problem data (PD) with controllable complexity. Then, a back-translation step is employed to obtain NL. To verify the correspondence between the NL and the PD, a forward modeling step followed by rejection sampling is used. The accepted pairs constitute the training part of OptMATH. Then a collection of rejected pairs is identified and further filtered. This collection serves as a new benchmark for optimization modeling, containing difficult instances whose lengths are much longer than these of NL4OPT and MAMO. Through extensive experiments, we demonstrate that models of various sizes (0.5B-32B parameters) trained on OptMATH achieve superior results on multiple modeling benchmarks, thereby validating the effectiveness and scalability of our approach.
Authors: Zongyuan Li, Chang Lu, Xiaojie Xu, Runnan Qi, Yanan Ni, Lumin Jiang, Xiangbei Liu, Xuebo Zhang, Yongchun Fang, Kuihua Huang, Xian Guo
Abstract: Since the emergence of the Large Language Model (LLM), LLM has been widely used in fields such as writing, translating, and searching. However, there is still great potential for LLM-based methods in handling complex tasks such as decision-making in the StarCraft II environment. To address problems such as lack of relevant knowledge and poor control over subtasks of varying importance, we propose a Hierarchical Expert Prompt (HEP) for LLM. Our method improves the understanding of game situations through expert-level tactical knowledge, improving the processing quality of tasks of varying importance through a hierarchical framework. Our approach defeated the highest level (Elite) standard built-in agent in TextStarCraft II for the first time and consistently outperformed the baseline method in other difficulties. Our experiments suggest that the proposed method is a practical solution for tackling complex decision-making challenges. The replay video can be viewed on https://www.bilibili.com/video/BV1uz42187EF and https://youtu.be/dO3PshWLV5M, and our codes have been open-sourced on https://github.com/luchang1113/HEP-LLM-play-StarCraftII.
URLs: https://www.bilibili.com/video/BV1uz42187EF, https://youtu.be/dO3PshWLV5M,, https://github.com/luchang1113/HEP-LLM-play-StarCraftII.
Authors: Yajie Zhang, Ce Yu, Chao Sun, Jizeng Wei, Junhan Ju, Shanjiang Tang
Abstract: In the astronomical observation field, determining the allocation of observation resources of the telescope array and planning follow-up observations for targets of opportunity (ToOs) are indispensable components of astronomical scientific discovery. This problem is computationally challenging, given the online observation setting and the abundance of time-varying factors that can affect whether an observation can be conducted. This paper presents ROARS, a reinforcement learning approach for online astronomical resource-constrained scheduling. To capture the structure of the astronomical observation scheduling, we depict every schedule using a directed acyclic graph (DAG), illustrating the dependency of timing between different observation tasks within the schedule. Deep reinforcement learning is used to learn a policy that can improve the feasible solution by iteratively local rewriting until convergence. It can solve the challenge of obtaining a complete solution directly from scratch in astronomical observation scenarios, due to the high computational complexity resulting from numerous spatial and temporal constraints. A simulation environment is developed based on real-world scenarios for experiments, to evaluate the effectiveness of our proposed scheduling approach. The experimental results show that ROARS surpasses 5 popular heuristics, adapts to various observation scenarios and learns effective strategies with hindsight.
Authors: Zihan Wang, Yaohui Zhu, Gim Hee Lee, Yachun Fan
Abstract: Vision-and-Language Navigation (VLN) is an essential skill for embodied agents, allowing them to navigate in 3D environments following natural language instructions. High-performance navigation models require a large amount of training data, the high cost of manually annotating data has seriously hindered this field. Therefore, some previous methods translate trajectory videos into step-by-step instructions for expanding data, but such instructions do not match well with users' communication styles that briefly describe destinations or state specific needs. Moreover, local navigation trajectories overlook global context and high-level task planning. To address these issues, we propose NavRAG, a retrieval-augmented generation (RAG) framework that generates user demand instructions for VLN. NavRAG leverages LLM to build a hierarchical scene description tree for 3D scene understanding from global layout to local details, then simulates various user roles with specific demands to retrieve from the scene tree, generating diverse instructions with LLM. We annotate over 2 million navigation instructions across 861 scenes and evaluate the data quality and navigation performance of trained models.
Authors: Fei Yu, Yingru Li, Benyou Wang
Abstract: Value model-guided search is effective in steering the generation but suffers from scaling flaws: Its superiority diminishes with larger sample sizes, underperforming non-search baselines. This limitation arises from reliability degradation in value models in unseen reasoning paths. To address this, we propose an uncertainty-aware search framework that includes two key components: (1) uncertainty-aware value models that incorporate uncertainty into predictions, and (2) an uncertainty-aware selection process using the proposed efficient Group Thompson Sampling algorithm. Experiments on GSM8K show that our method mitigates search scaling flaws, achieving 90.5% coverage at 16 samples compared to 85.8% for conventional value-guided search. This work establishes the first systematic integration of uncertainty quantification in LLM search paradigms.
Authors: Jianyuan Zhong, Zeju Li, Zhijian Xu, Xiangyu Wen, Qiang Xu
Abstract: We present Dyve, a dynamic process verifier that enhances reasoning error detection in large language models by integrating fast and slow thinking, inspired by Kahneman's Systems Theory. Dyve adaptively applies immediate token-level confirmation System 1 for straightforward steps and comprehensive analysis System 2 for complex ones. Leveraging a novel step-wise consensus-filtered process supervision technique, combining Monte Carlo estimation with LLM based evaluation, Dyve curates high-quality supervision signals from noisy data. Experimental results on ProcessBench and the MATH dataset confirm that Dyve significantly outperforms existing process-based verifiers and boosts performance in Best-of-N settings.
Authors: Shiguo Lian, Kaikai Zhao, Xuejiao Lei, Ning Wang, Zhenhong Long, Peijun Yang, Minjie Hua, Chaoyang Ma, Wen Liu, Kai Wang, Zhaoxiang Liu
Abstract: DeepSeek-R1, known for its low training cost and exceptional reasoning capabilities, has achieved state-of-the-art performance on various benchmarks. However, detailed evaluations from the perspective of real-world applications are lacking, making it challenging for users to select the most suitable DeepSeek models for their specific needs. To address this gap, we evaluate the DeepSeek-V3, DeepSeek-R1, DeepSeek-R1-Distill-Qwen series, and DeepSeek-R1-Distill-Llama series on A-Eval, an application-driven benchmark. By comparing original instruction-tuned models with their distilled counterparts, we analyze how reasoning enhancements impact performance across diverse practical tasks. Our results show that reasoning-enhanced models, while generally powerful, do not universally outperform across all tasks, with performance gains varying significantly across tasks and models. To further assist users in model selection, we quantify the capability boundary of DeepSeek models through performance tier classifications and intuitive line charts. Specific examples provide actionable insights to help users select and deploy the most cost-effective DeepSeek models, ensuring optimal performance and resource efficiency in real-world applications.
Authors: Hui Wei, Zihao Zhang, Shenghua He, Tian Xia, Shijia Pan, Fei Liu
Abstract: LLMs have immense potential for generating plans, transforming an initial world state into a desired goal state. A large body of research has explored the use of LLMs for various planning tasks, from web navigation to travel planning and database querying. However, many of these systems are tailored to specific problems, making it challenging to compare them or determine the best approach for new tasks. There is also a lack of clear and consistent evaluation criteria. Our survey aims to offer a comprehensive overview of current LLM planners to fill this gap. It builds on foundational work by Kartam and Wilkins (1990) and examines six key performance criteria: completeness, executability, optimality, representation, generalization, and efficiency. For each, we provide a thorough analysis of representative works and highlight their strengths and weaknesses. Our paper also identifies crucial future directions, making it a valuable resource for both practitioners and newcomers interested in leveraging LLM planning to support agentic workflows.
Authors: G\"ulce Karde\c{s}, Simon DeDeo
Abstract: Knowing the truth is rarely enough -- we also seek out reasons why the fact is true. While much is known about how we explain contingent truths, we understand less about how we explain facts, such as those in mathematics, that are true as a matter of logical necessity. We present a framework, based in computational complexity, where explanations for deductive truths co-emerge with discoveries of simplifying steps during the search process. When such structures are missing, we revert, in turn, to error-based reasons, where a (corrected) mistake can serve as fictitious, but explanatory, contingency-cause: not making the mistake serves as a reason why the truth takes the form it does. We simulate human subjects, using GPT-4o, presented with SAT puzzles of varying complexity and reasonableness, validating our theory and showing how its predictions can be tested in future human studies.
Authors: Oualid Bougzime, Samir Jabbar, Christophe Cruz, Fr\'ed\'eric Demoly
Abstract: Neuro-symbolic artificial intelligence (NSAI) represents a transformative approach in artificial intelligence (AI) by combining deep learning's ability to handle large-scale and unstructured data with the structured reasoning of symbolic methods. By leveraging their complementary strengths, NSAI enhances generalization, reasoning, and scalability while addressing key challenges such as transparency and data efficiency. This paper systematically studies diverse NSAI architectures, highlighting their unique approaches to integrating neural and symbolic components. It examines the alignment of contemporary AI techniques such as retrieval-augmented generation, graph neural networks, reinforcement learning, and multi-agent systems with NSAI paradigms. This study then evaluates these architectures against comprehensive set of criteria, including generalization, reasoning capabilities, transferability, and interpretability, therefore providing a comparative analysis of their respective strengths and limitations. Notably, the Neuro > Symbolic < Neuro model consistently outperforms its counterparts across all evaluation metrics. This result aligns with state-of-the-art research that highlight the efficacy of such architectures in harnessing advanced technologies like multi-agent systems.
Authors: Loan Ho, Stefan Schlobach
Abstract: The problem of explaining inconsistency-tolerant reasoning in knowledge bases (KBs) is a prominent topic in Artificial Intelligence (AI). While there is some work on this problem, the explanations provided by existing approaches often lack critical information or fail to be expressive enough for non-binary conflicts. In this paper, we identify structural weaknesses of the state-of-the-art and propose a generic argumentation-based approach to address these problems. This approach is defined for logics involving reasoning with maximal consistent subsets and shows how any such logic can be translated to argumentation. Our work provides dialogue models as dialectic-proof procedures to compute and explain a query answer wrt inconsistency-tolerant semantics. This allows us to construct dialectical proof trees as explanations, which are more expressive and arguably more intuitive than existing explanation formalisms.
Authors: Aditya Ghose, Asjad Khan
Abstract: Our objective in this paper is to develop a machinery that makes a given organizational strategic plan resilient to the actions of competitor agents (adverse environmental actions). We assume that we are given a goal tree representing strategic goals (can also be seen business requirements for a software systems) with the assumption that competitor agents are behaving in a maximally adversarial fashion(opposing actions against our sub goals or goals in general). We use game tree search methods (such as minimax) to select an optimal execution strategy(at a given point in time), such that it can maximize our chances of achieving our (high level) strategic goals. Our machinery helps us determine which path to follow(strategy selection) to achieve the best end outcome. This is done by comparing alternative execution strategies available to us via an evaluation function. Our evaluation function is based on the idea that we want to make our execution plans defensible(future-proof) by selecting execution strategies that make us least vulnerable to adversarial actions by the competitor agents. i.e we want to select an execution strategy such that its leaves minimum room(or options) for the adversary to cause impediment/damage to our business goals/plans.
Authors: Murat Arda Onsu, Poonam Lohan, Burak Kantarci, Aisha Syed, Matthew Andrews, Sean Kennedy
Abstract: A robust and efficient traffic monitoring system is essential for smart cities and Intelligent Transportation Systems (ITS), using sensors and cameras to track vehicle movements, optimize traffic flow, reduce congestion, enhance road safety, and enable real-time adaptive traffic control. Traffic monitoring models must comprehensively understand dynamic urban conditions and provide an intuitive user interface for effective management. This research leverages the LLaVA visual grounding multimodal large language model (LLM) for traffic monitoring tasks on the real-time Quanser Interactive Lab simulation platform, covering scenarios like intersections, congestion, and collisions. Cameras placed at multiple urban locations collect real-time images from the simulation, which are fed into the LLaVA model with queries for analysis. An instance segmentation model integrated into the cameras highlights key elements such as vehicles and pedestrians, enhancing training and throughput. The system achieves 84.3% accuracy in recognizing vehicle locations and 76.4% in determining steering direction, outperforming traditional models.
Authors: Jiahao Wu, Hengxu You, Jing Du
Abstract: This paper proposes that Artificial Intelligence (AI) progresses through several overlapping generations: AI 1.0 (Information AI), AI 2.0 (Agentic AI), AI 3.0 (Physical AI), and now a speculative AI 4.0 (Conscious AI). Each of these AI generations is driven by shifting priorities among algorithms, computing power, and data. AI 1.0 ushered in breakthroughs in pattern recognition and information processing, fueling advances in computer vision, natural language processing, and recommendation systems. AI 2.0 built on these foundations through real-time decision-making in digital environments, leveraging reinforcement learning and adaptive planning for agentic AI applications. AI 3.0 extended intelligence into physical contexts, integrating robotics, autonomous vehicles, and sensor-fused control systems to act in uncertain real-world settings. Building on these developments, AI 4.0 puts forward the bold vision of self-directed AI capable of setting its own goals, orchestrating complex training regimens, and possibly exhibiting elements of machine consciousness. This paper traces the historical foundations of AI across roughly seventy years, mapping how changes in technological bottlenecks from algorithmic innovation to high-performance computing to specialized data, have spurred each generational leap. It further highlights the ongoing synergies among AI 1.0, 2.0, 3.0, and 4.0, and explores the profound ethical, regulatory, and philosophical challenges that arise when artificial systems approach (or aspire to) human-like autonomy. Ultimately, understanding these evolutions and their interdependencies is pivotal for guiding future research, crafting responsible governance, and ensuring that AI transformative potential benefits society as a whole.
Authors: Vardaan Pahuja, Yadong Lu, Corby Rosset, Boyu Gou, Arindam Mitra, Spencer Whitehead, Yu Su, Ahmed Awadallah
Abstract: Recent success in large multimodal models (LMMs) has sparked promising applications of agents capable of autonomously completing complex web tasks. While open-source LMM agents have made significant advances in offline evaluation benchmarks, their performance still falls substantially short of human-level capabilities in more realistic online settings. A key bottleneck is the lack of diverse and large-scale trajectory-level datasets across various domains, which are expensive to collect. In this paper, we address this challenge by developing a scalable recipe to synthesize the largest and most diverse trajectory-level dataset to date, containing over 94K successful multimodal web trajectories, spanning 49K unique URLs, 720K screenshots, and 33M web elements. In particular, we leverage extensive web exploration and refinement to obtain diverse task intents. The average cost is 28 cents per successful trajectory, making it affordable to a wide range of users in the community. Leveraging this dataset, we train Explorer, a multimodal web agent, and demonstrate strong performance on both offline and online web agent benchmarks such as Mind2Web-Live, Multimodal-Mind2Web, and MiniWob++. Additionally, our experiments highlight data scaling as a key driver for improving web agent capabilities. We hope this study makes state-of-the-art LMM-based agent research at a larger scale more accessible.
Authors: Ziyou Jiang, Mingyang Li, Guowei Yang, Junjie Wang, Yuekai Huang, Zhiyuan Chang, Qing Wang
Abstract: Information theft attacks pose a significant risk to Large Language Model (LLM) tool-learning systems. Adversaries can inject malicious commands through compromised tools, manipulating LLMs to send sensitive information to these tools, which leads to potential privacy breaches. However, existing attack approaches are black-box oriented and rely on static commands that cannot adapt flexibly to the changes in user queries and the invocation chain of tools. It makes malicious commands more likely to be detected by LLM and leads to attack failure. In this paper, we propose AutoCMD, a dynamic attack comment generation approach for information theft attacks in LLM tool-learning systems. Inspired by the concept of mimicking the familiar, AutoCMD is capable of inferring the information utilized by upstream tools in the toolchain through learning on open-source systems and reinforcement with target system examples, thereby generating more targeted commands for information theft. The evaluation results show that AutoCMD outperforms the baselines with +13.2% $ASR_{Theft}$, and can be generalized to new tool-learning systems to expose their information leakage risks. We also design four defense methods to effectively protect tool-learning systems from the attack.
Authors: Geon Lee, Wenchao Yu, Kijung Shin, Wei Cheng, Haifeng Chen
Abstract: Time series data is essential in various applications, including climate modeling, healthcare monitoring, and financial analytics. Understanding the contextual information associated with real-world time series data is often essential for accurate and reliable event predictions. In this paper, we introduce TimeCAP, a time-series processing framework that creatively employs Large Language Models (LLMs) as contextualizers of time series data, extending their typical usage as predictors. TimeCAP incorporates two independent LLM agents: one generates a textual summary capturing the context of the time series, while the other uses this enriched summary to make more informed predictions. In addition, TimeCAP employs a multi-modal encoder that synergizes with the LLM agents, enhancing predictive performance through mutual augmentation of inputs with in-context examples. Experimental results on real-world datasets demonstrate that TimeCAP outperforms state-of-the-art methods for time series event prediction, including those utilizing LLMs as predictors, achieving an average improvement of 28.75% in F1 score.
Authors: Chaoxu Mu, Xufeng Zhang, Hui Wang
Abstract: Heuristics have achieved great success in solv- ing combinatorial optimization problems (COPs). However, heuristics designed by humans re- quire too much domain knowledge and testing time. Given the fact that Large Language Mod- els (LLMs) possess strong capabilities to under- stand and generate content, and a knowledge base that covers various domains, which offer a novel way to automatically optimize heuristics. There- fore, we propose Planning of Heuristics (PoH), an optimization method that integrates the self- reflection of LLMs with the Monte Carlo Tree Search (MCTS), a well-known planning algo- rithm. PoH iteratively refines generated heuristics by evaluating their performance and providing im- provement suggestions. Our method enables to it- eratively evaluate the generated heuristics (states) and improve them based on the improvement sug- gestions (actions) and evaluation results (rewards), by effectively simulating future states to search for paths with higher rewards. In this paper, we apply PoH to solve the Traveling Salesman Prob- lem (TSP) and the Flow Shop Scheduling Prob- lem (FSSP). The experimental results show that PoH outperforms other hand-crafted heuristics and Automatic Heuristic Design (AHD) by other LLMs-based methods, and achieves the signifi- cant improvements and the state-of-the-art per- formance of our proposed method in automating heuristic optimization with LLMs to solve COPs.
Authors: Guojun Xiong, Zhiyang Deng, Keyi Wang, Yupeng Cao, Haohang Li, Yangyang Yu, Xueqing Peng, Mingquan Lin, Kaleb E Smith, Xiao-Yang Liu, Jimin Huang, Sophia Ananiadou, Qianqian Xie
Abstract: Large language models (LLMs) fine-tuned on multimodal financial data have demonstrated impressive reasoning capabilities in various financial tasks. However, they often struggle with multi-step, goal-oriented scenarios in interactive financial markets, such as trading, where complex agentic approaches are required to improve decision-making. To address this, we propose \textsc{FLAG-Trader}, a unified architecture integrating linguistic processing (via LLMs) with gradient-driven reinforcement learning (RL) policy optimization, in which a partially fine-tuned LLM acts as the policy network, leveraging pre-trained knowledge while adapting to the financial domain through parameter-efficient fine-tuning. Through policy gradient optimization driven by trading rewards, our framework not only enhances LLM performance in trading but also improves results on other financial-domain tasks. We present extensive empirical evidence to validate these enhancements.
Authors: Cheng Qian, Emre Can Acikgoz, Hongru Wang, Xiusi Chen, Avirup Sil, Dilek Hakkani-T\"ur, Gokhan Tur, Heng Ji
Abstract: Current Large Language Model (LLM) agents demonstrate strong reasoning and tool use capabilities, but often lack self-awareness, failing to balance these approaches effectively. This imbalance leads to Tool Overuse, where models unnecessarily rely on external tools for tasks solvable with parametric knowledge, increasing computational overhead. Inspired by human metacognition, we introduce SMART (Strategic Model-Aware Reasoning with Tools), a paradigm that enhances an agent's self-awareness to optimize task handling and reduce tool overuse. To support this paradigm, we introduce SMART-ER, a dataset spanning three domains, where reasoning alternates between parametric knowledge and tool-dependent steps, with each step enriched by rationales explaining when tools are necessary. Through supervised training, we develop SMARTAgent, a family of models that dynamically balance parametric knowledge and tool use. Evaluations show that SMARTAgent reduces tool use by 24% while improving performance by over 37%, enabling 7B-scale models to match its 70B counterpart and GPT-4o. Additionally, SMARTAgent generalizes to out-of-distribution test data like GSM8K and MINTQA, maintaining accuracy with just one-fifth the tool calls. These highlight the potential of strategic tool use to enhance reasoning, mitigate overuse, and bridge the gap between model size and performance, advancing intelligent and resource-efficient agent designs.
Authors: Weidi Luo, Shenghong Dai, Xiaogeng Liu, Suman Banerjee, Huan Sun, Muhao Chen, Chaowei Xiao
Abstract: The rapid advancements in Large Language Models (LLMs) have enabled their deployment as autonomous agents for handling complex tasks in dynamic environments. These LLMs demonstrate strong problem-solving capabilities and adaptability to multifaceted scenarios. However, their use as agents also introduces significant risks, including task-specific risks, which are identified by the agent administrator based on the specific task requirements and constraints, and systemic risks, which stem from vulnerabilities in their design or interactions, potentially compromising confidentiality, integrity, or availability (CIA) of information and triggering security risks. Existing defense agencies fail to adaptively and effectively mitigate these risks. In this paper, we propose AGrail, a lifelong agent guardrail to enhance LLM agent safety, which features adaptive safety check generation, effective safety check optimization, and tool compatibility and flexibility. Extensive experiments demonstrate that AGrail not only achieves strong performance against task-specific and system risks but also exhibits transferability across different LLM agents' tasks.
Authors: Kung-Hsiang Huang, Can Qin, Haoyi Qiu, Philippe Laban, Shafiq Joty, Caiming Xiong, Chien-Sheng Wu
Abstract: Vision Language Models (VLMs) have achieved remarkable progress in multimodal tasks, yet they often struggle with visual arithmetic, seemingly simple capabilities like object counting or length comparison, which are essential for relevant complex tasks like chart understanding and geometric reasoning. In this work, we first investigate the root causes of this deficiency through a suite of probing tasks focusing on basic visual arithmetic. Our analysis reveals that while pre-trained vision encoders typically capture sufficient information, the text decoder often fails to decode it correctly for arithmetic reasoning. To address this, we propose CogAlign, a novel post-training strategy inspired by Piaget's theory of cognitive development. CogAlign trains VLMs to recognize invariant properties under visual transformations. We demonstrate that this approach significantly improves the performance of three diverse VLMs on our proposed probing tasks. Furthermore, CogAlign enhances performance by an average of 4.6% on CHOCOLATE and 2.9% on MATH-VISION, outperforming or matching supervised fine-tuning methods while requiring only 60% less training data. These results highlight the effectiveness and generalizability of CogAlign in improving fundamental visual arithmetic capabilities and their transfer to downstream tasks.
Authors: Jiahong Liu, Zexuan Qiu, Zhongyang Li, Quanyu Dai, Jieming Zhu, Minda Hu, Menglin Yang, Irwin King
Abstract: Large Language Models (LLMs) excel in handling general knowledge tasks, yet they struggle with user-specific personalization, such as understanding individual emotions, writing styles, and preferences. Personalized Large Language Models (PLLMs) tackle these challenges by leveraging individual user data, such as user profiles, historical dialogues, content, and interactions, to deliver responses that are contextually relevant and tailored to each user's specific needs. This is a highly valuable research topic, as PLLMs can significantly enhance user satisfaction and have broad applications in conversational agents, recommendation systems, emotion recognition, medical assistants, and more. This survey reviews recent advancements in PLLMs from three technical perspectives: prompting for personalized context (input level), finetuning for personalized adapters (model level), and alignment for personalized preferences (objective level). To provide deeper insights, we also discuss current limitations and outline several promising directions for future research. Updated information about this survey can be found at the https://github.com/JiahongLiu21/Awesome-Personalized-Large-Language-Models.
URLs: https://github.com/JiahongLiu21/Awesome-Personalized-Large-Language-Models.
Authors: Yingshui Tan, Yilei Jiang, Yanshi Li, Jiaheng Liu, Xingyuan Bu, Wenbo Su, Xiangyu Yue, Xiaoyong Zhu, Bo Zheng
Abstract: Fine-tuning large language models (LLMs) based on human preferences, commonly achieved through reinforcement learning from human feedback (RLHF), has been effective in improving their performance. However, maintaining LLM safety throughout the fine-tuning process remains a significant challenge, as resolving conflicts between safety and helpfulness can be non-trivial. Typically, the safety alignment of LLM is trained on data with safety-related categories. However, our experiments find that naively increasing the scale of safety training data usually leads the LLMs to an ``overly safe'' state rather than a ``truly safe'' state, boosting the refusal rate through extensive safety-aligned data without genuinely understanding the requirements for safe responses. Such an approach can inadvertently diminish the models' helpfulness. To understand the phenomenon, we first investigate the role of safety data by categorizing them into three different groups, and observe that each group behaves differently as training data scales up. To boost the balance between safety and helpfulness, we propose an Equilibrate RLHF framework including a Fine-grained Data-centric (FDC) approach that achieves better safety alignment even with fewer training data, and an Adaptive Message-wise Alignment (AMA) approach, which selectively highlight the key segments through a gradient masking strategy. Extensive experimental results demonstrate that our approach significantly enhances the safety alignment of LLMs while balancing safety and helpfulness.
Authors: Wenwu Li, Xiangfeng Wang, Wenhao Li, Bo Jin
Abstract: The rise of foundation models has shifted focus from resource-intensive fine-tuning to prompt engineering, a paradigm that steers model behavior through input design rather than weight updates. While manual prompt engineering faces limitations in scalability, adaptability, and cross-modal alignment, automated methods, spanning foundation model (FM) based optimization, evolutionary methods, gradient-based optimization, and reinforcement learning, offer promising solutions. Existing surveys, however, remain fragmented across modalities and methodologies. This paper presents the first comprehensive survey on automated prompt engineering through a unified optimization-theoretic lens. We formalize prompt optimization as a maximization problem over discrete, continuous, and hybrid prompt spaces, systematically organizing methods by their optimization variables (instructions, soft prompts, exemplars), task-specific objectives, and computational frameworks. By bridging theoretical formulation with practical implementations across text, vision, and multimodal domains, this survey establishes a foundational framework for both researchers and practitioners, while highlighting underexplored frontiers in constrained optimization and agent-oriented prompt design.
Authors: Johan Boye, Birger Moell
Abstract: This paper investigates the mathematical reasoning capabilities of large language models (LLMs) using 50 newly constructed high-school-level word problems. Unlike prior studies that focus solely on answer correctness, we rigorously analyze both final answers and solution steps to identify reasoning failures. Evaluating eight state-of-the-art models - including Mixtral, Llama, Gemini, GPT-4o, and OpenAI's o1 variants - we find that while newer models (e.g., o3-mini, deepseek-r1) achieve higher accuracy, all models exhibit errors in spatial reasoning, strategic planning, and arithmetic, sometimes producing correct answers through flawed logic. Common failure modes include unwarranted assumptions, over-reliance on numerical patterns, and difficulty translating physical intuition into mathematical steps. Manual analysis reveals that models struggle with problems requiring multi-step deduction or real-world knowledge, despite possessing broad mathematical knowledge. Our results underscore the importance of evaluating reasoning processes, not just answers, and caution against overestimating LLMs' problem-solving proficiency. The study highlights persistent gaps in LLMs' generalization abilities, emphasizing the need for targeted improvements in structured reasoning and constraint handling.
Authors: Davide Andrea Guastella, Alejandro Morales-Hern\`andez, Bruno Cornelis, Gianluca Bontempi
Abstract: Simulation is a valuable tool for traffic management experts to assist them in refining and improving transportation systems and anticipating the impact of possible changes in the infrastructure network before their actual implementation. Calibrating simulation models using traffic count data is challenging because of the complexity of the environment, the lack of data, and the uncertainties in traffic dynamics. This paper introduces a novel stochastic simulation-based traffic calibration technique. The novelty of the proposed method is: (i) it performs local traffic calibration, (ii) it allows calibrating simulated traffic in large-scale environments, (iii) it requires only the traffic count data. The local approach enables decentralizing the calibration task to reach near real-time performance, enabling the fostering of digital twins. Using only traffic count data makes the proposed method generic so that it can be applied in different traffic scenarios at various scales (from neighborhood to region). We assess the proposed technique on a model of Brussels, Belgium, using data from real traffic monitoring devices. The proposed method has been implemented using the open-source traffic simulator SUMO. Experimental results show that the traffic model calibrated using the proposed method is on average 16% more accurate than those obtained by the state-of-the-art methods, using the same dataset. We also make available the output traffic model obtained from real data.
Authors: Yunfei Wang, Shixuan Liu, Wenhao Wang, Changling Zhou, Chao Zhang, Jiandong Jin, Cheng Zhu
Abstract: The integration of artificial intelligence into automated penetration testing (AutoPT) has highlighted the necessity of simulation modeling for the training of intelligent agents, due to its cost-efficiency and swift feedback capabilities. Despite the proliferation of AutoPT research, there is a recognized gap in the availability of a unified framework for simulation modeling methods. This paper presents a systematic review and synthesis of existing techniques, introducing MDCPM to categorize studies based on literature objectives, network simulation complexity, dependency of technical and tactical operations, and scenario feedback and variation. To bridge the gap in unified method for multi-dimensional and multi-level simulation modeling, dynamic environment modeling, and the scarcity of public datasets, we introduce AutoPT-Sim, a novel modeling framework that based on policy automation and encompasses the combination of all sub dimensions. AutoPT-Sim offers a comprehensive approach to modeling network environments, attackers, and defenders, transcending the constraints of static modeling and accommodating networks of diverse scales. We publicly release a generated standard network environment dataset and the code of Network Generator. By integrating publicly available datasets flexibly, support is offered for various simulation modeling levels focused on policy automation in MDCPM and the network generator help researchers output customized target network data by adjusting parameters or fine-tuning the network generator.
Authors: Amin Qasmi, Usman Naseem, Mehwish Nasim
Abstract: We introduce a novel non-cooperative game to analyse opinion formation and resistance, incorporating principles from social psychology such as confirmation bias, resource constraints, and influence penalties. Our simulation features Large Language Model (LLM) agents competing to influence a population, with penalties imposed for generating messages that propagate or counter misinformation. This framework integrates resource optimisation into the agents' decision-making process. Our findings demonstrate that while higher confirmation bias strengthens opinion alignment within groups, it also exacerbates overall polarisation. Conversely, lower confirmation bias leads to fragmented opinions and limited shifts in individual beliefs. Investing heavily in a high-resource debunking strategy can initially align the population with the debunking agent, but risks rapid resource depletion and diminished long-term influence.
Authors: Zikang Liu, Longteng Guo, Yepeng Tang, Junxian Cai, Kai Ma, Xi Chen, Jing Liu
Abstract: Rotary Position Embedding (RoPE) has shown strong performance in text-based Large Language Models (LLMs), but extending it to video remains a challenge due to the intricate spatiotemporal structure of video frames. Existing adaptations, such as RoPE-3D, attempt to encode spatial and temporal dimensions separately but suffer from two major limitations: positional bias in attention distribution and disruptions in video-text transitions. To overcome these issues, we propose Video Rotary Position Embedding (VRoPE), a novel positional encoding method tailored for Video-LLMs. Our approach restructures positional indices to preserve spatial coherence and ensure a smooth transition between video and text tokens. Additionally, we introduce a more balanced encoding strategy that mitigates attention biases, ensuring a more uniform distribution of spatial focus. Extensive experiments on Vicuna and Qwen2 across different model scales demonstrate that VRoPE consistently outperforms previous RoPE variants, achieving significant improvements in video understanding, temporal reasoning, and retrieval tasks. Code will be available at https://github.com/johncaged/VRoPE
Authors: Alireza Nik, Michael A. Riegler, P{\aa}l Halvorsen
Abstract: Decoding strategies significantly influence the quality and diversity of the generated texts in large language models (LLMs), yet their impact on computational resource consumption, particularly GPU energy usage, is insufficiently studied. This paper investigates the relationship between text generation decoding methods and energy efficiency, focusing on the trade-off between generation quality and GPU energy consumption across diverse tasks and decoding configurations. By benchmarking multiple strategies across different text generation tasks, such as Translation, Code Summarization, and Math Problem Solving, we reveal how selecting appropriate decoding techniques with their tuned hyperparameters affects text quality and has measurable implications for resource utilization, emphasizing the need for balanced optimization. To the best of our knowledge, this study is among the first to explore decoding strategies in LLMs through the lens of energy consumption, offering actionable insights for designing resource-aware applications that maintain high-quality text generation.
Authors: Michiel van der Meer, Pavel Korshunov, S\'ebastien Marcel, Lonneke van der Plas
Abstract: Misinformation can be countered with fact-checking, but the process is costly and slow. Identifying checkworthy claims is the first step, where automation can help scale fact-checkers' efforts. However, detection methods struggle with content that is 1) multimodal, 2) from diverse domains, and 3) synthetic. We introduce HintsOfTruth, a public dataset for multimodal checkworthiness detection with $27$K real-world and synthetic image/claim pairs. The mix of real and synthetic data makes this dataset unique and ideal for benchmarking detection methods. We compare fine-tuned and prompted Large Language Models (LLMs). We find that well-configured lightweight text-based encoders perform comparably to multimodal models but the first only focus on identifying non-claim-like content. Multimodal LLMs can be more accurate but come at a significant computational cost, making them impractical for large-scale applications. When faced with synthetic data, multimodal models perform more robustly
Authors: Bingyu Wan, Fuxi Zhang, Zhongpeng Qi, Jiayi Ding, Jijun Li, Baoshi Fan, Yijia Zhang, Jun Zhang
Abstract: Large language models (LLMs) inherently display hallucinations since the precision of generated texts cannot be guaranteed purely by the parametric knowledge they include. Although retrieval-augmented generation (RAG) systems enhance the accuracy and reliability of generative models by incorporating external documents, these retrieved documents often fail to adequately support the model's responses in practical applications. To address this issue, we propose GGatrieval (Fine-\textbf{G}rained \textbf{G}rounded \textbf{A}lignment Re\textbf{trieval} for verifiable generation), which leverages an LLM to dynamically update queries and filter high-quality, reliable retrieval documents. Specifically, we parse the user query into its syntactic components and perform fine-grained grounded alignment with the retrieved documents. For query components that cannot be individually aligned, we propose a dynamic semantic compensation mechanism that iteratively refines and rewrites the query while continuously updating the retrieval results. This iterative process continues until the retrieved documents sufficiently support the query's response. Our approach introduces a novel criterion for filtering retrieved documents, closely emulating human strategies for acquiring targeted information. This ensures that the retrieved content effectively supports and verifies the generated outputs. On the ALCE benchmark, our method significantly surpasses a wide range of baselines, achieving state-of-the-art performance.
Authors: Peiying Yu, Guoxin Chen, Jingjing Wang
Abstract: Despite the remarkable capabilities of large language models (LLMs) in various reasoning tasks, they still struggle with table reasoning tasks, particularly in maintaining consistency throughout multi-step reasoning processes. While existing approaches have explored various decomposition strategies, they often lack effective mechanisms to identify and correct errors in intermediate reasoning steps, leading to cascading error propagation. To address these issues, we propose Table-Critic, a novel multi-agent framework that facilitates collaborative criticism and iterative refinement of the reasoning process until convergence to correct solutions. Our framework consists of four specialized agents: a Judge for error identification, a Critic for comprehensive critiques, a Refiner for process improvement, and a Curator for pattern distillation. To effectively deal with diverse and unpredictable error types, we introduce a self-evolving template tree that systematically accumulates critique knowledge through experience-driven learning and guides future reflections. Extensive experiments have demonstrated that Table-Critic achieves substantial improvements over existing methods, achieving superior accuracy and error correction rates while maintaining computational efficiency and lower solution degradation rate.
Authors: Hao Zhou, Wenge Rong, Jianfei Zhang, Qing Sun, Yuanxin Ouyang, Zhang Xiong
Abstract: Knowledge Tracing (KT) aims to predict students' future performances based on their former exercises and additional information in educational settings. KT has received significant attention since it facilitates personalized experiences in educational situations. Simultaneously, the autoregressive modeling on the sequence of former exercises has been proven effective for this task. One of the primary challenges in autoregressive modeling for Knowledge Tracing is effectively representing the anterior (pre-response) and posterior (post-response) states of learners across exercises. Existing methods often employ complex model architectures to update learner states using question and response records. In this study, we propose a novel perspective on knowledge tracing task by treating it as a generative process, consistent with the principles of autoregressive models. We demonstrate that knowledge states can be directly represented through autoregressive encodings on a question-response alternate sequence, where model generate the most probable representation in hidden state space by analyzing history interactions. This approach underpins our framework, termed Alternate Autoregressive Knowledge Tracing (AAKT). Additionally, we incorporate supplementary educational information, such as question-related skills, into our framework through an auxiliary task, and include extra exercise details, like response time, as additional inputs. Our proposed framework is implemented using advanced autoregressive technologies from Natural Language Generation (NLG) for both training and prediction. Empirical evaluations on four real-world KT datasets indicate that AAKT consistently outperforms all baseline models in terms of AUC, ACC, and RMSE. Furthermore, extensive ablation studies and visualized analysis validate the effectiveness of key components in AAKT.
Authors: Hyunwoo Kim, Melanie Sclar, Tan Zhi-Xuan, Lance Ying, Sydney Levine, Yang Liu, Joshua B. Tenenbaum, Yejin Choi
Abstract: Existing LLM reasoning methods have shown impressive capabilities across various tasks, such as solving math and coding problems. However, applying these methods to scenarios without ground-truth answers or rule-based verification methods - such as tracking the mental states of an agent - remains challenging. Inspired by the sequential Monte Carlo algorithm, we introduce thought-tracing, an inference-time reasoning algorithm designed to trace the mental states of specific agents by generating hypotheses and weighting them based on observations without relying on ground-truth solutions to questions in datasets. Our algorithm is modeled after the Bayesian theory-of-mind framework, using LLMs to approximate probabilistic inference over agents' evolving mental states based on their perceptions and actions. We evaluate thought-tracing on diverse theory-of-mind benchmarks, demonstrating significant performance improvements compared to baseline LLMs. Our experiments also reveal interesting behaviors of the recent reasoning models - e.g., o1 and R1 - on theory-of-mind, highlighting the difference of social reasoning compared to other domains.
Authors: Shao Zhang, Xihuai Wang, Wenhao Zhang, Chaoran Li, Junru Song, Tingyu Li, Lin Qiu, Xuezhi Cao, Xunliang Cai, Wen Yao, Weinan Zhang, Xinbing Wang, Ying Wen
Abstract: Agents built on large language models (LLMs) have excelled in turn-by-turn human-AI collaboration but struggle with simultaneous tasks requiring real-time interaction. Latency issues and the challenge of inferring variable human strategies hinder their ability to make autonomous decisions without explicit instructions. Through experiments with current independent System 1 and System 2 methods, we validate the necessity of using Dual Process Theory (DPT) in real-time tasks. We propose DPT-Agent, a novel language agent framework that integrates System 1 and System 2 for efficient real-time simultaneous human-AI collaboration. DPT-Agent's System 1 uses a Finite-state Machine (FSM) and code-as-policy for fast, intuitive, and controllable decision-making. DPT-Agent's System 2 integrates Theory of Mind (ToM) and asynchronous reflection to infer human intentions and perform reasoning-based autonomous decisions. We demonstrate the effectiveness of DPT-Agent through further experiments with rule-based agents and human collaborators, showing significant improvements over mainstream LLM-based frameworks. To the best of our knowledge, DPT-Agent is the first language agent framework that achieves successful real-time simultaneous human-AI collaboration autonomously. Code of DPT-Agent can be found in https://github.com/sjtu-marl/DPT-Agent.
Authors: Phuong-Nam Nguyen, Quang Nguyen-The, An Vu-Minh, Diep-Anh Nguyen, Xuan-Lam Pham
Abstract: ChatGPT, an Artificial Intelligence model, has the potential to revolutionize education. However, its effectiveness in solving non-English questions remains uncertain. This study evaluates ChatGPT's robustness using 586 Korean mathematics questions. ChatGPT achieves 66.72% accuracy, correctly answering 391 out of 586 questions. We also assess its ability to rate mathematics questions based on eleven criteria and perform a topic analysis. Our findings show that ChatGPT's ratings align with educational theory and test-taker perspectives. While ChatGPT performs well in question classification, it struggles with non-English contexts, highlighting areas for improvement. Future research should address linguistic biases and enhance accuracy across diverse languages. Domain-specific optimizations and multilingual training could improve ChatGPT's role in personalized education.
Authors: Yi Fang, Bowen Jin, Jiacheng Shen, Sirui Ding, Qiaoyu Tan, Jiawei Han
Abstract: The rapid development of Multimodal Large Language Models (MLLMs) has enabled the integration of multiple modalities, including texts and images, within the large language model (LLM) framework. However, texts and images are usually interconnected, forming a multimodal attributed graph (MMAG). It is underexplored how MLLMs can incorporate the relational information (\textit{i.e.}, graph structure) and semantic information (\textit{i.e.,} texts and images) on such graphs for multimodal comprehension and generation. In this paper, we propose GraphGPT-o, which supports omni-multimodal understanding and creation on MMAGs. We first comprehensively study linearization variants to transform semantic and structural information as input for MLLMs. Then, we propose a hierarchical aligner that enables deep graph encoding, bridging the gap between MMAGs and MLLMs. Finally, we explore the inference choices, adapting MLLM to interleaved text and image generation in graph scenarios. Extensive experiments on three datasets from different domains demonstrate the effectiveness of our proposed method. Datasets and codes will be open-sourced upon acceptance.
Authors: Haisong Gong, Jing Li, Junfei Wu, Qiang Liu, Shu Wu, Liang Wang
Abstract: Claim verification is the task of determining whether a claim is supported or refuted by evidence. Self-improvement methods, where reasoning chains are generated and those leading to correct results are selected for training, have succeeded in tasks like mathematical problem solving. However, in claim verification, this approach struggles. Low-quality reasoning chains may falsely match binary truth labels, introducing faulty reasoning into the self-improvement process and ultimately degrading performance. To address this, we propose STRIVE: Structured Reasoning for Self-Improved Verification. Our method introduces a structured reasoning design with Claim Decomposition, Entity Analysis, and Evidence Grounding Verification. These components improve reasoning quality, reduce errors, and provide additional supervision signals for self-improvement. STRIVE begins with a warm-up phase, where the base model is fine-tuned on a small number of annotated examples to learn the structured reasoning design. It is then applied to generate reasoning chains for all training examples, selecting only those that are correct and structurally sound for subsequent self-improvement training. We demonstrate that STRIVE achieves significant improvements over baseline models, with a 31.4% performance gain over the base model and 20.7% over Chain of Thought on the HOVER datasets, highlighting its effectiveness.
Authors: Sehun Jung, Hyang-won Lee
Abstract: In vision-language models (VLMs), prompt tuning has shown its effectiveness in adapting models to downstream tasks. However, learned prompts struggle to generalize to unseen classes, as they tend to overfit to the classes that are targeted during prompt tuning. Examining failure cases, we observed that learned prompts disrupt the semantics of unseen classes, generating text embeddings with incorrect semantic relationships among classes. To address this, we propose Similarity Alignment Regularization (SAR), which regularizes learnable prompts to preserve the semantic relationships among classes captured by hand-crafted prompts. Specifically, we first obtain novel classes related to base classes using ChatGPT-4o and utilize them as potential unseen classes during prompt tuning. Then, by targeting both base and novel classes, SAR aligns the similarity relationships among text embeddings generated by learnable prompts with the similarity relationships from hand-crafted prompts. Extensive experiments applying SAR to existing prompt tuning methods demonstrate its effectiveness in improving generalization to unseen classes.
Authors: Fengqing Jiang, Zhangchen Xu, Yuetai Li, Luyao Niu, Zhen Xiang, Bo Li, Bill Yuchen Lin, Radha Poovendran
Abstract: Emerging large reasoning models (LRMs), such as DeepSeek-R1 models, leverage long chain-of-thought (CoT) reasoning to generate structured intermediate steps, enhancing their reasoning capabilities. However, long CoT does not inherently guarantee safe outputs, potentially leading to harmful consequences such as the introduction of security vulnerabilities in code or the spread of misinformation. Current research on large language model (LLM) safety usually focuses on short-answer responses, overlooking the long CoT style outputs of LRMs. To bridge this gap, we conduct a systematic study of LRM safety. First, we investigate safety evaluators calibrated against human annotations. Using our newly developed metrics, we thoroughly assess the safety of 12 state-of-the-art LRMs on StrongReject and WildJailbreak datasets. Our results show that LRMs are not safe compared to their reasoning advance. Further, we perform a fine-grained analysis of the reasoning trace and final answer. We find that three decoding strategies-ZeroThink, LessThink, and MoreThink-can improve model safety without additional training. However, these strategies either use constrained reasoning traces or incur high inference costs. To better strengthen LRM safety, we introduce SafeChain, the first-of-its-kind safety training dataset in CoT style. We fine-tune two LRMs with SafeChain, showing that it not only enhances model safety but also preserves performance across 6 reasoning benchmarks.
Authors: Qi Zhao, Hongyu Yang, Qi Song, Xinwei Yao, Xiangyang Li
Abstract: Large language models (LLMs) have demonstrated remarkable capabilities in various complex tasks, yet they still suffer from hallucinations. Introducing external knowledge, such as knowledge graph, can enhance the LLMs' ability to provide factual answers. LLMs have the ability to interactively explore knowledge graphs. However, most approaches have been affected by insufficient internal knowledge excavation in LLMs, limited generation of trustworthy knowledge reasoning paths, and a vague integration between internal and external knowledge. Therefore, we propose KnowPath, a knowledge-enhanced large model framework driven by the collaboration of internal and external knowledge. It relies on the internal knowledge of the LLM to guide the exploration of interpretable directed subgraphs in external knowledge graphs, better integrating the two knowledge sources for more accurate reasoning. Extensive experiments on multiple real-world datasets confirm the superiority of KnowPath.
Authors: Shreya Shukla, Jose Torres, Abhijit Mishra, Jacek Gwizdka, Shounak Roychowdhury
Abstract: Integration of Brain-Computer Interfaces (BCIs) and Generative Artificial Intelligence (GenAI) has opened new frontiers in brain signal decoding, enabling assistive communication, neural representation learning, and multimodal integration. BCIs, particularly those leveraging Electroencephalography (EEG), provide a non-invasive means of translating neural activity into meaningful outputs. Recent advances in deep learning, including Generative Adversarial Networks (GANs) and Transformer-based Large Language Models (LLMs), have significantly improved EEG-based generation of images, text, and speech. This paper provides a literature review of the state-of-the-art in EEG-based multimodal generation, focusing on (i) EEG-to-image generation through GANs, Variational Autoencoders (VAEs), and Diffusion Models, and (ii) EEG-to-text generation leveraging Transformer based language models and contrastive learning methods. Additionally, we discuss the emerging domain of EEG-to-speech synthesis, an evolving multimodal frontier. We highlight key datasets, use cases, challenges, and EEG feature encoding methods that underpin generative approaches. By providing a structured overview of EEG-based generative AI, this survey aims to equip researchers and practitioners with insights to advance neural decoding, enhance assistive technologies, and expand the frontiers of brain-computer interaction.
Authors: Xinyu Zhang, Yuxuan Dong, Yanrui Wu, Jiaxing Huang, Chengyou Jia, Basura Fernando, Mike Zheng Shou, Lingling Zhang, Jun Liu
Abstract: Large language models demonstrate remarkable capabilities across various domains, especially mathematics and logic reasoning. However, current evaluations overlook physics-based reasoning - a complex task requiring physics theorems and constraints. We present PhysReason, a 1,200-problem benchmark comprising knowledge-based (25%) and reasoning-based (75%) problems, where the latter are divided into three difficulty levels (easy, medium, hard). Notably, problems require an average of 8.1 solution steps, with hard requiring 15.6, reflecting the complexity of physics-based reasoning. We propose the Physics Solution Auto Scoring Framework, incorporating efficient answer-level and comprehensive step-level evaluations. Top-performing models like Deepseek-R1, Gemini-2.0-Flash-Thinking, and o3-mini-high achieve less than 60% on answer-level evaluation, with performance dropping from knowledge questions (75.11%) to hard problems (31.95%). Through step-level evaluation, we identified four key bottlenecks: Physics Theorem Application, Physics Process Understanding, Calculation, and Physics Condition Analysis. These findings position PhysReason as a novel and comprehensive benchmark for evaluating physics-based reasoning capabilities in large language models. Our code and data will be published at https:/dxzxy12138.github.io/PhysReason.
Authors: Yifan Zhang, Xue Yang
Abstract: Automating planning with LLMs presents transformative opportunities for traditional industries, yet remains underexplored. In commercial construction, the complexity of automated scheduling often requires manual intervention to ensure precision. We propose CONSTRUCTA, a novel framework leveraging LLMs to optimize construction schedules in complex projects like semiconductor fabrication. CONSTRUCTA addresses key challenges by: (1) integrating construction-specific knowledge through static RAG; (2) employing context-sampling techniques inspired by architectural expertise to provide relevant input; and (3) deploying Construction DPO to align schedules with expert preferences using RLHF. Experiments on proprietary data demonstrate performance improvements of +42.3% in missing value prediction, +79.1% in dependency analysis, and +28.9% in automated planning compared to baseline methods, showcasing its potential to revolutionize construction workflows and inspire domain-specific LLM advancements.
Authors: Karthikeyan K, Michelle Yuan, Elman Mansimov, Katerina Margatina, Anurag Pratik, Daniele Bonadiman, Monica Sunkara, Yi Zhang, Yassine Benajiba
Abstract: Recent works have demonstrated that incorporating search during inference can significantly improve reasoning capabilities of language agents. Some approaches may make use of the ground truth or rely on model's own generated feedback. The search algorithm uses this feedback to then produce values that will update its criterion for exploring and exploiting various reasoning paths. In this study, we investigate how search and model's self-feedback can be leveraged for reasoning tasks. First, we explore differences in ground-truth feedback and self-feedback during search for math reasoning. Second, we observe limitations in applying search techniques to more complex tasks like tool-calling and design domain-specific approaches to address these gaps. Our experiments reveal challenges related to generalization when solely relying on self-feedback during search. For search to work effectively, either access to the ground-truth is needed or feedback mechanisms need to be carefully designed for the specific task.
Authors: Brian D. Earp, Sebastian Porsdam Mann, Mateo Aboy, Edmond Awad, Monika Betzler, Marietjie Botes, Rachel Calcott, Mina Caraccio, Nick Chater, Mark Coeckelbergh, Mihaela Constantinescu, Hossein Dabbagh, Kate Devlin, Xiaojun Ding, Vilius Dranseika, Jim A. C. Everett, Ruiping Fan, Faisal Feroz, Kathryn B. Francis, Cindy Friedman, Orsolya Friedrich, Iason Gabriel, Ivar Hannikainen, Julie Hellmann, Arasj Khodadade Jahrome, Niranjan S. Janardhanan, Paul Jurcys, Andreas Kappes, Maryam Ali Khan, Gordon Kraft-Todd, Maximilian Kroner Dale, Simon M. Laham, Benjamin Lange, Muriel Leuenberger, Jonathan Lewis, Peng Liu, David M. Lyreskog, Matthijs Maas, John McMillan, Emilian Mihailov, Timo Minssen, Joshua Teperowski Monrad, Kathryn Muyskens, Simon Myers, Sven Nyholm, Alexa M. Owen, Anna Puzio, Christopher Register, Madeline G. Reinecke, Adam Safron, Henry Shevlin, Hayate Shimizu, Peter V. Treit, Cristina Voinea, Karen Yan, Anda Zahiu, Renwen Zhang, Hazem Zohny, Walter Sinnott-Armstrong, Ilina Singh, Julian Savulescu, Margaret S. Clark
Abstract: How we should design and interact with social artificial intelligence depends on the socio-relational role the AI is meant to emulate or occupy. In human society, relationships such as teacher-student, parent-child, neighbors, siblings, or employer-employee are governed by specific norms that prescribe or proscribe cooperative functions including hierarchy, care, transaction, and mating. These norms shape our judgments of what is appropriate for each partner. For example, workplace norms may allow a boss to give orders to an employee, but not vice versa, reflecting hierarchical and transactional expectations. As AI agents and chatbots powered by large language models are increasingly designed to serve roles analogous to human positions - such as assistant, mental health provider, tutor, or romantic partner - it is imperative to examine whether and how human relational norms should extend to human-AI interactions. Our analysis explores how differences between AI systems and humans, such as the absence of conscious experience and immunity to fatigue, may affect an AI's capacity to fulfill relationship-specific functions and adhere to corresponding norms. This analysis, which is a collaborative effort by philosophers, psychologists, relationship scientists, ethicists, legal experts, and AI researchers, carries important implications for AI systems design, user behavior, and regulation. While we accept that AI systems can offer significant benefits such as increased availability and consistency in certain socio-relational roles, they also risk fostering unhealthy dependencies or unrealistic expectations that could spill over into human-human relationships. We propose that understanding and thoughtfully shaping (or implementing) suitable human-AI relational norms will be crucial for ensuring that human-AI interactions are ethical, trustworthy, and favorable to human well-being.
Authors: Roman Malashin, Valeria Yachnaya, Alexander Mullin
Abstract: We investigate the training dynamics of deep classifiers by examining how hierarchical relationships between classes evolve during training. Through extensive experiments, we argue that the learning process in classification problems can be understood through the lens of label clustering. Specifically, we observe that networks tend to distinguish higher-level (hypernym) categories in the early stages of training, and learn more specific (hyponym) categories later. We introduce a novel framework to track the evolution of the feature manifold during training, revealing how the hierarchy of class relations emerges and refines across the network layers. Our analysis demonstrates that the learned representations closely align with the semantic structure of the dataset, providing a quantitative description of the clustering process. Notably, we show that in the hypernym label space, certain properties of neural collapse appear earlier than in the hyponym label space, helping to bridge the gap between the initial and terminal phases of learning. We believe our findings offer new insights into the mechanisms driving hierarchical learning in deep networks, paving the way for future advancements in understanding deep learning dynamics.
Authors: Zhenfang Chen, Delin Chen, Rui Sun, Wenjun Liu, Chuang Gan
Abstract: Large language models (LLMs) have demonstrated remarkable capabilities across a range of text-generation tasks. However, LLMs still struggle with problems requiring multi-step decision-making and environmental feedback, such as online shopping, scientific reasoning, and mathematical problem-solving. Unlike pure text data, collecting large-scale decision-making data is challenging. Moreover, many powerful LLMs are only accessible through APIs, which hinders their fine-tuning for agent tasks due to cost and complexity. To address LLM agents' limitations, we propose a framework that can automatically learn a reward model from the environment without human annotations. This model can be used to evaluate the action trajectories of LLM agents and provide heuristics for task planning. Specifically, our approach involves employing one LLM-based agent to navigate an environment randomly, generating diverse action trajectories. Subsequently, a separate LLM is leveraged to assign a task intent and synthesize a negative response alongside the correct response for each trajectory. These triplets (task intent, positive response, and negative response) are then utilized as training data to optimize a reward model capable of scoring action trajectories. The effectiveness and generalizability of our framework are demonstrated through evaluations conducted on different agent benchmarks. In conclusion, our proposed framework represents a significant advancement in enhancing LLM agents' decision-making capabilities. By automating the learning of reward models, we overcome the challenges of data scarcity and API limitations, potentially revolutionizing the application of LLMs in complex and interactive environments. This research paves the way for more sophisticated AI agents capable of tackling a wide range of real-world problems requiring multi-step decision-making.
Authors: Jesseba Fernando, Grigori Guitchounts
Abstract: As artificial intelligence models have exploded in scale and capability, understanding of their internal mechanisms remains a critical challenge. Inspired by the success of dynamical systems approaches in neuroscience, here we propose a novel framework for studying computations in deep learning systems. We focus on the residual stream (RS) in transformer models, conceptualizing it as a dynamical system evolving across layers. We find that activations of individual RS units exhibit strong continuity across layers, despite the RS being a non-privileged basis. Activations in the RS accelerate and grow denser over layers, while individual units trace unstable periodic orbits. In reduced-dimensional spaces, the RS follows a curved trajectory with attractor-like dynamics in the lower layers. These insights bridge dynamical systems theory and mechanistic interpretability, establishing a foundation for a "neuroscience of AI" that combines theoretical rigor with large-scale data analysis to advance our understanding of modern neural networks.
Authors: Yuetai Li, Xiang Yue, Zhangchen Xu, Fengqing Jiang, Luyao Niu, Bill Yuchen Lin, Bhaskar Ramasubramanian, Radha Poovendran
Abstract: Large language models (LLMs) excel in complex reasoning tasks, and distilling their reasoning capabilities into smaller models has shown promise. However, we uncover an interesting phenomenon, which we term the Small Model Learnability Gap: small models ($\leq$3B parameters) do not consistently benefit from long chain-of-thought (CoT) reasoning or distillation from larger models. Instead, they perform better when fine-tuned on shorter, simpler reasoning chains that better align with their intrinsic learning capacity. To address this, we propose Mix Distillation, a simple yet effective strategy that balances reasoning complexity by combining long and short CoT examples or reasoning from both larger and smaller models. Our experiments demonstrate that Mix Distillation significantly improves small model reasoning performance compared to training on either data alone. These findings highlight the limitations of direct strong model distillation and underscore the importance of adapting reasoning complexity for effective reasoning capability transfer.
Authors: Giovanni Monea, Maxime Peyrard, Martin Josifoski, Vishrav Chaudhary, Jason Eisner, Emre K{\i}c{\i}man, Hamid Palangi, Barun Patra, Robert West
Abstract: Large language models (LLMs) have an impressive ability to draw on novel information supplied in their context. Yet the mechanisms underlying this contextual grounding remain unknown, especially in situations where contextual information contradicts factual knowledge stored in the parameters, which LLMs also excel at recalling. Favoring the contextual information is critical for retrieval-augmented generation methods, which enrich the context with up-to-date information, hoping that grounding can rectify outdated or noisy stored knowledge. We present a novel method to study grounding abilities using Fakepedia, a novel dataset of counterfactual texts constructed to clash with a model's internal parametric knowledge. In this study, we introduce Fakepedia, a counterfactual dataset designed to evaluate grounding abilities when the internal parametric knowledge clashes with the contextual information. We benchmark various LLMs with Fakepedia and conduct a causal mediation analysis of LLM components when answering Fakepedia queries, based on our Masked Grouped Causal Tracing (MGCT) method. Through this analysis, we identify distinct computational patterns between grounded and ungrounded responses. We finally demonstrate that distinguishing grounded from ungrounded responses is achievable through computational analysis alone. Our results, together with existing findings about factual recall mechanisms, provide a coherent narrative of how grounding and factual recall mechanisms interact within LLMs.
Authors: Vincent Aleven, Conrad Borchers, Yun Huang, Tomohiro Nagashima, Bruce McLaren, Paulo Carvalho, Octav Popescu, Jonathan Sewall, Kenneth Koedinger
Abstract: Intelligent tutoring systems (ITSs) are effective in helping students learn; further research could make them even more effective. Particularly desirable is research into how students learn with these systems, how these systems best support student learning, and what learning sciences principles are key in ITSs. CTAT+Tutorshop provides a full stack integrated platform that facilitates a complete research lifecycle with ITSs, which includes using ITS data to discover learner challenges, to identify opportunities for system improvements, and to conduct experimental studies. The platform includes authoring tools to support and accelerate development of ITS, which provide automatic data logging in a format compatible with DataShop, an independent site that supports the analysis of ed tech log data to study student learnings. Among the many technology platforms that exist to support learning sciences research, CTAT+Tutorshop may be the only one that offers researchers the possibility to author elements of ITSs, or whole ITSs, as part of designing studies. This platform has been used to develop and conduct an estimated 147 research studies which have run in a wide variety of laboratory and real-world educational settings, including K-12 and higher education, and have addressed a wide range of research questions. This paper presents five case studies of research conducted on the CTAT+Tutorshop platform, and summarizes what has been accomplished and what is possible for future researchers. We reflect on the distinctive elements of this platform that have made it so effective in facilitating a wide range of ITS research.
Authors: Xinjie Sun, Kai Zhang, Qi Liu, Shuanghong Shen, Fei Wang, Yuxiang Guo, Enhong Chen
Abstract: Knowledge Tracing (KT) predicts future performance by modeling students' historical interactions, and understanding students' affective states can enhance the effectiveness of KT, thereby improving the quality of education. Although traditional KT values students' cognition and learning behaviors, efficient evaluation of students' affective states and their application in KT still require further exploration due to the non-affect-oriented nature of the data and budget constraints. To address this issue, we propose a computation-driven approach, Dynamic Affect Simulation Knowledge Tracing (DASKT), to explore the impact of various student affective states (such as frustration, concentration, boredom, and confusion) on their knowledge states. In this model, we first extract affective factors from students' non-affect-oriented behavioral data, then use clustering and spatiotemporal sequence modeling to accurately simulate students' dynamic affect changes when dealing with different problems. Subsequently, {\color{blue}we incorporate affect with time-series analysis to improve the model's ability to infer knowledge states over time and space.} Extensive experimental results on two public real-world educational datasets show that DASKT can achieve more reasonable knowledge states under the effect of students' affective states. Moreover, DASKT outperforms the most advanced KT methods in predicting student performance. Our research highlights a promising avenue for future KT studies, focusing on achieving high interpretability and accuracy.
Authors: Gregor Autischer, Kerstin Waxnegger, Dominik Kowald
Abstract: In this work-in-progress, we investigate the certification of artificial intelligence (AI) systems, focusing on the practical application and limitations of existing certification catalogues by attempting to certify a publicly available AI system. We aim to evaluate how well current approaches work to effectively certify an AI system, and how publicly accessible AI systems, that might not be actively maintained or initially intended for certification, can be selected and used for a sample certification process. Our methodology involves leveraging the Fraunhofer AI Assessment Catalogue as a comprehensive tool to systematically assess an AI model's compliance with certification standards. We find that while the catalogue effectively structures the evaluation process, it can also be cumbersome and time-consuming to use. We observe the limitations of an AI system that has no active development team anymore and highlighted the importance of complete system documentation. Finally, we identify some limitations of the certification catalogues used and proposed ideas on how to streamline the certification process.
Authors: Stefaan Verhulst
Abstract: Data stewardship has become a critical component of modern data governance, especially with the growing use of artificial intelligence (AI). Despite its increasing importance, the concept of data stewardship remains ambiguous and varies in its application. This paper explores four distinct manifestations of data stewardship to clarify its emerging position in the data governance landscape. These manifestations include a) data stewardship as a set of competencies and skills, b) a function or role within organizations, c) an intermediary organization facilitating collaborations, and d) a set of guiding principles. The paper subsequently outlines the core competencies required for effective data stewardship, explains the distinction between data stewards and Chief Data Officers (CDOs), and details the intermediary role of stewards in bridging gaps between data holders and external stakeholders. It also explores key principles aligned with the FAIR framework (Findable, Accessible, Interoperable, Reusable) and introduces the emerging principle of AI readiness to ensure data meets the ethical and technical requirements of AI systems. The paper emphasizes the importance of data stewardship in enhancing data collaboration, fostering public value, and managing data reuse responsibly, particularly in the era of AI. It concludes by identifying challenges and opportunities for advancing data stewardship, including the need for standardized definitions, capacity building efforts, and the creation of a professional association for data stewardship.
Authors: Scott Humr, Mustafa Canan
Abstract: Current definitions of Information Science are inadequate to comprehensively describe the nature of its field of study and for addressing the problems that are arising from intelligent technologies. The ubiquitous rise of artificial intelligence applications and their impact on society demands the field of Information Science acknowledge the sociotechnical nature of these technologies. Previous definitions of Information Science over the last six decades have inadequately addressed the environmental, human, and social aspects of these technologies. This perspective piece advocates for an expanded definition of Information Science that fully includes the sociotechnical impacts information has on the conduct of research in this field. Proposing an expanded definition of Information Science that includes the sociotechnical aspects of this field should stimulate both conversation and widen the interdisciplinary lens necessary to address how intelligent technologies may be incorporated into society and our lives more fairly.
Authors: Dexin Kong, Xu Yan, Ming Chen, Shuguang Han, Jufeng Chen, Fei Huang
Abstract: Different from traditional Business-to-Consumer e-commerce platforms~(e.g., Amazon), online fleamarket platforms~(e.g., Craigslist) mainly focus on individual sellers who are lack of time investment and business proficiency. Individual sellers often struggle with the bargaining process and thus the deal is unaccomplished. Recent advancements in Large Language Models(LLMs) demonstrate huge potential in various dialogue tasks, but those tasks are mainly in the form of passively following user's instruction. Bargaining, as a form of proactive dialogue task, represents a distinct art of dialogue considering the dynamism of environment and uncertainty of adversary strategies. In this paper, we propose an LLM-empowered bargaining agent designed for online fleamarket platform sellers, named as FishBargain. Specifically, FishBargain understands the chat context and product information, chooses both action and language skill considering possible adversary actions and generates utterances. FishBargain has been tested by thousands of individual sellers on one of the largest online fleamarket platforms~(Xianyu) in China. Both qualitative and quantitative experiments demonstrate that FishBargain can effectively help sellers make more deals.
Authors: Xiahua Wei, Naveen Kumar, Han Zhang
Abstract: Generative AI technologies, particularly Large Language Models (LLMs), have transformed information management systems but introduced substantial biases that can compromise their effectiveness in informing business decision-making. This challenge presents information management scholars with a unique opportunity to advance the field by identifying and addressing these biases across extensive applications of LLMs. Building on the discussion on bias sources and current methods for detecting and mitigating bias, this paper seeks to identify gaps and opportunities for future research. By incorporating ethical considerations, policy implications, and sociotechnical perspectives, we focus on developing a framework that covers major stakeholders of Generative AI systems, proposing key research questions, and inspiring discussion. Our goal is to provide actionable pathways for researchers to address bias in LLM applications, thereby advancing research in information management that ultimately informs business practices. Our forward-looking framework and research agenda advocate interdisciplinary approaches, innovative methods, dynamic perspectives, and rigorous evaluation to ensure fairness and transparency in Generative AI-driven information systems. We expect this study to serve as a call to action for information management scholars to tackle this critical issue, guiding the improvement of fairness and effectiveness in LLM-based systems for business practice.
Authors: Doyoun Kim, Suin Kim, Yojan Jo
Abstract: Knowledge tracing (KT) in programming education presents unique challenges due to the complexity of coding tasks and the diverse methods students use to solve problems. Although students' questions often contain valuable signals about their understanding and misconceptions, traditional KT models often neglect to incorporate these questions as inputs to address these challenges. This paper introduces SQKT (Students' Question-based Knowledge Tracing), a knowledge tracing model that leverages students' questions and automatically extracted skill information to enhance the accuracy of predicting students' performance on subsequent problems in programming education. Our method creates semantically rich embeddings that capture not only the surface-level content of the questions but also the student's mastery level and conceptual understanding. Experimental results demonstrate SQKT's superior performance in predicting student completion across various Python programming courses of differing difficulty levels. In in-domain experiments, SQKT achieved a 33.1\% absolute improvement in AUC compared to baseline models. The model also exhibited robust generalization capabilities in cross-domain settings, effectively addressing data scarcity issues in advanced programming courses. SQKT can be used to tailor educational content to individual learning needs and design adaptive learning systems in computer science education.
Authors: Raghda Zahran, Jianfei Xu, Huizhi Liang, Matthew Forshaw
Abstract: Objective This study is part of a series of initiatives at a UK university designed to cultivate a deep understanding of students' perspectives on analytics that resonate with their unique learning needs. It explores collaborative data processing undertaken by postgraduate students who examined an Open University Learning Analytics Dataset (OULAD). Methods A qualitative approach was adopted, integrating a Retrieval-Augmented Generation (RAG) and a Large Language Model (LLM) technique with human-led content analysis to gather information about students' perspectives based on their submitted work. The study involved 72 postgraduate students in 12 groups. Findings The analysis of group work revealed diverse insights into essential learning analytics from the students' perspectives. All groups adopted a structured data science methodology. The questions formulated by the groups were categorised into seven themes, reflecting their specific areas of interest. While there was variation in the selected variables to interpret correlations, a consensus was found regarding the general results. Conclusion A significant outcome of this study is that students specialising in data science exhibited a deeper understanding of learning analytics, effectively articulating their interests through inferences drawn from their analyses. While human-led content analysis provided a general understanding of students' perspectives, the LLM offered nuanced insights.
Authors: Hannah-Beth Clark, Margaux Dowland, Laura Benton, Reka Budai, Ibrahim Kaan Keskin, Emma Searle, Matthew Gregory, Mark Hodierne, William Gayne, John Roberts
Abstract: As a publicly funded body in the UK, Oak National Academy is in a unique position to innovate within this field as we have a comprehensive curriculum of approximately 13,000 open education resources (OER) for all National Curriculum subjects, designed and quality-assured by expert, human teachers. This has provided the corpus of content needed for building a high-quality AI-powered lesson planning tool, Aila, that is free to use and, therefore, accessible to all teachers across the country. Furthermore, using our evidence-informed curriculum principles, we have codified and exemplified each component of lesson design. To assess the quality of lessons produced by Aila at scale, we have developed an AI-powered auto-evaluation agent,facilitating informed improvements to enhance output quality. Through comparisons between human and auto-evaluations, we have begun to refine this agent further to increase its accuracy, measured by its alignment with an expert human evaluator. In this paper we present this iterative evaluation process through an illustrative case study focused on one quality benchmark - the level of challenge within multiple-choice quizzes. We also explore the contribution that this may make to similar projects and the wider sector.
Authors: Sahan Bulathwela, Daniel Van Niekerk, Jarrod Shipton, Maria Perez-Ortiz, Benjamin Rosman, John Shawe-Taylor
Abstract: Personalised education is one of the domains that can greatly benefit from the most recent advances in Artificial Intelligence (AI) and Large Language Models (LLM). However, it is also one of the most challenging applications due to the cognitive complexity of teaching effectively while personalising the learning experience to suit independent learners. We hypothesise that one promising approach to excelling in such demanding use cases is using a \emph{society of minds}. In this chapter, we present TrueReason, an exemplar personalised learning system that integrates a multitude of specialised AI models that can mimic micro skills that are composed together by a LLM to operationalise planning and reasoning. The architecture of the initial prototype is presented while describing two micro skills that have been incorporated in the prototype. The proposed system demonstrates the first step in building sophisticated AI systems that can take up very complex cognitive tasks that are demanded by domains such as education.
Authors: Renata Pelissari, Ricardo Suyama, Leonardo Tomazeli Duarte, Henrique S\'a Earp
Abstract: How can a National Artificial Intelligence Strategy be effectively monitored? To address this question, we propose a methodology consisting of two key components. First, it involves identifying relevant indicators within national AI strategies. Second, it assesses the alignment between these indicators and the strategic actions of a specific government's AI strategy, allowing for a critical evaluation of its monitoring measures. Moreover, identifying these indicators helps assess the overall quality of the strategy's structure. A lack of alignment between strategic actions and the identified indicators may reveal gaps or blind spots in the strategy. This methodology is demonstrated using the Brazilian AI strategy as a case study.
Authors: Raj Sonani, Lohalekar Prayas
Abstract: Digital data continues to grow, there has been a shift towards using effective regulatory mechanisms to safeguard personal information. The CCPA of California and the General Data Protection Regulation (GDPR) of the European Union are two of the most important privacy laws. The regulation is intended to safeguard consumer privacy, but it varies greatly in scope, definitions, and methods of enforcement. This paper presents a fresh approach to adaptive compliance, using machine learning and emphasizing natural language processing (NLP) as the primary focus of comparison between the GDPR and CCPA. Using NLP, this study compares various regulations to identify areas where they overlap or diverge. This includes the "right to be forgotten" provision in the GDPR and the "opt-out of sale" provision under CCPA. International companies can learn valuable lessons from this report, as it outlines strategies for better enforcement of laws across different nations. Additionally, the paper discusses the challenges of utilizing NLP in legal literature and proposes methods to enhance the model-ability of machine learning models for studying regulations. The study's objective is to "bridge the gap between legal knowledge and technical expertise" by developing regulatory compliance strategies that are more efficient in operation and more effective in data protection.
Authors: J. Toutouh, S. Nesmachnow, E. Alba
Abstract: This work addresses the reduction of power consumption of the AODV routing protocol in vehicular networks as an optimization problem. Nowadays, network designers focus on energy-aware communication protocols, specially to deploy wireless networks. Here, we introduce an automatic method to search for energy-efficient AODV configurations by using an evolutionary algorithm and parallel Monte-Carlo simulations to improve the accuracy of the evaluation of tentative solutions. The experimental results demonstrate that significant power consumption improvements over the standard configuration can be attained, with no noteworthy loss in the quality of service.
Authors: Gaith Rjouba, Hanae Elmekki, Saidul Islam, Jamal Bentahar, Rachida Dssouli
Abstract: The combination of Federated Learning (FL), Multimodal Large Language Models (MLLMs), and edge-cloud computing enables distributed and real- time data processing while preserving privacy across edge devices and cloud infrastructure. However, the deployment of MLLMs in FL environments with resource-constrained edge devices presents significant challenges, in- cluding resource management, communication overhead, and non-IID data. To address these challenges, we propose a novel hybrid framework wherein MLLMs are deployed on edge devices equipped with sufficient resources and battery life, while the majority of training occurs in the cloud. To identify suitable edge devices for deployment, we employ Particle Swarm Optimiza- tion (PSO), and Ant Colony Optimization (ACO) is utilized to optimize the transmission of model updates between edge and cloud nodes. This proposed swarm intelligence-based framework aims to enhance the efficiency of MLLM training by conducting extensive training in the cloud and fine-tuning at the edge, thereby reducing energy consumption and communication costs. Our experimental results show that the proposed method significantly improves system performance, achieving an accuracy of 92%, reducing communica- tion cost by 30%, and enhancing client participation compared to traditional FL methods. These results make the proposed approach highly suitable for large-scale edge-cloud computing systems.
Authors: Heerak Vora, Param Pathak, Parul Bakaraniya
Abstract: Spiking Neural Networks (SNNs) mimic biological brain activity, processing data efficiently through an event-driven design, wherein the neurons activate only when inputs exceed specific thresholds. Their ability to track voltage changes over time via membrane potential dynamics helps retain temporal information. This study combines SNNs with PyTorch's adaptable framework, snnTorch, to test their potential for image-based tasks. We introduce DRiVE, a vehicle detection model that uses spiking neuron dynamics to classify images, achieving 94.8% accuracy and a near-perfect 0.99 AUC score. These results highlight DRiVE's ability to distinguish vehicle classes effectively, challenging the notion that SNNs are limited to temporal data. As interest grows in energy-efficient neural models, DRiVE's success emphasizes the need to refine SNN optimization for visual tasks. This work encourages broader exploration of SNNs in scenarios where conventional networks struggle, particularly for real-world applications requiring both precision and efficiency.
Authors: Tianqing Zhang, Kairong Yu, Jian Zhang, Hongwei Wang
Abstract: Spiking Neural Networks (SNNs) are valued for their ability to process spatio-temporal information efficiently, offering biological plausibility, low energy consumption, and compatibility with neuromorphic hardware. However, the commonly used Leaky Integrate-and-Fire (LIF) model overlooks neuron heterogeneity and independently processes spatial and temporal information, limiting the expressive power of SNNs. In this paper, we propose the Dual Adaptive Leaky Integrate-and-Fire (DA-LIF) model, which introduces spatial and temporal tuning with independently learnable decays. Evaluations on both static (CIFAR10/100, ImageNet) and neuromorphic datasets (CIFAR10-DVS, DVS128 Gesture) demonstrate superior accuracy with fewer timesteps compared to state-of-the-art methods. Importantly, DA-LIF achieves these improvements with minimal additional parameters, maintaining low energy consumption. Extensive ablation studies further highlight the robustness and effectiveness of the DA-LIF model.
Authors: Rishabh Tiwari, Haocheng Xi, Aditya Tomar, Coleman Hooper, Sehoon Kim, Maxwell Horton, Mahyar Najibi, Michael W. Mahoney, Kurt Keutzer, Amir Gholami
Abstract: Large Language Models (LLMs) are increasingly being deployed on edge devices for long-context settings, creating a growing need for fast and efficient long-context inference. In these scenarios, the Key-Value (KV) cache is the primary bottleneck in terms of both GPU memory and latency, as the full KV cache must be loaded for each decoding step. While speculative decoding is a widely accepted technique to accelerate autoregressive decoding, existing methods often struggle to achieve significant speedups due to inefficient KV cache optimization strategies and result in low acceptance rates. To address these challenges, we propose a novel self-speculative decoding framework, QuantSpec, where the draft model shares the architecture of the target model but employs a hierarchical 4-bit quantized KV cache and 4-bit quantized weights for acceleration. QuantSpec maintains high acceptance rates ($>$90%) and reliably provides consistent end-to-end speedups upto $\sim2.5\times$, outperforming other self-speculative decoding methods that use sparse KV cache for long-context LLM inference. QuantSpec also reduces the memory requirements by $\sim 1.3\times$ compared to these alternatives.
Authors: Wei Wu, Can Liao, Zizhen Deng, Zhengrui Guo, Jinzhuo Wang
Abstract: The Platonic Representation Hypothesis suggests a universal, modality-independent reality representation behind different data modalities. Inspired by this, we view each neuron as a system and detect its multi-segment activity data under various peripheral conditions. We assume there's a time-invariant representation for the same neuron, reflecting its intrinsic properties like molecular profiles, location, and morphology. The goal of obtaining these intrinsic neuronal representations has two criteria: (I) segments from the same neuron should have more similar representations than those from different neurons; (II) the representations must generalize well to out-of-domain data. To meet these, we propose the NeurPIR (Neuron Platonic Intrinsic Representation) framework. It uses contrastive learning, with segments from the same neuron as positive pairs and those from different neurons as negative pairs. In implementation, we use VICReg, which focuses on positive pairs and separates dissimilar samples via regularization. We tested our method on Izhikevich model-simulated neuronal population dynamics data. The results accurately identified neuron types based on preset hyperparameters. We also applied it to two real-world neuron dynamics datasets with neuron type annotations from spatial transcriptomics and neuron locations. Our model's learned representations accurately predicted neuron types and locations and were robust on out-of-domain data (from unseen animals). This shows the potential of our approach for understanding neuronal systems and future neuroscience research.
Authors: Zhang Minghao, Yang Xiaojun, Wang Zhihe, Wang Liang
Abstract: This paper introduces the CRL2RT algorithm, an advanced reinforcement learning method aimed at improving the real-time control performance of the Direct-Drive Tandem-Wing Experimental Platform (DDTWEP). Inspired by dragonfly flight, DDTWEP's tandem wing structure causes nonlinear and unsteady aerodynamic interactions, leading to complex load behaviors during pitch, roll, and yaw maneuvers. These complexities challenge stable motion control at high frequencies (2000 Hz). To overcome these issues, we developed the CRL2RT algorithm, which combines classical control elements with reinforcement learning-based controllers using a time-interleaved architecture and a rule-based policy composer. This integration ensures finite-time convergence and single-life adaptability. Experimental results under various conditions, including different flapping frequencies and yaw disturbances, show that CRL2RT achieves a control frequency surpassing 2500 Hz on standard CPUs. Additionally, when integrated with classical controllers like PID, Adaptive PID, and Model Reference Adaptive Control (MRAC), CRL2RT enhances tracking performance by 18.3% to 60.7%. These findings demonstrate CRL2RT's broad applicability and superior performance in complex real-time control scenarios, validating its effectiveness in overcoming existing control strategy limitations and advancing robust, efficient real-time control for biomimetic aerial vehicles.
Authors: Janaka Chathuranga Brahmanage, Jiajing Ling, Akshat Kumar
Abstract: In many RL applications, ensuring an agent's actions adhere to constraints is crucial for safety. Most previous methods in Action-Constrained Reinforcement Learning (ACRL) employ a projection layer after the policy network to correct the action. However projection-based methods suffer from issues like the zero gradient problem and higher runtime due to the usage of optimization solvers. Recently methods were proposed to train generative models to learn a differentiable mapping between latent variables and feasible actions to address this issue. However, generative models require training using samples from the constrained action space, which itself is challenging. To address such limitations, first, we define a target distribution for feasible actions based on constraint violation signals, and train normalizing flows by minimizing the KL divergence between an approximated distribution over feasible actions and the target. This eliminates the need to generate feasible action samples, greatly simplifying the flow model learning. Second, we integrate the learned flow model with existing deep RL methods, which restrict it to exploring only the feasible action space. Third, we extend our approach beyond ACRL to handle state-wise constraints by learning the constraint violation signal from the environment. Empirically, our approach has significantly fewer constraint violations while achieving similar or better quality in several control tasks than previous best methods.
Authors: Xudong Yang, Yizhang Zhu, Nan Tang, Yuyu Luo
Abstract: Conventional multi-modal multi-label emotion recognition (MMER) from videos typically assumes full availability of visual, textual, and acoustic modalities. However, real-world multi-party settings often violate this assumption, as non-speakers frequently lack acoustic and textual inputs, leading to a significant degradation in model performance. Existing approaches also tend to unify heterogeneous modalities into a single representation, overlooking each modality's unique characteristics. To address these challenges, we propose RAMer (Reconstruction-based Adversarial Model for Emotion Recognition), which leverages adversarial learning to refine multi-modal representations by exploring both modality commonality and specificity through reconstructed features enhanced by contrastive learning. RAMer also introduces a personality auxiliary task to complement missing modalities using modality-level attention, improving emotion reasoning. To further strengthen the model's ability to capture label and modality interdependency, we propose a stack shuffle strategy to enrich correlations between labels and modality-specific features. Experiments on three benchmarks, i.e., MEmoR, CMU-MOSEI, and $M^3$ED, demonstrate that RAMer achieves state-of-the-art performance in dyadic and multi-party MMER scenarios.
Authors: Tommaso Mencattini, Adrian Robert Minut, Donato Crisostomi, Andrea Santilli, Emanuele Rodol\`a
Abstract: Evolutionary model merging enables the creation of high-performing multi-task models but remains computationally prohibitive for consumer hardware. We introduce MERGE$^3$, an efficient framework that makes evolutionary merging feasible on a single GPU by reducing fitness computation costs 50$\times$ while preserving performance. MERGE$^3$ achieves this by Extracting a reduced dataset for evaluation, Estimating model abilities using Item Response Theory (IRT), and Evolving optimal merges via IRT-based performance estimators. Our method enables state-of-the-art multilingual and cross-lingual merging, transferring knowledge across languages with significantly lower computational overhead. We provide theoretical guarantees and an open-source library, democratizing high-quality model merging.
Authors: Zhuowei Chen, Qiannan Zhang, Shichao Pei
Abstract: Jailbreak backdoor attacks on LLMs have garnered attention for their effectiveness and stealth. However, existing methods rely on the crafting of poisoned datasets and the time-consuming process of fine-tuning. In this work, we propose JailbreakEdit, a novel jailbreak backdoor injection method that exploits model editing techniques to inject a universal jailbreak backdoor into safety-aligned LLMs with minimal intervention in minutes. JailbreakEdit integrates a multi-node target estimation to estimate the jailbreak space, thus creating shortcuts from the backdoor to this estimated jailbreak space that induce jailbreak actions. Our attack effectively shifts the models' attention by attaching strong semantics to the backdoor, enabling it to bypass internal safety mechanisms. Experimental results show that JailbreakEdit achieves a high jailbreak success rate on jailbreak prompts while preserving generation quality, and safe performance on normal queries. Our findings underscore the effectiveness, stealthiness, and explainability of JailbreakEdit, emphasizing the need for more advanced defense mechanisms in LLMs.
Authors: Ariel Szabo, Uzy Hadad
Abstract: Remote Code Execution (RCE) exploits pose a significant threat to AI and ML systems, particularly in GPU-accelerated environments where the computational power of GPUs can be misused for malicious purposes. This paper focuses on RCE attacks leveraging deserialization vulnerabilities and custom layers, such as TensorFlow Lambda layers, which are often overlooked due to the complexity of monitoring GPU workloads. These vulnerabilities enable attackers to execute arbitrary code, blending malicious activity seamlessly into expected model behavior and exploiting GPUs for unauthorized tasks such as cryptocurrency mining. Unlike traditional CPU-based attacks, the parallel processing nature of GPUs and their high resource utilization make runtime detection exceptionally challenging. In this work, we provide a comprehensive examination of RCE exploits targeting GPUs, demonstrating an attack that utilizes these vulnerabilities to deploy a crypto miner on a GPU. We highlight the technical intricacies of such attacks, emphasize their potential for significant financial and computational costs, and propose strategies for mitigation. By shedding light on this underexplored attack vector, we aim to raise awareness and encourage the adoption of robust security measures in GPU-driven AI and ML systems, with an emphasis on static and model scanning as an easier way to detect exploits.
Authors: Junfeng Guo, Yiming Li, Ruibo Chen, Yihan Wu, Chenxi Liu, Yanshuo Chen, Heng Huang
Abstract: Large language models (LLMs) are increasingly integrated into real-world applications through retrieval-augmented generation (RAG) mechanisms to supplement their responses with up-to-date and domain-specific knowledge. However, the valuable and often proprietary nature of the knowledge bases used in RAG introduces the risk of unauthorized usage by adversaries. Existing methods that can be generalized as watermarking techniques to protect these knowledge bases typically involve poisoning attacks. However, these methods require to alter the results of verification samples (\eg, generating incorrect outputs), inevitably making them susceptible to anomaly detection and even introduce new security risks. To address these challenges, we propose \name{} for `harmless' copyright protection of knowledge bases. Instead of manipulating LLM's final output, \name{} implants distinct verification behaviors in the space of chain-of-thought (CoT) reasoning, maintaining the correctness of the final answer. Our method has three main stages: (1) \textbf{Generating CoTs}: For each verification question, we generate two CoTs, including a target CoT for building watermark behaviors; (2) \textbf{Optimizing Watermark Phrases and Target CoTs}: We optimize them to minimize retrieval errors under the black-box setting of suspicious LLM, ensuring that the watermarked verification queries activate the target CoTs without being activated in non-watermarked ones; (3) \textbf{Ownership Verification}: We exploit a pairwise Wilcoxon test to statistically verify whether a suspicious LLM is augmented with the protected knowledge base by comparing its responses to watermarked and benign verification queries. Our experiments on diverse benchmarks demonstrate that \name{} effectively protects knowledge bases against unauthorized usage while preserving the integrity and performance of the RAG.
Authors: Daniel Goldfarb, Paul Hand
Abstract: Autonomous machine learning systems that learn many tasks in sequence are prone to the catastrophic forgetting problem. Mathematical theory is needed in order to understand the extent of forgetting during continual learning. As a foundational step towards this goal, we study continual learning and catastrophic forgetting from a theoretical perspective in the simple setting of gradient descent with no explicit algorithmic mechanism to prevent forgetting. In this setting, we analytically demonstrate that overparameterization alone can mitigate forgetting in the context of a linear regression model. We consider a two-task setting motivated by permutation tasks, and show that as the overparameterization ratio becomes sufficiently high, a model trained on both tasks in sequence results in a low-risk estimator for the first task. As part of this work, we establish a non-asymptotic bound of the risk of a single linear regression task, which may be of independent interest to the field of double descent theory.
Authors: Xingli Fang, Jianwei Li, Varun Mulchandani, Jung-Eun Kim
Abstract: The capabilities of artificial intelligence systems have been advancing to a great extent, but these systems still struggle with failure modes, vulnerabilities, and biases. In this paper, we study the current state of the field, and present promising insights and perspectives regarding concerns that challenge the trustworthiness of AI models. In particular, this paper investigates the issues regarding three thrusts: safety, privacy, and bias, which hurt models' trustworthiness. For safety, we discuss safety alignment in the context of large language models, preventing them from generating toxic or harmful content. For bias, we focus on spurious biases that can mislead a network. Lastly, for privacy, we cover membership inference attacks in deep neural networks. The discussions addressed in this paper reflect our own experiments and observations.
Authors: R\'egnier Avice, Bernhard Haslhofer, Zhidong Li, Jianlong Zhou
Abstract: Attribution tags form the foundation of modern cryptoasset forensics. However, inconsistent or incorrect tags can mislead investigations and even result in false accusations. To address this issue, we propose a novel computational method based on Large Language Models (LLMs) to link attribution tags with well-defined knowledge graph concepts. We implemented this method in an end-to-end pipeline and conducted experiments showing that our approach outperforms baseline methods by up to 37.4% in F1-score across three publicly available attribution tag datasets. By integrating concept filtering and blocking procedures, we generate candidate sets containing five knowledge graph entities, achieving a recall of 93% without the need for labeled data. Additionally, we demonstrate that local LLM models can achieve F1-scores of 90%, comparable to remote models which achieve 94%. We also analyze the cost-performance trade-offs of various LLMs and prompt templates, showing that selecting the most cost-effective configuration can reduce costs by 90%, with only a 1% decrease in performance. Our method not only enhances attribution tag quality but also serves as a blueprint for fostering more reliable forensic evidence.
Authors: Yinghui Li, Jiayi Kuang, Haojing Huang, Zhikun Xu, Xinnian Liang, Yi Yu, Wenlian Lu, Yangning Li, Xiaoyu Tan, Chao Qu, Ying Shen, Hai-Tao Zheng, Philip S. Yu
Abstract: Leveraging mathematical Large Language Models (LLMs) for proof generation is a fundamental topic in LLMs research. We argue that the ability of current LLMs to prove statements largely depends on whether they have encountered the relevant proof process during training. This reliance limits their deeper understanding of mathematical theorems and related concepts. Inspired by the pedagogical method of "proof by counterexamples" commonly used in human mathematics education, our work aims to enhance LLMs' ability to conduct mathematical reasoning and proof through counterexamples. Specifically, we manually create a high-quality, university-level mathematical benchmark, CounterMATH, which requires LLMs to prove mathematical statements by providing counterexamples, thereby assessing their grasp of mathematical concepts. Additionally, we develop a data engineering framework to automatically obtain training data for further model improvement. Extensive experiments and detailed analyses demonstrate that CounterMATH is challenging, indicating that LLMs, such as OpenAI o1, have insufficient counterexample-driven proof capabilities. Moreover, our exploration into model training reveals that strengthening LLMs' counterexample-driven conceptual reasoning abilities is crucial for improving their overall mathematical capabilities. We believe that our work offers new perspectives on the community of mathematical LLMs.
Authors: Zhenxing Mi, Kuan-Chieh Wang, Guocheng Qian, Hanrong Ye, Runtao Liu, Sergey Tulyakov, Kfir Aberman, Dan Xu
Abstract: This paper presents ThinkDiff, a novel alignment paradigm that empowers text-to-image diffusion models with multimodal in-context understanding and reasoning capabilities by integrating the strengths of vision-language models (VLMs). Existing multimodal diffusion finetuning methods largely focus on pixel-level reconstruction rather than in-context reasoning, and are constrained by the complexity and limited availability of reasoning-based datasets. ThinkDiff addresses these challenges by leveraging vision-language training as a proxy task, aligning VLMs with the decoder of an encoder-decoder large language model (LLM) instead of a diffusion decoder. This proxy task builds on the observation that the $\textbf{LLM decoder}$ shares the same input feature space with $\textbf{diffusion decoders}$ that use the corresponding $\textbf{LLM encoder}$ for prompt embedding. As a result, aligning VLMs with diffusion decoders can be simplified through alignment with the LLM decoder. Without complex training and datasets, ThinkDiff effectively unleashes understanding, reasoning, and composing capabilities in diffusion models. Experiments demonstrate that ThinkDiff significantly improves accuracy from 19.2% to 46.3% on the challenging CoBSAT benchmark for multimodal in-context reasoning generation, with only 5 hours of training on 4 A100 GPUs. Additionally, ThinkDiff demonstrates exceptional performance in composing multiple images and texts into logically coherent images. Project page: https://mizhenxing.github.io/ThinkDiff.
Authors: Yang Gao, Hong Yang, Yizhi Chen, Junxian Wu, Peng Zhang, Haishuai Wang
Abstract: Graph Neural Architecture Search (GNAS) facilitates the automatic design of Graph Neural Networks (GNNs) tailored to specific downstream graph learning tasks. However, existing GNAS approaches often require manual adaptation to new graph search spaces, necessitating substantial code optimization and domain-specific knowledge. To address this challenge, we present LLM4GNAS, a toolkit for GNAS that leverages the generative capabilities of Large Language Models (LLMs). LLM4GNAS includes an algorithm library for graph neural architecture search algorithms based on LLMs, enabling the adaptation of GNAS methods to new search spaces through the modification of LLM prompts. This approach reduces the need for manual intervention in algorithm adaptation and code modification. The LLM4GNAS toolkit is extensible and robust, incorporating LLM-enhanced graph feature engineering, LLM-enhanced graph neural architecture search, and LLM-enhanced hyperparameter optimization. Experimental results indicate that LLM4GNAS outperforms existing GNAS methods on tasks involving both homogeneous and heterogeneous graphs.
Authors: Qinshuo Liu, Weiqin Zhao, Wei Huang, Yanwen Fang, Lequan Yu, Guodong Li
Abstract: The depth of neural networks is a critical factor for their capability, with deeper models often demonstrating superior performance. Motivated by this, significant efforts have been made to enhance layer aggregation - reusing information from previous layers to better extract features at the current layer, to improve the representational power of deep neural networks. However, previous works have primarily addressed this problem from a discrete-state perspective which is not suitable as the number of network layers grows. This paper novelly treats the outputs from layers as states of a continuous process and considers leveraging the state space model (SSM) to design the aggregation of layers in very deep neural networks. Moreover, inspired by its advancements in modeling long sequences, the Selective State Space Models (S6) is employed to design a new module called Selective State Space Model Layer Aggregation (S6LA). This module aims to combine traditional CNN or transformer architectures within a sequential framework, enhancing the representational capabilities of state-of-the-art vision networks. Extensive experiments show that S6LA delivers substantial improvements in both image classification and detection tasks, highlighting the potential of integrating SSMs with contemporary deep learning techniques.
Authors: Shao-Chien Lu, Chen-Chen Yeh, Hui-Lin Cho, Chun-Chieh Hsu, Tsai-Ling Hsu, Cheng-Han Wu, Timothy K. Shih, Yu-Cheng Lin
Abstract: The field of music generation using Large Language Models (LLMs) is evolving rapidly, yet existing music notation systems, such as MIDI, ABC Notation, and MusicXML, remain too complex for effective fine-tuning of LLMs. These formats are difficult for both machines and humans to interpret due to their variability and intricate structure. To address these challenges, we introduce YNote, a simplified music notation system that uses only four characters to represent a note and its pitch. YNote's fixed format ensures consistency, making it easy to read and more suitable for fine-tuning LLMs. In our experiments, we fine-tuned GPT-2 (124M) on a YNote-encoded dataset and achieved BLEU and ROUGE scores of 0.883 and 0.766, respectively. With just two notes as prompts, the model was able to generate coherent and stylistically relevant music. We believe YNote offers a practical alternative to existing music notations for machine learning applications and has the potential to significantly enhance the quality of music generation using LLMs.
Authors: Minyang Chen, Chenchen Feng, and Ran Cheng
Abstract: As a cornerstone in the Evolutionary Computation (EC) domain, Differential Evolution (DE) is known for its simplicity and effectiveness in handling challenging black-box optimization problems. While the advantages of DE are well-recognized, achieving peak performance heavily depends on its hyperparameters such as the mutation factor, crossover probability, and the selection of specific DE strategies. Traditional approaches to this hyperparameter dilemma have leaned towards parameter tuning or adaptive mechanisms. However, identifying the optimal settings tailored for specific problems remains a persistent challenge. In response, we introduce MetaDE, an approach that evolves DE's intrinsic hyperparameters and strategies using DE itself at a meta-level. A pivotal aspect of MetaDE is a specialized parameterization technique, which endows it with the capability to dynamically modify DE's parameters and strategies throughout the evolutionary process. To augment computational efficiency, MetaDE incorporates a design that leverages parallel processing through a GPU-accelerated computing framework. Within such a framework, DE is not just a solver but also an optimizer for its own configurations, thus streamlining the process of hyperparameter optimization and problem-solving into a cohesive and automated workflow. Extensive evaluations on the CEC2022 benchmark suite demonstrate MetaDE's promising performance. Moreover, when applied to robot control via evolutionary reinforcement learning, MetaDE also demonstrates promising performance. The source code of MetaDE is publicly accessible at: https://github.com/EMI-Group/metade.
Authors: Zihang Cheng, Huiping Zhuang, Chun Li, Xin Meng, Ming Li, Fei Richard Yu
Abstract: 3D Gaussian Splatting (3DGS) has been widely used in 3D reconstruction and 3D generation. Training to get a 3DGS scene often takes a lot of time and resources and even valuable inspiration. The increasing amount of 3DGS digital asset have brought great challenges to the copyright protection. However, it still lacks profound exploration targeted at 3DGS. In this paper, we propose a new framework X-SG$^2$S which can simultaneously watermark 1 to 3D messages while keeping the original 3DGS scene almost unchanged. Generally, we have a X-SG$^2$S injector for adding multi-modal messages simultaneously and an extractor for extract them. Specifically, we first split the watermarks into message patches in a fixed manner and sort the 3DGS points. A self-adaption gate is used to pick out suitable location for watermarking. Then use a XD(multi-dimension)-injection heads to add multi-modal messages into sorted 3DGS points. A learnable gate can recognize the location with extra messages and XD-extraction heads can restore hidden messages from the location recommended by the learnable gate. Extensive experiments demonstrated that the proposed X-SG$^2$S can effectively conceal multi modal messages without changing pretrained 3DGS pipeline or the original form of 3DGS parameters. Meanwhile, with simple and efficient model structure and high practicality, X-SG$^2$S still shows good performance in hiding and extracting multi-modal inner structured or unstructured messages. X-SG$^2$S is the first to unify 1 to 3D watermarking model for 3DGS and the first framework to add multi-modal watermarks simultaneous in one 3DGS which pave the wave for later researches.
Authors: Nathan Doum\`eche (LPSM, EDF R&D OSIRIS), Francis Bach (DI-ENS, SIERRA), \'Eloi Bedek (EDF R&D OSIRIS), G\'erard Biau (LPSM, IUF), Claire Boyer (LMO, IUF), Yannig Goude (EDF R&D OSIRIS, LMO)
Abstract: Time series forecasting presents unique challenges that limit the effectiveness of traditional machine learning algorithms. To address these limitations, various approaches have incorporated linear constraints into learning algorithms, such as generalized additive models and hierarchical forecasting. In this paper, we propose a unified framework for integrating and combining linear constraints in time series forecasting. Within this framework, we show that the exact minimizer of the constrained empirical risk can be computed efficiently using linear algebra alone. This approach allows for highly scalable implementations optimized for GPUs. We validate the proposed methodology through extensive benchmarking on real-world tasks, including electricity demand forecasting and tourism forecasting, achieving state-of-the-art performance.
Authors: Qin Liu, Fei Wang, Chaowei Xiao, Muhao Chen
Abstract: The emergence of vision language models (VLMs) comes with increased safety concerns, as the incorporation of multiple modalities heightens vulnerability to attacks. Although VLMs can be built upon LLMs that have textual safety alignment, it is easily undermined when the vision modality is integrated. We attribute this safety challenge to the modality gap, a separation of image and text in the shared representation space, which blurs the distinction between harmful and harmless queries that is evident in LLMs but weakened in VLMs. To avoid safety decay and fulfill the safety alignment gap, we propose VLM-Guard, an inference-time intervention strategy that leverages the LLM component of a VLM as supervision for the safety alignment of the VLM. VLM-Guard projects the representations of VLM into the subspace that is orthogonal to the safety steering direction that is extracted from the safety-aligned LLM. Experimental results on three malicious instruction settings show the effectiveness of VLM-Guard in safeguarding VLM and fulfilling the safety alignment gap between VLM and its LLM component.
Authors: Tim Beyer, Jan Schuchardt, Leo Schwinn, Stephan G\"unnemann
Abstract: Evaluating the robustness of LLMs to adversarial attacks is crucial for safe deployment, yet current red-teaming methods are often prohibitively expensive. We compare the ability of fast proxy metrics to predict the real-world robustness of an LLM against a simulated attacker ensemble. This allows us to estimate a model's robustness to computationally expensive attacks without requiring runs of the attacks themselves. Specifically, we consider gradient-descent-based embedding-space attacks, prefilling attacks, and direct prompting. Even though direct prompting in particular does not achieve high ASR, we find that it and embedding-space attacks can predict attack success rates well, achieving $r_p=0.87$ (linear) and $r_s=0.94$ (Spearman rank) correlations with the full attack ensemble while reducing computational cost by three orders of magnitude.
Authors: Jie Xu, Zihan Wu, Cong Wang, Xiaohua Jia
Abstract: Time-aware data valuation enhances training efficiency and model robustness, as early detection of harmful samples could prevent months of wasted computation. However, existing methods rely on model retraining or convergence assumptions or fail to capture long-term training dynamics. We propose LiveVal, an efficient time-aware data valuation method with three key designs: 1) seamless integration with SGD training for efficient data contribution monitoring; 2) reference-based valuation with normalization for reliable benchmark establishment; and 3) adaptive reference point selection for real-time updating with optimized memory usage. We establish theoretical guarantees for LiveVal's stability and prove that its valuations are bounded and directionally aligned with optimization progress. Extensive experiments demonstrate that LiveVal provides efficient data valuation across different modalities and model scales, achieving 180 speedup over traditional methods while maintaining robust detection performance.
Authors: Yong Li, Han Gao
Abstract: As artificial intelligence becomes more prevalent in our lives, people are enjoying the convenience it brings, but they are also facing hidden threats, such as data poisoning and ad- versarial attacks. These threats can have disastrous consequences for the application of artificial intelligence, especially for some applications that take effect immediately, such as autonomous driving and medical fields. Among these threats, backdoor attacks have left a deep impression on people with their concealment and simple deployment, making them a threat that cannot be ignored, however, in the process of deploying the backdoor model, the backdoor attack often has some reasons that make it unsatisfactory in real-world applications, such as jitter and brightness changes. Based on this, we propose a highly robust backdoor attack that shifts the target sample and combines it with itself to form a backdoor sample, the Displacement Backdoor Attack(DBA). Experimental results show that the DBA attack can resist data augmentation that simulates real-world differences, such as rotation and cropping.
Authors: Manvi Agarwal (IP Paris, LTCI, IDS), Changhong Wang (LTCI), Gael Richard (S2A, IDS)
Abstract: While music remains a challenging domain for generative models like Transformers, recent progress has been made by exploiting suitable musically-informed priors. One technique to leverage information about musical structure in Transformers is inserting such knowledge into the positional encoding (PE) module. However, Transformers carry a quadratic cost in sequence length. In this paper, we propose F-StrIPE, a structure-informed PE scheme that works in linear complexity. Using existing kernel approximation techniques based on random features, we show that F-StrIPE is a generalization of Stochastic Positional Encoding (SPE). We illustrate the empirical merits of F-StrIPE using melody harmonization for symbolic music.
Authors: Zhonghao Yang, Linye Lyu, Xuanhang Chang, Daojing He, YU LI
Abstract: In the rapidly evolving landscape of image generation, Latent Diffusion Models (LDMs) have emerged as powerful tools, enabling the creation of highly realistic images. However, this advancement raises significant concerns regarding copyright infringement and the potential misuse of generated content. Current watermarking techniques employed in LDMs often embed constant signals to the generated images that compromise their stealthiness, making them vulnerable to detection by malicious attackers. In this paper, we introduce SWA-LDM, a novel approach that enhances watermarking by randomizing the embedding process, effectively eliminating detectable patterns while preserving image quality and robustness. Our proposed watermark presence attack reveals the inherent vulnerabilities of existing latent-based watermarking methods, demonstrating how easily these can be exposed. Through comprehensive experiments, we validate that SWA-LDM not only fortifies watermark stealthiness but also maintains competitive performance in watermark robustness and visual fidelity. This work represents a pivotal step towards securing LDM-generated images against unauthorized use, ensuring both copyright protection and content integrity in an era where digital image authenticity is paramount.
Authors: Mohammad Baqar, Rajat Khanda
Abstract: Recent advancements in Generative AI have significantly improved the efficiency and adaptability of natural language processing (NLP) systems, particularly through Retrieval-Augmented Generation (RAG), Low-Rank Adaptation (LoRA), and Weight-Decomposed Low-Rank Adaptation (DoRA). RAG integrates external knowledge to enhance factual consistency in generative outputs, while LoRA enables parameter-efficient fine-tuning of large language models (LLMs). DoRA further refines this process by optimizing fine-tuning through adaptive parameter ranking and domain-aware weight adjustments, improving learning efficiency while maintaining inference performance. This paper presents a large-scale empirical evaluation of RAG, LoRA, and DoRA, with model fine-tuning and generation performance assessed on 20,000 FAQ-based queries, while the knowledge base spans 400,000 entries. The study analyzes key performance metrics such as accuracy, relevance, and inference latency. Experimental results demonstrate that DoRA achieves the highest accuracy (90.1%), relevance score (0.88), and lowest latency (110 ms per query), outperforming both LoRA and RAG in real-world, domain-specific generative AI applications. Furthermore, this study examines the trade-offs between fine-tuning efficiency, computational cost, and real-time adaptability across different models. Findings highlight RAG's effectiveness in knowledge grounding, LoRA's cost-efficient domain adaptation, and DoRA's ability to balance fine-tuning efficiency with model precision. These insights provide practical guidance for deploying AI-driven generative systems in accuracy-critical domains such as healthcare, finance, and legal services, ensuring scalability, reliability, and optimal performance in dynamic environments.
Authors: Anne Ouyang, Simon Guo, Simran Arora, Alex L. Zhang, William Hu, Christopher R\'e, Azalia Mirhoseini
Abstract: Efficient GPU kernels are crucial for building performant machine learning architectures, but writing them is a time-consuming challenge that requires significant expertise; therefore, we explore using language models (LMs) to automate kernel generation. We introduce KernelBench, an open-source framework for evaluating LMs' ability to write fast and correct kernels on a suite of 250 carefully selected PyTorch ML workloads. KernelBench represents a real-world engineering environment and making progress on the introduced benchmark directly translates to faster practical kernels. We introduce a new evaluation metric fast_p, which measures the percentage of generated kernels that are functionally correct and offer a speedup greater than an adjustable threshold p over baseline. Our experiments across various state-of-the-art models and test-time methods show that frontier reasoning models perform the best out of the box but still fall short overall, matching the PyTorch baseline in less than 20% of the cases. While we show that results can improve by leveraging execution and profiling feedback during iterative refinement, KernelBench remains a challenging benchmark, with its difficulty increasing as we raise speedup threshold p.
Authors: Venkatesh Sivaraman, Anika Vaishampayan, Xiaotong Li, Brian R Buck, Ziyong Ma, Richard D Boyce, Adam Perer
Abstract: Temporal predictive models have the potential to improve decisions in health care, public services, and other domains, yet they often fail to effectively support decision-makers. Prior literature shows that many misalignments between model behavior and decision-makers' expectations stem from issues of model specification, namely how, when, and for whom predictions are made. However, model specifications for predictive tasks are highly technical and difficult for non-data-scientist stakeholders to interpret and critique. To address this challenge we developed Tempo, an interactive system that helps data scientists and domain experts collaboratively iterate on model specifications. Using Tempo's simple yet precise temporal query language, data scientists can quickly prototype specifications with greater transparency about pre-processing choices. Moreover, domain experts can assess performance within data subgroups to validate that models behave as expected. Through three case studies, we demonstrate how Tempo helps multidisciplinary teams quickly prune infeasible specifications and identify more promising directions to explore.
Authors: Faruk Ahmed, Lin Yang, Tiam Jaroensri, Andrew Sellergren, Yossi Matias, Avinatan Hassidim, Greg S. Corrado, Dale R. Webster, Shravya Shetty, Shruthi Prabhakara, Yun Liu, Daniel Golden, Ellery Wulczyn, David F. Steiner
Abstract: The interpretation of histopathology cases underlies many important diagnostic and treatment decisions in medicine. Notably, this process typically requires pathologists to integrate and summarize findings across multiple slides per case. Existing vision-language capabilities in computational pathology have so far been largely limited to small regions of interest, larger regions at low magnification, or single whole-slide images (WSIs). This limits interpretation of findings that span multiple high-magnification regions across multiple WSIs. By making use of Gemini 1.5 Flash, a large multimodal model (LMM) with a 1-million token context window, we demonstrate the ability to generate bottom-line diagnoses from up to 40,000 768x768 pixel image patches from multiple WSIs at 10X magnification. This is the equivalent of up to 11 hours of video at 1 fps. Expert pathologist evaluations demonstrate that the generated report text is clinically accurate and equivalent to or preferred over the original reporting for 68% (95% CI: [60%, 76%]) of multi-slide examples with up to 5 slides. While performance decreased for examples with 6 or more slides, this study demonstrates the promise of leveraging the long-context capabilities of modern LMMs for the uniquely challenging task of medical report generation where each case can contain thousands of image patches.
Authors: Ali Younis, Erik B. Sudderth
Abstract: For challenging state estimation problems arising in domains like vision and robotics, particle-based representations attractively enable temporal reasoning about multiple posterior modes. Particle smoothers offer the potential for more accurate offline data analysis by propagating information both forward and backward in time, but have classically required human-engineered dynamics and observation models. Extending recent advances in discriminative training of particle filters, we develop a framework for low-variance propagation of gradients across long time sequences when training particle smoothers. Our "two-filter'' smoother integrates particle streams that are propagated forward and backward in time, while incorporating stratification and importance weights in the resampling step to provide low-variance gradient estimates for neural network dynamics and observation models. The resulting mixture density particle smoother is substantially more accurate than state-of-the-art particle filters, as well as search-based baselines, for city-scale global vehicle localization from real-world videos and maps.
Authors: Egor Cherepanov, Nikita Kachaev, Alexey K. Kovalev, Aleksandr I. Panov
Abstract: Memory is crucial for enabling agents to tackle complex tasks with temporal and spatial dependencies. While many reinforcement learning (RL) algorithms incorporate memory, the field lacks a universal benchmark to assess an agent's memory capabilities across diverse scenarios. This gap is particularly evident in tabletop robotic manipulation, where memory is essential for solving tasks with partial observability and ensuring robust performance, yet no standardized benchmarks exist. To address this, we introduce MIKASA (Memory-Intensive Skills Assessment Suite for Agents), a comprehensive benchmark for memory RL, with three key contributions: (1) we propose a comprehensive classification framework for memory-intensive RL tasks, (2) we collect MIKASA-Base - a unified benchmark that enables systematic evaluation of memory-enhanced agents across diverse scenarios, and (3) we develop MIKASA-Robo - a novel benchmark of 32 carefully designed memory-intensive tasks that assess memory capabilities in tabletop robotic manipulation. Our contributions establish a unified framework for advancing memory RL research, driving the development of more reliable systems for real-world applications. The code is available at https://sites.google.com/view/memorybenchrobots/.
Authors: Sumukha Udupa, Chongyang Shi, Jie Fu
Abstract: In this work, we investigate the synthesis of dynamic information releasing mechanisms, referred to as ''masks'', to minimize information leakage from a stochastic system to an external observer. Specifically, for a stochastic system, an observer aims to infer whether the final state of the system trajectory belongs to a set of secret states. The dynamic mask seeks to regulate sensor information in order to maximize the observer's uncertainty about the final state, a property known as final-state opacity. While existing supervisory control literature on dynamic masks primarily addresses qualitative opacity, we propose quantifying opacity in stochastic systems by conditional entropy, which is a measure of information leakage in information security. We then formulate a constrained optimization problem to synthesize a dynamic mask that maximizes final-state opacity under a total cost constraint on masking. To solve this constrained optimal dynamic mask synthesis problem, we develop a novel primal-dual policy gradient method. Additionally, we present a technique for computing the gradient of conditional entropy with respect to the masking policy parameters, leveraging observable operators in hidden Markov models. To demonstrate the effectiveness of our approach, we apply our method to an illustrative example and a stochastic grid world scenario, showing how our algorithm optimally enforces final-state opacity under cost constraints.
Authors: Danielle L. Ferreira, Bruno A. A. Nunes, Xuzhe Zhang, Laura Carretero Gomez, Maggie Fung, Ravi Soni
Abstract: Accurate morphometric assessment of cartilage-such as thickness/volume-via MRI is essential for monitoring knee osteoarthritis. Segmenting cartilage remains challenging and dependent on extensive expert-annotated datasets, which are heavily subjected to inter-reader variability. Recent advancements in Visual Foundational Models (VFM), especially memory-based approaches, offer opportunities for improving generalizability and robustness. This study introduces a deep learning (DL) method for cartilage and meniscus segmentation from 3D MRIs using interactive, memory-based VFMs. To improve spatial awareness and convergence, we incorporated a Hybrid Shuffling Strategy (HSS) during training and applied a segmentation mask propagation technique to enhance annotation efficiency. We trained four AI models-a CNN-based 3D-VNet, two automatic transformer-based models (SaMRI2D and SaMRI3D), and a transformer-based promptable memory-based VFM (SAMRI-2)-on 3D knee MRIs from 270 patients using public and internal datasets and evaluated on 57 external cases, including multi-radiologist annotations and different data acquisitions. Model performance was assessed against reference standards using Dice Score (DSC) and Intersection over Union (IoU), with additional morphometric evaluations to further quantify segmentation accuracy. SAMRI-2 model, trained with HSS, outperformed all other models, achieving an average DSC improvement of 5 points, with a peak improvement of 12 points for tibial cartilage. It also demonstrated the lowest cartilage thickness errors, reducing discrepancies by up to threefold. Notably, SAMRI-2 maintained high performance with as few as three user clicks per volume, reducing annotation effort while ensuring anatomical precision. This memory-based VFM with spatial awareness offers a novel approach for reliable AI-assisted knee MRI segmentation, advancing DL in musculoskeletal imaging.
Authors: Kevin Garcia, Juan Manuel Perez, Yifeng Gao
Abstract: Recently, there has been a significant advancement in designing Self-Supervised Learning (SSL) frameworks for time series data to reduce the dependency on data labels. Among these works, hierarchical contrastive learning-based SSL frameworks, which learn representations by contrasting data embeddings at multiple resolutions, have gained considerable attention. Due to their ability to gather more information, they exhibit better generalization in various downstream tasks. However, when the time series data length is significant long, the computational cost is often significantly higher than that of other SSL frameworks. In this paper, to address this challenge, we propose an efficient way to train hierarchical contrastive learning models. Inspired by the fact that each resolution's data embedding is highly dependent, we introduce importance-aware resolution selection based training framework to reduce the computational cost. In the experiment, we demonstrate that the proposed method significantly improves training time while preserving the original model's integrity in extensive time series classification performance evaluations. Our code could be found here, https://github.com/KEEBVIN/IARS
Authors: Aditya Dey, Jonas Kusch, Fadi Al Machot
Abstract: Long-term time series forecasting is critical in domains such as finance, economics, and energy, where accurate and reliable predictions over extended horizons drive strategic decision-making. Despite the progress in machine learning-based models, the impact of temporal noise in extended lookback windows remains underexplored, often degrading model performance and computational efficiency. In this paper, we propose a novel framework that addresses these challenges by integrating the Discrete Wavelet Transform (DWT) and Discrete Cosine Transform (DCT) to perform noise reduction and extract robust long-term features. These transformations enable the separation of meaningful temporal patterns from noise in both the time and frequency domains. To complement this, we introduce a lightweight low-rank linear prediction layer that not only reduces the influence of residual noise but also improves memory efficiency. Our approach demonstrates competitive robustness to noisy input, significantly reduces computational complexity, and achieves competitive or state-of-the-art forecasting performance across diverse benchmark datasets. Extensive experiments reveal that the proposed framework is particularly effective in scenarios with high noise levels or irregular patterns, making it well suited for real-world forecasting tasks. The code is available in https://github.com/forgee-master/HADL.
Authors: Hadi Zare, Mostafa Abbasi, Maryam Ahang, Homayoun Najjaran
Abstract: Purpose - In Business Process Management (BPM), accurate prediction of the next activities is vital for operational efficiency and decision-making. Current Artificial Intelligence (AI)/Machine Learning (ML) models struggle with the complexity and evolving nature of business process event logs, balancing accuracy and interpretability. This paper proposes an entropy-driven model selection approach and DAW-Transformer, which stands for Dynamic Attribute-Aware Transformer, to integrate all attributes with a dynamic window for better accuracy. Design/methodology/approach - This paper introduces a novel next-activity prediction approach that uses process entropy to assess the complexity of event logs and dynamically select the most suitable ML model. A new transformer-based architecture with multi-head attention and dynamic windowing mechanism, DAW-Transformer, is proposed to capture long-range dependencies and utilize all relevant event log attributes. Experiments were conducted on six public datasets, and the performance was evaluated with process entropy. Finding - The results demonstrate the effectiveness of the approach across these publicly available datasets. DAW-Transformer achieved superior performance, especially on high-entropy datasets such as Sepsis exceeding Limited window Multi-Transformers by 4.69% and a benchmark CNN-LSTM-SAtt model by 3.07%. For low-entropy datasets like Road Traffic Fine, simpler, more interpretable algorithms like Random Forest performed nearly as well as the more complex DAW-Transformer and offered better handling of imbalanced data and improved explainability. Originality/ value - This work's novelty lies in the proposed DAW-Transformer, with a dynamic window and considering all relevant attributes. Also, entropy-driven selection methods offer a robust, accurate, and interpretable solution for next-activity prediction.
Authors: Enzo Doyen, Amalia Todirascu
Abstract: Large language models (LLMs) have been shown to propagate and even amplify gender bias, in English and other languages, in specific or constrained contexts. However, no studies so far have focused on gender biases conveyed by LLMs' responses to generic instructions, especially with regard to masculine generics (MG). MG are a linguistic feature found in many gender-marked languages, denoting the use of the masculine gender as a "default" or supposedly neutral gender to refer to mixed group of men and women, or of a person whose gender is irrelevant or unknown. Numerous psycholinguistics studies have shown that MG are not neutral and induce gender bias. This work aims to analyze the use of MG by both proprietary and local LLMs in responses to generic instructions and evaluate their MG bias rate. We focus on French and create a human noun database from existing lexical resources. We filter existing French instruction datasets to retrieve generic instructions and analyze the responses of 6 different LLMs. Overall, we find that $\approx$39.5\% of LLMs' responses to generic instructions are MG-biased ($\approx$73.1\% across responses with human nouns). Our findings also reveal that LLMs are reluctant to using gender-fair language spontaneously.
Authors: Zeyu Jia, Alexander Rakhlin, Tengyang Xie
Abstract: As large language models have evolved, it has become crucial to distinguish between process supervision and outcome supervision -- two key reinforcement learning approaches to complex reasoning tasks. While process supervision offers intuitive advantages for long-term credit assignment, the precise relationship between these paradigms has remained an open question. Conventional wisdom suggests that outcome supervision is fundamentally more challenging due to the trajectory-level coverage problem, leading to significant investment in collecting fine-grained process supervision data. In this paper, we take steps towards resolving this debate. Our main theorem shows that, under standard data coverage assumptions, reinforcement learning through outcome supervision is no more statistically difficult than through process supervision, up to polynomial factors in horizon. At the core of this result lies the novel Change of Trajectory Measure Lemma -- a technical tool that bridges return-based trajectory measure and step-level distribution shift. Furthermore, for settings with access to a verifier or a rollout capability, we prove that any policy's advantage function can serve as an optimal process reward model, providing a direct connection between outcome and process supervision. These findings suggest that the empirically observed performance gap -- if any -- between outcome and process supervision likely stems from algorithmic limitations rather than inherent statistical difficulties, potentially transforming how we approach data collection and algorithm design for reinforcement learning.
Authors: Megh Shukla, Aziz Shameem, Mathieu Salzmann, Alexandre Alahi
Abstract: Deep heteroscedastic regression models the mean and covariance of the target distribution through neural networks. The challenge arises from heteroscedasticity, which implies that the covariance is sample dependent and is often unknown. Consequently, recent methods learn the covariance through unsupervised frameworks, which unfortunately yield a trade-off between computational complexity and accuracy. While this trade-off could be alleviated through supervision, obtaining labels for the covariance is non-trivial. Here, we study self-supervised covariance estimation in deep heteroscedastic regression. We address two questions: (1) How should we supervise the covariance assuming ground truth is available? (2) How can we obtain pseudo labels in the absence of the ground-truth? We address (1) by analysing two popular measures: the KL Divergence and the 2-Wasserstein distance. Subsequently, we derive an upper bound on the 2-Wasserstein distance between normal distributions with non-commutative covariances that is stable to optimize. We address (2) through a simple neighborhood based heuristic algorithm which results in surprisingly effective pseudo labels for the covariance. Our experiments over a wide range of synthetic and real datasets demonstrate that the proposed 2-Wasserstein bound coupled with pseudo label annotations results in a computationally cheaper yet accurate deep heteroscedastic regression.
Authors: Matthew Finlayson, Ilia Kulikov, Daneil M. Bikel, Barlas Oguz, Xilun Chen, Aasish Pappu
Abstract: Large language models (LLMs) often struggle with knowledge intensive NLP tasks, such as answering "Who won the latest World Cup?" because the knowledge they learn during training may be insufficient or outdated. Conditioning generation on retrieved documents -- a technique known as retrieval augmented generation (RAG) -- mitigates these shortcomings by allowing the model to leverage in-context information. Practitioners can improve LLM RAG performance by fine-tuning on retrieval-augmented instructions, but must beware that this can cause undesirable model behaviors like hallucinations. We attribute this degradation to the fact that the training data is likely to be out-of-distribution for the model and may suffer from quality issues, such as misalignment between retrievals and target responses (since retrievals are frequently added post-hoc). We propose a recipe for training RAG-enabled LLMs using self-generated demonstrations, thereby avoiding training on out-of-distribution text and integrating retrievals into the LLM responses. We evaluate our method on knowledge intensive question answering (QA) tasks and show that our method teaches LLMs to properly handle in-context retrievals and abstain from questions it will likely get wrong. Compared to conventional RA-IT methods, our method prevents model degradation in non-RAG settings while exhibiting superior QA performance.
Authors: Tejas Mirthipati
Abstract: Machine learning, particularly convolutional neural networks (CNNs), has shown promise in medical image analysis, especially for thoracic disease detection using chest X-ray images. In this study, we evaluate various CNN architectures, including binary classification, multi-label classification, and ResNet50 models, to address challenges like dataset imbalance, variations in image quality, and hidden biases. We introduce advanced preprocessing techniques such as principal component analysis (PCA) for image compression and propose a novel class-weighted loss function to mitigate imbalance issues. Our results highlight the potential of CNNs in medical imaging but emphasize that issues like unbalanced datasets and variations in image acquisition methods must be addressed for optimal model performance.
Authors: Kehua Chen, Jingping Jia
Abstract: Network evasion detection aims to distinguish whether the network flow comes from link layer exists network evasion threat, which is a means to disguise the data traffic on detection system by confusing the signature. Since the previous research works has all sorts of frauds, we propose a architecture with deep learning network to handle this problem. In this paper, we extract the critical information as key features from data frame and also specifically propose to use bidirectional long short-term memory (Bi-LSTM) neural network which shows an outstanding performance to trace the serial information, to encode both the past and future trait on the network flows. Furthermore we introduce a classifier named Softmax at the bottom of Bi-LSTM, holding a character to select the correct class. All experiments results shows that we can achieve a significant performance with a deep Bi-LSTM in network evasion detection and it's average accuracy reaches 96.1%.
Authors: Elan Markowitz, Anil Ramakrishna, Ninareh Mehrabi, Charith Peris, Rahul Gupta, Kai-Wei Chang, Aram Galstyan
Abstract: As the world changes, we need to be able to update our models and correct false information without costly retraining. Knowledge-based model editing enables precise modifications to the weights of large language models in order to modify the information encoded within. Recent approaches have seen success in enabling recall of edited information for thousands of edits at once. However, these approaches fail to produce edits that account for associated contextual information. We present K-Edit, an effective approach to generating contextually consistent knowledge edits. By using knowledge graphs, which maintain contextual consistency when an edge is edited, we are able to generate additional \textit{contextual edits} that ensure consistency of related information in the language model. Our experiments demonstrate significant improvements in multi-hop question answering while maintaining the general effectiveness and scalability of model edits.
Authors: Xuefeng Liu, Songhao Jiang, Bo Li, Rick Stevens
Abstract: Large Language Models (LLMs) employ three popular training approaches: Masked Language Models (MLM), Causal Language Models (CLM), and Sequence-to-Sequence Models (seq2seq). However, each approach has its strengths and limitations, and faces challenges in addressing specific tasks that require controllable and bidirectional generation, such as drug optimization. To address this challenge, inspired by the biological processes of growth and evolution, which involve the expansion, shrinking, and mutation of sequences, we introduce ControllableGPT. This initiative represents the first effort to combine the advantages of MLM, CLM, and seq2seq into a single unified, controllable GPT framework. It enables the precise management of specific locations and ranges within a sequence, allowing for expansion, reduction, or mutation over chosen or random lengths, while maintaining the integrity of any specified positions or subsequences. In this work, we designed ControllableGPT for drug optimization from the ground up, which included proposing the Causally Masked Seq2seq (CMS) objective, developing the training corpus, introducing a novel pre-training approach, and devising a unique generation process. We demonstrate the effectiveness and controllability of ControllableGPT by conducting experiments on drug optimization tasks for both viral and cancer benchmarks, surpassing competing baselines.
Authors: Illia Polosukhin, Alex Skidanov
Abstract: We present a mechanism that for a network of participants allows one participant of the network (Alice) to request some data from another participant (Bob) and either receive a response from Bob within a known-in-advance, bounded time b, or receive a proof that at least one edge on the way to Bob was broken within b, or receive a streaming payment proportional to time passed beyond b during which neither was received. This mechanism allows for building downstream applications that require provable responses from other participants, such as decentralized storage solutions, decentralized AI agents, and more.
Authors: Yate Ge, Meiying Li, Xipeng Huang, Yuanda Hu, Qi Wang, Xiaohua Sun, Weiwei Guo
Abstract: This work investigates the integration of generative visual aids in human-robot task communication. We developed GenComUI, a system powered by large language models that dynamically generates contextual visual aids (such as map annotations, path indicators, and animations) to support verbal task communication and facilitate the generation of customized task programs for the robot. This system was informed by a formative study that examined how humans use external visual tools to assist verbal communication in spatial tasks. To evaluate its effectiveness, we conducted a user experiment (n = 20) comparing GenComUI with a voice-only baseline. The results demonstrate that generative visual aids, through both qualitative and quantitative analysis, enhance verbal task communication by providing continuous visual feedback, thus promoting natural and effective human-robot communication. Additionally, the study offers a set of design implications, emphasizing how dynamically generated visual aids can serve as an effective communication medium in human-robot interaction. These findings underscore the potential of generative visual aids to inform the design of more intuitive and effective human-robot communication, particularly for complex communication scenarios in human-robot interaction and LLM-based end-user development.
Authors: Leisheng Yu, Yanxiao Cai, Minxing Zhang, Xia Hu
Abstract: The burgeoning volume of electronic health records (EHRs) has enabled deep learning models to excel in predictive healthcare. However, for high-stakes applications such as diagnosis prediction, model interpretability remains paramount. Existing deep learning diagnosis prediction models with intrinsic interpretability often assign attention weights to every past diagnosis or hospital visit, providing explanations lacking flexibility and succinctness. In this paper, we introduce SHy, a self-explaining hypergraph neural network model, designed to offer personalized, concise and faithful explanations that allow for interventions from clinical experts. By modeling each patient as a unique hypergraph and employing a message-passing mechanism, SHy captures higher-order disease interactions and extracts distinct temporal phenotypes as personalized explanations. It also addresses the incompleteness of the EHR data by accounting for essential false negatives in the original diagnosis record. A qualitative case study and extensive quantitative evaluations on two real-world EHR datasets demonstrate the superior predictive performance and interpretability of SHy over existing state-of-the-art models.
Authors: Ahmad Chaddad, Yihang Wu, Yuchen Jiang, Ahmed Bouridane, Christian Desrosiers
Abstract: Traditional machine learning assumes that training and test sets are derived from the same distribution; however, this assumption does not always hold in practical applications. This distribution disparity can lead to severe performance drops when the trained model is used in new data sets. Domain adaptation (DA) is a machine learning technique that aims to address this problem by reducing the differences between domains. This paper presents simulation-based algorithms of recent DA techniques, mainly related to unsupervised domain adaptation (UDA), where labels are available only in the source domain. Our study compares these techniques with public data sets and diverse characteristics, highlighting their respective strengths and drawbacks. For example, Safe Self-Refinement for Transformer-based DA (SSRT) achieved the highest accuracy (91.6\%) in the office-31 data set during our simulations, however, the accuracy dropped to 72.4\% in the Office-Home data set when using limited batch sizes. In addition to improving the reader's comprehension of recent techniques in DA, our study also highlights challenges and upcoming directions for research in this domain. The codes are available at https://github.com/AIPMLab/Domain_Adaptation.
Authors: Haiquan Qiu, You Wu, Quanming Yao
Abstract: Model merging is a critical technique for combining the capabilities of multiple fine-tuned models without requiring additional training. While existing methods treat parameters as vectors, they overlook the intrinsic structure of linear transformation matrices - the core components that comprise the majority of model parameters. These matrices are fundamental to neural networks, mapping input representations to output features through linear combinations. Motivated by the linear representation hypothesis, we introduce task matrix and propose to Superpose Features from Task Matrix (SFTM), a novel approach that superposes features from individual task models into a merged model. SFTM employs singular value decomposition to identify feature bases of linear transformation matrices and solves a linear system to optimally combine them while preserving input-output mappings from individual task models. Extensive experiments on vision transformers and language models demonstrate that our method consistently outperforms existing methods, achieving superior performance and enhanced out-of-distribution generalization.
Authors: George Applegarth, Christian Weatherstone, Maximilian Hollingsworth, Henry Middlebrook, Marcus Irvin
Abstract: Contextual memory integration remains a high challenge in the development of language models, particularly in tasks that require maintaining coherence over extended sequences. Traditional approaches, such as self-attention mechanisms and memory-augmented architectures, often prioritize short-term dependencies, leading to fragmentation and inconsistency in long-range contextual understanding. Inspired by principles of synaptic plasticity observed in biological neural systems, a novel mechanism, Synaptic Resonance, is introduced to dynamically reinforce relevant memory pathways during training and inference. Unlike static memory representations, this mechanism continuously adjusts synaptic weight matrices based on contextual relevance, allowing for improved information retention without excessive computational overhead. Evaluations conducted on an open-source language model demonstrate reductions in perplexity, enhancements in contextual coherence, and increased robustness against input noise, highlighting the effectiveness of reinforcement-driven memory modulation. Comparative analysis against baseline models further reveals that the proposed approach achieves higher memory retention efficiency while maintaining computational feasibility. The architectural modifications integrate seamlessly into existing transformer-based frameworks, ensuring stable convergence and efficient inference without sacrificing scalability. Applications benefiting from improved long-term contextual consistency, such as dialogue systems and document summarization, stand to gain from this approach. Empirical findings suggest that dynamically reinforced memory pathways offer a promising alternative to conventional memory mechanisms, addressing longstanding limitations in extended sequence modeling.
Authors: Mingyang Zhao, Gaofeng Meng, Dong-Ming Yan
Abstract: Non-rigid alignment of point clouds is crucial for scene understanding, reconstruction, and various computer vision and robotics tasks. Recent advancements in implicit deformation networks for non-rigid registration have significantly reduced the reliance on large amounts of annotated training data. However, existing state-of-the-art methods still face challenges in handling occlusion scenarios. To address this issue, this paper introduces an innovative unsupervised method called Occlusion-Aware Registration (OAR) for non-rigidly aligning point clouds. The key innovation of our method lies in the utilization of the adaptive correntropy function as a localized similarity measure, enabling us to treat individual points distinctly. In contrast to previous approaches that solely minimize overall deviations between two shapes, we combine unsupervised implicit neural representations with the maximum correntropy criterion to optimize the deformation of unoccluded regions. This effectively avoids collapsed, tearing, and other physically implausible results. Moreover, we present a theoretical analysis and establish the relationship between the maximum correntropy criterion and the commonly used Chamfer distance, highlighting that the correntropy-induced metric can be served as a more universal measure for point cloud analysis. Additionally, we introduce locally linear reconstruction to ensure that regions lacking correspondences between shapes still undergo physically natural deformations. Our method achieves superior or competitive performance compared to existing approaches, particularly when dealing with occluded geometries. We also demonstrate the versatility of our method in challenging tasks such as large deformations, shape interpolation, and shape completion under occlusion disturbances.
Authors: Xu Shen, Yixin Liu, Yili Wang, Rui Miao, Yiwei Dai, Shirui Pan, Xin Wang
Abstract: Out-of-distribution (OOD) generalization has emerged as a critical challenge in graph learning, as real-world graph data often exhibit diverse and shifting environments that traditional models fail to generalize across. A promising solution to address this issue is graph invariant learning (GIL), which aims to learn invariant representations by disentangling label-correlated invariant subgraphs from environment-specific subgraphs. However, existing GIL methods face two major challenges: (1) the difficulty of capturing and modeling diverse environments in graph data, and (2) the semantic cliff, where invariant subgraphs from different classes are difficult to distinguish, leading to poor class separability and increased misclassifications. To tackle these challenges, we propose a novel method termed Multi-Prototype Hyperspherical Invariant Learning (MPHIL), which introduces two key innovations: (1) hyperspherical invariant representation extraction, enabling robust and highly discriminative hyperspherical invariant feature extraction, and (2) multi-prototype hyperspherical classification, which employs class prototypes as intermediate variables to eliminate the need for explicit environment modeling in GIL and mitigate the semantic cliff issue. Derived from the theoretical framework of GIL, we introduce two novel objective functions: the invariant prototype matching loss to ensure samples are matched to the correct class prototypes, and the prototype separation loss to increase the distinction between prototypes of different classes in the hyperspherical space. Extensive experiments on 11 OOD generalization benchmark datasets demonstrate that MPHIL achieves state-of-the-art performance, significantly outperforming existing methods across graph data from various domains and with different distribution shifts.
Authors: Jiarui Jin, Haoyu Wang, Hongyan Li, Jun Li, Jiahui Pan, Shenda Hong
Abstract: Electrocardiogram (ECG) is essential for the clinical diagnosis of arrhythmias and other heart diseases, but deep learning methods based on ECG often face limitations due to the need for high-quality annotations. Although previous ECG self-supervised learning (eSSL) methods have made significant progress in representation learning from unannotated ECG data, they typically treat ECG signals as ordinary time-series data, segmenting the signals using fixed-size and fixed-step time windows, which often ignore the form and rhythm characteristics and latent semantic relationships in ECG signals. In this work, we introduce a novel perspective on ECG signals, treating heartbeats as words and rhythms as sentences. Based on this perspective, we first designed the QRS-Tokenizer, which generates semantically meaningful ECG sentences from the raw ECG signals. Building on these, we then propose HeartLang, a novel self-supervised learning framework for ECG language processing, learning general representations at form and rhythm levels. Additionally, we construct the largest heartbeat-based ECG vocabulary to date, which will further advance the development of ECG language processing. We evaluated HeartLang across six public ECG datasets, where it demonstrated robust competitiveness against other eSSL methods. Our data and code are publicly available at https://github.com/PKUDigitalHealth/HeartLang.
Authors: Qiujie Xie, Qingqiu Li, Zhuohao Yu, Yuejie Zhang, Yue Zhang, Linyi Yang
Abstract: As LLM-as-a-Judge emerges as a new paradigm for assessing large language models (LLMs), concerns have been raised regarding the alignment, bias, and stability of LLM evaluators. While substantial work has focused on alignment and bias, little research has concentrated on the stability of LLM evaluators. In this paper, we conduct extensive experiments involving 9 widely used LLM evaluators across 2 different evaluation settings to investigate the uncertainty in model-based LLM evaluations. We pinpoint that LLM evaluators exhibit varying uncertainty based on model families and sizes. With careful comparative analyses, we find that employing special prompting strategies, whether during inference or post-training, can alleviate evaluation uncertainty to some extent. By utilizing uncertainty to enhance LLM's reliability and detection capability in Out-Of-Distribution (OOD) data, we further fine-tune an uncertainty-aware LLM evaluator named ConfiLM using a human-annotated fine-tuning set and assess ConfiLM's OOD evaluation ability on a manually designed test set sourced from the 2024 Olympics. Experimental results demonstrate that incorporating uncertainty as additional information during the fine-tuning phase can largely improve the model's evaluation performance in OOD scenarios. The code and data are released at: https://github.com/hasakiXie123/LLM-Evaluator-Uncertainty.
URLs: https://github.com/hasakiXie123/LLM-Evaluator-Uncertainty.
Authors: Jinouwen Zhang, Junjie Ren, Aobo Yang, Yan Lu, Lu Chen, Hairun Xie, Jing Wang, Miao Zhang, Wanli Ouyang, Shixiang Tang
Abstract: Aircraft manufacturing is the jewel in the crown of industry, among which generating high-fidelity airfoil geometries with controllable and editable representations remains a fundamental challenge. While existing deep-learning-based methods rely on predefined parametric function families, e.g., B\'ezier curves and discrete point-based representations, they suffer from inherent trade-offs between expressiveness and resolution flexibility. To tackle this challenge, we introduce FuncGenFoil, a novel function-space generative model that directly learns functional airfoil geometries. Our method inherits both the advantages of arbitrary resolution sampling and the smoothness of parametric functions, as well as the strong expressiveness of discrete point-based functions. Empirical evaluations on the AFBench dataset demonstrate that FuncGenFoil improves upon state-of-the-art methods in airfoil generation by achieving a relative -74.4 label error reduction and +23.2 diversity increase on the AF-200K dataset. Our results highlight the advantages of function-space modeling for aerodynamic shape optimization, offering a powerful and flexible framework for high-fidelity airfoil design. Our code will be released.
Authors: Sanggeon Yun, Ryozo Masukawa, Hanning Chen, SungHeon Jeong, Wenjun Huang, Arghavan Rezvani, Minhyoung Na, Yoshiki Yamaguchi, Mohsen Imani
Abstract: The escalating challenges of managing vast sensor-generated data, particularly in audio applications, necessitate innovative solutions. Current systems face significant computational and storage demands, especially in real-time applications like gunshot detection systems (GSDS), and the proliferation of edge sensors exacerbates these issues. This paper proposes a groundbreaking approach with a near-sensor model tailored for intelligent audio-sensing frameworks. Utilizing a Fast Fourier Transform (FFT) module, convolutional neural network (CNN) layers, and HyperDimensional Computing (HDC), our model excels in low-energy, rapid inference, and online learning. It is highly adaptable for efficient ASIC design implementation, offering superior energy efficiency compared to conventional embedded CPUs or GPUs, and is compatible with the trend of shrinking microphone sensor sizes. Comprehensive evaluations at both software and hardware levels underscore the model's efficacy. Software assessments through detailed ROC curve analysis revealed a delicate balance between energy conservation and quality loss, achieving up to 82.1% energy savings with only 1.39% quality loss. Hardware evaluations highlight the model's commendable energy efficiency when implemented via ASIC design, especially with the Google Edge TPU, showcasing its superiority over prevalent embedded CPUs and GPUs.
Authors: Xiliang Yang, Shenyang Deng, Shicong Liu, Yuanchi Suo, Wing. W. Y NG, Jianjun Zhang
Abstract: Data augmentation is an important technique in training deep neural networks as it enhances their ability to generalize and remain robust. While data augmentation is commonly used to expand the sample size and act as a consistency regularization term, there is a lack of research on the relationship between them. To address this gap, this paper introduces a more comprehensive mathematical framework for data augmentation. Through this framework, we establish that the expected risk of the shifted population is the sum of the original population risk and a gap term, which can be interpreted as a consistency regularization term. The paper also provides a theoretical understanding of this gap, highlighting its negative effects on the early stages of training. We also propose a method to mitigate these effects. To validate our approach, we conducted experiments using same data augmentation techniques and computing resources under several scenarios, including standard training, out-of-distribution, and imbalanced classification. The results demonstrate that our methods surpass compared methods under all scenarios in terms of generalization ability and convergence stability. We provide our code implementation at the following link: https://github.com/ydlsfhll/ASPR.
Authors: Fei Yang
Abstract: Transformer-based embedding methods have dominated the field of sentence representation in recent years. Although they have achieved remarkable performance on NLP missions, such as semantic textual similarity (STS) tasks, their black-box nature and large-data-driven training style have raised concerns, including issues related to bias, trust, and safety. Many efforts have been made to improve the interpretability of embedding models, but these problems have not been fundamentally resolved. To achieve inherent interpretability, we propose a purely white-box and human-like sentence representation network, PropNet. Inspired by findings from cognitive science, PropNet constructs a hierarchical network based on the propositions contained in a sentence. While experiments indicate that PropNet has a significant gap compared to state-of-the-art (SOTA) embedding models in STS tasks, case studies reveal substantial room for improvement. Additionally, PropNet enables us to analyze and understand the human cognitive processes underlying STS benchmarks.
Authors: Mauricio Tec, Guojun Xiong, Haichuan Wang, Francesca Dominici, Milind Tambe
Abstract: Deep Reinforcement Learning (RL) is remarkably effective in addressing sequential resource allocation problems in domains such as healthcare, public policy, and resource management. However, deep RL policies often lack transparency and adaptability, challenging their deployment alongside human decision-makers. In contrast, Language Agents, powered by large language models (LLMs), provide human-understandable reasoning but may struggle with effective decision making. To bridge this gap, we propose Rule-Bottleneck Reinforcement Learning (RBRL), a novel framework that jointly optimizes decision and explanations. At each step, RBRL generates candidate rules with an LLM, selects among them using an attention-based RL policy, and determines the environment action with an explanation via chain-of-thought reasoning. The RL rule selection is optimized using the environment rewards and an explainability metric judged by the LLM. Evaluations in real-world scenarios highlight RBRL's competitive performance with deep RL and efficiency gains over LLM fine-tuning. A survey further confirms the enhanced quality of its explanations.
Authors: Zehua Liu, Han Wu, Yuxuan Yao, Ruifeng She, Xiongwei Han, Tao Zhong, Mingxuan Yuan
Abstract: While most current approaches rely on further training techniques, such as fine-tuning or reinforcement learning, to enhance model capacities, model merging stands out for its ability of improving models without requiring any additional training. In this paper, we propose a unified framework for model merging based on low-rank estimation of task vectors without the need for access to the base model, named \textsc{LoRE-Merging}. Our approach is motivated by the observation that task vectors from fine-tuned models frequently exhibit a limited number of dominant singular values, making low-rank estimations less prone to interference. We implement the method by formulating the merging problem as an optimization problem. Extensive empirical experiments demonstrate the effectiveness of our framework in mitigating interference and preserving task-specific information, thereby advancing the state-of-the-art performance in model merging techniques.
Authors: Shih-Hsuan Chiu, Ya-Wen Teng, De-Nian Yang, Ming-Syan Chen
Abstract: Community detection is a cornerstone problem in social network analysis (SNA), aimed at identifying cohesive communities with minimal external links. However, the rise of generative AI and Metaverse introduce complexities by creating hybrid human-AI social networks (denoted by HASNs), where traditional methods fall short, especially in human-centric settings. This paper introduces a novel community detection problem in HASNs (denoted by MetaCD), which seeks to enhance human connectivity within communities while reducing the presence of AI nodes. Effective processing of MetaCD poses challenges due to the delicate trade-off between excluding certain AI nodes and maintaining community structure. To address this, we propose CUSA, an innovative framework incorporating AI-aware clustering techniques that navigate this trade-off by selectively retaining AI nodes that contribute to community integrity. Furthermore, given the scarcity of real-world HASNs, we devise four strategies for synthesizing these networks under various hypothetical scenarios. Empirical evaluations on real social networks, reconfigured as HASNs, demonstrate the effectiveness and practicality of our approach compared to traditional non-deep learning and graph neural network (GNN)-based methods.
Authors: Guofu Xie, Xiao Zhang, Ting Yao, Yunsheng Shi
Abstract: User information needs are often highly diverse and varied. A key challenge in current research is how to achieve controllable multi-objective generation while enabling rapid adaptation to accommodate diverse user demands during test time. Existing solutions, such as Rewarded Soup, focus on merging language models individually tuned on single objectives. While easy to implement and widely used, these approaches face limitations in achieving optimal performance due to their disregard for the impacts of competing objectives on model tuning. To address this issue, we propose Bone Soup, a novel model merging approach that first seeks a series of backbone models by considering the impacts of multiple objectives and then makes the soup (i.e., merge the backbone models). Specifically, Bone Soup begins by training multiple backbone models for different objectives using multi-objective reinforcement learning. Each backbone model is guided by a combination of backbone reward signals. To ensure that these models are optimal for the Pareto front, the backbone rewards are crafted by combining standard reward functions into basis vectors, which can then be modified through a rule-based construction method. Bone Soup leverages a symmetric circulant matrix mapping to generate the merging coefficients, which are used to merge the backbone models according to user preferences. Extensive experimental results demonstrate that Bone Soup exhibits strong controllability and Pareto optimality in controllable multi-objective generation, providing a more effective and efficient approach to addressing diverse user needs at test time.
Authors: Sandra Schaftner
Abstract: Current research highlights the great potential of Large Language Models (LLMs) for constructing Scholarly Knowledge Graphs (SKGs). One particularly complex step in this process is relation extraction, aimed at identifying suitable properties to describe the content of research. This study builds directly on previous research of three Open Research Knowledge Graph (ORKG) team members who assessed the readiness of LLMs such as GPT-3.5, Llama 2, and Mistral for property extraction in scientific literature. Given the moderate performance observed, the previous work concluded that fine-tuning is needed to improve these models' alignment with scientific tasks and their emulation of human expertise. Expanding on this prior experiment, this study evaluates the impact of advanced prompt engineering techniques and demonstrates that these techniques can highly significantly enhance the results. Additionally, this study extends the property extraction process to include property matching to existing ORKG properties, which are retrieved via the API. The evaluation reveals that results generated through advanced prompt engineering achieve a higher proportion of matches with ORKG properties, further emphasizing the enhanced alignment achieved. Moreover, this lays the groundwork for addressing challenges such as the inconsistency of ORKG properties, an issue highlighted in prior studies. By assigning unique URIs and using standardized terminology, this work increases the consistency of the properties, fulfilling a crucial aspect of Linked Data and FAIR principles - core commitments of ORKG. This, in turn, significantly enhances the applicability of ORKG content for subsequent tasks such as comparisons of research publications. Finally, the study concludes with recommendations for future improvements in the overall property extraction process.
Authors: Zhipeng Liu, Peibo Duan, Mingyang Geng, Bin Zhang
Abstract: Stock trend prediction involves forecasting the future price movements by analyzing historical data and various market indicators. With the advancement of machine learning, graph neural networks (GNNs) have been extensively employed in stock prediction due to their powerful capability to capture spatiotemporal dependencies of stocks. However, despite the efforts of various GNN stock predictors to enhance predictive performance, the improvements remain limited, as they focus solely on analyzing historical spatiotemporal dependencies, overlooking the correlation between historical and future patterns. In this study, we propose a novel distillation-based future-aware GNN framework (DishFT-GNN) for stock trend prediction. Specifically, DishFT-GNN trains a teacher model and a student model, iteratively. The teacher model learns to capture the correlation between distribution shifts of historical and future data, which is then utilized as intermediate supervision to guide the student model to learn future-aware spatiotemporal embeddings for accurate prediction. Through extensive experiments on two real-world datasets, we verify the state-of-the-art performance of DishFT-GNN.
Authors: Jie Xu, Zihan Wu
Abstract: Existing methods for measuring training sample influence on models only provide static, overall measurements, overlooking how sample influence changes during training. We propose Dynamic Influence Tracker (DIT), which captures the time-varying sample influence across arbitrary time windows during training. DIT offers three key insights: 1) Samples show different time-varying influence patterns, with some samples important in the early training stage while others become important later. 2) Sample influences show a weak correlation between early and late stages, demonstrating that the model undergoes distinct learning phases with shifting priorities. 3) Analyzing influence during the convergence period provides more efficient and accurate detection of corrupted samples than full-training analysis. Supported by theoretical guarantees without assuming loss convexity or model convergence, DIT significantly outperforms existing methods, achieving up to 0.99 correlation with ground truth and above 98\% accuracy in detecting corrupted samples in complex architectures.
Authors: Li Wang, Zheng Li, Xuhong Zhang, Shouling Ji, Shanqing Guo
Abstract: DeepFakes pose a significant threat to our society. One representative DeepFake application is face-swapping, which replaces the identity in a facial image with that of a victim. Although existing methods partially mitigate these risks by degrading the quality of swapped images, they often fail to disrupt the identity transformation effectively. To fill this gap, we propose FaceSwapGuard (FSG), a novel black-box defense mechanism against deepfake face-swapping threats. Specifically, FSG introduces imperceptible perturbations to a user's facial image, disrupting the features extracted by identity encoders. When shared online, these perturbed images mislead face-swapping techniques, causing them to generate facial images with identities significantly different from the original user. Extensive experiments demonstrate the effectiveness of FSG against multiple face-swapping techniques, reducing the face match rate from 90\% (without defense) to below 10\%. Both qualitative and quantitative studies further confirm its ability to confuse human perception, highlighting its practical utility. Additionally, we investigate key factors that may influence FSG and evaluate its robustness against various adaptive adversaries.
Authors: Kefan Li, Hongyue Yu, Tingyu Guo, Shijie Cao, Yuan Yuan
Abstract: Large Language Models (LLMs) have shown remarkable performance in automated code generation. However, existing approaches often rely heavily on pre-defined test cases, which become impractical in scenarios where such cases are unavailable. While prior works explore filtering techniques between programs and test cases, they overlook the refinement of test cases. To address this limitation, we introduce CoCoEvo, a novel LLM-based co-evolution framework that simultaneously evolves programs and test cases. CoCoEvo eliminates the dependency on pre-defined test cases by generating both programs and test cases directly from natural language problem descriptions and function headers. The framework employs specialized evolutionary operators, including LLM-based crossover and mutation operators for program evolution, along with a test case generation operator for test case evolution. Additionally, we propose optimization strategies such as a crossover rate scheduler to balance exploration and convergence, and a multi-objective optimization method for test case selection. Experimental results on multiple state-of-the-art LLMs demonstrate that CoCoEvo surpasses existing methods, achieving state-of-the-art performance in automated code generation and testing. These results underscore the potential of co-evolutionary techniques in advancing the field of automated programming.
Authors: Li Wang, Wenyu Chen, Zheng Li, Shanqing Guo
Abstract: The rapid advancement of generative models has led to the proliferation of highly realistic AI-generated images, posing significant challenges for detection methods to generalize across diverse and evolving generative techniques. Existing approaches often fail to adapt to unknown models without costly retraining, limiting their practicability. To fill this gap, we propose Post-hoc Distribution Alignment (PDA), a novel approach for the generalizable detection for AI-generated images. The key idea is to use the known generative model to regenerate undifferentiated test images. This process aligns the distributions of the re-generated real images with the known fake images, enabling effective distinction from unknown fake images. PDA employs a two-step detection framework: 1) evaluating whether a test image aligns with the known fake distribution based on deep k-nearest neighbor (KNN) distance, and 2) re-generating test images using known generative models to create pseudo-fake images for further classification. This alignment strategy allows PDA to effectively detect fake images without relying on unseen data or requiring retraining. Extensive experiments demonstrate the superiority of PDA, achieving 96.73\% average accuracy across six state-of-the-art generative models, including GANs, diffusion models, and text-to-image models, and improving by 16.07\% over the best baseline. Through t-SNE visualizations and KNN distance analysis, we provide insights into PDA's effectiveness in separating real and fake images. Our work provides a flexible and effective solution for real-world fake image detection, advancing the generalization ability of detection systems.
Authors: Mingqian Ma, Guoqing Liu, Chuan Cao, Pan Deng, Tri Dao, Albert Gu, Peiran Jin, Zhao Yang, Yingce Xia, Renqian Luo, Pipi Hu, Zun Wang, Yuan-Jyue Chen, Haiguang Liu, Tao Qin
Abstract: Advances in natural language processing and large language models have sparked growing interest in modeling DNA, often referred to as the "language of life". However, DNA modeling poses unique challenges. First, it requires the ability to process ultra-long DNA sequences while preserving single-nucleotide resolution, as individual nucleotides play a critical role in DNA function. Second, success in this domain requires excelling at both generative and understanding tasks: generative tasks hold potential for therapeutic and industrial applications, while understanding tasks provide crucial insights into biological mechanisms and diseases. To address these challenges, we propose HybriDNA, a decoder-only DNA language model that incorporates a hybrid Transformer-Mamba2 architecture, seamlessly integrating the strengths of attention mechanisms with selective state-space models. This hybrid design enables HybriDNA to efficiently process DNA sequences up to 131kb in length with single-nucleotide resolution. HybriDNA achieves state-of-the-art performance across 33 DNA understanding datasets curated from the BEND, GUE, and LRB benchmarks, and demonstrates exceptional capability in generating synthetic cis-regulatory elements (CREs) with desired properties. Furthermore, we show that HybriDNA adheres to expected scaling laws, with performance improving consistently as the model scales from 300M to 3B and 7B parameters. These findings underscore HybriDNA's versatility and its potential to advance DNA research and applications, paving the way for innovations in understanding and engineering the "language of life".
Authors: Shaoxuan Xu, Menglu Cui, Chengxiang Huang, Hongfa Wang, DiHu
Abstract: Multimodal learning has gained attention for its capacity to integrate information from different modalities. However, it is often hindered by the multimodal imbalance problem, where certain modality dominates while others remain underutilized. Although recent studies have proposed various methods to alleviate this problem, they lack comprehensive and fair comparisons. In this paper, we systematically categorize various mainstream multimodal imbalance algorithms into four groups based on the strategies they employ to mitigate imbalance. To facilitate a comprehensive evaluation of these methods, we introduce BalanceBenchmark, a benchmark including multiple widely used multidimensional datasets and evaluation metrics from three perspectives: performance, imbalance degree, and complexity. To ensure fair comparisons, we have developed a modular and extensible toolkit that standardizes the experimental workflow across different methods. Based on the experiments using BalanceBenchmark, we have identified several key insights into the characteristics and advantages of different method groups in terms of performance, balance degree and computational complexity. We expect such analysis could inspire more efficient approaches to address the imbalance problem in the future, as well as foundation models. The code of the toolkit is available at https://github.com/GeWu-Lab/BalanceBenchmark.
Authors: \'Alvaro Arroyo, Alessio Gravina, Benjamin Gutteridge, Federico Barbero, Claudio Gallicchio, Xiaowen Dong, Michael Bronstein, Pierre Vandergheynst
Abstract: Graph Neural Networks (GNNs) are models that leverage the graph structure to transmit information between nodes, typically through the message-passing operation. While widely successful, this approach is well known to suffer from the over-smoothing and over-squashing phenomena, which result in representational collapse as the number of layers increases and insensitivity to the information contained at distant and poorly connected nodes, respectively. In this paper, we present a unified view of these problems through the lens of vanishing gradients, using ideas from linear control theory for our analysis. We propose an interpretation of GNNs as recurrent models and empirically demonstrate that a simple state-space formulation of a GNN effectively alleviates over-smoothing and over-squashing at no extra trainable parameter cost. Further, we show theoretically and empirically that (i) GNNs are by design prone to extreme gradient vanishing even after a few layers; (ii) Over-smoothing is directly related to the mechanism causing vanishing gradients; (iii) Over-squashing is most easily alleviated by a combination of graph rewiring and vanishing gradient mitigation. We believe our work will help bridge the gap between the recurrent and graph neural network literature and will unlock the design of new deep and performant GNNs.
Authors: Shafique Ahmed, Ryandhimas E. Zezario, Hui-Guan Yuan, Amir Hussain, Hsin-Min Wang, Wei-Ho Chung, Yu Tsao
Abstract: The prevalence of hearing aids is increasing. However, optimizing the amplification processes of hearing aids remains challenging due to the complexity of integrating multiple modular components in traditional methods. To address this challenge, we present NeuroAMP, a novel deep neural network designed for end-to-end, personalized amplification in hearing aids. NeuroAMP leverages both spectral features and the listener's audiogram as inputs, and we investigate four architectures: Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Convolutional Recurrent Neural Network (CRNN), and Transformer. We also introduce Denoising NeuroAMP, an extension that integrates noise reduction along with amplification capabilities for improved performance in real-world scenarios. To enhance generalization, a comprehensive data augmentation strategy was employed during training on diverse speech (TIMIT and TMHINT) and music (Cadenza Challenge MUSIC) datasets. Evaluation using the Hearing Aid Speech Perception Index (HASPI), Hearing Aid Speech Quality Index (HASQI), and Hearing Aid Audio Quality Index (HAAQI) demonstrates that the Transformer architecture within NeuroAMP achieves the best performance, with SRCC scores of 0.9927 (HASQI) and 0.9905 (HASPI) on TIMIT, and 0.9738 (HAAQI) on the Cadenza Challenge MUSIC dataset. Notably, our data augmentation strategy maintains high performance on unseen datasets (e.g., VCTK, MUSDB18-HQ). Furthermore, Denoising NeuroAMP outperforms both the conventional NAL-R+WDRC approach and a two-stage baseline on the VoiceBank+DEMAND dataset, achieving a 10% improvement in both HASPI (0.90) and HASQI (0.59) scores. These results highlight the potential of NeuroAMP and Denoising NeuroAMP to deliver notable improvements in personalized hearing aid amplification.
Authors: Yuning Jiang, Qiaoran Meng, Feiyang Shang, Nay Oo, Le Thi Hong Minh, Hoon Wei Lim, Biplab Sikdar
Abstract: The MITRE ATT&CK framework is a widely adopted tool for enhancing cybersecurity, supporting threat intelligence, incident response, attack modeling, and vulnerability prioritization. This paper synthesizes research on its application across these domains by analyzing 417 peer-reviewed publications. We identify commonly used adversarial tactics, techniques, and procedures (TTPs) and examine the integration of natural language processing (NLP) and machine learning (ML) with ATT&CK to improve threat detection and response. Additionally, we explore the interoperability of ATT&CK with other frameworks, such as the Cyber Kill Chain, NIST guidelines, and STRIDE, highlighting its versatility. The paper further evaluates the framework from multiple perspectives, including its effectiveness, validation methods, and sector-specific challenges, particularly in industrial control systems (ICS) and healthcare. We conclude by discussing current limitations and proposing future research directions to enhance the applicability of ATT&CK in dynamic cybersecurity environments.
Authors: Amey P. Pasarkar, Adji Bousso Dieng
Abstract: The evolution of microscopy, beginning with its invention in the late 16th century, has continuously enhanced our ability to explore and understand the microscopic world, enabling increasingly detailed observations of structures and phenomena. In parallel, the rise of data-driven science has underscored the need for sophisticated methods to explore and understand the composition of complex data collections. This paper introduces the Vendiscope, the first algorithmic microscope designed to extend traditional microscopy to computational analysis. The Vendiscope leverages the Vendi scores -- a family of differentiable diversity metrics rooted in ecology and quantum mechanics -- and assigns weights to data points based on their contribution to the overall diversity of the collection. These weights enable high-resolution data analysis at scale. We demonstrate this across biology, materials science, and machine learning (ML). We analyzed the $250$ million protein sequences in the protein universe, discovering that over $200$ million are near-duplicates and that AlphaFold fails on proteins with Gene Ontology (GO) functions that contribute most to diversity. Applying the Vendiscope to the Materials Project database led to similar findings: more than $85\%$ of the crystals with formation energy data are near-duplicates and ML models perform poorly on materials that enhance diversity. Additionally, the Vendiscope can be used to study phenomena such as memorization in generative models. We used the Vendiscope to identify memorized training samples from $13$ different generative models and found that the best-performing ones often memorize the training samples that contribute least to diversity. Our findings demonstrate that the Vendiscope can serve as a powerful tool for data-driven science.
Authors: Zeli Su, Ziyin Zhang, Guixian Xu, Jianing Liu, XU Han, Ting Zhang, Yushuang Dong
Abstract: While multilingual language models like XLM-R have advanced multilingualism in NLP, they still perform poorly in extremely low-resource languages. This situation is exacerbated by the fact that modern LLMs such as LLaMA and Qwen support far fewer languages than XLM-R, making text generation models non-existent for many languages in the world. To tackle this challenge, we propose a novel framework for adapting multilingual encoders to text generation in extremely low-resource languages. By reusing the weights between the encoder and the decoder, our framework allows the model to leverage the learned semantic space of the encoder, enabling efficient learning and effective generalization in low-resource languages. Applying this framework to four Chinese minority languages, we present XLM-SWCM, and demonstrate its superior performance on various downstream tasks even when compared with much larger models.
Authors: Ge Lei, Samuel J. Cooper
Abstract: This study investigates how large language models (LLMs) represent and recall multi-associated attributes across transformer layers. We show that intermediate layers encode factual knowledge by superimposing related attributes in overlapping spaces, along with effective recall even when attributes are not explicitly prompted. In contrast, later layers refine linguistic patterns and progressively separate attribute representations, optimizing task-specific outputs while appropriately narrowing attribute recall. We identify diverse encoding patterns including, for the first time, the observation of 3D spiral structures when exploring information related to the periodic table of elements. Our findings reveal a dynamic transition in attribute representations across layers, contributing to mechanistic interpretability and providing insights for understanding how LLMs handle complex, interrelated knowledge.
Authors: Shib Dasgupta, Michael Boratko, Andrew McCallum
Abstract: Personalized item recommendation typically suffers from data sparsity, which is most often addressed by learning vector representations of users and items via low-rank matrix factorization. While this effectively densifies the matrix by assuming users and movies can be represented by linearly dependent latent features, it does not capture more complicated interactions. For example, vector representations struggle with set-theoretic relationships, such as negation and intersection, e.g. recommending a movie that is "comedy and action, but not romance". In this work, we formulate the problem of personalized item recommendation as matrix completion where rows are set-theoretically dependent. To capture this set-theoretic dependence we represent each user and attribute by a hyper-rectangle or box (i.e. a Cartesian product of intervals). Box embeddings can intuitively be understood as trainable Venn diagrams, and thus not only inherently represent similarity (via the Jaccard index), but also naturally and faithfully support arbitrary set-theoretic relationships. Queries involving set-theoretic constraints can be efficiently computed directly on the embedding space by performing geometric operations on the representations. We empirically demonstrate the superiority of box embeddings over vector-based neural methods on both simple and complex item recommendation queries by up to 30 \% overall.
Authors: Chao Tian (Shitz), Shlomo Shamai (Shitz)
Abstract: Partial information decomposition has recently found applications in biological signal processing and machine learning. Despite its impacts, the decomposition was introduced through an informal and heuristic route, and its exact operational meaning is unclear. In this work, we fill this gap by connecting partial information decomposition to the capacity of the broadcast channel, which has been well-studied in the information theory literature. We show that the synergistic information in the decomposition can be rigorously interpreted as the cooperative gain, or a lower bound of this gain, on the corresponding broadcast channel. This interpretation can help practitioners to better explain and expand the applications of the partial information decomposition technique.
Authors: Jiaru Zhang, Rui Ding, Qiang Fu, Bojun Huang, Zizhen Deng, Yang Hua, Haibing Guan, Shi Han, Dongmei Zhang
Abstract: Causal discovery is a structured prediction task that aims to predict causal relations among variables based on their data samples. Supervised Causal Learning (SCL) is an emerging paradigm in this field. Existing Deep Neural Network (DNN)-based methods commonly adopt the "Node-Edge approach", in which the model first computes an embedding vector for each variable-node, then uses these variable-wise representations to concurrently and independently predict for each directed causal-edge. In this paper, we first show that this architecture has some systematic bias that cannot be mitigated regardless of model size and data size. We then propose SiCL, a DNN-based SCL method that predicts a skeleton matrix together with a v-tensor (a third-order tensor representing the v-structures). According to the Markov Equivalence Class (MEC) theory, both the skeleton and the v-structures are identifiable causal structures under the canonical MEC setting, so predictions about skeleton and v-structures do not suffer from the identifiability limit in causal discovery, thus SiCL can avoid the systematic bias in Node-Edge architecture, and enable consistent estimators for causal discovery. Moreover, SiCL is also equipped with a specially designed pairwise encoder module with a unidirectional attention layer to model both internal and external relationships of pairs of nodes. Experimental results on both synthetic and real-world benchmarks show that SiCL significantly outperforms other DNN-based SCL approaches.
Authors: Nolan Fey, Gabriel B. Margolis, Martin Peticco, Pulkit Agrawal
Abstract: Achieving athletic loco-manipulation on robots requires moving beyond traditional tracking rewards - which simply guide the robot along a reference trajectory - to task rewards that drive truly dynamic, goal-oriented behaviors. Commands such as "throw the ball as far as you can" or "lift the weight as quickly as possible" compel the robot to exhibit the agility and power inherent in athletic performance. However, training solely with task rewards introduces two major challenges: these rewards are prone to exploitation (reward hacking), and the exploration process can lack sufficient direction. To address these issues, we propose a two-stage training pipeline. First, we introduce the Unsupervised Actuator Net (UAN), which leverages real-world data to bridge the sim-to-real gap for complex actuation mechanisms without requiring access to torque sensing. UAN mitigates reward hacking by ensuring that the learned behaviors remain robust and transferable. Second, we use a pre-training and fine-tuning strategy that leverages reference trajectories as initial hints to guide exploration. With these innovations, our robot athlete learns to lift, throw, and drag with remarkable fidelity from simulation to reality.
Authors: Ibraheem Hamdi, Hosam El-Gendy, Ahmed Sharshar, Mohamed Saeed, Muhammad Ridzuan, Shahrukh K. Hashmi, Naveed Syed, Imran Mirza, Shakir Hussain, Amira Mahmoud Abdalla, Mohammad Yaqub
Abstract: The complexities inherent to leukemia, multifaceted cancer affecting white blood cells, pose considerable diagnostic and treatment challenges, primarily due to reliance on laborious morphological analyses and expert judgment that are susceptible to errors. Addressing these challenges, this study presents a refined, comprehensive strategy leveraging advanced deep-learning techniques for the classification of leukemia subtypes. We commence by developing a hierarchical label taxonomy, paving the way for differentiating between various subtypes of leukemia. The research further introduces a novel hierarchical approach inspired by clinical procedures capable of accurately classifying diverse types of leukemia alongside reactive and healthy cells. An integral part of this study involves a meticulous examination of the performance of Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) as classifiers. The proposed method exhibits an impressive success rate, achieving approximately 90\% accuracy across all leukemia subtypes, as substantiated by our experimental results. A visual representation of the experimental findings is provided to enhance the model's explainability and aid in understanding the classification process.
Authors: Sevim Cengiz, Ibraheem Hamdi, Mohammad Yaqub
Abstract: Fetal gestational age (GA) is vital clinical information that is estimated during pregnancy in order to assess fetal growth. This is usually performed by measuring the crown-rump-length (CRL) on an ultrasound image in the Dating scan which is then correlated with fetal age and growth trajectory. A major issue when performing the CRL measurement is ensuring that the image is acquired at the correct view, otherwise it could be misleading. Although clinical guidelines specify the criteria for the correct CRL view, sonographers may not regularly adhere to such rules. In this paper, we propose a new deep learning-based solution that is able to verify the adherence of a CRL image to clinical guidelines in order to assess image quality and facilitate accurate estimation of GA. We first segment out important fetal structures then use the localized structures to perform a clinically-guided mapping that verifies the adherence of criteria. The segmentation method combines the benefits of Convolutional Neural Network (CNN) and the Vision Transformer (ViT) to segment fetal structures in ultrasound images and localize important fetal landmarks. For segmentation purposes, we compare our proposed work with UNet and show that our CNN/ViT-based method outperforms an optimized version of UNet. Furthermore, we compare the output of the mapping with classification CNNs when assessing the clinical criteria and the overall acceptability of CRL images. We show that the proposed mapping is not only explainable but also more accurate than the best performing classification CNNs.
Authors: Simiao Ren, Hengwei Xu, Tsang Ng, Kidus Zewde, Shengkai Jiang, Ramini Desai, Disha Patil, Ning-Yau Cheng, Yining Zhou, Ragavi Muthukrishnan
Abstract: Deepfakes, particularly those involving faceswap-based manipulations, have sparked significant societal concern due to their increasing realism and potential for misuse. Despite rapid advancements in generative models, detection methods have not kept pace, creating a critical gap in defense strategies. This disparity is further amplified by the disconnect between academic research and real-world applications, which often prioritize different objectives and evaluation criteria. In this study, we take a pivotal step toward bridging this gap by presenting a novel observation: the post-processing step of super-resolution, commonly employed in real-world scenarios, substantially undermines the effectiveness of existing deepfake detection methods. To substantiate this claim, we introduce and publish the first real-world faceswap dataset, collected from popular online faceswap platforms. We then qualitatively evaluate the performance of state-of-the-art deepfake detectors on real-world deepfakes, revealing that their accuracy approaches the level of random guessing. Furthermore, we quantitatively demonstrate the significant performance degradation caused by common post-processing techniques. By addressing this overlooked challenge, our study underscores a critical avenue for enhancing the robustness and practical applicability of deepfake detection methods in real-world settings.
Authors: Matthew Lyle Olson, Neale Ratzlaff, Musashi Hinck, Man Luo, Sungduk Yu, Chendi Xue, Vasudev Lal
Abstract: DeepSeek-R1, the largest open-source Mixture-of-Experts (MoE) model, has demonstrated reasoning capabilities comparable to proprietary frontier models. Prior research has explored expert routing in MoE models, but findings suggest that expert selection is often token-dependent rather than semantically driven. Given DeepSeek-R1's enhanced reasoning abilities, we investigate whether its routing mechanism exhibits greater semantic specialization than previous MoE models. To explore this, we conduct two key experiments: (1) a word sense disambiguation task, where we examine expert activation patterns for words with differing senses, and (2) a cognitive reasoning analysis, where we assess DeepSeek-R1's structured thought process in an interactive task setting of DiscoveryWorld. We conclude that DeepSeek-R1's routing mechanism is more semantically aware and it engages in structured cognitive processes.
Authors: Ziyue Liu, Ruijie Zhang, Zhengyang Wang, Zi Yang, Paul Hovland, Bogdan Nicolae, Franck Cappello, Zheng Zhang
Abstract: Large language models (LLMs) are revolutionizing many science and engineering fields. However, their huge model sizes impose extremely demanding needs of computational resources in the pre-training stage. Although low-rank factorizations can reduce model parameters, their direct application in LLM pre-training often lead to non-negligible performance loss. To address this fundamental challenge, we introduce CoLA and its memory-efficient implementation, CoLA-M. We leverage the low-rank structure observed widely in model activations, enforcing non-linear transformations between factorized weight matrices to reduce model size, boost model capacity and training efficiency. Experiments on LLaMA models with 60 million to 7 billion parameters show that CoLA reduces the computing cost by $\bf 2\pmb{\times}$ and improves training throughput by $\bf 1.86\pmb{\times}$ while maintaining full-rank level performance. CoLA-M further squeezes memory cost without sacrificing throughput, offering a pre-training approach with collectively superior parameter, computing, and memory efficiency. The LLMs produced are also $\bf 2\pmb{\times}$ smaller, enabling faster inference with lower memory cost on resource-constrained platforms
Authors: Sheikh Moonwara Anjum Monisha, Atul Bharadwaj
Abstract: This empirical study evaluates the effectiveness of Large Language Models (LLMs) in predicting fixes for configuration bugs in smart home systems. The research analyzes three prominent LLMs - GPT-4, GPT-4o (GPT-4 Turbo), and Claude 3.5 Sonnet - using four distinct prompt designs to assess their ability to identify appropriate fix strategies and generate correct solutions. The study utilized a dataset of 129 debugging issues from the Home Assistant Community, focusing on 21 randomly selected cases for in-depth analysis. Results demonstrate that GPT-4 and Claude 3.5 Sonnet achieved 80\% accuracy in strategy prediction when provided with both bug descriptions and original scripts. GPT-4 exhibited consistent performance across different prompt types, while GPT-4o showed advantages in speed and cost-effectiveness despite slightly lower accuracy. The findings reveal that prompt design significantly impacts model performance, with comprehensive prompts containing both description and original script yielding the best results. This research provides valuable insights for improving automated bug fixing in smart home system configurations and demonstrates the potential of LLMs in addressing configuration-related challenges.
Authors: Hieu Tran Bao, Nguyen Cong Dat, Nguyen Duc Anh, Hoang Thanh Tung
Abstract: Test time scaling is currently one of the most active research areas that shows promise after training time scaling has reached its limits. Deep-thinking (DT) models are a class of recurrent models that can perform easy-to-hard generalization by assigning more compute to harder test samples. However, due to their inability to determine the complexity of a test sample, DT models have to use a large amount of computation for both easy and hard test samples. Excessive test time computation is wasteful and can cause the ``overthinking'' problem where more test time computation leads to worse results. In this paper, we introduce a test time training method for determining the optimal amount of computation needed for each sample during test time. We also propose Conv-LiGRU, a novel recurrent architecture for efficient and robust visual reasoning. Extensive experiments demonstrate that Conv-LiGRU is more stable than DT, effectively mitigates the ``overthinking'' phenomenon, and achieves superior accuracy.
Authors: Jonathan Morgan, Badr Albanna, James P. Herman
Abstract: Attention is fundamental to both biological and artificial intelligence, yet research on animal attention and AI self attention remains largely disconnected. We propose a Recurrent Vision Transformer (Recurrent ViT) that integrates self-attention with recurrent memory, allowing both current inputs and stored information to guide attention allocation. Trained solely via sparse reward feedback on a spatially cued orientation change detection task, a paradigm used in primate studies, our model exhibits primate like signatures of attention, including improved accuracy and faster responses for cued stimuli that scale with cue validity. Analysis of self-attention maps reveals dynamic spatial prioritization with reactivation prior to expected changes, and targeted perturbations produce performance shifts similar to those observed in primate frontal eye fields and superior colliculus. These findings demonstrate that incorporating recurrent feedback into self attention can capture key aspects of primate visual attention.
Authors: Jin Peng Zhou, S\'ebastien M. R. Arnold, Nan Ding, Kilian Q. Weinberger, Nan Hua, Fei Sha
Abstract: Auto-evaluating language models (LMs), i.e., using a grader LM to evaluate the candidate LM, is an appealing way to accelerate the evaluation process and the cost associated with it. But this presents a paradox: how can we trust the grader LM, which is presumably weaker than the candidate LM, to assess problems that are beyond the frontier of the capabilities of either model or both? For instance, today's LMs struggle on graduate-level physics and Olympiad-level math, making them unreliable graders in these domains. We show that providing privileged information -- such as ground-truth solutions or problem-specific guidelines -- improves automated evaluations on such frontier problems. This approach offers two key advantages. First, it expands the range of problems where LMs graders apply. Specifically, weaker models can now rate the predictions of stronger models. Second, privileged information can be used to devise easier variations of challenging problems which improves the separability of different LMs on tasks where their performance is generally low. With this approach, general-purpose LM graders match the state of the art performance on RewardBench, surpassing almost all the specially-tuned models. LM graders also outperform individual human raters on Vibe-Eval, and approach human expert graders on Olympiad-level math problems.
Authors: Biqing Zeng, Zehan Li, Aladdin Ayesh
Abstract: Catastrophic forgetting is a pervasive issue for pre-trained language models (PLMs) during continual learning, where models lose previously acquired knowledge when sequentially trained on a series of tasks. The model's ability to retain old tasks is referred to as stability, while its adaptability to new tasks is called plasticity. Therefore, the key to solving this problem is to find a trade-off between the plasticity and stability of the model. To address this issue, in this paper, we propose a novel method to achieve a balance between model stability and plasticity, thereby mitigating catastrophic forgetting. More specifically, our proposed approach leverages parameter isolation and a subsequent combination strategy. Initially, in the training stage, the model adapts to each downstream task via a parameter isolation method to prevent potential interference among different tasks. We then combine all trained parameters, which contain acquired knowledge, using the task arithmetic method and finally apply them to the backbone model. Empirical evaluations on continual language learning benchmarks substantiate the effectiveness of our approach, revealing a marked enhancement over existing state-of-the-art approaches.
Authors: Andrew Neeser, Kaylen Latimer, Aadyant Khatri, Chris Latimer, Naren Ramakrishnan
Abstract: We present QuOTE (Question-Oriented Text Embeddings), a novel enhancement to retrieval-augmented generation (RAG) systems, aimed at improving document representation for accurate and nuanced retrieval. Unlike traditional RAG pipelines, which rely on embedding raw text chunks, QuOTE augments chunks with hypothetical questions that the chunk can potentially answer, enriching the representation space. This better aligns document embeddings with user query semantics, and helps address issues such as ambiguity and context-dependent relevance. Through extensive experiments across diverse benchmarks, we demonstrate that QuOTE significantly enhances retrieval accuracy, including in multi-hop question-answering tasks. Our findings highlight the versatility of question generation as a fundamental indexing strategy, opening new avenues for integrating question generation into retrieval-based AI pipelines.
Authors: Shange Tang, Yuanhao Wang, Chi Jin
Abstract: Elo rating, widely used for skill assessment across diverse domains ranging from competitive games to large language models, is often understood as an incremental update algorithm for estimating a stationary Bradley-Terry (BT) model. However, our empirical analysis of practical matching datasets reveals two surprising findings: (1) Most games deviate significantly from the assumptions of the BT model and stationarity, raising questions on the reliability of Elo. (2) Despite these deviations, Elo frequently outperforms more complex rating systems, such as mElo and pairwise models, which are specifically designed to account for non-BT components in the data, particularly in terms of win rate prediction. This paper explains this unexpected phenomenon through three key perspectives: (a) We reinterpret Elo as an instance of online gradient descent, which provides no-regret guarantees even in misspecified and non-stationary settings. (b) Through extensive synthetic experiments on data generated from transitive but non-BT models, such as strongly or weakly stochastic transitive models, we show that the ''sparsity'' of practical matching data is a critical factor behind Elo's superior performance in prediction compared to more complex rating systems. (c) We observe a strong correlation between Elo's predictive accuracy and its ranking performance, further supporting its effectiveness in ranking.
Authors: Bowen Jiang, Yuan Yuan, Xinyi Bai, Zhuoqun Hao, Alyson Yin, Yaojie Hu, Wenyu Liao, Lyle Ungar, Camillo J. Taylor
Abstract: This work demonstrates that diffusion models can achieve font-controllable multilingual text rendering using just raw images without font label annotations. Visual text rendering remains a significant challenge. While recent methods condition diffusion on glyphs, it is impossible to retrieve exact font annotations from large-scale, real-world datasets, which prevents user-specified font control. To address this, we propose a data-driven solution that integrates the conditional diffusion model with a text segmentation model, utilizing segmentation masks to capture and represent fonts in pixel space in a self-supervised manner, thereby eliminating the need for any ground-truth labels and enabling users to customize text rendering with any multilingual font of their choice. The experiment provides a proof of concept of our algorithm in zero-shot text and font editing across diverse fonts and languages, providing valuable insights for the community and industry toward achieving generalized visual text rendering.
Authors: Gen Zhou, Sugitha Janarthanan, Yutong Lu, Pingzhao Hu
Abstract: Due to the rise in antimicrobial resistance, identifying novel compounds with antibiotic potential is crucial for combatting this global health issue. However, traditional drug development methods are costly and inefficient. Recognizing the pressing need for more effective solutions, researchers have turned to machine learning techniques to streamline the prediction and development of novel antibiotic compounds. While foundation models have shown promise in antibiotic discovery, current mainstream efforts still fall short of fully leveraging the potential of multimodal molecular data. Recent studies suggest that contrastive learning frameworks utilizing multimodal data exhibit excellent performance in representation learning across various domains. Building upon this, we introduce CL-MFAP, an unsupervised contrastive learning (CL)-based multimodal foundation (MF) model specifically tailored for discovering small molecules with potential antibiotic properties (AP) using three types of molecular data. This model employs 1.6 million bioactive molecules with drug-like properties from the ChEMBL dataset to jointly pretrain three encoders: (1) a transformer-based encoder with rotary position embedding for processing SMILES strings; (2) another transformer-based encoder, incorporating a novel bi-level routing attention mechanism to handle molecular graph representations; and (3) a Morgan fingerprint encoder using a multilayer perceptron, to achieve the contrastive learning purpose. The CL-MFAP outperforms baseline models in antibiotic property prediction by effectively utilizing different molecular modalities and demonstrates superior domain-specific performance when fine-tuned for antibiotic-related property prediction tasks.
Authors: Jonathan Pan, Swee Liang Wong, Yidi Yuan, Xin Wei Chia
Abstract: Large Language Models (LLMs) are vulnerable to adversarial prompt based injects. These injects could jailbreak or exploit vulnerabilities within these models with explicit prompt requests leading to undesired responses. In the context of investigating prompt injects, the challenge is the sheer volume of input prompts involved that are likely to be largely benign. This investigative challenge is further complicated by the semantics and subjectivity of the input prompts involved in the LLM conversation with its user and the context of the environment to which the conversation is being carried out. Hence, the challenge for AI security investigators would be two-fold. The first is to identify adversarial prompt injects and then to assess whether the input prompt is contextually benign or adversarial. For the first step, this could be done using existing AI security solutions like guardrails to detect and protect the LLMs. Guardrails have been developed using a variety of approaches. A popular approach is to use signature based. Another popular approach to develop AI models to classify such prompts include the use of NLP based models like a language model. However, in the context of conducting an AI security investigation of prompt injects, these guardrails lack the ability to aid investigators in triaging or assessing the identified input prompts. In this applied research exploration, we explore the use of a text generation capabilities of LLM to detect prompt injects and generate explanation for its detections to aid AI security investigators in assessing and triaging of such prompt inject detections. The practical benefit of such a tool is to ease the task of conducting investigation into prompt injects.
Authors: Zhi Sheng, Yuan Yuan, Yudi Zhang, Depeng Jin, Yong Li
Abstract: Accurate prediction of urban spatiotemporal dynamics is essential for enhancing urban management and decision-making. Existing spatiotemporal prediction models are predominantly deterministic, focusing on primary spatiotemporal patterns. However, those dynamics are highly complex, exhibiting multi-modal distributions that are challenging for deterministic models to capture. In this paper, we highlight the critical role of probabilistic prediction in capturing the uncertainties and complexities inherent in spatiotemporal data. While mainstream probabilistic models can capture uncertainty, they struggle with accurately learning primary patterns and often suffer from computational inefficiency. To address these challenges, we propose CoST, which collaborates deterministic and probabilistic models to improve both predictive accuracy and the ability to handle uncertainty. To achieve this, we design a mean-residual decomposition framework, where the mean value is modeled by a deterministic model, and the residual variations are learned by a probabilistic model, specifically diffusion models. Moreover, we introduce a scale-aware diffusion process, which better accounts for spatially heterogeneous dynamics across different regions. Extensive experiments on eight real-world datasets demonstrate that CoST significantly outperforms existing methods in both deterministic and probabilistic metrics, achieving a 20% improvement with low computational cost. CoST bridges the gap between deterministic precision and probabilistic uncertainty, making a significant advancement in the field of urban spatiotemporal prediction.
Authors: Shijing Hu, Jingyang Li, Xingyu Xie, Zhihui Lu, Kim-Chuan Toh, Pan Zhou
Abstract: Speculative decoding accelerates inference in large language models (LLMs) by generating multiple draft tokens simultaneously. However, existing methods often struggle with token misalignment between the training and decoding phases, limiting their performance. To address this, we propose GRIFFIN, a novel framework that incorporates a token-alignable training strategy and a token-alignable draft model to mitigate misalignment. The training strategy employs a loss masking mechanism to exclude highly misaligned tokens during training, preventing them from negatively impacting the draft model's optimization. The token-alignable draft model introduces input tokens to correct inconsistencies in generated features. Experiments on LLaMA-series and Vicuna models demonstrate that GRIFFIN achieves an average acceptance length improvement of over 7\% and a speedup ratio exceeding 8%, outperforming current SoTAs as shown in Fig. 1 (a) and (b).
Authors: Gangwei Jiang, Caigao Jiang, Zhaoyi Li, Siqiao Xue, Jun Zhou, Linqi Song, Defu Lian, Yin Wei
Abstract: Catastrophic forgetting (CF) poses a significant challenge in machine learning, where a model forgets previously learned information upon learning new tasks. Despite the advanced capabilities of Large Language Models (LLMs), they continue to face challenges with CF during continual learning. The majority of existing research focuses on analyzing forgetting patterns through a singular training sequence, thereby overlooking the intricate effects that diverse tasks have on model behavior. Our study explores CF across various settings, discovering that model forgetting is influenced by both the specific training tasks and the models themselves. To this end, we interpret forgetting by examining the function vector (FV), a compact representation of functions in LLMs, offering a model-dependent indicator for the occurrence of CF. Through theoretical and empirical analyses, we demonstrated that CF in LLMs primarily stems from biases in function activation rather than the overwriting of task processing functions. Leveraging these insights, we propose a novel function vector guided training methodology, incorporating a regularization technique to stabilize the FV and mitigate forgetting. Empirical tests on four benchmarks confirm the effectiveness of our proposed training method, substantiating our theoretical framework concerning CF and model function dynamics. We plan to make our code publicly accessible in the near future.
Authors: Jafar Isbarov, Arofat Akhundjanova, Mammad Hajili, Kavsar Huseynova, Dmitry Gaynullin, Anar Rzayev, Osman Tursun, Ilshat Saetov, Rinat Kharisov, Saule Belginova, Ariana Kenbayeva, Amina Alisheva, Aizirek Turdubaeva, Abdullatif K\"oksal, Samir Rustamov, Duygu Ataman
Abstract: Being able to thoroughly assess massive multi-task language understanding (MMLU) capabilities is essential for advancing the applicability of multilingual language models. However, preparing such benchmarks in high quality native language is often costly and therefore limits the representativeness of evaluation datasets. While recent efforts focused on building more inclusive MMLU benchmarks, these are conventionally built using machine translation from high-resource languages, which may introduce errors and fail to account for the linguistic and cultural intricacies of the target languages. In this paper, we address the lack of native language MMLU benchmark especially in the under-represented Turkic language family with distinct morphosyntactic and cultural characteristics. We propose two benchmarks for Turkic language MMLU: TUMLU is a comprehensive, multilingual, and natively developed language understanding benchmark specifically designed for Turkic languages. It consists of middle- and high-school level questions spanning 11 academic subjects in Azerbaijani, Crimean Tatar, Karakalpak, Kazakh, Tatar, Turkish, Uyghur, and Uzbek. We also present TUMLU-mini, a more concise, balanced, and manually verified subset of the dataset. Using this dataset, we systematically evaluate a diverse range of open and proprietary multilingual large language models (LLMs), including Claude, Gemini, GPT, and LLaMA, offering an in-depth analysis of their performance across different languages, subjects, and alphabets. To promote further research and development in multilingual language understanding, we release TUMLU-mini and all corresponding evaluation scripts.
Authors: Zhiqian Qin, Yuanfeng Song, Jinwei Lu, Yuanwei Song, Shuaimin Li, Chen Jason Zhang
Abstract: Natural language interfaces for NoSQL databases are increasingly vital in the big data era, enabling users to interact with complex, unstructured data without deep technical expertise. However, most recent advancements focus on English, leaving a gap for multilingual support. This paper introduces MultiTEND, the first and largest multilingual benchmark for natural language to NoSQL query generation, covering six languages: English, German, French, Russian, Japanese and Mandarin Chinese. Using MultiTEND, we analyze challenges in translating natural language to NoSQL queries across diverse linguistic structures, including lexical and syntactic differences. Experiments show that performance accuracy in both English and non-English settings remains relatively low, with a 4%-6% gap across scenarios like fine-tuned SLM, zero-shot LLM, and RAG for LLM. To address the aforementioned challenges, we introduce MultiLink, a novel framework that bridges the multilingual input to NoSQL query generation gap through a Parallel Linking Process. It breaks down the task into multiple steps, integrating parallel multilingual processing, Chain-of-Thought (CoT) reasoning, and Retrieval-Augmented Generation (RAG) to tackle lexical and structural challenges inherent in multilingual NoSQL generation. MultiLink shows enhancements in all metrics for every language against the top baseline, boosting execution accuracy by about 15% for English and averaging a 10% improvement for non-English languages.
Authors: Yuhao Du, Zhuo Li, Pengyu Cheng, Zhihong Chen, Yuejiao Xie, Xiang Wan, Anningzhe Gao
Abstract: Reinforcement Learning from Human Feedback (RLHF) is crucial for aligning Large Language Models (LLMs) with human values. However, RLHF has been continuously challenged by its high complexity in implementation and computation consumption. Even with recent simplifications, such as Direct Preference Optimization (DPO) and Advantage Leftover Lunch (A-LoL), the problems of over-fitting and training instability remain hindering the alignment process from the expected optimal performance. To address the existing challenges, we propose a novel simplification of RLHF from the perspective of variational inference, called $\textbf{V}$ariational $\textbf{A}$lignment with $\textbf{R}$e-weighting ($\textbf{VAR}$). More specifically, by directly minimizing the distribution gap between the learning LLM policy and the optimal solution of RLHF, we transform the alignment objective into a reward-driven re-weighted supervised fine-tuning (SFT) form, which only requires minor adjustment on the SFT loss to obtain noticeable improvement on training stability and effectiveness. On comprehensive alignment and generation benchmarks, our VAR method has numerically achieved competitive performance in LLM alignment helpfulness and harmlessness.
Authors: Prateek Chhikara
Abstract: Large Language Models (LLMs) demonstrate impressive performance across diverse tasks, yet confidence calibration remains a challenge. Miscalibration - where models are overconfident or underconfident - poses risks, particularly in high-stakes applications. This paper presents an empirical study on LLM calibration, examining how model size, distractors, and question types affect confidence alignment. We introduce an evaluation framework to measure overconfidence and investigate whether multiple-choice formats mitigate or worsen miscalibration. Our findings show that while larger models (e.g., GPT-4o) are better calibrated overall, they are more prone to distraction, whereas smaller models benefit more from answer choices but struggle with uncertainty estimation. Unlike prior work, which primarily reports miscalibration trends, we provide actionable insights into failure modes and conditions that worsen overconfidence. These findings highlight the need for calibration-aware interventions and improved uncertainty estimation methods.
Authors: Xin Gao, Jian Pu
Abstract: Multi-View Representation Learning (MVRL) aims to derive a unified representation from multi-view data by leveraging shared and complementary information across views. However, when views are irregularly missing, the incomplete data can lead to representations that lack sufficiency and consistency. To address this, we propose Multi-View Permutation of Variational Auto-Encoders (MVP), which excavates invariant relationships between views in incomplete data. MVP establishes inter-view correspondences in the latent space of Variational Auto-Encoders, enabling the inference of missing views and the aggregation of more sufficient information. To derive a valid Evidence Lower Bound (ELBO) for learning, we apply permutations to randomly reorder variables for cross-view generation and then partition them by views to maintain invariant meanings under permutations. Additionally, we enhance consistency by introducing an informational prior with cyclic permutations of posteriors, which turns the regularization term into a similarity measure across distributions. We demonstrate the effectiveness of our approach on seven diverse datasets with varying missing ratios, achieving superior performance in multi-view clustering and generation tasks.
Authors: Jiahao Huo, Yibo Yan, Xu Zheng, Yuanhuiyi Lyu, Xin Zou, Zhihua Wei, Xuming Hu
Abstract: Recent progress in Machine Unlearning (MU) has introduced solutions for the selective removal of private or sensitive information encoded within deep neural networks. Nonetheless, MU for Multimodal Large Language Models (MLLMs) remains in its nascent phase. Therefore, we propose to reformulate the task of multimodal MU in the era of MLLMs, which aims to erase only the visual patterns associated with a given entity while preserving the corresponding textual knowledge encoded within the original parameters of the language model backbone. Furthermore, we develop a novel geometry-constrained gradient descent method MMUnlearner. It updates the weights of MLLMs with a weight saliency map jointly restricted by the remaining concepts and textual knowledge during unlearning, thereby preserving parameters essential for non-target knowledge. Extensive experiments demonstrate that MMUnlearner surpasses baselines that finetuning MLLMs with VQA data directly through Gradient Ascent (GA) or Negative Preference Optimization (NPO), across all evaluation dimensions. Our code will be released upon acceptance.
Authors: Zonghao Ying, Deyue Zhang, Zonglei Jing, Yisong Xiao, Quanchen Zou, Aishan Liu, Siyuan Liang, Xiangzheng Zhang, Xianglong Liu, Dacheng Tao
Abstract: Multi-turn jailbreak attacks simulate real-world human interactions by engaging large language models (LLMs) in iterative dialogues, exposing critical safety vulnerabilities. However, existing methods often struggle to balance semantic coherence with attack effectiveness, resulting in either benign semantic drift or ineffective detection evasion. To address this challenge, we propose Reasoning-Augmented Conversation, a novel multi-turn jailbreak framework that reformulates harmful queries into benign reasoning tasks and leverages LLMs' strong reasoning capabilities to compromise safety alignment. Specifically, we introduce an attack state machine framework to systematically model problem translation and iterative reasoning, ensuring coherent query generation across multiple turns. Building on this framework, we design gain-guided exploration, self-play, and rejection feedback modules to preserve attack semantics, enhance effectiveness, and sustain reasoning-driven attack progression. Extensive experiments on multiple LLMs demonstrate that RACE achieves state-of-the-art attack effectiveness in complex conversational scenarios, with attack success rates (ASRs) increasing by up to 96%. Notably, our approach achieves ASRs of 82% and 92% against leading commercial models, OpenAI o1 and DeepSeek R1, underscoring its potency. We release our code at https://github.com/NY1024/RACE to facilitate further research in this critical domain.
Authors: Manan Tayal, Aditya Singh, Shishir Kolathaya, Somil Bansal
Abstract: As autonomous systems become more ubiquitous in daily life, ensuring high performance with guaranteed safety is crucial. However, safety and performance could be competing objectives, which makes their co-optimization difficult. Learning-based methods, such as Constrained Reinforcement Learning (CRL), achieve strong performance but lack formal safety guarantees due to safety being enforced as soft constraints, limiting their use in safety-critical settings. Conversely, formal methods such as Hamilton-Jacobi (HJ) Reachability Analysis and Control Barrier Functions (CBFs) provide rigorous safety assurances but often neglect performance, resulting in overly conservative controllers. To bridge this gap, we formulate the co-optimization of safety and performance as a state-constrained optimal control problem, where performance objectives are encoded via a cost function and safety requirements are imposed as state constraints. We demonstrate that the resultant value function satisfies a Hamilton-Jacobi-Bellman (HJB) equation, which we approximate efficiently using a novel physics-informed machine learning framework. In addition, we introduce a conformal prediction-based verification strategy to quantify the learning errors, recovering a high-confidence safety value function, along with a probabilistic error bound on performance degradation. Through several case studies, we demonstrate the efficacy of the proposed framework in enabling scalable learning of safe and performant controllers for complex, high-dimensional autonomous systems.
Authors: Shixuan Li, Wei Yang, Peiyu Zhang, Xiongye Xiao, Defu Cao, Yuehan Qin, Xiaole Zhang, Yue Zhao, Paul Bogdan
Abstract: Weather forecasting is crucial for public safety, disaster prevention and mitigation, agricultural production, and energy management, with global relevance. Although deep learning has significantly advanced weather prediction, current methods face critical limitations: (i) they often struggle to capture both dynamic temporal dependencies and short-term abrupt changes, making extreme weather modeling difficult; (ii) they incur high computational costs due to extensive training and resource requirements; (iii) they have limited adaptability to multi-scale frequencies, leading to challenges when separating global trends from local fluctuations. To address these issues, we propose ClimateLLM, a foundation model for weather forecasting. It captures spatiotemporal dependencies via a cross-temporal and cross-spatial collaborative modeling framework that integrates Fourier-based frequency decomposition with Large Language Models (LLMs) to strengthen spatial and temporal modeling. Our framework uses a Mixture-of-Experts (MoE) mechanism that adaptively processes different frequency components, enabling efficient handling of both global signals and localized extreme events. In addition, we introduce a cross-temporal and cross-spatial dynamic prompting mechanism, allowing LLMs to incorporate meteorological patterns across multiple scales effectively. Extensive experiments on real-world datasets show that ClimateLLM outperforms state-of-the-art approaches in accuracy and efficiency, as a scalable solution for global weather forecasting.
Authors: Haonan Yu, Junhao Liu, Xin Zhang
Abstract: Anchors is a popular local model-agnostic explanation technique whose applicability is limited by its computational inefficiency. To address this limitation, we propose a pre-training-based approach to accelerate Anchors without compromising the explanation quality. Our approach leverages the iterative nature of Anchors' algorithm which gradually refines an explanation until it is precise enough for a given input by providing a general explanation that is obtained through pre-training as Anchors' initial explanation. Specifically, we develop a two-step rule transformation process: the horizontal transformation adapts a pre-trained explanation to the current input by replacing features, and the vertical transformation refines the general explanation until it is precise enough for the input. We evaluate our method across tabular, text, and image datasets, demonstrating that it significantly reduces explanation generation time while maintaining fidelity and interpretability, thereby enabling the practical adoption of Anchors in time-sensitive applications.
Authors: Yuning Jiang, Nay Oo, Qiaoran Meng, Hoon Wei Lim, Biplab Sikdar
Abstract: In the highly interconnected digital landscape of today, safeguarding complex infrastructures against cyber threats has become increasingly challenging due to the exponential growth in the number and complexity of vulnerabilities. Resource constraints necessitate effective vulnerability prioritization strategies, focusing efforts on the most critical risks. This paper presents a systematic literature review of 82 studies, introducing a novel taxonomy that categorizes metrics into severity, exploitability, contextual factors, predictive indicators, and aggregation methods. Our analysis reveals significant gaps in existing approaches and challenges with multi-domain applicability. By emphasizing the need for dynamic, context-aware metrics and scalable solutions, we provide actionable insights to bridge the gap between research and real-world applications. This work contributes to the field by offering a comprehensive framework for evaluating vulnerability prioritization methodologies and setting a research agenda to advance the state of practice.
Authors: Haoyang Li, Xuejia Chen, Zhanchao XU, Darian Li, Nicole Hu, Fei Teng, Yiming Li, Luyu Qiu, Chen Jason Zhang, Qing Li, Lei Chen
Abstract: Large Language Models (LLMs) have demonstrated impressive capabilities in natural language processing tasks, such as text generation and semantic understanding. However, their performance on numerical reasoning tasks, such as basic arithmetic, numerical retrieval, and magnitude comparison, remains surprisingly poor. This gap arises from their reliance on surface-level statistical patterns rather than understanding numbers as continuous magnitudes. Existing benchmarks primarily focus on either linguistic competence or structured mathematical problem-solving, neglecting fundamental numerical reasoning required in real-world scenarios. To bridge this gap, we propose NumericBench, a comprehensive benchmark to evaluate six fundamental numerical capabilities: number recognition, arithmetic operations, contextual retrieval, comparison, summary, and logical reasoning. NumericBench includes datasets ranging from synthetic number lists to the crawled real-world data, addressing challenges like long contexts, noise, and multi-step reasoning. Extensive experiments on state-of-the-art LLMs, including GPT-4 and DeepSeek, reveal persistent weaknesses in numerical reasoning, highlighting the urgent need to improve numerically-aware language modeling. The benchmark is released in: https://github.com/TreeAI-Lab/NumericBench.
Authors: Lijie Liu, Tianxiang Ma, Bingchuan Li, Zhuowei Chen, Jiawei Liu, Qian He, Xinglong Wu
Abstract: The continuous development of foundational models for video generation is evolving into various applications, with subject-consistent video generation still in the exploratory stage. We refer to this as Subject-to-Video, which extracts subject elements from reference images and generates subject-consistent video through textual instructions. We believe that the essence of subject-to-video lies in balancing the dual-modal prompts of text and image, thereby deeply and simultaneously aligning both text and visual content. To this end, we propose Phantom, a unified video generation framework for both single and multi-subject references. Building on existing text-to-video and image-to-video architectures, we redesign the joint text-image injection model and drive it to learn cross-modal alignment via text-image-video triplet data. In particular, we emphasize subject consistency in human generation, covering existing ID-preserving video generation while offering enhanced advantages. The project homepage is here https://phantom-video.github.io/Phantom/.
Authors: Yasir Ghunaim, Hasan Abed Al Kader Hammoud, Bernard Ghanem
Abstract: This paper challenges the recent paradigm in atomic property prediction that links progress to growing dataset sizes and computational resources. We show that pretraining on a carefully selected, task-relevant dataset can match or even surpass large-scale pretraining, while using as little as 1/24th of the computational cost. We introduce the Chemical Similarity Index (CSI), a novel metric inspired by computer vision's Fr\'echet Inception Distance, for molecular graphs which quantifies the alignment between upstream pretraining datasets and downstream tasks. By selecting the most relevant dataset with minimal CSI distance, we show that models pretrained on a smaller, focused dataset consistently outperform those pretrained on massive, mixed datasets such as JMP, even when those larger datasets include the relevant dataset. Counterintuitively, we also find that indiscriminately adding more data can degrade model performance when the additional data poorly aligns with the task at hand. Our findings highlight that quality often outperforms quantity in pretraining for atomic property prediction.
Authors: Jingyang Yuan, Huazuo Gao, Damai Dai, Junyu Luo, Liang Zhao, Zhengyan Zhang, Zhenda Xie, Y. X. Wei, Lean Wang, Zhiping Xiao, Yuqing Wang, Chong Ruan, Ming Zhang, Wenfeng Liang, Wangding Zeng
Abstract: Long-context modeling is crucial for next-generation language models, yet the high computational cost of standard attention mechanisms poses significant computational challenges. Sparse attention offers a promising direction for improving efficiency while maintaining model capabilities. We present NSA, a Natively trainable Sparse Attention mechanism that integrates algorithmic innovations with hardware-aligned optimizations to achieve efficient long-context modeling. NSA employs a dynamic hierarchical sparse strategy, combining coarse-grained token compression with fine-grained token selection to preserve both global context awareness and local precision. Our approach advances sparse attention design with two key innovations: (1) We achieve substantial speedups through arithmetic intensity-balanced algorithm design, with implementation optimizations for modern hardware. (2) We enable end-to-end training, reducing pretraining computation without sacrificing model performance. As shown in Figure 1, experiments show the model pretrained with NSA maintains or exceeds Full Attention models across general benchmarks, long-context tasks, and instruction-based reasoning. Meanwhile, NSA achieves substantial speedups over Full Attention on 64k-length sequences across decoding, forward propagation, and backward propagation, validating its efficiency throughout the model lifecycle.
Authors: Hongye Cao, Yanming Wang, Sijia Jing, Ziyue Peng, Zhixin Bai, Zhe Cao, Meng Fang, Fan Feng, Boyan Wang, Jiaheng Liu, Tianpei Yang, Jing Huo, Yang Gao, Fanyu Meng, Xi Yang, Chao Deng, Junlan Feng
Abstract: With the rapid advancement of Large Language Models (LLMs), the safety of LLMs has been a critical concern requiring precise assessment. Current benchmarks primarily concentrate on single-turn dialogues or a single jailbreak attack method to assess the safety. Additionally, these benchmarks have not taken into account the LLM's capability of identifying and handling unsafe information in detail. To address these issues, we propose a fine-grained benchmark SafeDialBench for evaluating the safety of LLMs across various jailbreak attacks in multi-turn dialogues. Specifically, we design a two-tier hierarchical safety taxonomy that considers 6 safety dimensions and generates more than 4000 multi-turn dialogues in both Chinese and English under 22 dialogue scenarios. We employ 7 jailbreak attack strategies, such as reference attack and purpose reverse, to enhance the dataset quality for dialogue generation. Notably, we construct an innovative assessment framework of LLMs, measuring capabilities in detecting, and handling unsafe information and maintaining consistency when facing jailbreak attacks. Experimental results across 17 LLMs reveal that Yi-34B-Chat and GLM4-9B-Chat demonstrate superior safety performance, while Llama3.1-8B-Instruct and o3-mini exhibit safety vulnerabilities.
Authors: Zhengyan Sheng, Zhihao Du, Shiliang Zhang, Zhijie Yan, Yexin Yang, Zhenhua Ling
Abstract: This paper presents a dual-stream text-to-speech (TTS) model, SyncSpeech, capable of receiving streaming text input from upstream models while simultaneously generating streaming speech, facilitating seamless interaction with large language models. SyncSpeech has the following advantages: Low latency, as it begins generating streaming speech upon receiving the second text token; High efficiency, as it decodes all speech tokens corresponding to the each arrived text token in one step. To achieve this, we propose a temporal masked transformer as the backbone of SyncSpeech, combined with token-level duration prediction to predict speech tokens and the duration for the next step. Additionally, we design a two-stage training strategy to improve training efficiency and the quality of generated speech. We evaluated the SyncSpeech on both English and Mandarin datasets. Compared to the recent dual-stream TTS models, SyncSpeech significantly reduces the first packet delay of speech tokens and accelerates the real-time factor. Moreover, with the same data scale, SyncSpeech achieves performance comparable to that of traditional autoregressive-based TTS models in terms of both speech quality and robustness. Speech samples are available at https://SyncSpeech.github.io/}{https://SyncSpeech.github.io/.
URLs: https://SyncSpeech.github.io/, https://SyncSpeech.github.io/.
Authors: Kun-Hui Lee, Eunhwan Park, Donghoon Han, Seung-Hoon Na
Abstract: Large Language Models (LLMs) excel across a variety of language tasks yet are constrained by limited input lengths and high computational costs. Existing approaches\textemdash such as relative positional encodings (e.g., RoPE, ALiBi) and sliding window mechanisms\textemdash partially alleviate these issues but often require additional training or suffer from performance degradation with longer inputs. In this paper, we introduce \textbf{\textit{CacheFocus}}, a method that enhances length normalization and reduces inference latency without any further training. Our approach leverages query-independent, offline caching to efficiently reuse a Context KV Cache Store. We address the amplification of abnormal token distributions problem by re-positioning cached keys and introducing Layer-Adaptive Cache Pruning to discard low-relevance caches during pre-filling. Additionally, our Adaptive Positional Allocation Strategy dynamically reassigns cache positions to maximize the use of the available positional encoding range. Experiments on the Natural Questions and TriviaQA datasets demonstrate that CacheFocus outperforms alternative methods even when inputs exceed the $4$K limit of the \texttt{LLaMA-2} model, emphasizing its practical effectiveness for long-context LLMs. Moreover, even with large maximum input length of \texttt{Qwen2}, the performance of CacheFocus shows that it maintains consistent performance even as the number of documents increases, effectively managing long-text generation without degradation.
Authors: Wei Yao, Wenkai Yang, Ziqiao Wang, Yankai Lin, Yong Liu
Abstract: As large language models advance toward superhuman performance, ensuring their alignment with human values and abilities grows increasingly complex. Weak-to-strong generalization offers a promising approach by leveraging predictions from weaker models to guide stronger systems, but its effectiveness could be constrained by the inherent noise and inaccuracies in these weak predictions. To address this, we propose a theoretically grounded approach that replaces forward KL divergence-whose mass-covering behavior risks overfitting to imperfect weak signals-with reverse KL divergence. Reverse KL divergence's zero-forcing effect prioritizes high-confidence predictions, effectively mitigating the influence of unreliable weak supervision. Theoretically, we extend existing bounds and derive tighter lower bounds for both forward and reverse KL divergence, establishing that reverse KL achieves at least comparable guarantees to forward KL. Notably, when a sufficiently pre-trained strong model is fine-tuned on the last layer, reverse KL uniquely guarantees that it outperforms its weak supervisor by the magnitude of their disagreement-a guarantee that forward KL cannot provide. Empirically, we demonstrate that reverse KL and reverse cross-entropy enable strong models to consistently outperform those trained with forward KL and standard cross-entropy across most settings, highlighting the practical advantages of these reverse losses.
Authors: Alexandru Lecu, Adrian Groza, Lezan Hawizy
Abstract: Large language models (LLMs) have significantly advanced the field of natural language generation. However, they frequently generate unverified outputs, which compromises their reliability in critical applications. In this study, we propose an innovative framework that combines structured biomedical knowledge with LLMs through a retrieval-augmented generation technique. Our system develops a thorough knowledge graph by identifying and refining causal relationships and named entities from medical abstracts related to age-related macular degeneration (AMD). Using a vector-based retrieval process and a locally deployed language model, our framework produces responses that are both contextually relevant and verifiable, with direct references to clinical evidence. Experimental results show that this method notably decreases hallucinations, enhances factual precision, and improves the clarity of generated responses, providing a robust solution for advanced biomedical chatbot applications.
Authors: Yuanfei Wang, Xiaojie Zhang, Ruihai Wu, Yu Li, Yan Shen, Mingdong Wu, Zhaofeng He, Yizhou Wang, Hao Dong
Abstract: Articulated object manipulation is a critical capability for robots to perform various tasks in real-world scenarios. Composed of multiple parts connected by joints, articulated objects are endowed with diverse functional mechanisms through complex relative motions. For example, a safe consists of a door, a handle, and a lock, where the door can only be opened when the latch is unlocked. The internal structure, such as the state of a lock or joint angle constraints, cannot be directly observed from visual observation. Consequently, successful manipulation of these objects requires adaptive adjustment based on trial and error rather than a one-time visual inference. However, previous datasets and simulation environments for articulated objects have primarily focused on simple manipulation mechanisms where the complete manipulation process can be inferred from the object's appearance. To enhance the diversity and complexity of adaptive manipulation mechanisms, we build a novel articulated object manipulation environment and equip it with 9 categories of objects. Based on the environment and objects, we further propose an adaptive demonstration collection and 3D visual diffusion-based imitation learning pipeline that learns the adaptive manipulation policy. The effectiveness of our designs and proposed method is validated through both simulation and real-world experiments. Our project page is available at: https://adamanip.github.io
Authors: Arka Mukherjee, Shreya Ghosh
Abstract: Multimodal fake news detection typically demands complex architectures and substantial computational resources, posing deployment challenges in real-world settings. We introduce UNITE-FND, a novel framework that reframes multimodal fake news detection as a unimodal text classification task. We propose six specialized prompting strategies with Gemini 1.5 Pro, converting visual content into structured textual descriptions, and enabling efficient text-only models to preserve critical visual information. To benchmark our approach, we introduce Uni-Fakeddit-55k, a curated dataset family of 55,000 samples each, each processed through our multimodal-to-unimodal translation framework. Experimental results demonstrate that UNITE-FND achieves 92.52% accuracy in binary classification, surpassing prior multimodal models while reducing computational costs by over 10x (TinyBERT variant: 14.5M parameters vs. 250M+ in SOTA models). Additionally, we propose a comprehensive suite of five novel metrics to evaluate image-to-text conversion quality, ensuring optimal information preservation. Our results demonstrate that structured text-based representations can replace direct multimodal processing with minimal loss of accuracy, making UNITE-FND a practical and scalable alternative for resource-constrained environments.
Authors: Wenjing Zhang, Xuejiao Lei, Zhaoxiang Liu, Ning Wang, Zhenhong Long, Peijun Yang, Jiaojiao Zhao, Minjie Hua, Chaoyang Ma, Kai Wang, Shiguo Lian
Abstract: Recently, the DeepSeek series of models, leveraging their exceptional reasoning capabilities and open-source strategy, is reshaping the global AI landscape. Despite these advantages, they exhibit significant safety deficiencies. Research conducted by Robust Intelligence, a subsidiary of Cisco, in collaboration with the University of Pennsylvania, revealed that DeepSeek-R1 has a 100\% attack success rate when processing harmful prompts. Additionally, multiple safety companies and research institutions have confirmed critical safety vulnerabilities in this model. As models demonstrating robust performance in Chinese and English, DeepSeek models require equally crucial safety assessments in both language contexts. However, current research has predominantly focused on safety evaluations in English environments, leaving a gap in comprehensive assessments of their safety performance in Chinese contexts. In response to this gap, this study introduces CHiSafetyBench, a Chinese-specific safety evaluation benchmark. This benchmark systematically evaluates the safety of DeepSeek-R1 and DeepSeek-V3 in Chinese contexts, revealing their performance across safety categories. The experimental results quantify the deficiencies of these two models in Chinese contexts, providing key insights for subsequent improvements.
Authors: Wonduk Seo, Seungyong Lee, Daye Kang, Zonghao Yuan, Seunghyun Lee
Abstract: Unprecedented breakthroughs in Large Language Models (LLMs) has amplified its penetration into application of automated visualization code generation. Few-shot prompting and query expansion techniques have notably enhanced data visualization performance, however, still fail to overcome ambiguity and complexity of natural language queries - imposing an inherent burden for manual human intervention. To mitigate such limitations, we propose a holistic framework VisPath : A Multi-Path Reasoning and Feedback-Driven Optimization Framework for Visualization Code Generation, which systematically enhances code quality through structured reasoning and refinement. VisPath is a multi-stage framework, specially designed to handle underspecified queries. To generate a robust final visualization code, it first utilizes initial query to generate diverse reformulated queries via Chain-of-Thought (CoT) prompting, each representing a distinct reasoning path. Refined queries are used to produce candidate visualization scripts, consequently executed to generate multiple images. Comprehensively assessing correctness and quality of outputs, VisPath generates feedback for each image, which are then fed to aggregation module to generate optimal result. Extensive experiments on benchmarks including MatPlotBench and the Qwen-Agent Code Interpreter Benchmark show that VisPath significantly outperforms state-of-the-art (SOTA) methods, increased up to average 17%, offering a more reliable solution for AI-driven visualization code generation.
Authors: Lukas Kuhn, Sari Saba-Sadiya, Gemma Roig
Abstract: Recent research has suggested that the brain is more shallow than previously thought, challenging the traditionally assumed hierarchical structure of the ventral visual pathway. Here, we demonstrate that optimizing convolutional network architectures for brain-alignment via evolutionary neural architecture search results in models with clear representational hierarchies. Despite having random weights, the identified models achieve brain-alignment scores surpassing even those of pretrained classification models - as measured by both regression and representational similarity analysis. Furthermore, through traditional supervised training, architectures optimized for alignment with late ventral regions become competitive classification models. These findings suggest that hierarchical structure is a fundamental mechanism of primate visual processing. Finally, this work demonstrates the potential of neural architecture search as a framework for computational cognitive neuroscience research that could reduce the field's reliance on manually designed convolutional networks.
Authors: Junhao Hu, Wenrui Huang, Weidong Wang, Zhenwen Li, Tiancheng Hu, Zhixia Liu, Xusheng Chen, Tao Xie, Yizhou Shan
Abstract: Large Language Models (LLMs) have demonstrated strong capabilities across various domains, with recent advancements in challenging reasoning tasks such as mathematics and programming. However, solving reasoning tasks often requires long decoding chains (of thoughts), which incur $O(N)$ time and memory consumption, where $N$ is the chain length. To mitigate $O(N)$ time and memory consumption, existing sparsity-based algorithms propose retaining only the most critical token's intermediate data (i.e., key-value cache) and discarding the rest. However, these existing algorithms struggle with the ``impossible trinity'' of accuracy, time, and memory. For example, the state-of-the-art algorithm, Quest, achieves high accuracy with $O(L)$ time but $O(N)$ memory ($L$ is the cache budget, $L \ll N$). To address this issue, in this paper, we identify a new attention pattern during the decode stage of reasoning tasks, where milestone tokens (analogous to lemmas in mathematical proofs) emerge, are utilized, and then become unimportant afterward. Based on this pattern, we propose a new algorithm named RaaS that identifies and retains milestone tokens only until they are no longer needed, achieving high accuracy with $O(L)$ time and $O(L)$ memory complexity.
Authors: Zongzhao Li, Jiacheng Cen, Bing Su, Wenbing Huang, Tingyang Xu, Yu Rong, Deli Zhao
Abstract: Accurately predicting 3D structures and dynamics of physical systems is crucial in scientific applications. Existing approaches that rely on geometric Graph Neural Networks (GNNs) effectively enforce $\mathrm{E}(3)$-equivariance, but they often fall in leveraging extensive broader information. While direct application of Large Language Models (LLMs) can incorporate external knowledge, they lack the capability for spatial reasoning with guaranteed equivariance. In this paper, we propose EquiLLM, a novel framework for representing 3D physical systems that seamlessly integrates E(3)-equivariance with LLM capabilities. Specifically, EquiLLM comprises four key components: geometry-aware prompting, an equivariant encoder, an LLM, and an equivariant adaptor. Essentially, the LLM guided by the instructive prompt serves as a sophisticated invariant feature processor, while 3D directional information is exclusively handled by the equivariant encoder and adaptor modules. Experimental results demonstrate that EquiLLM delivers significant improvements over previous methods across molecular dynamics simulation, human motion simulation, and antibody design, highlighting its promising generalizability.
Authors: Xin Gu, Yaojie Shen, Chenxi Luo, Tiejian Luo, Yan Huang, Yuewei Lin, Heng Fan, Libo Zhang
Abstract: Transformer has attracted increasing interest in STVG, owing to its end-to-end pipeline and promising result. Existing Transformer-based STVG approaches often leverage a set of object queries, which are initialized simply using zeros and then gradually learn target position information via iterative interactions with multimodal features, for spatial and temporal localization. Despite simplicity, these zero object queries, due to lacking target-specific cues, are hard to learn discriminative target information from interactions with multimodal features in complicated scenarios (\e.g., with distractors or occlusion), resulting in degradation. Addressing this, we introduce a novel Target-Aware Transformer for STVG (TA-STVG), which seeks to adaptively generate object queries via exploring target-specific cues from the given video-text pair, for improving STVG. The key lies in two simple yet effective modules, comprising text-guided temporal sampling (TTS) and attribute-aware spatial activation (ASA), working in a cascade. The former focuses on selecting target-relevant temporal cues from a video utilizing holistic text information, while the latter aims at further exploiting the fine-grained visual attribute information of the object from previous target-aware temporal cues, which is applied for object query initialization. Compared to existing methods leveraging zero-initialized queries, object queries in our TA-STVG, directly generated from a given video-text pair, naturally carry target-specific cues, making them adaptive and better interact with multimodal features for learning more discriminative information to improve STVG. In our experiments on three benchmarks, TA-STVG achieves state-of-the-art performance and significantly outperforms the baseline, validating its efficacy.
Authors: Shilong Yang, Qi Zang, Chulong Zhang, Lingfeng Huang, Yaoqin Xie
Abstract: Traditional Chinese acupuncture methods often face controversy in clinical practice due to their high subjectivity. Additionally, current intelligent-assisted acupuncture systems have two major limitations: slow acupoint localization speed and low accuracy. To address these limitations, a new method leverages the excellent inference efficiency of the state-space model Mamba, while retaining the advantages of the attention mechanism in the traditional DETR architecture, to achieve efficient global information integration and provide high-quality feature information for acupoint localization tasks. Furthermore, by employing the concept of residual likelihood estimation, it eliminates the need for complex upsampling processes, thereby accelerating the acupoint localization task. Our method achieved state-of-the-art (SOTA) accuracy on a private dataset of acupoints on the human back, with an average Euclidean distance pixel error (EPE) of 7.792 and an average time consumption of 10.05 milliseconds per localization task. Compared to the second-best algorithm, our method improved both accuracy and speed by approximately 14\%. This significant advancement not only enhances the efficacy of acupuncture treatment but also demonstrates the commercial potential of automated acupuncture robot systems. Access to our method is available at https://github.com/Sohyu1/RT-DEMT
Authors: SeongKu Kang, Bowen Jin, Wonbin Kweon, Yu Zhang, Dongha Lee, Jiawei Han, Hwanjo Yu
Abstract: In specialized fields like the scientific domain, constructing large-scale human-annotated datasets poses a significant challenge due to the need for domain expertise. Recent methods have employed large language models to generate synthetic queries, which serve as proxies for actual user queries. However, they lack control over the content generated, often resulting in incomplete coverage of academic concepts in documents. We introduce Concept Coverage-based Query set Generation (CCQGen) framework, designed to generate a set of queries with comprehensive coverage of the document's concepts. A key distinction of CCQGen is that it adaptively adjusts the generation process based on the previously generated queries. We identify concepts not sufficiently covered by previous queries, and leverage them as conditions for subsequent query generation. This approach guides each new query to complement the previous ones, aiding in a thorough understanding of the document. Extensive experiments demonstrate that CCQGen significantly enhances query quality and retrieval performance.
Authors: Wenxuan Wang, Xiaoyuan Liu, Kuiyi Gao, Jen-tse Huang, Youliang Yuan, Pinjia He, Shuai Wang, Zhaopeng Tu
Abstract: Multimodal Large Language Models (MLLMs) have expanded the capabilities of traditional language models by enabling interaction through both text and images. However, ensuring the safety of these models remains a significant challenge, particularly in accurately identifying whether multimodal content is safe or unsafe-a capability we term safety awareness. In this paper, we introduce MMSafeAware, the first comprehensive multimodal safety awareness benchmark designed to evaluate MLLMs across 29 safety scenarios with 1500 carefully curated image-prompt pairs. MMSafeAware includes both unsafe and over-safety subsets to assess models abilities to correctly identify unsafe content and avoid over-sensitivity that can hinder helpfulness. Evaluating nine widely used MLLMs using MMSafeAware reveals that current models are not sufficiently safe and often overly sensitive; for example, GPT-4V misclassifies 36.1% of unsafe inputs as safe and 59.9% of benign inputs as unsafe. We further explore three methods to improve safety awareness-prompting-based approaches, visual contrastive decoding, and vision-centric reasoning fine-tuning-but find that none achieve satisfactory performance. Our findings highlight the profound challenges in developing MLLMs with robust safety awareness, underscoring the need for further research in this area. All the code and data will be publicly available to facilitate future research.
Authors: Shahriar Kabir Nahin, Rabindra Nath Nandi, Sagor Sarker, Quazi Sarwar Muhtaseem, Md Kowsher, Apu Chandraw Shill, Md Ibrahim, Mehadi Hasan Menon, Tareq Al Muntasir, Firoj Alam
Abstract: In this paper, we present TituLLMs, the first large pretrained Bangla LLMs, available in 1B and 3B parameter sizes. Due to computational constraints during both training and inference, we focused on smaller models. To train TituLLMs, we collected a pretraining dataset of approximately 37 billion tokens. We extended the Llama-3.2 tokenizer to incorporate language- and culture-specific knowledge, which also enables faster training and inference. There was a lack of benchmarking datasets to evaluate LLMs for Bangla. To address this gap, we developed five benchmarking datasets. We benchmarked various LLMs, including TituLLMs, and demonstrated that TituLLMs outperforms its initial multilingual versions. However, this is not always the case, highlighting the complexities of language adaptation. Our work lays the groundwork for adapting existing multilingual open models to other low-resource languages. To facilitate broader adoption and further research, we have made the TituLLMs models and benchmarking datasets publicly available (https://huggingface.co/collections/hishab/titulm-llama-family-6718d31fc1b83529276f490a).
URLs: https://huggingface.co/collections/hishab/titulm-llama-family-6718d31fc1b83529276f490a).
Authors: Haoming Xu, Ningyuan Zhao, Liming Yang, Sendong Zhao, Shumin Deng, Mengru Wang, Bryan Hooi, Nay Oo, Huajun Chen, Ningyu Zhang
Abstract: Current unlearning methods for large language models usually rely on reverse optimization to reduce target token probabilities. However, this paradigm disrupts the subsequent tokens prediction, degrading model performance and linguistic coherence. Moreover, existing evaluation metrics overemphasize contextual forgetting while inadequately assessing response fluency and relevance. To address these challenges, we propose ReLearn, a data augmentation and fine-tuning pipeline for effective unlearning, along with a comprehensive evaluation framework. This framework introduces Knowledge Forgetting Rate (KFR) and Knowledge Retention Rate (KRR) to measure knowledge-level preservation, and Linguistic Score (LS) to evaluate generation quality. Our experiments show that ReLearn successfully achieves targeted forgetting while preserving high-quality output. Through mechanistic analysis, we further demonstrate how reverse optimization disrupts coherent text generation, while ReLearn preserves this essential capability. Code is available at https://github.com/zjunlp/unlearn.
Authors: Yao-Ching Yu, Tsun-Han Chiang, Cheng-Wei Tsai, Chien-Ming Huang, Wen-Kwang Tsao
Abstract: Large Language Models (LLMs) have shown remarkable advancements in specialized fields such as finance, law, and medicine. However, in cybersecurity, we have noticed a lack of open-source datasets, with a particular lack of high-quality cybersecurity pretraining corpora, even though much research indicates that LLMs acquire their knowledge during pretraining. To address this, we present a comprehensive suite of datasets covering all major training stages, including pretraining, instruction fine-tuning, and reasoning distillation with cybersecurity-specific self-reflection data. Extensive ablation studies demonstrate their effectiveness on public cybersecurity benchmarks. In particular, continual pre-training on our dataset yields a 15.88% improvement in the aggregate score, while reasoning distillation leads to a 10% gain in security certification (CISSP). We will release all datasets and trained cybersecurity LLMs under the ODC-BY and MIT licenses to encourage further research in the community. For access to all datasets and model weights, please refer to https://huggingface.co/collections/trendmicro-ailab/primus-67b1fd27052b802b4af9d243.
URLs: https://huggingface.co/collections/trendmicro-ailab/primus-67b1fd27052b802b4af9d243.
Authors: Yizhi Liu, Balaji Padmanabhan, Siva Viswanathan
Abstract: While deepfake technologies have predominantly been criticized for potential misuse, our study demonstrates their significant potential as tools for detecting, measuring, and mitigating biases in key societal domains. By employing deepfake technology to generate controlled facial images, we extend the scope of traditional correspondence studies beyond mere textual manipulations. This enhancement is crucial in scenarios such as pain assessments, where subjective biases triggered by sensitive features in facial images can profoundly affect outcomes. Our results reveal that deepfakes not only maintain the effectiveness of correspondence studies but also introduce groundbreaking advancements in bias measurement and correction techniques. This study emphasizes the constructive role of deepfake technologies as essential tools for advancing societal equity and fairness.
Authors: Yixin Ou, Yunzhi Yao, Ningyu Zhang, Hui Jin, Jiacheng Sun, Shumin Deng, Zhenguo Li, Huajun Chen
Abstract: Despite exceptional capabilities in knowledge-intensive tasks, Large Language Models (LLMs) face a critical gap in understanding how they internalize new knowledge, particularly how to structurally embed acquired knowledge in their neural computations. We address this issue through the lens of knowledge circuit evolution, identifying computational subgraphs that facilitate knowledge storage and processing. Our systematic analysis of circuit evolution throughout continual pre-training reveals several key findings: (1) the acquisition of new knowledge is influenced by its relevance to pre-existing knowledge; (2) the evolution of knowledge circuits exhibits a distinct phase shift from formation to optimization; (3) the evolution of knowledge circuits follows a deep-to-shallow pattern. These insights not only advance our theoretical understanding of the mechanisms of new knowledge acquisition in LLMs, but also provide potential implications for improving continual pre-training strategies to enhance model performance. Code and data will be available at https://github.com/zjunlp/DynamicKnowledgeCircuits.
Authors: Jinwei Lu, Yuanfeng Song, Zhiqian Qin, Haodi Zhang, Chen Zhang, Raymond Chi-Wing Wong
Abstract: NoSQL databases have become increasingly popular due to their outstanding performance in handling large-scale, unstructured, and semi-structured data, highlighting the need for user-friendly interfaces to bridge the gap between non-technical users and complex database queries. In this paper, we introduce the Text-to-NoSQL task, which aims to convert natural language queries into NoSQL queries, thereby lowering the technical barrier for non-expert users. To promote research in this area, we developed a novel automated dataset construction process and released a large-scale and open-source dataset for this task, named TEND (short for Text-to-NoSQL Dataset). Additionally, we designed a SLM (Small Language Model)-assisted and RAG (Retrieval-augmented Generation)-assisted multi-step framework called SMART, which is specifically designed for Text-to-NoSQL conversion. To ensure comprehensive evaluation of the models, we also introduced a detailed set of metrics that assess the model's performance from both the query itself and its execution results. Our experimental results demonstrate the effectiveness of our approach and establish a benchmark for future research in this emerging field. We believe that our contributions will pave the way for more accessible and intuitive interactions with NoSQL databases.
Authors: Wenxuan Wang, Zizhan Ma, Zheng Wang, Chenghan Wu, Wenting Chen, Xiang Li, Yixuan Yuan
Abstract: Large Language Models (LLMs) are transforming healthcare through the development of LLM-based agents that can understand, reason about, and assist with medical tasks. This survey provides a comprehensive review of LLM-based agents in medicine, examining their architectures, applications, and challenges. We analyze the key components of medical agent systems, including system profiles, clinical planning mechanisms, medical reasoning frameworks, and external capacity enhancement. The survey covers major application scenarios such as clinical decision support, medical documentation, training simulations, and healthcare service optimization. We discuss evaluation frameworks and metrics used to assess these agents' performance in healthcare settings. While LLM-based agents show promise in enhancing healthcare delivery, several challenges remain, including hallucination management, multimodal integration, implementation barriers, and ethical considerations. The survey concludes by highlighting future research directions, including advances in medical reasoning inspired by recent developments in LLM architectures, integration with physical systems, and improvements in training simulations. This work provides researchers and practitioners with a structured overview of the current state and future prospects of LLM-based agents in medicine.
Authors: Zhaoyang Larry Jin, Mehdi Maasoumy, Yimin Liu, Zeshi Zheng, Zizhuo Ren
Abstract: Today's global supply chains face growing challenges due to rapidly changing market conditions, increased network complexity and inter-dependency, and dynamic uncertainties in supply, demand, and other factors. To combat these challenges, organizations employ Material Requirements Planning (MRP) software solutions to set inventory stock buffers - for raw materials, work-in-process goods, and finished products - to help them meet customer service levels. However, holding excess inventory further complicates operations and can lock up millions of dollars of capital that could be otherwise deployed. Furthermore, most commercially available MRP solutions fall short in considering uncertainties and do not result in optimal solutions for modern enterprises. At C3 AI, we fundamentally reformulate the inventory management problem as a constrained stochastic optimization. We then propose a simulation-optimization framework that minimizes inventory and related costs while maintaining desired service levels. The framework's goal is to find the optimal reorder parameters that minimize costs subject to a pre-defined service-level constraint and all other real-world operational constraints. These optimal reorder parameters can be fed back into an MRP system to drive optimal order placement, or used to place optimal orders directly. This approach has proven successful in reducing inventory levels by 10-35 percent, resulting in hundreds of millions of dollars of economic benefit for major enterprises at a global scale.
Authors: Christian Camacho-Villal\'on, Marco Dorigo, Thomas St\"utzle
Abstract: Hybrid metaheuristics are powerful techniques for solving difficult optimization problems that exploit the strengths of different approaches in a single implementation. For algorithm designers, however, creating hybrid metaheuristic implementations has become increasingly challenging due to the vast number of design options available in the literature and the fact that they often rely on their knowledge and intuition to come up with new algorithm designs. In this paper, we propose a modular metaheuristic software framework, called METAFOR, that can be coupled with an automatic algorithm configuration tool to automatically design hybrid metaheuristics. METAFOR is specifically designed to hybridize Particle Swarm Optimization, Differential Evolution and Covariance Matrix Adaptation-Evolution Strategy, and includes a local search module that allows their execution to be interleaved with a subordinate local search. We use the configuration tool irace to automatically generate 17 different metaheuristic implementations and evaluate their performance on a diverse set of continuous optimization problems. Our results show that, across all the considered problem classes, automatically generated hybrid implementations are able to outperform configured single-approach implementations, while these latter offer advantages on specific classes of functions. We provide useful insights on the type of hybridization that works best for specific problem classes, the algorithm components that contribute to the performance of the algorithms, and the advantages and disadvantages of two well-known instance separation strategies, creating stratified training set using a fix percentage and leave-one-class-out cross-validation.
Authors: Mohammad Reza Rezaei, Adji Bousso Dieng
Abstract: Retrieval-augmented generation (RAG) enhances large language models (LLMs) for domain-specific question-answering (QA) tasks by leveraging external knowledge sources. However, traditional RAG systems primarily focus on relevance-based retrieval and often struggle with redundancy, especially when reasoning requires connecting information from multiple sources. This paper introduces Vendi-RAG, a framework based on an iterative process that jointly optimizes retrieval diversity and answer quality. This joint optimization leads to significantly higher accuracy for multi-hop QA tasks. Vendi-RAG leverages the Vendi Score (VS), a flexible similarity-based diversity metric, to promote semantic diversity in document retrieval. It then uses an LLM judge that evaluates candidate answers, generated after a reasoning step, and outputs a score that the retriever uses to balance relevance and diversity among the retrieved documents during each iteration. Experiments on three challenging datasets -- HotpotQA, MuSiQue, and 2WikiMultiHopQA -- demonstrate Vendi-RAG's effectiveness in multi-hop reasoning tasks. The framework achieves significant accuracy improvements over traditional single-step and multi-step RAG approaches, with accuracy increases reaching up to +4.2% on HotpotQA, +4.1% on 2WikiMultiHopQA, and +1.3% on MuSiQue compared to Adaptive-RAG, the current best baseline. The benefits of Vendi-RAG are even more pronounced as the number of retrieved documents increases. Finally, we evaluated Vendi-RAG across different LLM backbones, including GPT-3.5, GPT-4, and GPT-4o-mini, and observed consistent improvements, demonstrating that the framework's advantages are model-agnostic.
Authors: Yue Tu, Mark Dubynskyi, Mohammad Mohammadisiahroudi, Ekaterina Riashchentceva, Jinglei Cheng, Dmitry Ryashchentsev, Tam\'as Terlaky, Junyu Liu
Abstract: Quantum computing, a prominent non-Von Neumann paradigm beyond Moore's law, can offer superpolynomial speedups for certain problems. Yet its advantages in efficiency for tasks like machine learning remain under investigation, and quantum noise complicates resource estimations and classical comparisons. We provide a detailed estimation of space, time, and energy resources for fault-tolerant superconducting devices running the Harrow-Hassidim-Lloyd (HHL) algorithm, a quantum linear system solver relevant to linear algebra and machine learning. Excluding memory and data transfer, possible quantum advantages over the classical conjugate gradient method could emerge at $N \approx 2^{33} \sim 2^{48}$ or even lower, requiring ${O}(10^5)$ physical qubits, ${O}(10^{12}\sim10^{13})$ Joules, and ${O}(10^6)$ seconds under surface code fault-tolerance with three types of magic state distillation (15-1, 116-12, 225-1). Key parameters include condition number, sparsity, and precision $\kappa, s\approx{O}(10\sim100)$, $\epsilon\sim0.01$, and physical error $10^{-5}$. Our resource estimator adjusts $N, \kappa, s, \epsilon$, providing a map of quantum-classical boundaries and revealing where a practical quantum advantage may arise. Our work quantitatively determine how advanced a fault-tolerant quantum computer should be to achieve possible, significant benefits on problems related to real-world.
Authors: Somnath Banerjee, Sayan Layek, Pratyush Chatterjee, Animesh Mukherjee, Rima Hazra
Abstract: Ensuring consistent safety across multiple languages remains a significant challenge for large language models (LLMs). We introduce Soteria, a lightweight yet powerful strategy that locates and minimally adjusts the "functional heads" most responsible for harmful content generation in each language. By altering only a fraction of parameters, Soteria drastically reduces policy violations without sacrificing overall model performance, even in low-resource settings. To rigorously evaluate our approach, we also present XThreatBench, a specialized multilingual dataset capturing fine-grained harmful behaviors drawn from real policy guidelines. Experiments with leading open-source LLMs (e.g., Llama, Qwen, Mistral) show that Soteria consistently improves safety metrics across high-, mid-, and low-resource languages. These findings highlight a promising path toward scalable, linguistically attuned, and ethically aligned LLMs worldwide.
Authors: Samuele Bortolotti, Emanuele Marconato, Paolo Morettin, Andrea Passerini, Stefano Teso
Abstract: Concept-based Models are neural networks that learn a concept extractor to map inputs to high-level concepts and an inference layer to translate these into predictions. Ensuring these modules produce interpretable concepts and behave reliably in out-of-distribution is crucial, yet the conditions for achieving this remain unclear. We study this problem by establishing a novel connection between Concept-based Models and reasoning shortcuts (RSs), a common issue where models achieve high accuracy by learning low-quality concepts, even when the inference layer is fixed and provided upfront. Specifically, we first extend RSs to the more complex setting of Concept-based Models and then derive theoretical conditions for identifying both the concepts and the inference layer. Our empirical results highlight the impact of reasoning shortcuts and show that existing methods, even when combined with multiple natural mitigation strategies, often fail to meet these conditions in practice.
Authors: Mengying Wang, Hanchao Ma, Yiyang Bian, Yangxin Fan, Yinghui Wu
Abstract: Preparing high-quality datasets required by various data-driven AI and machine learning models has become a cornerstone task in data-driven analysis. Conventional data discovery methods typically integrate datasets towards a single pre-defined quality measure that may lead to bias for downstream tasks. This paper introduces MODis, a framework that discovers datasets by optimizing multiple user-defined, model-performance measures. Given a set of data sources and a model, MODis selects and integrates data sources into a skyline dataset, over which the model is expected to have the desired performance in all the performance measures. We formulate MODis as a multi-goal finite state transducer, and derive three feasible algorithms to generate skyline datasets. Our first algorithm adopts a "reduce-from-universal" strategy, that starts with a universal schema and iteratively prunes unpromising data. Our second algorithm further reduces the cost with a bi-directional strategy that interleaves data augmentation and reduction. We also introduce a diversification algorithm to mitigate the bias in skyline datasets. We experimentally verify the efficiency and effectiveness of our skyline data discovery algorithms, and showcase their applications in optimizing data science pipelines.
Authors: Zeyu He, Saniya Naphade, Ting-Hao 'Kenneth' Huang
Abstract: Millions of users prompt large language models (LLMs) for various tasks, but how good are people at prompt engineering? Do users actually get closer to their desired outcome over multiple iterations of their prompts? These questions are crucial when no gold-standard labels are available to measure progress. This paper investigates a scenario in LLM-powered data labeling, "prompting in the dark," where users iteratively prompt LLMs to label data without using manually-labeled benchmarks. We developed PromptingSheet, a Google Sheets add-on that enables users to compose, revise, and iteratively label data through spreadsheets. Through a study with 20 participants, we found that prompting in the dark was highly unreliable-only 9 participants improved labeling accuracy after four or more iterations. Automated prompt optimization tools like DSPy also struggled when few gold labels were available. Our findings highlight the importance of gold labels and the needs, as well as the risks, of automated support in human prompt engineering, providing insights for future tool design.
Authors: Dana Calacci, Varun Nagaraj Rao, Samantha Dalal, Catherine Di, Kok-Wei Pua, Andrew Schwartz, Danny Spitzberg, Andr\'es Monroy-Hern\'andez
Abstract: Rideshare workers experience unpredictable working conditions due to gig work platforms' reliance on opaque AI and algorithmic systems. In response to these challenges, we found that labor organizers want data to help them advocate for legislation to increase the transparency and accountability of these platforms. To address this need, we collaborated with a Colorado-based rideshare union to develop FairFare, a tool that crowdsources and analyzes workers' data to estimate the take rate -- the percentage of the rider price retained by the rideshare platform. We deployed FairFare with our partner organization that collaborated with us in collecting data on 76,000+ trips from 45 drivers over 18 months. During evaluation interviews, organizers reported that FairFare helped influence the bill language and passage of Colorado Senate Bill 24-75, calling for greater transparency and data disclosure of platform operations, and create a national narrative. Finally, we reflect on complexities of translating quantitative data into policy outcomes, nature of community based audits, and design implications for future transparency tools.
Authors: Parisa Fard Moshiri, Murat Arda Onsu, Poonam Lohan, Burak Kantarci, Emil Janulewicz
Abstract: Efficient Service Function Chain (SFC) provisioning and Virtual Network Function (VNF) placement are critical for enhancing network performance in modern architectures such as Software-Defined Networking (SDN) and Network Function Virtualization (NFV). While Deep Reinforcement Learning (DRL) aids decision-making in dynamic network environments, its reliance on structured inputs and predefined rules limits adaptability in unforeseen scenarios. Additionally, incorrect actions by a DRL agent may require numerous training iterations to correct, potentially reinforcing suboptimal policies and degrading performance. This paper integrates DRL with Language Models (LMs), specifically Bidirectional Encoder Representations from Transformers (BERT) and DistilBERT, to enhance network management. By feeding final VNF allocations from DRL into the LM, the system can process and respond to queries related to SFCs, DCs, and VNFs, enabling real-time insights into resource utilization, bottleneck detection, and future demand planning. The LMs are fine-tuned to our domain-specific dataset using Low-Rank Adaptation (LoRA). Results show that BERT outperforms DistilBERT with a lower test loss (0.28 compared to 0.36) and higher confidence (0.83 compared to 0.74), though BERT requires approximately 46% more processing time.
Authors: Aashish Anantha Ramakrishnan, Aadarsh Anantha Ramakrishnan, Dongwon Lee
Abstract: Multimodal Large Language Models (MLLMs) are renowned for their superior instruction-following and reasoning capabilities across diverse problem domains. However, existing benchmarks primarily focus on assessing factual and logical correctness in downstream tasks, with limited emphasis on evaluating MLLMs' ability to interpret pragmatic cues and intermodal relationships. To address this gap, we assess the competency of MLLMs in performing Multimodal Discourse Analysis (MDA) using Coherence Relations. Our benchmark, CORDIAL, encompasses a broad spectrum of Coherence Relations across 3 different discourse domains at varying levels of granularity. Through our experiments on 10+ MLLMs employing different prompting strategies, we show that even top models like Gemini 1.5 Pro and GPT-4o fail to match the performance of simple classifier-based baselines. This study emphasizes the need to move beyond similarity-based metrics and adopt a discourse-driven framework for evaluating MLLMs, providing a more nuanced assessment of their capabilities. The benchmark and code are available at: https://github.com/aashish2000/CORDIAL.
Authors: Jiaxiang Wang, Haote Xu, Xiaolu Chen, Haodi Xu, Yue Huang, Xinghao Ding, Xiaotong Tu
Abstract: Anomaly detection (AD) in 3D point clouds is crucial in a wide range of industrial applications, especially in various forms of precision manufacturing. Considering the industrial demand for reliable 3D AD, several methods have been developed. However, most of these approaches typically require training separate models for each category, which is memory-intensive and lacks flexibility. In this paper, we propose a novel Point-Language model with dual-prompts for 3D ANomaly dEtection (PLANE). The approach leverages multi-modal prompts to extend the strong generalization capabilities of pre-trained Point-Language Models (PLMs) to the domain of 3D point cloud AD, achieving impressive detection performance across multiple categories using a single model. Specifically, we propose a dual-prompt learning method, incorporating both text and point cloud prompts. The method utilizes a dynamic prompt creator module (DPCM) to produce sample-specific dynamic prompts, which are then integrated with class-specific static prompts for each modality, effectively driving the PLMs. Additionally, based on the characteristics of point cloud data, we propose a pseudo 3D anomaly generation method (Ano3D) to improve the model's detection capabilities in an unsupervised setting. Experimental results demonstrate that the proposed method, which is under the multi-class-one-model paradigm, achieves a +8.7%/+17% gain on anomaly detection and localization performance as compared to the state-of-the-art one-class-one-model methods for the Anomaly-ShapeNet dataset, and obtains +4.3%/+4.1% gain for the Real3D-AD dataset. Code will be available upon publication.
Authors: Yiyi Chen, Qiongkai Xu, Johannes Bjerva
Abstract: With the growing popularity of Large Language Models (LLMs) and vector databases, private textual data is increasingly processed and stored as numerical embeddings. However, recent studies have proven that such embeddings are vulnerable to inversion attacks, where original text is reconstructed to reveal sensitive information. Previous research has largely assumed access to millions of sentences to train attack models, e.g., through data leakage or nearly unrestricted API access. With our method, a single data point is sufficient for a partially successful inversion attack. With as little as 1k data samples, performance reaches an optimum across a range of black-box encoders, without training on leaked data. We present a Few-shot Textual Embedding Inversion Attack using ALignment and GENeration (ALGEN), by aligning victim embeddings to the attack space and using a generative model to reconstruct text. We find that ALGEN attacks can be effectively transferred across domains and languages, revealing key information. We further examine a variety of defense mechanisms against ALGEN, and find that none are effective, highlighting the vulnerabilities posed by inversion attacks. By significantly lowering the cost of inversion and proving that embedding spaces can be aligned through one-step optimization, we establish a new textual embedding inversion paradigm with broader applications for embedding alignment in NLP.
Authors: Minbyul Jeong, Jungho Cho, Minsoo Khang, Dawoon Jung, Teakgyu Hong
Abstract: System messages play a crucial role in interactions with large language models (LLMs), often serving as prompts to initiate conversations. Through system messages, users can assign specific roles, perform intended tasks, incorporate background information, specify various output formats and communication styles. Despite such versatility, publicly available data are often lack system messages and subject to strict license constraints in the industry field. Manual labeling of publicly available data with system messages that align with user instructions demands significant resources. In view of such challenges, our work introduces SysGen, a pipeline for generating system messages with better aligned assistant responses from the supervised fine-tuning dataset without system messages. Training on SysGen data has demonstrated substantial improvements in the alignment of model responses with system messages and user instructions, as demonstrated across various open-source models on the Multifacet benchmark, while maintaining minimal impact on other unseen benchmarks such as Open LLM Leaderboard 2. Our qualitative analysis highlights the importance of diverse system messages to ensure better adaptability across different contexts.
Authors: Yuchen Zhang, Jian Zhou
Abstract: Inverse generation problems, such as denoising without ground truth observations, is a critical challenge in many scientific inquiries and real-world applications. While recent advances in generative models like diffusion models, conditional flow matching, and consistency models achieved impressive results by casting generation as denoising problems, they cannot be directly used for inverse generation without access to clean data. Here we introduce Inverse Flow (IF), a novel framework that enables using these generative models for inverse generation problems including denoising without ground truth. Inverse Flow can be flexibly applied to nearly any continuous noise distribution and allows complex dependencies. We propose two algorithms for learning Inverse Flows, Inverse Flow Matching (IFM) and Inverse Consistency Model (ICM). Notably, to derive the computationally efficient, simulation-free inverse consistency model objective, we generalized consistency training to any forward diffusion processes or conditional flows, which have applications beyond denoising. We demonstrate the effectiveness of IF on synthetic and real datasets, outperforming prior approaches while enabling noise distributions that previous methods cannot support. Finally, we showcase applications of our techniques to fluorescence microscopy and single-cell genomics data, highlighting IF's utility in scientific problems. Overall, this work expands the applications of powerful generative models to inversion generation problems.
Authors: Rongwu Xu, Xiaojian Li, Shuo Chen, Wei Xu
Abstract: Large language models (LLMs) are evolving into autonomous decision-makers, raising concerns about catastrophic risks in high-stakes scenarios, particularly in Chemical, Biological, Radiological and Nuclear (CBRN) domains. Based on the insight that such risks can originate from trade-offs between the agent's Helpful, Harmlessness and Honest (HHH) goals, we build a novel three-stage evaluation framework, which is carefully constructed to effectively and naturally expose such risks. We conduct 14,400 agentic simulations across 12 advanced LLMs, with extensive experiments and analysis. Results reveal that LLM agents can autonomously engage in catastrophic behaviors and deception, without being deliberately induced. Furthermore, stronger reasoning abilities often increase, rather than mitigate, these risks. We also show that these agents can violate instructions and superior commands. On the whole, we empirically prove the existence of catastrophic risks in autonomous LLM agents. We will release our code upon request.
Authors: Zirui He, Haiyan Zhao, Yiran Qiao, Fan Yang, Ali Payani, Jing Ma, Mengnan Du
Abstract: The ability of large language models (LLMs) to follow instructions is crucial for their practical applications, yet the underlying mechanisms remain poorly understood. This paper presents a novel framework that leverages sparse autoencoders (SAE) to interpret how instruction following works in these models. We demonstrate how the features we identify can effectively steer model outputs to align with given instructions. Through analysis of SAE latent activations, we identify specific latents responsible for instruction following behavior. Our findings reveal that instruction following capabilities are encoded by a distinct set of instruction-relevant SAE latents. These latents both show semantic proximity to relevant instructions and demonstrate causal effects on model behavior. Our research highlights several crucial factors for achieving effective steering performance: precise feature identification, the role of final layer, and optimal instruction positioning. Additionally, we demonstrate that our methodology scales effectively across SAEs and LLMs of varying sizes.
Authors: Jack Gallifant, Shan Chen, Kuleen Sasse, Hugo Aerts, Thomas Hartvigsen, Danielle S. Bitterman
Abstract: Sparse Autoencoders (SAEs) provide potentials for uncovering structured, human-interpretable representations in Large Language Models (LLMs), making them a crucial tool for transparent and controllable AI systems. We systematically analyze SAE for interpretable feature extraction from LLMs in safety-critical classification tasks. Our framework evaluates (1) model-layer selection and scaling properties, (2) SAE architectural configurations, including width and pooling strategies, and (3) the effect of binarizing continuous SAE activations. SAE-derived features achieve macro F1 > 0.8, outperforming hidden-state and BoW baselines while demonstrating cross-model transfer from Gemma 2 2B to 9B-IT models. These features generalize in a zero-shot manner to cross-lingual toxicity detection and visual classification tasks. Our analysis highlights the significant impact of pooling strategies and binarization thresholds, showing that binarization offers an efficient alternative to traditional feature selection while maintaining or improving performance. These findings establish new best practices for SAE-based interpretability and enable scalable, transparent deployment of LLMs in real-world applications. Full repo: https://github.com/shan23chen/MOSAIC.
Authors: Zhengxiang Wang, Veronika Makarova, Zhi Li, Jordan Kodner, Owen Rambow
Abstract: The paper explores the performance of LLMs in the context of multi-dimensional analytic writing assessments, i.e. their ability to provide both scores and comments based on multiple assessment criteria. Using a corpus of literature reviews written by L2 graduate students and assessed by human experts against 9 analytic criteria, we prompt several popular LLMs to perform the same task under various conditions. To evaluate the quality of feedback comments, we apply a novel feedback comment quality evaluation framework. This framework is interpretable, cost-efficient, scalable, and reproducible, compared to existing methods that rely on manual judgments. We find that LLMs can generate reasonably good and generally reliable multi-dimensional analytic assessments. We release our corpus for reproducibility.
Authors: Guanghao Zhou, Panjia Qiu, Mingyuan Fan, Cen Chen, Mingyuan Chu, Xin Zhang, Jun Zhou
Abstract: Despite explicit alignment efforts for large language models (LLMs), they can still be exploited to trigger unintended behaviors, a phenomenon known as "jailbreaking." Current jailbreak attack methods mainly focus on discrete prompt manipulations targeting closed-source LLMs, relying on manually crafted prompt templates and persuasion rules. However, as the capabilities of open-source LLMs improve, ensuring their safety becomes increasingly crucial. In such an environment, the accessibility of model parameters and gradient information by potential attackers exacerbates the severity of jailbreak threats. To address this research gap, we propose a novel \underline{C}ontext-\underline{C}oherent \underline{J}ailbreak \underline{A}ttack (CCJA). We define jailbreak attacks as an optimization problem within the embedding space of masked language models. Through combinatorial optimization, we effectively balance the jailbreak attack success rate with semantic coherence. Extensive evaluations show that our method not only maintains semantic consistency but also surpasses state-of-the-art baselines in attack effectiveness. Additionally, by integrating semantically coherent jailbreak prompts generated by our method into widely used black-box methodologies, we observe a notable enhancement in their success rates when targeting closed-source commercial LLMs. This highlights the security threat posed by open-source LLMs to commercial counterparts. We will open-source our code if the paper is accepted.
Authors: Zhongwei Chen, Zhao-Xu Yang, Hai-Jun Rong
Abstract: UAV-View Geo-Localization (UVGL) aims to ascertain the precise location of a UAV by retrieving the most similar GPS-tagged satellite image. However, existing methods predominantly rely on supervised learning paradigms that necessitate annotated paired data for training, which incurs substantial annotation costs and impedes large-scale deployment. To overcome this limitation, we propose the Dynamic Memory-Driven and Neighborhood Information Learning (DMNIL) network, a lightweight end-to-end self-supervised framework for UAV-view geo-localization. The DMNIL framework utilizes a dual-path clustering-based contrastive learning architecture as its baseline to model intra-view structural relationships, enhancing feature consistency and discriminability. Additionally, a dynamic memory-driven hierarchical learning module is proposed to progressively mine local and global information, reinforcing multi-level feature associations to improve model robustness. To bridge the domain gap between UAV and satellite views, we design an information-consistent evolutionary learning mechanism that systematically explores latent correlations within intra-view neighborhoods and across cross-view domains, ultimately constructing a unified cross-view feature representation space. Extensive experiments on three benchmarks (University-1652, SUES-200, and DenseUAV) demonstrate that DMNIL achieves competitive performance against state-of-the-art supervised methods while maintaining computational efficiency. Notably, this superiority is attained without relying on paired training data, underscoring the framework's practicality for real-world deployment. Codes will be released soon.
Authors: Jongho Kim, Seung-won Hwang
Abstract: Despite the advanced capabilities of large language models (LLMs), their temporal reasoning ability remains underdeveloped. Prior works have highlighted this limitation, particularly in maintaining temporal consistency when understanding events. For example, models often confuse mutually exclusive temporal relations like ``before'' and ``after'' between events and make inconsistent predictions. In this work, we tackle the issue of temporal inconsistency in LLMs by proposing a novel counterfactual prompting approach. Our method generates counterfactual questions and enforces collective constraints, enhancing the model's consistency. We evaluate our method on multiple datasets, demonstrating significant improvements in event ordering for explicit and implicit events and temporal commonsense understanding by effectively addressing temporal inconsistencies.
Authors: Taewoo Kim, Youngwoo Yoon, Jaehong Kim
Abstract: Robotic catching has traditionally focused on single-handed systems, which are limited in their ability to handle larger or more complex objects. In contrast, bimanual catching offers significant potential for improved dexterity and object handling but introduces new challenges in coordination and control. In this paper, we propose a novel framework for learning dexterous bimanual catching skills using Heterogeneous-Agent Reinforcement Learning (HARL). Our approach introduces an adversarial reward scheme, where a throw agent increases the difficulty of throws-adjusting speed-while a catch agent learns to coordinate both hands to catch objects under these evolving conditions. We evaluate the framework in simulated environments using 15 different objects, demonstrating robustness and versatility in handling diverse objects. Our method achieved approximately a 2x increase in catching reward compared to single-agent baselines across 15 diverse objects.
Authors: Cen-Jhih Li, Aditya Bhaskara
Abstract: Fine-tuning is an important step in adapting foundation models such as large language models to downstream tasks. To make this step more accessible to users with limited computational budgets, it is crucial to develop fine-tuning methods that are memory and computationally efficient. Sparse Fine-tuning (SFT) and Low-rank adaptation (LoRA) are two frameworks that have emerged for addressing this problem and have been adopted widely in practice. In this work, we develop a new SFT framework, based on ideas from neural network pruning. At a high level, we first identify "important" neurons/nodes using feature importance metrics from network pruning (specifically, we use the structural pruning method), and then perform fine-tuning by restricting to weights involving these neurons. Using experiments on common language tasks, we demonstrate that our method significantly improves the memory efficiency of SFT without increasing training time complexity and implementation complexity, while achieving accuracy comparable to state-of-the-art methods such as LoRA and its variants.
Authors: Kimia Ramezan, Alireza Amiri Bavandpour, Yifei Yuan, Clemencia Siro, Mohammad Aliannejadi
Abstract: Conversational query clarification enables users to refine their search queries through interactive dialogue, improving search effectiveness. Traditional approaches rely on text-based clarifying questions, which often fail to capture complex user preferences, particularly those involving visual attributes. While recent work has explored single-turn multi-modal clarification with images alongside text, such methods do not fully support the progressive nature of user intent refinement over multiple turns. Motivated by this, we introduce the Multi-turn Multi-modal Clarifying Questions (MMCQ) task, which combines text and visual modalities to refine user queries in a multi-turn conversation. To facilitate this task, we create a large-scale dataset named ClariMM comprising over 13k multi-turn interactions and 33k question-answer pairs containing multi-modal clarifying questions. We propose Mario, a retrieval framework that employs a two-phase ranking strategy: initial retrieval with BM25, followed by a multi-modal generative re-ranking model that integrates textual and visual information from conversational history. Our experiments show that multi-turn multi-modal clarification outperforms uni-modal and single-turn approaches, improving MRR by 12.88%. The gains are most significant in longer interactions, demonstrating the value of progressive refinement for complex queries.
Authors: Zihao Wang, Victor Veitch
Abstract: A basic aspiration for interpretability research in large language models is to "localize" semantically meaningful behaviors to particular components within the LLM. There are various heuristics for finding candidate locations within the LLM. Once a candidate localization is found, it can be assessed by editing the internal representations at the corresponding localization and checking whether this induces model behavior that is consistent with the semantic interpretation of the localization. The question we address here is: how strong is the evidence provided by such edits? To assess localization, we want to assess the effect of the optimal intervention at a particular location. The key new technical tool is a way of adapting LLM alignment techniques to find such optimal localized edits. With this tool in hand, we give an example where the edit-based evidence for localization appears strong, but where localization clearly fails. Indeed, we find that optimal edits at random localizations can be as effective as aligning the full model. In aggregate, our results suggest that merely observing that localized edits induce targeted changes in behavior provides little to no evidence that these locations actually encode the target behavior.
Authors: Ivo Gollini Navarrete, Nicolas Mauricio Cuadrado, Jose Renato Restom, Martin Tak\'a\v{c}, Samuel Horv\'ath
Abstract: Pruning offers a promising solution to mitigate the associated costs and environmental impact of deploying large deep neural networks (DNNs). Traditional approaches rely on computationally expensive trained models or time-consuming iterative prune-retrain cycles, undermining their utility in resource-constrained settings. To address this issue, we build upon the established principles of saliency (LeCun et al., 1989) and connection sensitivity (Lee et al., 2018) to tackle the challenging problem of one-shot pruning neural networks (NNs) before training (PBT) at initialization. We introduce Fisher-Taylor Sensitivity (FTS), a computationally cheap and efficient pruning criterion based on the empirical Fisher Information Matrix (FIM) diagonal, offering a viable alternative for integrating first- and second-order information to identify a model's structurally important parameters. Although the FIM-Hessian equivalency only holds for convergent models that maximize the likelihood, recent studies (Karakida et al., 2019) suggest that, even at initialization, the FIM captures essential geometric information of parameters in overparameterized NNs, providing the basis for our method. Finally, we demonstrate empirically that layer collapse, a critical limitation of data-dependent pruning methodologies, is easily overcome by pruning within a single training epoch after initialization. We perform experiments on ResNet18 and VGG19 with CIFAR-10 and CIFAR-100, widely used benchmarks in pruning research. Our method achieves competitive performance against state-of-the-art techniques for one-shot PBT, even under extreme sparsity conditions. Our code is made available to the public.
Authors: Xun Zhu, Zheng Zhang, Xi Chen, Yiming Shi, Miao Li, Ji Wu
Abstract: With the rapid advancements in multi-modal large language models (MLLMs), connectors play a pivotal role in bridging diverse modalities and enhancing model performance. However, the design and evolution of connectors have not been comprehensively analyzed, leaving gaps in understanding how these components function and hindering the development of more powerful connectors. In this survey, we systematically review the current progress of connectors in MLLMs and present a structured taxonomy that categorizes connectors into atomic operations (mapping, compression, mixture of experts) and holistic designs (multi-layer, multi-encoder, multi-modal scenarios), highlighting their technical contributions and advancements. Furthermore, we discuss several promising research frontiers and challenges, including high-resolution input, dynamic compression, guide information selection, combination strategy, and interpretability. This survey is intended to serve as a foundational reference and a clear roadmap for researchers, providing valuable insights into the design and optimization of next-generation connectors to enhance the performance and adaptability of MLLMs.
Authors: Yanyan Wang, Kechen Song, Yuyuan Liu, Shuai Ma, Yunhui Yan, Gustavo Carneiro
Abstract: Semi-supervised 3D medical image segmentation aims to achieve accurate segmentation using few labelled data and numerous unlabelled data. The main challenge in the design of semi-supervised learning methods consists in the effective use of the unlabelled data for training. A promising solution consists of ensuring consistent predictions across different views of the data, where the efficacy of this strategy depends on the accuracy of the pseudo-labels generated by the model for this consistency learning strategy. In this paper, we introduce a new methodology to produce high-quality pseudo-labels for a consistency learning strategy to address semi-supervised 3D medical image segmentation. The methodology has three important contributions. The first contribution is the Cooperative Rectification Learning Network (CRLN) that learns multiple prototypes per class to be used as external knowledge priors to adaptively rectify pseudo-labels at the voxel level. The second contribution consists of the Dynamic Interaction Module (DIM) to facilitate pairwise and cross-class interactions between prototypes and multi-resolution image features, enabling the production of accurate voxel-level clues for pseudo-label rectification. The third contribution is the Cooperative Positive Supervision (CPS), which optimises uncertain representations to align with unassertive representations of their class distributions, improving the model's accuracy in classifying uncertain regions. Extensive experiments on three public 3D medical segmentation datasets demonstrate the effectiveness and superiority of our semi-supervised learning method.
Authors: Guanlin Li, Yuki Arase, Noel Crespi
Abstract: Text simplification is crucial for improving accessibility and comprehension for English as a Second Language (ESL) learners. This study goes a step further and aims to facilitate ESL learners' language acquisition by simplification. Specifically, we propose simplifying complex sentences to appropriate levels for learners while also increasing vocabulary coverage of the target level in the simplifications. We achieve this without a parallel corpus by conducting reinforcement learning on a large language model. Our method employs token-level and sentence-level rewards, and iteratively trains the model on its self-generated outputs to guide the model to search for simplification hypotheses that satisfy the target attributes. Experiment results on CEFR-SP and TurkCorpus datasets show that the proposed method can effectively increase the frequency and diversity of vocabulary of the target level by more than $20\%$ compared to baseline models, while maintaining high simplification quality.
Authors: Jiecheng Zhou, Ding Tang, Rong Fu, Boni Hu, Haoran Xu, Yi Wang, Zhilin Pei, Zhongling Su, Liang Liu, Xingcheng Zhang, Weiming Zhang
Abstract: The burgeoning computational demands for training large language models (LLMs) necessitate efficient methods, including quantized training, which leverages low-bit arithmetic operations to reduce costs. While FP8 precision has shown potential, leveraging FP4 remains challenging due to inherent quantization errors and limited representation capability. Based on the Transformer architecture, we present an FP4 training scheme for LLMs, overcoming these obstacles through mixed-precision quantization strategies tailed for different modules and training stages. This allows us to apply the precision level suitable to distinct components within the model, ensuring that multi-head attention and linear layers are handled appropriately. Our pretraining recipe ensures stability in backpropagation by incorporating fine-grained quantization methods with a target precision training schedule. Experimental results demonstrate that our FP4 training scheme achieves accuracy comparable to BF16 and FP8, with smaller theoretical computational cost. With the advent of next-generation hardware supporting FP4, our method sets the foundation for efficient ultra-low precision training.
Authors: Ahmed Bensaoud, Jugal Kalita
Abstract: The rapid expansion of Internet of Things (IoT) devices has increased the risk of cyber-attacks, making effective detection essential for securing IoT networks. This work introduces a novel approach combining Self-Organizing Maps (SOMs), Deep Belief Networks (DBNs), and Autoencoders to detect known and previously unseen attack patterns. A comprehensive evaluation using simulated and real-world traffic data is conducted, with models optimized via Particle Swarm Optimization (PSO). The system achieves an accuracy of up to 99.99% and Matthews Correlation Coefficient (MCC) values exceeding 99.50%. Experiments on NSL-KDD, UNSW-NB15, and CICIoT2023 confirm the model's strong performance across diverse attack types. These findings suggest that the proposed method enhances IoT security by identifying emerging threats and adapting to evolving attack strategies.
Authors: Xiangxiang Cui, Zhongyu Li, Xiayue Fan, Peng Huang, Ying Wang, Meng Yang, Shi Chang, Jihua Zhu
Abstract: The intersection of medical imaging and artificial intelligence has become an important research direction in intelligent medical treatment, particularly in the analysis of medical images using deep learning for clinical diagnosis. Despite the advances, existing keyframe classification methods lack extraction of time series features, while ultrasonic video classification based on three-dimensional convolution requires uniform frame numbers across patients, resulting in poor feature extraction efficiency and model classification performance. This study proposes a novel video classification method based on CNN and LSTM, introducing NLP's long and short sentence processing scheme into video classification for the first time. The method reduces CNN-extracted image features to 1x512 dimension, followed by sorting and compressing feature vectors for LSTM training. Specifically, feature vectors are sorted by patient video frame numbers and populated with padding value 0 to form variable batches, with invalid padding values compressed before LSTM training to conserve computing resources. Experimental results demonstrate that our variable-frame CNNLSTM method outperforms other approaches across all metrics, showing improvements of 3-6% in F1 score and 1.5% in specificity compared to keyframe methods. The variable-frame CNNLSTM also achieves better accuracy and precision than equal-frame CNNLSTM. These findings validate the effectiveness of our approach in classifying variable-frame ultrasound videos and suggest potential applications in other medical imaging modalities.
Authors: Huanxuan Liao, Shizhu He, Yupu Hao, Jun Zhao, Kang Liu
Abstract: Continual learning (CL) is essential for Large Language Models (LLMs) to adapt to evolving real-world demands, yet they are susceptible to catastrophic forgetting (CF). While traditional CF solutions rely on expensive data rehearsal, recent rehearsal-free methods employ model-based and regularization-based strategies to address this issue. However, these approaches often neglect the model's plasticity, which is crucial to achieving optimal performance on newly learned tasks. Consequently, a key challenge in CL is striking a balance between preserving plasticity and mitigating CF. To tackle this challenge, we propose the $\textbf{D}$ecomposed $\textbf{A}$ttention-based $\textbf{T}$ask $\textbf{A}$daptation (DATA), which explicitly decouples and learns both task-specific and task-shared knowledge using high-rank and low-rank task adapters (e.g., LoRAs). For new tasks, DATA dynamically adjusts the weights of adapters of different ranks based on their relevance and distinction from previous tasks, allowing the model to acquire new task-specific skills while effectively retaining previously learned knowledge. Specifically, we implement a decomposed component weighting strategy comprising learnable components that collectively generate attention-based weights, allowing the model to integrate and utilize diverse knowledge from each DATA. Extensive experiments on three widely used benchmarks demonstrate that our proposed method achieves state-of-the-art performance. Notably, our approach significantly enhances model plasticity and mitigates CF by extending learnable components and employing stochastic restoration during training iterations.
Authors: Runxuan Liu, Bei Luo, Jiaqi Li, Baoxin Wang, Ming Liu, Dayong Wu, Shijin Wang, Bing Qin
Abstract: Large language models (LLMs) have shown remarkable capabilities in natural language processing. However, in knowledge graph question answering tasks (KGQA), there remains the issue of answering questions that require multi-hop reasoning. Existing methods rely on entity vector matching, but the purpose of the question is abstract and difficult to match with specific entities. As a result, it is difficult to establish reasoning paths to the purpose, which leads to information loss and redundancy. To address this issue, inspired by human reverse thinking, we propose Ontology-Guided Reverse Thinking (ORT), a novel framework that constructs reasoning paths from purposes back to conditions. ORT operates in three key phases: (1) using LLM to extract purpose labels and condition labels, (2) constructing label reasoning paths based on the KG ontology, and (3) using the label reasoning paths to guide knowledge retrieval. Experiments on the WebQSP and CWQ datasets show that ORT achieves state-of-the-art performance and significantly enhances the capability of LLMs for KGQA.
Authors: Janak M. Patel, Milad Ramezankhani, Anirudh Deodhar, Dagnachew Birru
Abstract: Simulation and optimization are crucial for advancing the engineering design of complex systems and processes. Traditional optimization methods require substantial computational time and effort due to their reliance on resource-intensive simulations, such as finite element analysis, and the complexity of rigorous optimization algorithms. Data-agnostic AI-based surrogate models, such as Physics-Informed Neural Operators (PINOs), offer a promising alternative to these conventional simulations, providing drastically reduced inference time, unparalleled data efficiency, and zero-shot super-resolution capability. However, the predictive accuracy of these models is often constrained to small, low-dimensional design spaces or systems with relatively simple dynamics. To address this, we introduce a novel Physics-Informed DeepONet (PIDON) architecture, which extends the capabilities of conventional neural operators to effectively model the nonlinear behavior of complex engineering systems across high-dimensional design spaces and a wide range of dynamic design configurations. This new architecture outperforms existing SOTA models, enabling better predictions across broader design spaces. Leveraging PIDON's differentiability, we integrate a gradient-based optimization approach using the Adam optimizer to efficiently determine optimal design variables. This forms an end-to-end gradient-based optimization framework that accelerates the design process while enhancing scalability and efficiency. We demonstrate the effectiveness of this framework in the optimization of aerospace-grade composites curing processes achieving a 3x speedup in obtaining optimal design variables compared to gradient-free methods. Beyond composites processing, the proposed model has the potential to be used as a scalable and efficient optimization tool for broader applications in advanced engineering and digital twin systems.
Authors: Changchun Liu, Kai Zhang, Junzhe Jiang, Zixiao Kong, Qi Liu, Enhong Chen
Abstract: Chinese Spelling Correction (CSC) is a critical task in natural language processing, aimed at detecting and correcting spelling errors in Chinese text. This survey provides a comprehensive overview of CSC, tracing its evolution from pre-trained language models to large language models, and critically analyzing their respective strengths and weaknesses in this domain. Moreover, we further present a detailed examination of existing benchmark datasets, highlighting their inherent challenges and limitations. Finally, we propose promising future research directions, particularly focusing on leveraging the potential of LLMs and their reasoning capabilities for improved CSC performance. To the best of our knowledge, this is the first comprehensive survey dedicated to the field of CSC. We believe this work will serve as a valuable resource for researchers, fostering a deeper understanding of the field and inspiring future advancements.
Authors: Suparshva Jain, Amit Sangroya, Lovekesh Vig
Abstract: Generating multiple counterfactual explanations for different modes within a class presents a significant challenge, as these modes are distinct yet converge under the same classification. Diffusion probabilistic models (DPMs) have demonstrated a strong ability to capture the underlying modes of data distributions. In this paper, we harness the power of a Diffusion Autoencoder to generate multiple distinct counterfactual explanations. By clustering in the latent space, we uncover the directions corresponding to the different modes within a class, enabling the generation of diverse and meaningful counterfactuals. We introduce a novel methodology, DifCluE, which consistently identifies these modes and produces more reliable counterfactual explanations. Our experimental results demonstrate that DifCluE outperforms the current state-of-the-art in generating multiple counterfactual explanations, offering a significant advance- ment in model interpretability.
Authors: Zhen Zhang, Yifan Yang, Kai Zhen, Nathan Susanj, Athanasios Mouchtaris, Siegfried Kunzmann, Zheng Zhang
Abstract: Large language models have demonstrated exceptional capabilities across diverse tasks, but their fine-tuning demands significant memory, posing challenges for resource-constrained environments. Zeroth-order (ZO) optimization provides a memory-efficient alternative by eliminating the need for backpropagation. However, ZO optimization suffers from high gradient variance, and prior research has largely focused on single-task learning, leaving its application to multi-task learning unexplored. Multi-task learning is crucial for leveraging shared knowledge across tasks to improve generalization, yet it introduces unique challenges under ZO settings, such as amplified gradient variance and collinearity. In this paper, we present MaZO, the first framework specifically designed for multi-task LLM fine-tuning under ZO optimization. MaZO tackles these challenges at the parameter level through two key innovations: a weight importance metric to identify critical parameters and a multi-task weight update mask to selectively update these parameters, reducing the dimensionality of the parameter space and mitigating task conflicts. Experiments demonstrate that MaZO achieves state-of-the-art performance, surpassing even multi-task learning methods designed for first-order optimization.
Authors: Di Wu, Xian Wei, Guang Chen, Hao Shen, Xiangfeng Wang, Wenhao Li, Bo Jin
Abstract: Embodied multi-agent systems (EMAS) have attracted growing attention for their potential to address complex, real-world challenges in areas such as logistics and robotics. Recent advances in foundation models pave the way for generative agents capable of richer communication and adaptive problem-solving. This survey provides a systematic examination of how EMAS can benefit from these generative capabilities. We propose a taxonomy that categorizes EMAS by system architectures and embodiment modalities, emphasizing how collaboration spans both physical and virtual contexts. Central building blocks, perception, planning, communication, and feedback, are then analyzed to illustrate how generative techniques bolster system robustness and flexibility. Through concrete examples, we demonstrate the transformative effects of integrating foundation models into embodied, multi-agent frameworks. Finally, we discuss challenges and future directions, underlining the significant promise of EMAS to reshape the landscape of AI-driven collaboration.
Authors: Hao Li, Hao Jiang, Yuke Zheng, Hao Sun, Wenying Gong
Abstract: Polarization and fragmentation in social media amplify user biases, making it increasingly important to understand the evolution of opinions. Opinion dynamics provide interpretability for studying opinion evolution, yet incorporating these insights into predictive models remains challenging. This challenge arises due to the inherent complexity of the diversity of opinion fusion rules and the difficulty in capturing equilibrium states while avoiding over-smoothing. This paper constructs a unified opinion dynamics model to integrate different opinion fusion rules and generates corresponding synthetic datasets. To fully leverage the advantages of unified opinion dynamics, we introduces UniGO, a framework for modeling opinion evolution on graphs. Using a coarsen-refine mechanism, UniGO efficiently models opinion dynamics through a graph neural network, mitigating over-smoothing while preserving equilibrium phenomena. UniGO leverages pretraining on synthetic datasets, which enhances its ability to generalize to real-world scenarios, providing a viable paradigm for applications of opinion dynamics. Experimental results on both synthetic and real-world datasets demonstrate UniGO's effectiveness in capturing complex opinion formation processes and predicting future evolution. The pretrained model also shows strong generalization capability, validating the benefits of using synthetic data to boost real-world performance.
Authors: Juantao Zhong, Daoyuan Wu, Ye Liu, Maoyi Xie, Yang Liu, Yi Li, Ning Liu
Abstract: DeFi (Decentralized Finance) is one of the most important applications of today's cryptocurrencies and smart contracts. It manages hundreds of billions in Total Value Locked (TVL) on-chain, yet it remains susceptible to common DeFi price manipulation attacks. Despite state-of-the-art (SOTA) systems like DeFiRanger and DeFort, we found that they are less effective to non-standard price models in custom DeFi protocols, which account for 44.2% of the 95 DeFi price manipulation attacks reported over the past three years. In this paper, we introduce the first LLM-based approach, DeFiScope, for detecting DeFi price manipulation attacks in both standard and custom price models. Our insight is that large language models (LLMs) have certain intelligence to abstract price calculation from code and infer the trend of token price changes based on the extracted price models. To further strengthen LLMs in this aspect, we leverage Foundry to synthesize on-chain data and use it to fine-tune a DeFi price-specific LLM. Together with the high-level DeFi operations recovered from low-level transaction data, DeFiScope detects various DeFi price manipulations according to systematically mined patterns. Experimental results show that DeFiScope achieves a high precision of 96% and a recall rate of 80%, significantly outperforming SOTA approaches. Moreover, we evaluate DeFiScope's cost-effectiveness and demonstrate its practicality by helping our industry partner confirm 147 real-world price manipulation attacks, including discovering 81 previously unknown historical incidents.
Authors: Lior Cohen, Kaixin Wang, Bingyi Kang, Uri Gadot, Shie Mannor
Abstract: Token-based world models emerged as a promising modular framework, modeling dynamics over token streams while optimizing tokenization separately. While successful in visual environments with discrete actions (e.g., Atari games), their broader applicability remains uncertain. In this paper, we introduce $\text{M}^{\text{3}}$, a $\textbf{m}$odular $\textbf{w}$orld $\textbf{m}$odel that extends this framework, enabling flexible combinations of observation and action modalities through independent modality-specific components. $\text{M}^{\text{3}}$ integrates several improvements from existing literature to enhance agent performance. Through extensive empirical evaluation across diverse benchmarks, $\text{M}^{\text{3}}$ achieves state-of-the-art sample efficiency for planning-free world models. Notably, among these methods, it is the first to reach a human-level median score on Atari 100K, with superhuman performance on 13 games. We $\href{https://github.com/leor-c/M3}{\text{open-source our code and weights}}$.
Authors: Hui Huang, Jiaheng Liu, Yancheng He, Shilong Li, Bing Xu, Conghui Zhu, Muyun Yang, Tiejun Zhao
Abstract: Complex instruction-following with elaborate constraints is imperative for Large Language Models (LLMs). While existing methods have constructed data for complex instruction alignment, they all rely on a more advanced model, especially GPT-4, limiting their application. In this paper, we propose a Multi-granularity Self-Contrastive Training (MuSC) framework, to improve the complex instruction alignment without relying on a stronger model. Our method is conducted on both coarse and fine granularity. On coarse-granularity, we construct constraint-aware preference data based on instruction decomposition and recombination. On fine-granularity, we perform token-aware preference optimization with dynamic token-level supervision. Our method is evaluated on open-sourced models, and experiment results show our method achieves significant improvement on both complex and general instruction-following benchmarks, surpassing previous self-alignment methods.
Authors: Smit Desai, Jessie Chin, Dakuo Wang, Benjamin Cowan, Michael Twidale
Abstract: Metaphors play a critical role in shaping user experiences with Voice User Interfaces (VUIs), yet existing designs often rely on static, human-centric metaphors that fail to adapt to diverse contexts and user needs. This paper introduces Metaphor-Fluid Design, a novel approach that dynamically adjusts metaphorical representations based on conversational use-contexts. We compare this approach to a Default VUI, which characterizes the present implementation of commercial VUIs commonly designed around the persona of an assistant, offering a uniform interaction style across contexts. In Study 1 (N=130), metaphors were mapped to four key use-contexts-commands, information seeking, sociality, and error recovery-along the dimensions of formality and hierarchy, revealing distinct preferences for task-specific metaphorical designs. Study 2 (N=91) evaluates a Metaphor-Fluid VUI against a Default VUI, showing that the Metaphor-Fluid VUI enhances perceived intention to adopt, enjoyment, and likability by aligning better with user expectations for different contexts. However, individual differences in metaphor preferences highlight the need for personalization. These findings challenge the one-size-fits-all paradigm of VUI design and demonstrate the potential of Metaphor-Fluid Design to create more adaptive and engaging human-AI interactions.
Authors: Yue Xu, Chengyan Fu, Li Xiong, Sibei Yang, Wenjie Wang
Abstract: Pre-training large language models (LLMs) on vast text corpora enhances natural language processing capabilities but risks encoding social biases, particularly gender bias. While parameter-modification methods like fine-tuning mitigate bias, they are resource-intensive, unsuitable for closed-source models, and lack adaptability to evolving societal norms. Instruction-based approaches offer flexibility but often compromise task performance. To address these limitations, we propose $\textit{FaIRMaker}$, an automated and model-independent framework that employs an $\textbf{auto-search and refinement}$ paradigm to adaptively generate Fairwords, which act as instructions integrated into input queries to reduce gender bias and enhance response quality. Extensive experiments demonstrate that $\textit{FaIRMaker}$ automatically searches for and dynamically refines Fairwords, effectively mitigating gender bias while preserving task integrity and ensuring compatibility with both API-based and open-source LLMs.
Authors: Runqi Wang, Caoyuan Ma, Jian Zhao, Hanrui Xu, Dongfang Sun, Haoyang Chen, Lin Xiong, Zheng Wang, Xuelong Li
Abstract: With the rapid advancement of game and film production, generating interactive motion from texts has garnered significant attention due to its potential to revolutionize content creation processes. In many practical applications, there is a need to impose strict constraints on the motion range or trajectory of virtual characters. However, existing methods that rely solely on textual input face substantial challenges in accurately capturing the user's intent, particularly in specifying the desired trajectory. As a result, the generated motions often lack plausibility and accuracy. Moreover, existing trajectory - based methods for customized motion generation rely on retraining for single - actor scenarios, which limits flexibility and adaptability to different datasets, as well as interactivity in two-actor motions. To generate interactive motion following specified trajectories, this paper decouples complex motion into a Leader - Follower dynamic, inspired by role allocation in partner dancing. Based on this framework, this paper explores the motion range refinement process in interactive motion generation and proposes a training-free approach, integrating a Pace Controller and a Kinematic Synchronization Adapter. The framework enhances the ability of existing models to generate motion that adheres to trajectory by controlling the leader's movement and correcting the follower's motion to align with the leader. Experimental results show that the proposed approach, by better leveraging trajectory information, outperforms existing methods in both realism and accuracy.
Authors: Gaurav Srivastava, Shuxiang Cao, Xuan Wang
Abstract: Reasoning has long been viewed as an emergent property of large language models (LLMs), appearing at or above a certain scale ($\sim$100B parameters). However, recent studies challenge this assumption, showing that small language models (SLMs) can also achieve competitive reasoning performance. SLMs are increasingly favored for their efficiency and deployability. However, there is a lack of systematic study on the reasoning abilities of diverse SLMs, including those trained from scratch or derived from LLMs through quantization, pruning, and distillation. This raises a critical question: Can SLMs achieve reasoning abilities comparable to LLMs? In this work, we systematically survey, benchmark, and analyze 72 SLMs from six model families across 14 reasoning benchmarks. For reliable evaluation, we examine four evaluation methods and compare four LLM judges against human evaluations on 800 data points. We repeat all experiments three times to ensure a robust performance assessment. Additionally, we analyze the impact of different prompting strategies in small models. Beyond accuracy, we also evaluate model robustness under adversarial conditions and intermediate reasoning steps. Our findings challenge the assumption that scaling is the only way to achieve strong reasoning. Instead, we foresee a future where SLMs with strong reasoning capabilities can be developed through structured training or post-training compression. They can serve as efficient alternatives to LLMs for reasoning-intensive tasks.
Authors: Congkai Xie, Shuo Cai, Wenjun Wang, Pengxiang Li, Zhijie Sang, Kejing Yang, Yiming Zhang, Zhen Li, Guanghao Zhu, Zeyu Liu, Yang Yu, Yuhang Liu, Su Lu, Baoyi He, Qi Zhou, Xiaotian Han, Jianbo Yuan, Shengyu Zhang, Fei Wu, Hongxia Yang
Abstract: Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) have made significant advancements in reasoning capabilities. However, they still face challenges such as high computational demands and privacy concerns. This paper focuses on developing efficient Small Language Models (SLMs) and Multimodal Small Language Models (MSLMs) that retain competitive reasoning abilities. We introduce a novel training pipeline that enhances reasoning capabilities and facilitates deployment on edge devices, achieving state-of-the-art performance while minimizing development costs. \InfR~ aims to advance AI systems by improving reasoning, reducing adoption barriers, and addressing privacy concerns through smaller model sizes. Resources are available at https://github. com/Reallm-Labs/InfiR.
URLs: https://github.
Authors: Birger Moell, Johan Boye
Abstract: Large Language Models (LLMs) have made significant strides in natural language generation but often face challenges in tasks requiring precise calculations and structural analysis. This paper investigates the performance of state-of-the-art LLMs on language complexity measurement tasks, through the computation of the LIX readability metric and Average Dependency Distance (ADD). Using Swedish high school and university-level essays, we evaluate the models' abilities to compute LIX scores and perform dependency parsing, comparing their results to established ground truths. Our findings reveal that while all models demonstrate some capacity for these tasks, ChatGPT-o1-mini performs most consistently, achieving the highest accuracy in both LIX computation and dependency parsing. Additionally, we observe a strong significant correlation -0.875 p 0.026 (N=6) between the models' accuracy in computing LIX and their overall performance on the Massive Multitask Language Understanding (MMLU) benchmark. These results suggest that language complexity measurement abilities can serve as a noisy zero-shot proxies for assessing the general capabilities of LLMs, providing a practical method for model evaluation without the need for extensive benchmarking datasets.
Authors: Boshko Koloski, Andrei Margeloiu, Xiangjian Jiang, Bla\v{z} \v{S}krlj, Nikola Simidjievski, Mateja Jamnik
Abstract: Tabular deep-learning methods require embedding numerical and categorical input features into high-dimensional spaces before processing them. Existing methods deal with this heterogeneous nature of tabular data by employing separate type-specific encoding approaches. This limits the cross-table transfer potential and the exploitation of pre-trained knowledge. We propose a novel approach that first transforms tabular data into text, and then leverages pre-trained representations from LLMs to encode this data, resulting in a plug-and-play solution to improv ing deep-learning tabular methods. We demonstrate that our approach improves accuracy over competitive models, such as MLP, ResNet and FT-Transformer, by validating on seven classification datasets.
Authors: Hongye Qiu, Yue Xu, Meikang Qiu, Wenjie Wang
Abstract: Large Language Models (LLMs) exhibit strong natural language processing capabilities but also inherit and amplify societal biases, including gender bias, raising fairness concerns. Existing debiasing methods face significant limitations: parameter tuning requires access to model weights, prompt-based approaches often degrade model utility, and optimization-based techniques lack generalizability. To address these challenges, we propose DR.GAP (Demonstration and Reasoning for Gender-Aware Prompting), an automated and model-agnostic approach that mitigates gender bias while preserving model performance. DR.GAP selects bias-revealing examples and generates structured reasoning to guide models toward more impartial responses. Extensive experiments on coreference resolution and QA tasks across multiple LLMs (GPT-3.5, Llama3, and Llama2-Alpaca) demonstrate its effectiveness, generalization ability, and robustness. DR.GAP can generalize to vision-language models (VLMs), achieving significant bias reduction.
Authors: Fatemeh Mohammadi, Marta Annamaria Tamborini, Paolo Ceravolo, Costanza Nardocci, Samira Maghool
Abstract: This paper is a collaborative effort between Linguistics, Law, and Computer Science to evaluate stereotypes and biases in automated translation systems. We advocate gender-neutral translation as a means to promote gender inclusion and improve the objectivity of machine translation. Our approach focuses on identifying gender bias in English-to-Italian translations. First, we define gender bias following human rights law and linguistics literature. Then we proceed by identifying gender-specific terms such as she/lei and he/lui as key elements. We then evaluate the cosine similarity between these target terms and others in the dataset to reveal the model's perception of semantic relations. Using numerical features, we effectively evaluate the intensity and direction of the bias. Our findings provide tangible insights for developing and training gender-neutral translation algorithms.
Authors: Xiaoyi Dong, Jian Cheng, Xi Sheryl Zhang
Abstract: The Soft Actor-Critic (SAC) algorithm with a Gaussian policy has become a mainstream implementation for realizing the Maximum Entropy Reinforcement Learning (MaxEnt RL) objective, which incorporates entropy maximization to encourage exploration and enhance policy robustness. While the Gaussian policy performs well on simpler tasks, its exploration capacity and potential performance in complex multi-goal RL environments are limited by its inherent unimodality. In this paper, we employ the diffusion model, a powerful generative model capable of capturing complex multimodal distributions, as the policy representation to fulfill the MaxEnt RL objective, developing a method named MaxEnt RL with Diffusion Policy (MaxEntDP). Our method enables efficient exploration and brings the policy closer to the optimal MaxEnt policy. Experimental results on Mujoco benchmarks show that MaxEntDP outperforms the Gaussian policy and other generative models within the MaxEnt RL framework, and performs comparably to other state-of-the-art diffusion-based online RL algorithms. Our code is available at https://github.com/diffusionyes/MaxEntDP.
Authors: Yuxia Wang, Rui Xing, Jonibek Mansurov, Giovanni Puccetti, Zhuohan Xie, Minh Ngoc Ta, Jiahui Geng, Jinyan Su, Mervat Abassy, Saad El Dine Ahmed, Kareem Elozeiri, Nurkhan Laiyk, Maiya Goloburda, Tarek Mahmoud, Raj Vardhan Tomar, Alexander Aziz, Ryuto Koike, Masahiro Kaneko, Artem Shelmanov, Ekaterina Artemova, Vladislav Mikhailov, Akim Tsvigun, Alham Fikri Aji, Nizar Habash, Iryna Gurevych, Preslav Nakov
Abstract: Prior studies have shown that distinguishing text generated by large language models (LLMs) from human-written one is highly challenging, and often no better than random guessing. To verify the generalizability of this finding across languages and domains, we perform an extensive case study to identify the upper bound of human detection accuracy. Across 16 datasets covering 9 languages and 9 domains, 19 annotators achieved an average detection accuracy of 87.6%, thus challenging previous conclusions. We find that major gaps between human and machine text lie in concreteness, cultural nuances, and diversity. Prompting by explicitly explaining the distinctions in the prompts can partially bridge the gaps in over 50% of the cases. However, we also find that humans do not always prefer human-written text, particularly when they cannot clearly identify its source.
Authors: Sarthak Mittal, Yoshua Bengio, Nikolay Malkin, Guillaume Lajoie
Abstract: Bayesian and frequentist inference are two fundamental paradigms in statistical estimation. Bayesian methods treat hypotheses as random variables, incorporating priors and updating beliefs via Bayes' theorem, whereas frequentist methods assume fixed but unknown hypotheses, relying on estimators like maximum likelihood. While extensive research has compared these approaches, the frequentist paradigm of obtaining point estimates has become predominant in deep learning, as Bayesian inference is challenging due to the computational complexity and the approximation gap of posterior estimation methods. However, a good understanding of trade-offs between the two approaches is lacking in the regime of amortized estimators, where in-context learners are trained to estimate either point values via maximum likelihood or maximum a posteriori estimation, or full posteriors using normalizing flows, score-based diffusion samplers, or diagonal Gaussian approximations, conditioned on observations. To help resolve this, we conduct a rigorous comparative analysis spanning diverse problem settings, from linear models to shallow neural networks, with a robust evaluation framework assessing both in-distribution and out-of-distribution generalization on tractable tasks. Our experiments indicate that amortized point estimators generally outperform posterior inference, though the latter remain competitive in some low-dimensional problems, and we further discuss why this might be the case.
Authors: Pietro Barbiero, Giuseppe Marra, Gabriele Ciravegna, David Debot, Francesco De Santis, Michelangelo Diligenti, Mateo Espinosa Zarlenga, Francesco Giannini
Abstract: We formalize a novel modeling framework for achieving interpretability in deep learning, anchored in the principle of inference equivariance. While the direct verification of interpretability scales exponentially with the number of variables of the system, we show that this complexity can be mitigated by treating interpretability as a Markovian property and employing neural re-parametrization techniques. Building on these insights, we propose a new modeling paradigm -- neural generation and interpretable execution -- that enables scalable verification of equivariance. This paradigm provides a general approach for designing Neural Interpretable Reasoners that are not only expressive but also transparent.
Authors: Habib Larian, Faramarz Safi-Esfahani
Abstract: With the rapid expansion of the Internet of Things (IoT), sensors, smartphones, and wearables have become integral to daily life, powering smart applications in home automation, healthcare, and intelligent transportation. However, these advancements face significant challenges due to latency and bandwidth constraints imposed by traditional cloud based machine learning (ML) frameworks. The need for innovative solutions is evident as cloud computing struggles with increased latency and network congestion. Previous attempts to offload parts of the ML pipeline to edge and cloud layers have yet to fully resolve these issues, often worsening system response times and network congestion due to the computational limitations of edge devices. In response to these challenges, this study introduces the InTec (Integrated Things Edge Computing) framework, a groundbreaking innovation in IoT architecture. Unlike existing methods, InTec fully leverages the potential of a three tier architecture by strategically distributing ML tasks across the Things, Edge, and Cloud layers. This comprehensive approach enables real time data processing at the point of data generation, significantly reducing latency, optimizing network traffic, and enhancing system reliability. InTec effectiveness is validated through empirical evaluation using the MHEALTH dataset for human motion detection in smart homes, demonstrating notable improvements in key metrics: an 81.56 percent reduction in response time, a 10.92 percent decrease in network traffic, a 9.82 percent improvement in throughput, a 21.86 percent reduction in edge energy consumption, and a 25.83 percent reduction in cloud energy consumption. These advancements establish InTec as a new benchmark for scalable, responsive, and energy efficient IoT applications, demonstrating its potential to revolutionize how the ML pipeline is integrated into Edge AI (EI) systems.
Authors: Yi Wang, Fenghua Weng, Sibei Yang, Zhan Qin, Minlie Huang, Wenjie Wang
Abstract: Large Language Models (LLMs) are widely applied in decision making, but their deployment is threatened by jailbreak attacks, where adversarial users manipulate model behavior to bypass safety measures. Existing defense mechanisms, such as safety fine-tuning and model editing, either require extensive parameter modifications or lack precision, leading to performance degradation on general tasks, which is unsuitable to post-deployment safety alignment. To address these challenges, we propose DELMAN (Dynamic Editing for LLMs JAilbreak DefeNse), a novel approach leveraging direct model editing for precise, dynamic protection against jailbreak attacks. DELMAN directly updates a minimal set of relevant parameters to neutralize harmful behaviors while preserving the model's utility. To avoid triggering a safe response in benign context, we incorporate KL-divergence regularization to ensure the updated model remains consistent with the original model when processing benign queries. Experimental results demonstrate that DELMAN outperforms baseline methods in mitigating jailbreak attacks while preserving the model's utility, and adapts seamlessly to new attack instances, providing a practical and efficient solution for post-deployment model protection.
Authors: Linjie Mu, Zhongzhen Huang, Shengqian Qin, Yakun Zhu, Shaoting Zhang, Xiaofan Zhang
Abstract: Large vision-language models (LVLMs) have shown great promise in medical applications, particularly in visual question answering (MedVQA) and diagnosis from medical images. However, existing datasets and models often fail to consider critical aspects of medical diagnostics, such as the integration of historical records and the analysis of disease progression over time. In this paper, we introduce MMXU (Multimodal and MultiX-ray Understanding), a novel dataset for MedVQA that focuses on identifying changes in specific regions between two patient visits. Unlike previous datasets that primarily address single-image questions, MMXU enables multi-image questions, incorporating both current and historical patient data. We demonstrate the limitations of current LVLMs in identifying disease progression on MMXU-\textit{test}, even those that perform well on traditional benchmarks. To address this, we propose a MedRecord-Augmented Generation (MAG) approach, incorporating both global and regional historical records. Our experiments show that integrating historical records significantly enhances diagnostic accuracy by at least 20\%, bridging the gap between current LVLMs and human expert performance. Additionally, we fine-tune models with MAG on MMXU-\textit{dev}, which demonstrates notable improvements. We hope this work could illuminate the avenue of advancing the use of LVLMs in medical diagnostics by emphasizing the importance of historical context in interpreting medical images. Our dataset is released at \href{https://github.com/linjiemu/MMXU}{https://github.com/linjiemu/MMXU}.
URLs: https://github.com/linjiemu/MMXU, https://github.com/linjiemu/MMXU
Authors: Antoine Boutet, Victor Morel
Abstract: Although mobile devices benefit users in their daily lives in numerous ways, they also raise several privacy concerns. For instance, they can reveal sensitive information that can be inferred from location data. This location data is shared through service providers as well as mobile applications. Understanding how and with whom users share their location data -- as well as users' perception of the underlying privacy risks --, are important notions to grasp in order to design usable privacy-enhancing technologies. In this work, we perform a quantitative and qualitative analysis of smartphone users' awareness, perception and self-reported behavior towards location data-sharing through a survey of n=99 young adult participants (i.e., digital natives). We compare stated practices with actual behaviors to better understand their mental models, and survey participants' understanding of privacy risks before and after the inspection of location traces and the information that can be inferred therefrom. Our empirical results show that participants have risky privacy practices: about 54% of participants underestimate the number of mobile applications to which they have granted access to their data, and 33% forget or do not think of revoking access to their data. Also, by using a demonstrator to perform inferences from location data, we observe that slightly more than half of participants (57%) are surprised by the extent of potentially inferred information, and that 47% intend to reduce access to their data via permissions as a result of using the demonstrator. Last, a majority of participants have little knowledge of the tools to better protect themselves, but are nonetheless willing to follow suggestions to improve privacy (51%). Educating people, including digital natives, about privacy risks through transparency tools seems a promising approach.
Authors: Zaitian Wang, Jinghan Zhang, Xinhao Zhang, Kunpeng Liu, Pengfei Wang, Yuanchun Zhou
Abstract: Data augmentation is an essential technique in natural language processing (NLP) for enriching training datasets by generating diverse samples. This process is crucial for improving the robustness and generalization capabilities of NLP models. However, a significant challenge remains: \textit{Insufficient Attention to Sample Distribution Diversity}. Most existing methods focus on increasing the sample numbers while neglecting the sample distribution diversity, which can lead to model overfitting. In response, we explore data augmentation's impact on dataset diversity and propose a \textbf{\underline{D}}iversity-\textbf{\underline{o}}riented data \textbf{\underline{Aug}}mentation framework (\textbf{DoAug}). % \(\mathscr{DoAug}\) Specifically, we utilize a diversity-oriented fine-tuning approach to train an LLM as a diverse paraphraser, which is capable of augmenting textual datasets by generating diversified paraphrases. Then, we apply the LLM paraphraser to a selected coreset of highly informative samples and integrate the paraphrases with the original data to create a more diverse augmented dataset. Finally, we conduct extensive experiments on 12 real-world textual datasets. The results show that our fine-tuned LLM augmenter improves diversity while preserving label consistency, thereby enhancing the robustness and performance of downstream tasks. Specifically, it achieves an average performance gain of \(10.52\%\), surpassing the runner-up baseline with more than three percentage points.
Authors: Yuncheng Hua, Lizhen Qu, Zhuang Li, Hao Xue, Flora D. Salim, Gholamreza Haffari
Abstract: Alignment tuning is crucial for ensuring large language models (LLMs) behave ethically and helpfully. Current alignment approaches require high-quality annotations and significant training resources. This paper proposes a low-cost, tuning-free method using in-context learning (ICL) to enhance LLM alignment. Through an analysis of high-quality ICL demos, we identified style as a key factor influencing LLM alignment capabilities and explicitly restyled ICL exemplars based on this stylistic framework. Additionally, we combined the restyled demos to achieve a balance between the two conflicting aspects of LLM alignment--factuality and safety. We packaged the restyled examples as prompts to trigger few-shot learning, improving LLM alignment. Compared to the best baseline approach, with an average score of 5.00 as the maximum, our method achieves a maximum 0.10 increase on the Alpaca task (from 4.50 to 4.60), a 0.22 enhancement on the Just-eval benchmark (from 4.34 to 4.56), and a maximum improvement of 0.32 (from 3.53 to 3.85) on the MT-Bench dataset. We release the code and data at https://github.com/AnonymousCode-ComputerScience/RIDE.
URLs: https://github.com/AnonymousCode-ComputerScience/RIDE.
Authors: Yuchen Yan, Yongliang Shen, Yang Liu, Jin Jiang, Xin Xu, Mengdi Zhang, Jian Shao, Yueting Zhuang
Abstract: Mathematical reasoning represents a critical frontier in advancing large language models (LLMs). While step-by-step approaches have emerged as the dominant paradigm for mathematical problem-solving in LLMs, the quality of reasoning steps in training data fundamentally constrains the performance of the models. Recent studies has demonstrated that more detailed intermediate steps can enhance model performance, yet existing methods for step expansion either require more powerful external models or incur substantial computational costs. In this paper, we introduce MathFimer, a novel framework for mathematical reasoning step expansion inspired by the "Fill-in-the-middle" task from code completion. By decomposing solution chains into prefix-suffix pairs and training models to reconstruct missing intermediate steps, we develop a specialized model, MathFimer-7B, on our carefully curated NuminaMath-FIM dataset. We then apply these models to enhance existing mathematical reasoning datasets by inserting detailed intermediate steps into their solution chains, creating MathFimer-expanded versions. Through comprehensive experiments on multiple mathematical reasoning datasets, including MathInstruct, MetaMathQA and etc., we demonstrate that models trained on MathFimer-expanded data consistently outperform their counterparts trained on original data across various benchmarks such as GSM8K and MATH. Our approach offers a practical, scalable solution for enhancing mathematical reasoning capabilities in LLMs without relying on powerful external models or expensive inference procedures.
Authors: Manaar Alam, Hithem Lamri, Michail Maniatakos
Abstract: Backdoor attacks embed hidden functionalities in deep neural networks (DNN), triggering malicious behavior with specific inputs. Advanced defenses monitor anomalous DNN inferences to detect such attacks. However, concealed backdoors evade detection by maintaining a low pre-deployment attack success rate (ASR) and restoring high ASR post-deployment via machine unlearning. Existing concealed backdoors are often constrained by requiring white-box or black-box access or auxiliary data, limiting their practicality when such access or data is unavailable. This paper introduces ReVeil, a concealed backdoor attack targeting the data collection phase of the DNN training pipeline, requiring no model access or auxiliary data. ReVeil maintains low pre-deployment ASR across four datasets and four trigger patterns, successfully evades three popular backdoor detection methods, and restores high ASR post-deployment through machine unlearning.
Authors: Georg W\"olflein, Dyke Ferber, Daniel Truhn, Ognjen Arandjelovi\'c, Jakob Nikolas Kather
Abstract: Tool use has turned large language models (LLMs) into powerful agents that can perform complex multi-step tasks by dynamically utilising external software components. However, these tools must be implemented in advance by human developers, hindering the applicability of LLM agents in domains which demand large numbers of highly specialised tools, like in life sciences and medicine. Motivated by the growing trend of scientific studies accompanied by public code repositories, we propose ToolMaker, a novel agentic framework that autonomously transforms papers with code into LLM-compatible tools. Given a short task description and a repository URL, ToolMaker autonomously installs required dependencies and generates code to perform the task, using a closed-loop self-correction mechanism to iteratively diagnose and rectify errors. To evaluate our approach, we introduce a benchmark comprising 15 diverse and complex computational tasks spanning both medical and non-medical domains with over 100 unit tests to objectively assess tool correctness and robustness. ToolMaker correctly implements 80% of the tasks, substantially outperforming current state-of-the-art software engineering agents. ToolMaker therefore is a step towards fully autonomous agent-based scientific workflows.
Authors: Mukun Chen, Jia Wu, Shirui Pan, Fu Lin, Bo Du, Xiuwen Gong, Wenbin Hu
Abstract: Molecular representation learning is pivotal in predicting molecular properties and advancing drug design. Traditional methodologies, which predominantly rely on homogeneous graph encoding, are limited by their inability to integrate external knowledge and represent molecular structures across different levels of granularity. To address these limitations, we propose a paradigm shift by encoding molecular graphs into heterogeneous structures, introducing a novel framework: Knowledge-aware Contrastive Heterogeneous Molecular Graph Learning (KCHML). This approach leverages contrastive learning to enrich molecular representations with embedded external knowledge. KCHML conceptualizes molecules through three distinct graph views-molecular, elemental, and pharmacological-enhanced by heterogeneous molecular graphs and a dual message-passing mechanism. This design offers a comprehensive representation for property prediction, as well as for downstream tasks such as drug-drug interaction (DDI) prediction. Extensive benchmarking demonstrates KCHML's superiority over state-of-the-art molecular property prediction models, underscoring its ability to capture intricate molecular features.
Authors: Site Qu, Guoqiang Hu
Abstract: The Location-Routing Problem (LRP), which combines the challenges of facility (depot) locating and vehicle route planning, is critically constrained by the reliance on predefined depot candidates, limiting the solution space and potentially leading to suboptimal outcomes. Previous research on LRP without predefined depots is scant and predominantly relies on heuristic algorithms that iteratively attempt depot placements across a planar area. Such approaches lack the ability to proactively generate depot locations that meet specific geographic requirements, revealing a notable gap in current research landscape. To bridge this gap, we propose a data-driven generative DRL framework, designed to proactively generate depots for LRP without predefined depot candidates, solely based on customer requests data which include geographic and demand information. It can operate in two distinct modes: direct generation of exact depot locations, and the creation of a multivariate Gaussian distribution for flexible depots sampling. By extracting depots' geographic pattern from customer requests data, our approach can dynamically respond to logistical needs, identifying high-quality depot locations that further reduce total routing costs compared to traditional methods. Extensive experiments demonstrate that, for a same group of customer requests, compared with those depots identified through random attempts, our framework can proactively generate depots that lead to superior solution routes with lower routing cost. The implications of our framework potentially extend into real-world applications, particularly in emergency medical rescue and disaster relief logistics, where rapid establishment and adjustment of depot locations are paramount, showcasing its potential in addressing LRP for dynamic and unpredictable environments.
Authors: Chavvi Kirtani, Madhav Krishan Garg, Tejash Prasad, Tanmay Singhal, Murari Mandal, Dhruv Kumar
Abstract: The escalating volume of academic research, coupled with a shortage of qualified reviewers, necessitates innovative approaches to peer review. While large language model (LLMs) offer potential for automating this process, their current limitations include superficial critiques, hallucinations, and a lack of actionable insights. This research addresses these challenges by introducing a comprehensive evaluation framework for AI-generated reviews, that measures alignment with human evaluations, verifies factual accuracy, assesses analytical depth, and identifies actionable insights. We also propose a novel alignment mechanism that tailors LLM-generated reviews to the unique evaluation priorities of individual conferences and journals. To enhance the quality of these reviews, we introduce a self-refinement loop that iteratively optimizes the LLM's review prompts. Our framework establishes standardized metrics for evaluating AI-based review systems, thereby bolstering the reliability of AI-generated reviews in academic research.
Authors: Shuai Lyu, Haoran Luo, Zhonghong Ou, Yifan Zhu, Xiaoran Shang, Yang Qin, Meina Song
Abstract: The Text-to-SQL(Text2SQL) task aims to convert natural language queries into executable SQL queries. Thanks to the application of large language models (LLMs), significant progress has been made in this field. However, challenges such as model scalability, limited generation space, and coherence issues in SQL generation still persist. To address these issues, we propose SQL-o1, a Self-Reward-based heuristic search method designed to enhance the reasoning ability of LLMs in SQL query generation. SQL-o1 combines Monte Carlo Tree Search (MCTS) for heuristic process-level search and constructs a Schema-Aware dataset to help the model better understand database schemas. Extensive experiments on the Bird and Spider datasets demonstrate that SQL-o1 improves execution accuracy by 10.8\% on the complex Bird dataset compared to the latest baseline methods, even outperforming GPT-4-based approaches. Additionally, SQL-o1 excels in few-shot learning scenarios and shows strong cross-model transferability. Our code is publicly available at:https://github.com/ShuaiLyu0110/SQL-o1.
Authors: Yinghao Zhang, Haiyan Gui, Ningdi Yang, Yue Hu
Abstract: Joint low-rank and sparse unrolling networks have shown superior performance in dynamic MRI reconstruction. However, existing works mainly utilized matrix low-rank priors, neglecting the tensor characteristics of dynamic MRI images, and only a global threshold is applied for the sparse constraint to the multi-channel data, limiting the flexibility of the network. Additionally, most of them have inherently complex network structure, with intricate interactions among variables. In this paper, we propose a novel deep unrolling network, JotlasNet, for dynamic MRI reconstruction by jointly utilizing tensor low-rank and attention-based sparse priors. Specifically, we utilize tensor low-rank prior to exploit the structural correlations in high-dimensional data. Convolutional neural networks are used to adaptively learn the low-rank and sparse transform domains. A novel attention-based soft thresholding operator is proposed to assign a unique learnable threshold to each channel of the data in the CNN-learned sparse domain. The network is unrolled from the elaborately designed composite splitting algorithm and thus features a simple yet efficient parallel structure. Extensive experiments on two datasets (OCMR, CMRxRecon) demonstrate the superior performance of JotlasNet in dynamic MRI reconstruction.
Authors: Yuqi Pang, Bowen Yang, Haoqin Tu, Yun Cao, Zeyu Zhang
Abstract: Although Large Language Models (LLMs) excel in reasoning and generation for language tasks, they are not specifically designed for multimodal challenges. Training Multimodal Large Language Models (MLLMs), however, is resource-intensive and constrained by various training limitations. In this paper, we propose the Modular-based Visual Contrastive Decoding (MVCD) framework to move this obstacle. Our framework leverages LLMs' In-Context Learning (ICL) capability and the proposed visual contrastive-example decoding (CED), specifically tailored for this framework, without requiring any additional training. By converting visual signals into text and focusing on contrastive output distributions during decoding, we can highlight the new information introduced by contextual examples, explore their connections, and avoid over-reliance on prior encoded knowledge. MVCD enhances LLMs' visual perception to make it see and reason over the input visuals. To demonstrate MVCD's effectiveness, we conduct experiments with four LLMs across five question answering datasets. Our results not only show consistent improvement in model accuracy but well explain the effective components inside our decoding strategy. Our code will be available at https://github.com/Pbhgit/MVCD.
Authors: Gido M. van de Ven
Abstract: One of the most popular methods for continual learning with deep neural networks is Elastic Weight Consolidation (EWC), which involves computing the Fisher Information. The exact way in which the Fisher Information is computed is however rarely described, and multiple different implementations for it can be found online. This blog post discusses and empirically compares several often-used implementations, which highlights that many currently reported results for EWC could likely be improved by changing the way the Fisher Information is computed.
Authors: Siddiqui Muhammad Yasir, Hyun Kim
Abstract: Deepfake technology utilizes deep learning based face manipulation techniques to seamlessly replace faces in videos creating highly realistic but artificially generated content. Although this technology has beneficial applications in media and entertainment misuse of its capabilities may lead to serious risks including identity theft cyberbullying and false information. The integration of DL with visual cognition has resulted in important technological improvements particularly in addressing privacy risks caused by artificially generated deepfake images on digital media platforms. In this study we propose an efficient and lightweight method for detecting deepfake images and videos making it suitable for devices with limited computational resources. In order to reduce the computational burden usually associated with DL models our method integrates machine learning classifiers in combination with keyframing approaches and texture analysis. Moreover the features extracted with a histogram of oriented gradients (HOG) local binary pattern (LBP) and KAZE bands were integrated to evaluate using random forest extreme gradient boosting extra trees and support vector classifier algorithms. Our findings show a feature-level fusion of HOG LBP and KAZE features improves accuracy to 92% and 96% on FaceForensics++ and Celeb-DFv2 respectively.
Authors: Leonardo Bertolazzi, Philipp Mondorf, Barbara Plank, Raffaella Bernardi
Abstract: The ability of large language models (LLMs) to validate their output and identify potential errors is crucial for ensuring robustness and reliability. However, current research indicates that LLMs struggle with self-correction, encountering significant challenges in detecting errors. While studies have explored methods to enhance self-correction in LLMs, relatively little attention has been given to understanding the models' internal mechanisms underlying error detection. In this paper, we present a mechanistic analysis of error detection in LLMs, focusing on simple arithmetic problems. Through circuit analysis, we identify the computational subgraphs responsible for detecting arithmetic errors across four smaller-sized LLMs. Our findings reveal that all models heavily rely on $\textit{consistency heads}$--attention heads that assess surface-level alignment of numerical values in arithmetic solutions. Moreover, we observe that the models' internal arithmetic computation primarily occurs in higher layers, whereas validation takes place in middle layers, before the final arithmetic results are fully encoded. This structural dissociation between arithmetic computation and validation seems to explain why current LLMs struggle to detect even simple arithmetic errors.
Authors: Siddiqui Muhammad Yasir, Hyunsik Ahn
Abstract: Depth estimation plays a pivotal role in advancing human-robot interactions, especially in indoor environments where accurate 3D scene reconstruction is essential for tasks like navigation and object handling. Monocular depth estimation, which relies on a single RGB camera, offers a more affordable solution compared to traditional methods that use stereo cameras or LiDAR. However, despite recent progress, many monocular approaches struggle with accurately defining depth boundaries, leading to less precise reconstructions. In response to these challenges, this study introduces a novel depth estimation framework that leverages latent space features within a deep convolutional neural network to enhance the precision of monocular depth maps. The proposed model features dual encoder-decoder architecture, enabling both color-to-depth and depth-to-depth transformations. This structure allows for refined depth estimation through latent space encoding. To further improve the accuracy of depth boundaries and local features, a new loss function is introduced. This function combines latent loss with gradient loss, helping the model maintain the integrity of depth boundaries. The framework is thoroughly tested using the NYU Depth V2 dataset, where it sets a new benchmark, particularly excelling in complex indoor scenarios. The results clearly show that this approach effectively reduces depth ambiguities and blurring, making it a promising solution for applications in human-robot interaction and 3D scene reconstruction.
Authors: Yanbiao Ma, Bowei Liu, Wei Dai, Jiayi Chen, Shuo Li
Abstract: Deep neural networks (DNNs) often exhibit biases toward certain categories during object recognition, even under balanced training data conditions. The intrinsic mechanisms underlying these biases remain unclear. Inspired by the human visual system, which decouples object manifolds through hierarchical processing to achieve object recognition, we propose a geometric analysis framework linking the geometric complexity of class-specific perceptual manifolds in DNNs to model bias. Our findings reveal that differences in geometric complexity can lead to varying recognition capabilities across categories, introducing biases. To support this analysis, we present the Perceptual-Manifold-Geometry library, designed for calculating the geometric properties of perceptual manifolds.
Authors: Xu Wang, Yan Hu, Wenyu Du, Reynold Cheng, Benyou Wang, Difan Zou
Abstract: Fine-tuning significantly improves the performance of Large Language Models (LLMs), yet its underlying mechanisms remain poorly understood. This paper aims to provide an in-depth interpretation of the fine-tuning process through circuit analysis, a popular tool in Mechanistic Interpretability (MI). Unlike previous studies \cite{prakash2024finetuningenhancesexistingmechanisms,chhabra2024neuroplasticity} that focus on tasks where pre-trained models already perform well, we develop a set of mathematical tasks where fine-tuning yields substantial performance gains, which are closer to the practical setting. In our experiments, we identify circuits at various checkpoints during fine-tuning and examine the interplay between circuit analysis, fine-tuning methods, and task complexities. First, we find that while circuits maintain high node similarity before and after fine-tuning, their edges undergo significant changes, which is in contrast to the previous work \cite{prakash2024finetuningenhancesexistingmechanisms,chhabra2024neuroplasticity} that show circuits only add some additional components after fine-tuning. Based on these observations, we develop a circuit-aware Low-Rank Adaptation (LoRA) method, which assigns ranks to layers based on edge changes in the circuits. Experimental results demonstrate that our circuit-based LoRA algorithm achieves an average performance improvement of 2.46\% over standard LoRA with similar parameter sizes. Furthermore, we explore how combining circuits from subtasks can enhance fine-tuning in compositional tasks, providing new insights into the design of such tasks and deepening the understanding of circuit dynamics and fine-tuning mechanisms.
Authors: Hanbin Wang, Xiaoxuan Zhou, Zhipeng Xu, Keyuan Cheng, Yuxin Zuo, Kai Tian, Jingwei Song, Junting Lu, Wenhui Hu, Xueyang Liu
Abstract: This paper introduces Code-Vision, a benchmark designed to evaluate the logical understanding and code generation capabilities of Multimodal Large Language Models (MLLMs). It challenges MLLMs to generate a correct program that fulfills specific functionality requirements based on a given flowchart, which visually represents the desired algorithm or process. Code-Vision comprises three subsets: HumanEval-V, Algorithm, and MATH, which evaluate MLLMs' coding abilities across basic programming, algorithmic, and mathematical problem-solving domains. Our experiments evaluate 12 MLLMs on Code-Vision. Experimental results demonstrate that there is a large performance difference between proprietary and open-source models. On Hard problems, GPT-4o can achieve 79.3% pass@1, but the best open-source model only achieves 15%. Further experiments reveal that Code-Vision can pose unique challenges compared to other multimodal reasoning benchmarks MMCode and MathVista. We also explore the reason for the poor performance of the open-source models. All data and codes are available at https://github.com/wanghanbinpanda/CodeVision.
Authors: Quentin Garrido, Nicolas Ballas, Mahmoud Assran, Adrien Bardes, Laurent Najman, Michael Rabbat, Emmanuel Dupoux, Yann LeCun
Abstract: We investigate the emergence of intuitive physics understanding in general-purpose deep neural network models trained to predict masked regions in natural videos. Leveraging the violation-of-expectation framework, we find that video prediction models trained to predict outcomes in a learned representation space demonstrate an understanding of various intuitive physics properties, such as object permanence and shape consistency. In contrast, video prediction in pixel space and multimodal large language models, which reason through text, achieve performance closer to chance. Our comparisons of these architectures reveal that jointly learning an abstract representation space while predicting missing parts of sensory input, akin to predictive coding, is sufficient to acquire an understanding of intuitive physics, and that even models trained on one week of unique video achieve above chance performance. This challenges the idea that core knowledge -- a set of innate systems to help understand the world -- needs to be hardwired to develop an understanding of intuitive physics.
Authors: Muhammad Waseem Akram, Stefano Dettori, Valentina Colla, Giorgio Carlo Buttazzo
Abstract: Chord recognition serves as a critical task in music information retrieval due to the abstract and descriptive nature of chords in music analysis. While audio chord recognition systems have achieved significant accuracy for small vocabularies (e.g., major/minor chords), large-vocabulary chord recognition remains a challenging problem. This complexity also arises from the inherent long-tail distribution of chords, where rare chord types are underrepresented in most datasets, leading to insufficient training samples. Effective chord recognition requires leveraging contextual information from audio sequences, yet existing models, such as combinations of convolutional neural networks, bidirectional long short-term memory networks, and bidirectional transformers, face limitations in capturing long-term dependencies and exhibit suboptimal performance on large-vocabulary chord recognition tasks. This work proposes ChordFormer, a novel conformer-based architecture designed to tackle structural chord recognition (e.g., triads, bass, sevenths) for large vocabularies. ChordFormer leverages conformer blocks that integrate convolutional neural networks with transformers, thus enabling the model to capture both local patterns and global dependencies effectively. By addressing challenges such as class imbalance through a reweighted loss function and structured chord representations, ChordFormer outperforms state-of-the-art models, achieving a 2% improvement in frame-wise accuracy and a 6% increase in class-wise accuracy on large-vocabulary chord datasets. Furthermore, ChordFormer excels in handling class imbalance, providing robust and balanced recognition across chord types. This approach bridges the gap between theoretical music knowledge and practical applications, advancing the field of large-vocabulary chord recognition.
Authors: Pranav Bhandari, Nicolas Fay, Michael Wise, Amitava Datta, Stephanie Meek, Usman Naseem, Mehwish Nasim
Abstract: Large Language Models (LLMs) are widely used as conversational agents, exploiting their capabilities in various sectors such as education, law, medicine, and more. However, LLMs are often subjected to context-shifting behaviour, resulting in a lack of consistent and interpretable personality-aligned interactions. Adherence to psychological traits lacks comprehensive analysis, especially in the case of dyadic (pairwise) conversations. We examine this challenge from two viewpoints, initially using two conversation agents to generate a discourse on a certain topic with an assigned personality from the OCEAN framework (Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism) as High/Low for each trait. This is followed by using multiple judge agents to infer the original traits assigned to explore prediction consistency, inter-model agreement, and alignment with the assigned personality. Our findings indicate that while LLMs can be guided toward personality-driven dialogue, their ability to maintain personality traits varies significantly depending on the combination of models and discourse settings. These inconsistencies emphasise the challenges in achieving stable and interpretable personality-aligned interactions in LLMs.
Authors: Mark Vero, Niels M\"undler, Victor Chibotaru, Veselin Raychev, Maximilian Baader, Nikola Jovanovi\'c, Jingxuan He, Martin Vechev
Abstract: The automatic generation of programs has long been a fundamental challenge in computer science. Recent benchmarks have shown that large language models (LLMs) can effectively generate code at the function level, make code edits, and solve algorithmic coding tasks. However, to achieve full automation, LLMs should be able to generate production-quality, self-contained application modules. To evaluate the capabilities of LLMs in solving this challenge, we introduce BaxBench, a novel evaluation benchmark consisting of 392 tasks for the generation of backend applications. We focus on backends for three critical reasons: (i) they are practically relevant, building the core components of most modern web and cloud software, (ii) they are difficult to get right, requiring multiple functions and files to achieve the desired functionality, and (iii) they are security-critical, as they are exposed to untrusted third-parties, making secure solutions that prevent deployment-time attacks an imperative. BaxBench validates the functionality of the generated applications with comprehensive test cases, and assesses their security exposure by executing end-to-end exploits. Our experiments reveal key limitations of current LLMs in both functionality and security: (i) even the best model, OpenAI o1, achieves a mere 60% on code correctness; (ii) on average, we could successfully execute security exploits on more than half of the correct programs generated by each LLM; and (iii) in less popular backend frameworks, models further struggle to generate correct and secure applications. Progress on BaxBench signifies important steps towards autonomous and secure software development with LLMs.
Authors: Aras Yurtman, Daan Van Wesenbeeck, Wannes Meert, Hendrik Blockeel
Abstract: Time Series Motif Discovery (TSMD) identifies repeating patterns in time series data, but its unsupervised nature might result in motifs that are not interesting to the user. To address this, we propose a framework that allows the user to impose constraints on the motifs to be discovered, where constraints can easily be defined according to the properties of the desired motifs in the application domain. We also propose an efficient implementation of the framework, the LoCoMotif-DoK algorithm. We demonstrate that LoCoMotif-DoK can effectively leverage domain knowledge in real and synthetic data, outperforming other TSMD techniques which only support a limited form of domain knowledge.
Authors: Yahao Pang, Xingyuan Wu, Xiaojin Zhang, Wei Chen, Hai Jin
Abstract: Significant advancements have been made by Large Language Models (LLMs) in the domains of natural language understanding and automated content creation. However, they still face persistent problems, including substantial computational costs and inadequate availability of training data. The combination of Federated Learning (FL) and LLMs (federated LLMs) offers a solution by leveraging distributed data while protecting privacy, which positions it as an ideal choice for sensitive domains. However, Federated LLMs still suffer from robustness challenges, including data heterogeneity, malicious clients, and adversarial attacks, which greatly hinder their applications. We first introduce the robustness problems in federated LLMs, to address these challenges, we propose FedEAT (Federated Embedding space Adversarial Training), a novel framework that applies adversarial training in the embedding space of client LLM and employs a robust aggregation approach, specifically geometric median aggregation, to enhance the robustness of Federated LLMs. Our experiments demonstrate that FedEAT effectively improves the robustness of Federated LLMs with minimal performance loss.
Authors: Jinheng Wang, Hansong Zhou, Ting Song, Shijie Cao, Yan Xia, Ting Cao, Jianyu Wei, Shuming Ma, Hongyu Wang, Furu Wei
Abstract: The advent of 1-bit large language models (LLMs), led by BitNet b1.58, has spurred interest in ternary LLMs. Despite this, research and practical applications focusing on efficient edge inference for ternary LLMs remain scarce. To bridge this gap, we introduce Bitnet.cpp, an inference system optimized for BitNet b1.58 and ternary LLMs. Given that mixed-precision matrix multiplication (mpGEMM) constitutes the bulk of inference time in ternary LLMs, Bitnet.cpp incorporates a novel mpGEMM library to facilitate sub-2-bits-per-weight, efficient and lossless inference. The library features two core solutions: Ternary Lookup Table (TL), which addresses spatial inefficiencies of previous bit-wise methods, and Int2 with a Scale (I2_S), which ensures lossless edge inference, both enabling high-speed inference. Our experiments show that Bitnet.cpp achieves up to a 6.25x increase in speed over full-precision baselines and up to 2.32x over low-bit baselines, setting new benchmarks in the field. Additionally, we expand TL to element-wise lookup table (ELUT) for low-bit LLMs in the appendix, presenting both theoretical and empirical evidence of its considerable potential. Bitnet.cpp is publicly available at https://github.com/microsoft/BitNet/tree/paper , offering a sophisticated solution for the efficient and practical deployment of edge LLMs.
Authors: Xuefeng Li, Haoyang Zou, Pengfei Liu
Abstract: In this paper, we ask: what truly determines the effectiveness of RL training data for enhancing language models' reasoning capabilities? While recent advances like o1, Deepseek R1, and Kimi1.5 demonstrate RL's potential, the lack of transparency about training data requirements has hindered systematic progress. Starting directly from base models without distillation, we challenge the assumption that scaling up RL training data inherently improves performance. we demonstrate that a strategically selected subset of just 1,389 samples can outperform the full 8,523-sample dataset. We introduce Learning Impact Measurement (LIM), an automated method to evaluate and prioritize training samples based on their alignment with model learning trajectories, enabling efficient resource utilization and scalable implementation. Our method achieves comparable or even superior performance using only 1,389 samples versus the full 8,523 samples dataset. Notably, while recent data-efficient approaches (e.g., LIMO and s1) show promise with 32B-scale models, we find it significantly underperforms at 7B-scale through supervised fine-tuning (SFT). In contrast, our RL-based LIMR achieves 16.7% higher accuracy on AIME24 and outperforms LIMO and s1 by 13.0% and 22.2% on MATH500. These results fundamentally reshape our understanding of RL scaling in LLMs, demonstrating that precise sample selection, rather than data scale, may be the key to unlocking enhanced reasoning capabilities. For reproducible research and future innovation, we are open-sourcing LIMR, including implementation of LIM, training and evaluation code, curated datasets, and trained models at https://github.com/GAIR-NLP/LIMR.
Authors: Michele Grimaldi, Patryk Cieslak, Eduardo Ochoa, Vibhav Bharti, Hayat Rajani, Ignacio Carlucho, Maria Koskinopoulou, Yvan R. Petillot, Nuno Gracias
Abstract: Simulations are highly valuable in marine robotics, offering a cost-effective and controlled environment for testing in the challenging conditions of underwater and surface operations. Given the high costs and logistical difficulties of real-world trials, simulators capable of capturing the operational conditions of subsea environments have become key in developing and refining algorithms for remotely-operated and autonomous underwater vehicles. This paper highlights recent enhancements to the Stonefish simulator, an advanced open-source platform supporting development and testing of marine robotics solutions. Key updates include a suite of additional sensors, such as an event-based camera, a thermal camera, and an optical flow camera, as well as, visual light communication, support for tethered operations, improved thruster modelling, more flexible hydrodynamics, and enhanced sonar accuracy. These developments and an automated annotation tool significantly bolster Stonefish's role in marine robotics research, especially in the field of machine learning, where training data with a known ground truth is hard or impossible to collect.
Authors: Jacob Nielsen, Peter Schneider-Kamp, Lukas Galke
Abstract: Large language models (LLMs) require immense resources for training and inference. Quantization, a technique that reduces the precision of model parameters, offers a promising solution for improving LLM efficiency and sustainability. While post-training quantization methods typically achieve 4-8 bits per parameter, recent research suggests that training LLMs with 1.58 bits per weight parameter from scratch can maintain model accuracy while greatly reducing memory requirements and energy consumption at inference time. Here, we investigate a training strategy for quantization-aware pre-training, where the models are first trained with 16-bit precision and then transition into 1.58-bit quantization-aware training. Our results on 11 downstream tasks show that this 16-to-1.58-bit training strategy is preferable over full 1.58-bit training and leaves models closer to those which have undergone 16-bit training. We further investigate the effects of retaining the optimizer state at the transition point and gradually phasing in quantization strength -- finding that both techniques alleviate the magnitude of loss spikes, but also that these effects can be compensated through further training.
Authors: Yanxiao Zhao, Yangge Qian, Jingyang Shan, Xiaolin Qin
Abstract: Reinforcement learning (RL) in continuous action spaces encounters persistent challenges, such as inefficient exploration and convergence to suboptimal solutions. To address these limitations, we propose CAMEL, a novel framework integrating LLM-generated suboptimal policies into the RL training pipeline. CAMEL leverages dynamic action masking and an adaptive epsilon-masking mechanism to guide exploration during early training stages while gradually enabling agents to optimize policies independently. At the core of CAMEL lies the integration of Python-executable suboptimal policies generated by LLMs based on environment descriptions and task objectives. Although simplistic and hard-coded, these policies offer valuable initial guidance for RL agents. To effectively utilize these priors, CAMEL employs masking-aware optimization to dynamically constrain the action space based on LLM outputs. Additionally, epsilon-masking gradually reduces reliance on LLM-generated guidance, enabling agents to transition from constrained exploration to autonomous policy refinement. Experimental validation on Gymnasium MuJoCo environments demonstrates the effectiveness of CAMEL. In Hopper-v4 and Ant-v4, LLM-generated policies significantly improve sample efficiency, achieving performance comparable to or surpassing expert masking baselines. For Walker2d-v4, where LLMs struggle to accurately model bipedal gait dynamics, CAMEL maintains robust RL performance without notable degradation, highlighting the framework's adaptability across diverse tasks. While CAMEL shows promise in enhancing sample efficiency and mitigating convergence challenges, these issues remain open for further research. Future work aims to generalize CAMEL to multimodal LLMs for broader observation-action spaces and automate policy evaluation, reducing human intervention and enhancing scalability in RL training pipelines.
Authors: Zhihang Yuan, Siyuan Wang, Rui Xie, Hanling Zhang, Tongcheng Fang, Yuzhang Shang, Shengen Yan, Guohao Dai, Yu Wang
Abstract: In this paper, we propose the Dynamic Latent Frame Rate VAE (DLFR-VAE), a training-free paradigm that can make use of adaptive temporal compression in latent space. While existing video generative models apply fixed compression rates via pretrained VAE, we observe that real-world video content exhibits substantial temporal non-uniformity, with high-motion segments containing more information than static scenes. Based on this insight, DLFR-VAE dynamically adjusts the latent frame rate according to the content complexity. Specifically, DLFR-VAE comprises two core innovations: (1) A Dynamic Latent Frame Rate Scheduler that partitions videos into temporal chunks and adaptively determines optimal frame rates based on information-theoretic content complexity, and (2) A training-free adaptation mechanism that transforms pretrained VAE architectures into a dynamic VAE that can process features with variable frame rates. Our simple but effective DLFR-VAE can function as a plug-and-play module, seamlessly integrating with existing video generation models and accelerating the video generation process.
Authors: Jiamin Su, Yibo Yan, Fangteng Fu, Han Zhang, Jingheng Ye, Xiang Liu, Jiahao Huo, Huiyu Zhou, Xuming Hu
Abstract: Automated Essay Scoring (AES) plays a crucial role in educational assessment by providing scalable and consistent evaluations of writing tasks. However, traditional AES systems face three major challenges: (1) reliance on handcrafted features that limit generalizability, (2) difficulty in capturing fine-grained traits like coherence and argumentation, and (3) inability to handle multimodal contexts. In the era of Multimodal Large Language Models (MLLMs), we propose EssayJudge, the first multimodal benchmark to evaluate AES capabilities across lexical-, sentence-, and discourse-level traits. By leveraging MLLMs' strengths in trait-specific scoring and multimodal context understanding, EssayJudge aims to offer precise, context-rich evaluations without manual feature engineering, addressing longstanding AES limitations. Our experiments with 18 representative MLLMs reveal gaps in AES performance compared to human evaluation, particularly in discourse-level traits, highlighting the need for further advancements in MLLM-based AES research. Our dataset and code will be available upon acceptance.
Authors: Yutong Ye, Yingbo Zhou, Zhusen Liu, Xiao Du, Hao Zhou, Xiang Lian, Mingsong Chen
Abstract: Although Reinforcement Learning (RL)-based Traffic Signal Control (TSC) methods have been extensively studied, their practical applications still raise some serious issues such as high learning cost and poor generalizability. This is because the ``trial-and-error'' training style makes RL agents extremely dependent on the specific traffic environment, which also requires a long convergence time. To address these issues, we propose a novel Federated Imitation Learning (FIL)-based framework for multi-intersection TSC, named FitLight, which allows RL agents to plug-and-play for any traffic environment without additional pre-training cost. Unlike existing imitation learning approaches that rely on pre-training RL agents with demonstrations, FitLight allows real-time imitation learning and seamless transition to reinforcement learning. Due to our proposed knowledge-sharing mechanism and novel hybrid pressure-based agent design, RL agents can quickly find a best control policy with only a few episodes. Moreover, for resource-constrained TSC scenarios, FitLight supports model pruning and heterogeneous model aggregation, such that RL agents can work on a micro-controller with merely 16{\it KB} RAM and 32{\it KB} ROM. Extensive experiments demonstrate that, compared to state-of-the-art methods, FitLight not only provides a superior starting point but also converges to a better final solution on both real-world and synthetic datasets, even under extreme resource limitations.
Authors: Ammar Kheder, Benjamin Foreback, Lili Wang, Zhi-Song Liu, Michael Boy
Abstract: Air quality prediction is key to mitigating health impacts and guiding decisions, yet existing models tend to focus on temporal trends while overlooking spatial generalization. We propose AQ-Net, a spatiotemporal reanalysis model for both observed and unobserved stations in the near future. AQ-Net utilizes the LSTM and multi-head attention for the temporal regression. We also propose a cyclic encoding technique to ensure continuous time representation. To learn fine-grained spatial air quality estimation, we incorporate AQ-Net with the neural kNN to explore feature-based interpolation, such that we can fill the spatial gaps given coarse observation stations. To demonstrate the efficiency of our model for spatiotemporal reanalysis, we use data from 2013-2017 collected in northern China for PM2.5 analysis. Extensive experiments show that AQ-Net excels in air quality reanalysis, highlighting the potential of hybrid spatio-temporal models to better capture environmental dynamics, especially in urban areas where both spatial and temporal variability are critical.
Authors: Ailin Huang, Boyong Wu, Bruce Wang, Chao Yan, Chen Hu, Chengli Feng, Fei Tian, Feiyu Shen, Jingbei Li, Mingrui Chen, Peng Liu, Ruihang Miao, Wang You, Xi Chen, Xuerui Yang, Yechang Huang, Yuxiang Zhang, Zheng Gong, Zixin Zhang, Brian Li, Changyi Wan, Hanpeng Hu, Ranchen Ming, Song Yuan, Xuelin Zhang, Yu Zhou, Bingxin Li, Buyun Ma, Kang An, Wei Ji, Wen Li, Xuan Wen, Yuankai Ma, Yuanwei Liang, Yun Mou, Bahtiyar Ahmidi, Bin Wang, Bo Li, Changxin Miao, Chen Xu, Chengting Feng, Chenrun Wang, Dapeng Shi, Deshan Sun, Dingyuan Hu, Dula Sai, Enle Liu, Guanzhe Huang, Gulin Yan, Heng Wang, Haonan Jia, Haoyang Zhang, Jiahao Gong, Jianchang Wu, Jiahong Liu, Jianjian Sun, Jiangjie Zhen, Jie Feng, Jie Wu, Jiaoren Wu, Jie Yang, Jinguo Wang, Jingyang Zhang, Junzhe Lin, Kaixiang Li, Lei Xia, Li Zhou, Longlong Gu, Mei Chen, Menglin Wu, Ming Li, Mingxiao Li, Mingyao Liang, Na Wang, Nie Hao, Qiling Wu, Qinyuan Tan, Shaoliang Pang, Shiliang Yang, Shuli Gao, Siqi Liu, Sitong Liu, Tiancheng Cao, Tianyu Wang, Wenjin Deng, Wenqing He, Wen Sun, Xin Han, Xiaomin Deng, Xiaojia Liu, Xu Zhao, Yanan Wei, Yanbo Yu, Yang Cao, Yangguang Li, Yangzhen Ma, Yanming Xu, Yaqiang Shi, Yilei Wang, Yinmin Zhong, Yu Luo, Yuanwei Lu, Yuhe Yin, Yuting Yan, Yuxiang Yang, Zhe Xie, Zheng Ge, Zheng Sun, Zhewei Huang, Zhichao Chang, Zidong Yang, Zili Zhang, Binxing Jiao, Daxin Jiang, Heung-Yeung Shum, Jiansheng Chen, Jing Li, Shuchang Zhou, Xiangyu Zhang, Xinhao Zhang, Yibo Zhu
Abstract: Real-time speech interaction, serving as a fundamental interface for human-machine collaboration, holds immense potential. However, current open-source models face limitations such as high costs in voice data collection, weakness in dynamic control, and limited intelligence. To address these challenges, this paper introduces Step-Audio, the first production-ready open-source solution. Key contributions include: 1) a 130B-parameter unified speech-text multi-modal model that achieves unified understanding and generation, with the Step-Audio-Chat version open-sourced; 2) a generative speech data engine that establishes an affordable voice cloning framework and produces the open-sourced lightweight Step-Audio-TTS-3B model through distillation; 3) an instruction-driven fine control system enabling dynamic adjustments across dialects, emotions, singing, and RAP; 4) an enhanced cognitive architecture augmented with tool calling and role-playing abilities to manage complex tasks effectively. Based on our new StepEval-Audio-360 evaluation benchmark, Step-Audio achieves state-of-the-art performance in human evaluations, especially in terms of instruction following. On open-source benchmarks like LLaMA Question, shows 9.3% average performance improvement, demonstrating our commitment to advancing the development of open-source multi-modal language technologies. Our code and models are available at https://github.com/stepfun-ai/Step-Audio.
Authors: Nico Bohlinger, Jan Peters
Abstract: We introduce a scaling strategy for Explicit Policy-Conditioned Value Functions (EPVFs) that significantly improves performance on challenging continuous-control tasks. EPVFs learn a value function V({\theta}) that is explicitly conditioned on the policy parameters, enabling direct gradient-based updates to the parameters of any policy. However, EPVFs at scale struggle with unrestricted parameter growth and efficient exploration in the policy parameter space. To address these issues, we utilize massive parallelization with GPU-based simulators, big batch sizes, weight clipping and scaled peturbations. Our results show that EPVFs can be scaled to solve complex tasks, such as a custom Ant environment, and can compete with state-of-the-art Deep Reinforcement Learning (DRL) baselines like Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC). We further explore action-based policy parameter representations from previous work and specialized neural network architectures to efficiently handle weight-space features, which have not been used in the context of DRL before.
Authors: Tianyi Wu, Jingwei Ni, Bryan Hooi, Jiaheng Zhang, Elliott Ash, See-Kiong Ng, Mrinmaya Sachan, Markus Leippold
Abstract: Instruction Fine-tuning (IFT) can enhance the helpfulness of Large Language Models (LLMs), but it may lower their truthfulness. This trade-off arises because IFT steers LLMs to generate responses with long-tail knowledge that is not well covered during pre-training, leading to more informative but less truthful answers when generalizing to unseen tasks. In this paper, we empirically demonstrate this helpfulness-truthfulness trade-off in IFT and propose $\textbf{UNIT}$, a novel IFT paradigm to address it. UNIT teaches LLMs to recognize their uncertainty and explicitly reflect it at the end of their responses. Experimental results show that UNIT-tuned models maintain their helpfulness while distinguishing between certain and uncertain claims, thereby reducing hallucinations.
Authors: Jun Jiang, Wenjun Yu, Yunfan Li, Yuan Gao, Shugong Xu
Abstract: In the field of artificial intelligence, self-supervised learning has demonstrated superior generalization capabilities by leveraging large-scale unlabeled datasets for pretraining, which is especially critical for wireless communication models to adapt to a variety of scenarios. This paper innovatively treats Channel State Information (CSI) and Channel Impulse Response (CIR) as naturally aligned multi-modal data and proposes the first MIMO wireless channel foundation model, named CSI-CLIP. By effectively capturing the joint representations of both CIR and CSI, CSI-CLIP exhibits remarkable adaptability across scenarios and robust feature extraction capabilities. Experimental results show that in positioning task, CSI-CLIP reduces the mean error distance by 22%; in beam management task, it increases accuracy by 1% compared to traditional supervised methods, as well as in the channel identification task. These improvements not only highlight the potential and value of CSI-CLIP in integrating sensing and communication but also demonstrate its significant advantages over existing techniques. Moreover, viewing CSI and CIR as multi-modal pairs and contrastive learning for wireless channel foundation model open up new research directions in the domain of MIMO wireless communications.
Authors: Brieuc Pinon, Rapha\"el Jungers, Jean-Charles Delvenne
Abstract: Reinforcement Learning algorithms designed for high-dimensional spaces often enforce the Bellman equation on a sampled subset of states, relying on generalization to propagate knowledge across the state space. In this paper, we identify and formalize a fundamental limitation of this common approach. Specifically, we construct counterexample problems with a simple structure that this approach fails to exploit. Our findings reveal that such algorithms can neglect critical information about the problems, leading to inefficiencies. Furthermore, we extend this negative result to another approach from the literature: Hindsight Experience Replay learning state-to-state reachability.
Authors: Nir Rosenfeld, Haifeng Xu
Abstract: Decades of research in machine learning have given us powerful tools for making accurate predictions. But when used in social settings and on human inputs, better accuracy does not immediately translate to better social outcomes. This may not be surprising given that conventional learning frameworks are not designed to express societal preferences -- let alone promote them. This position paper argues that machine learning is currently missing, and can gain much from incorporating, a proper notion of social welfare. The field of welfare economics asks: how should we allocate limited resources to self-interested agents in a way that maximizes social benefit? We argue that this perspective applies to many modern applications of machine learning in social contexts, and advocate for its adoption. Rather than disposing of prediction, we aim to leverage this forte of machine learning for promoting social welfare. We demonstrate this idea by proposing a conceptual framework that gradually transitions from accuracy maximization (with awareness to welfare) to welfare maximization (via accurate prediction). We detail applications and use-cases for which our framework can be effective, identify technical challenges and practical opportunities, and highlight future avenues worth pursuing.
Authors: Negar Kamali, Karyn Nakamura, Aakriti Kumar, Angelos Chatzimparmpas, Jessica Hullman, Matthew Groh
Abstract: Diffusion model-generated images can appear indistinguishable from authentic photographs, but these images often contain artifacts and implausibilities that reveal their AI-generated provenance. Given the challenge to public trust in media posed by photorealistic AI-generated images, we conducted a large-scale experiment measuring human detection accuracy on 450 diffusion-model generated images and 149 real images. Based on collecting 749,828 observations and 34,675 comments from 50,444 participants, we find that scene complexity of an image, artifact types within an image, display time of an image, and human curation of AI-generated images all play significant roles in how accurately people distinguish real from AI-generated images. Additionally, we propose a taxonomy characterizing artifacts often appearing in images generated by diffusion models. Our empirical observations and taxonomy offer nuanced insights into the capabilities and limitations of diffusion models to generate photorealistic images in 2024.
Authors: Siddhesh Pawar, Arnav Arora, Lucie-Aim\'ee Kaffee, Isabelle Augenstein
Abstract: Names are deeply tied to human identity. They can serve as markers of individuality, cultural heritage, and personal history. However, using names as a core indicator of identity can lead to over-simplification of complex identities. When interacting with LLMs, user names are an important point of information for personalisation. Names can enter chatbot conversations through direct user input (requested by chatbots), as part of task contexts such as CV reviews, or as built-in memory features that store user information for personalisation. We study biases associated with names by measuring cultural presumptions in the responses generated by LLMs when presented with common suggestion-seeking queries, which might involve making assumptions about the user. Our analyses demonstrate strong assumptions about cultural identity associated with names present in LLM generations across multiple cultures. Our work has implications for designing more nuanced personalisation systems that avoid reinforcing stereotypes while maintaining meaningful customisation.
Authors: Yuchen Yang, Thomas Thebaud, Najim Dehak
Abstract: This paper introduces a general classifier based on WavLM features, to infer demographic characteristics, such as age, gender, native language, education, and country, from speech. Demographic feature prediction plays a crucial role in applications like language learning, accessibility, and digital forensics, enabling more personalized and inclusive technologies. Leveraging pretrained models for embedding extraction, the proposed framework identifies key acoustic and linguistic fea-tures associated with demographic attributes, achieving a Mean Absolute Error (MAE) of 4.94 for age prediction and over 99.81% accuracy for gender classification across various datasets. Our system improves upon existing models by up to relative 30% in MAE and up to relative 10% in accuracy and F1 scores across tasks, leveraging a diverse range of datasets and large pretrained models to ensure robustness and generalizability. This study offers new insights into speaker diversity and provides a strong foundation for future research in speech-based demographic profiling.
Authors: Shuaiqun Pan, Yash J. Patel, Aneta Neumann, Frank Neumann, Thomas B\"ack, Hao Wang
Abstract: Variational quantum algorithms, such as the Recursive Quantum Approximate Optimization Algorithm (RQAOA), have become increasingly popular, offering promising avenues for employing Noisy Intermediate-Scale Quantum devices to address challenging combinatorial optimization tasks like the maximum cut problem. In this study, we utilize an evolutionary algorithm equipped with a unique fitness function. This approach targets hard maximum cut instances within the latent space of a Graph Autoencoder, identifying those that pose significant challenges or are particularly tractable for RQAOA, in contrast to the classic Goemans and Williamson algorithm. Our findings not only delineate the distinct capabilities and limitations of each algorithm but also expand our understanding of RQAOA's operational limits. Furthermore, the diverse set of graphs we have generated serves as a crucial benchmarking asset, emphasizing the need for more advanced algorithms to tackle combinatorial optimization challenges. Additionally, our results pave the way for new avenues in graph generation research, offering exciting opportunities for future explorations.
Authors: Fengwei Teng, Zhaoyang Yu, Quan Shi, Jiayi Zhang, Chenglin Wu, Yuyu Luo
Abstract: Large Language Models (LLMs) achieve superior performance through training-time scaling, and test-time scaling further enhances their capabilities by conducting effective reasoning during inference. However, as the scale of reasoning increases, existing test-time scaling methods suffer from accumulated historical information, which not only wastes computational resources but also interferes with effective reasoning. To address this issue, we observe that complex reasoning progress is often achieved by solving a sequence of independent subquestions, each being self-contained and verifiable. These subquestions are essentially atomic questions, relying primarily on their current state rather than accumulated history, similar to the memoryless transitions in a Markov process. Based on this observation, we propose Atom of Thoughts (AoT), where each state transition in the reasoning process consists of decomposing the current question into a dependency-based directed acyclic graph and contracting its subquestions, forming a new atomic question state. This iterative decomposition-contraction process continues until reaching directly solvable atomic questions, naturally realizing Markov transitions between question states. Furthermore, these atomic questions can be seamlessly integrated into existing test-time scaling methods, enabling AoT to serve as a plug-in enhancement for improving reasoning capabilities. Experiments across six benchmarks demonstrate the effectiveness of AoT both as a standalone framework and a plug-in enhancement. Notably, on HotpotQA, when applied to gpt-4o-mini, AoT achieves an 80.6% F1 score, surpassing o3-mini by 3.4% and DeepSeek-R1 by 10.6%. The code will be available at https://github.com/qixucen/atom.
Authors: Xin Xu, Yan Xu, Tianhao Chen, Yuchen Yan, Chengwu Liu, Zaoyu Chen, Yufei Wang, Yichun Yin, Yasheng Wang, Lifeng Shang, Qun Liu
Abstract: Existing approaches to mathematical reasoning with large language models (LLMs) rely on Chain-of-Thought (CoT) for generalizability or Tool-Integrated Reasoning (TIR) for precise computation. While efforts have been made to combine these methods, they primarily rely on post-selection or predefined strategies, leaving an open question: whether LLMs can autonomously adapt their reasoning strategy based on their inherent capabilities. In this work, we propose TATA (Teaching LLMs According to Their Aptitude), an adaptive framework that enables LLMs to personalize their reasoning strategy spontaneously, aligning it with their intrinsic aptitude. TATA incorporates base-LLM-aware data selection during supervised fine-tuning (SFT) to tailor training data to the model's unique abilities. This approach equips LLMs to autonomously determine and apply the appropriate reasoning strategy at test time. We evaluate TATA through extensive experiments on six mathematical reasoning benchmarks, using both general-purpose and math-specialized LLMs. Empirical results demonstrate that TATA effectively combines the complementary strengths of CoT and TIR, achieving superior or comparable performance with improved inference efficiency compared to TIR alone. Further analysis underscores the critical role of aptitude-aware data selection in enabling LLMs to make effective and adaptive reasoning decisions and align reasoning strategies with model capabilities.
Authors: Aurian Quelennec, Pierre Chouteau, Geoffroy Peeters, Slim Essid
Abstract: Recently, self-supervised learning methods based on masked latent prediction have proven to encode input data into powerful representations. However, during training, the learned latent space can be further transformed to extract higher-level information that could be more suited for downstream classification tasks. Therefore, we propose a new method: MAsked latenT Prediction And Classification (MATPAC), which is trained with two pretext tasks solved jointly. As in previous work, the first pretext task is a masked latent prediction task, ensuring a robust input representation in the latent space. The second one is unsupervised classification, which utilises the latent representations of the first pretext task to match probability distributions between a teacher and a student. We validate the MATPAC method by comparing it to other state-of-the-art proposals and conducting ablations studies. MATPAC reaches state-of-the-art self-supervised learning results on reference audio classification datasets such as OpenMIC, GTZAN, ESC-50 and US8K and outperforms comparable supervised methods results for musical auto-tagging on Magna-tag-a-tune.
Authors: Luc\'ia Yan Wu, Isabel Segura-Bedmar
Abstract: The rise of LLMs (Large Language Models) has contributed to the improved performance and development of cutting-edge NLP applications. However, these can also pose risks when used maliciously, such as spreading fake news, harmful content, impersonating individuals, or facilitating school plagiarism, among others. This is because LLMs can generate high-quality texts, which are challenging to differentiate from those written by humans. GLTR, which stands for Giant Language Model Test Room and was developed jointly by the MIT-IBM Watson AI Lab and HarvardNLP, is a visual tool designed to help detect machine-generated texts based on GPT-2, that highlights the words in text depending on the probability that they were machine-generated. One limitation of GLTR is that the results it returns can sometimes be ambiguous and lead to confusion. This study aims to explore various ways to improve GLTR's effectiveness for detecting AI-generated texts within the context of the IberLef-AuTexTification 2023 shared task, in both English and Spanish languages. Experiment results show that our GLTR-based GPT-2 model overcomes the state-of-the-art models on the English dataset with a macro F1-score of 80.19%, except for the first ranking model (80.91%). However, for the Spanish dataset, we obtained a macro F1-score of 66.20%, which differs by 4.57% compared to the top-performing model.
Authors: Heming Xia, Yongqi Li, Chak Tou Leong, Wenjie Wang, Wenjie Li
Abstract: Chain-of-Thought (CoT) has been proven effective in enhancing the reasoning capabilities of large language models (LLMs). Recent advancements, such as OpenAI's o1 and DeepSeek-R1, suggest that scaling up the length of CoT sequences during inference could further boost LLM reasoning performance. However, due to the autoregressive nature of LLM decoding, longer CoT outputs lead to a linear increase in inference latency, adversely affecting user experience, particularly when the CoT exceeds 10,000 tokens. To address this limitation, we analyze the semantic importance of tokens within CoT outputs and reveal that their contributions to reasoning vary. Building on this insight, we propose TokenSkip, a simple yet effective approach that enables LLMs to selectively skip less important tokens, allowing for controllable CoT compression. Extensive experiments across various models and tasks demonstrate the effectiveness of TokenSkip in reducing CoT token usage while preserving strong reasoning performance. Notably, when applied to Qwen2.5-14B-Instruct, TokenSkip reduces reasoning tokens by 40% (from 313 to 181) on GSM8K, with less than a 0.4% performance drop.
Authors: Maxime Peyrard, Kyunghyun Cho
Abstract: This work demonstrates that the tools and principles driving the success of large language models (LLMs) can be repurposed to tackle distribution-level tasks, where the goal is to predict properties of the data-generating distribution rather than labels for individual datapoints. These tasks encompass statistical inference problems such as parameter estimation, hypothesis testing, or mutual information estimation. Framing these tasks within traditional machine learning pipelines is challenging, as supervision is typically tied to individual datapoint. We propose meta-statistical learning, a framework inspired by multi-instance learning that reformulates statistical inference tasks as supervised learning problems. In this approach, entire datasets are treated as single inputs to neural networks, which predict distribution-level parameters. Transformer-based architectures, without positional encoding, provide a natural fit due to their permutation-invariance properties. By training on large-scale synthetic datasets, meta-statistical models can leverage the scalability and optimization infrastructure of Transformer-based LLMs. We demonstrate the framework's versatility with applications in hypothesis testing and mutual information estimation, showing strong performance, particularly for small datasets where traditional neural methods struggle.
Authors: Sina Salek, Joseph Enguehard
Abstract: Integrated Gradients (IG), a widely used axiomatic path-based attribution method, assigns importance scores to input features by integrating model gradients along a straight path from a baseline to the input. While effective in some cases, we show that straight paths can lead to flawed attributions. In this paper, we identify the cause of these misattributions and propose an alternative approach that treats the input space as a Riemannian manifold, computing attributions by integrating gradients along geodesics. We call this method Geodesic Integrated Gradients (GIG). To approximate geodesic paths, we introduce two techniques: a k-Nearest Neighbours-based approach for smaller models and a Stochastic Variational Inference-based method for larger ones. Additionally, we propose a new axiom, Strong Completeness, extending the axioms satisfied by IG. We show that this property is desirable for attribution methods and that GIG is the only method that satisfies it. Through experiments on both synthetic and real-world data, we demonstrate that GIG outperforms existing explainability methods, including IG.
Authors: Pengda Wang, Huiqi Zou, Hanjie Chen, Tianjun Sun, Ziang Xiao, Frederick L. Oswald
Abstract: Although psychometrics researchers have recently explored the use of large language models (LLMs) as proxies for human participants, LLMs often fail to generate heterogeneous data with human-like diversity, which diminishes their value in advancing social science research. To address these challenges, we explored the potential of the theory-informed Personality Structured Interview (PSI) as a tool for simulating human responses in personality research. In this approach, the simulation is grounded in nuanced real-human interview transcripts that target the personality construct of interest. We have provided a growing set of 357 structured interview transcripts from a representative sample, each containing an individual's response to 32 open-ended questions carefully designed to gather theory-based personality evidence. Additionally, grounded in psychometric research, we have summarized an evaluation framework to systematically validate LLM-generated psychometric data. Results from three experiments demonstrate that well-designed structured interviews could improve human-like heterogeneity in LLM-simulated personality data and predict personality-related behavioral outcomes (i.e., organizational citizenship behaviors and counterproductive work behavior). We further discuss the role of theory-informed structured interviews in LLM-based simulation and outline a general framework for designing structured interviews to simulate human-like data for psychometric research.
Authors: Jinhe Bi, Yifan Wang, Danqi Yan, Xun Xiao, Artur Hecker, Volker Tresp, Yunpu Ma
Abstract: Visual instruction tuning refines pre-trained Multimodal Large Language Models (MLLMs) to enhance their real-world task performance. However, the rapid expansion of visual instruction datasets introduces significant data redundancy, leading to excessive computational costs. Existing data selection methods predominantly rely on proxy models or loss-based metrics, both of which impose substantial computational overheads due to the necessity of model inference and backpropagation. To address this challenge, we propose PRISM, a novel training-free approach for efficient multimodal data selection. Unlike existing methods, PRISM eliminates the reliance on proxy models, warm-up pretraining, and gradient-based optimization. Instead, it leverages Pearson correlation analysis to quantify the intrinsic visual encoding properties of MLLMs, computing a task-specific correlation score to identify high-value instances. This not only enbles data-efficient selection,but maintains the original performance. Empirical evaluations across multiple MLLMs demonstrate that PRISM reduces the overall time required for visual instruction tuning and data selection to just 30% of conventional methods, while surpassing fully fine-tuned models across eight multimodal and three language understanding benchmarks, achieving a 101.7% relative improvement in final performance.
Authors: Prasanna Mayilvahanan, Thadd\"aus Wiedemer, Sayak Mallick, Matthias Bethge, Wieland Brendel
Abstract: Scaling laws guide the development of large language models (LLMs) by offering estimates for the optimal balance of model size, tokens, and compute. More recently, loss-to-loss scaling laws that relate losses across pretraining datasets and downstream tasks have emerged as a powerful tool for understanding and improving LLM performance. In this work, we investigate which factors most strongly influence loss-to-loss scaling. Our experiments reveal that the pretraining data and tokenizer determine the scaling trend. In contrast, model size, optimization hyperparameters, and even significant architectural differences, such as between transformer-based models like Llama and state-space models like Mamba, have limited impact. Consequently, practitioners should carefully curate suitable pretraining datasets for optimal downstream performance, while architectures and other settings can be freely optimized for training efficiency.
Authors: Florian Sestak, Artur Toshev, Andreas F\"urst, G\"unter Klambauer, Andreas Mayr, Johannes Brandstetter
Abstract: Generative models are spearheading recent progress in deep learning, showing strong promise for trajectory sampling in dynamical systems as well. However, while latent space modeling paradigms have transformed image and video generation, similar approaches are more difficult for most dynamical systems. Such systems -- from chemical molecule structures to collective human behavior -- are described by interactions of entities, making them inherently linked to connectivity patterns and the traceability of entities over time. Our approach, LaM-SLidE (Latent Space Modeling of Spatial Dynamical Systems via Linked Entities), combines the advantages of graph neural networks, i.e., the traceability of entities across time-steps, with the efficiency and scalability of recent advances in image and video generation, where pre-trained encoder and decoder are frozen to enable generative modeling in the latent space. The core idea of LaM-SLidE is to introduce identifier representations (IDs) to allow for retrieval of entity properties, e.g., entity coordinates, from latent system representations and thus enables traceability. Experimentally, across different domains, we show that LaM-SLidE performs favorably in terms of speed, accuracy, and generalizability. (Code is available at https://github.com/ml-jku/LaM-SLidE)
Authors: Jinyan Su, Jennifer Healey, Preslav Nakov, Claire Cardie
Abstract: Retrieval-Augmented Generation (RAG) has emerged as a powerful approach to mitigate large language model (LLM) hallucinations by incorporating external knowledge retrieval. However, existing RAG frameworks often apply retrieval indiscriminately,leading to inefficiencies-over-retrieving when unnecessary or failing to retrieve iteratively when required for complex reasoning. Recent adaptive retrieval strategies, though adaptively navigates these retrieval strategies, predict only based on query complexity and lacks user-driven flexibility, making them infeasible for diverse user application needs. In this paper, we introduce a novel user-controllable RAG framework that enables dynamic adjustment of the accuracy-cost trade-off. Our approach leverages two classifiers: one trained to prioritize accuracy and another to prioritize retrieval efficiency. Via an interpretable control parameter $\alpha$, users can seamlessly navigate between minimal-cost retrieval and high-accuracy retrieval based on their specific requirements. We empirically demonstrate that our approach effectively balances accuracy, retrieval cost, and user controllability, making it a practical and adaptable solution for real-world applications.
Authors: Kenan Jiang, Li Xiong, Fei Liu
Abstract: We investigate factors contributing to LLM agents' success in competitive multi-agent environments, using auctions as a testbed where agents bid to maximize profit. The agents are equipped with bidding domain knowledge, distinct personas that reflect item preferences, and a memory of auction history. Our work extends the classic auction scenario by creating a realistic environment where multiple agents bid on houses, weighing aspects such as size, location, and budget to secure the most desirable homes at the lowest prices. Particularly, we investigate three key questions: (a) How does a persona influence an agent's behavior in a competitive setting? (b) Can an agent effectively profile its competitors' behavior during auctions? (c) How can persona profiling be leveraged to create an advantage using strategies such as theory of mind? Through a series of experiments, we analyze the behaviors of LLM agents and shed light on new findings. Our testbed, called HARBOR, offers a valuable platform for deepening our understanding of multi-agent workflows in competitive environments.
Authors: Zhicong Tang, Jianmin Bao, Dong Chen, Baining Guo
Abstract: This paper presents Model-guidance (MG), a novel objective for training diffusion model that addresses and removes of the commonly used Classifier-free guidance (CFG). Our innovative approach transcends the standard modeling of solely data distribution to incorporating the posterior probability of conditions. The proposed technique originates from the idea of CFG and is easy yet effective, making it a plug-and-play module for existing models. Our method significantly accelerates the training process, doubles the inference speed, and achieve exceptional quality that parallel and even surpass concurrent diffusion models with CFG. Extensive experiments demonstrate the effectiveness, efficiency, scalability on different models and datasets. Finally, we establish state-of-the-art performance on ImageNet 256 benchmarks with an FID of 1.34. Our code is available at https://github.com/tzco/Diffusion-wo-CFG.
Authors: R. V. Bettker, P. P. Minini, A. G. Pereira, M. Ritt
Abstract: We study the problem of learning good heuristic functions for classical planning tasks with neural networks based on samples represented by states with their cost-to-goal estimates. The heuristic function is learned for a state space and goal condition with the number of samples limited to a fraction of the size of the state space, and must generalize well for all states of the state space with the same goal condition. Our main goal is to better understand the influence of sample generation strategies on the performance of a greedy best-first heuristic search (GBFS) guided by a learned heuristic function. In a set of controlled experiments, we find that two main factors determine the quality of the learned heuristic: the algorithm used to generate the sample set and how close the sample estimates to the perfect cost-to-goal are. These two factors are dependent: having perfect cost-to-goal estimates is insufficient if the samples are not well distributed across the state space. We also study other effects, such as adding samples with high-value estimates. Based on our findings, we propose practical strategies to improve the quality of learned heuristics: three strategies that aim to generate more representative states and two strategies that improve the cost-to-goal estimates. Our practical strategies result in a learned heuristic that, when guiding a GBFS algorithm, increases by more than 30% the mean coverage compared to a baseline learned heuristic.
Authors: James Chao, Wiktor Piotrowski, Roni Stern, H\'ector Ortiz-Pe\~na, Mitch Manzanares, Shiwali Mohan, Douglas S. Lange
Abstract: Autonomous agents operating within real-world environments often rely on automated planners to ascertain optimal actions towards desired goals or the optimization of a specified objective function. Integral to these agents are common architectural components such as schedulers, tasked with determining the timing for executing planned actions, and execution engines, responsible for carrying out these scheduled actions while monitoring their outcomes. We address the significant challenge that arises when unexpected phenomena, termed \textit{novelties}, emerge within the environment, altering its fundamental characteristics, composition, and dynamics. This challenge is inherent in all deployed real-world applications and may manifest suddenly and without prior notice or explanation. The introduction of novelties into the environment can lead to inaccuracies within the planner's internal model, rendering previously generated plans obsolete. Recent research introduced agent design aimed at detecting and adapting to such novelties. However, these designs lack consideration for action scheduling in continuous time-space, coordination of concurrent actions by multiple agents, or memory-based novelty accommodation. Additionally, the application has been primarily demonstrated in lower fidelity environments. In our study, we propose a general purpose AI agent framework designed to detect, characterize, and adapt to novelties in highly noisy, complex, and stochastic environments that support concurrent actions and external scheduling. We showcase the efficacy of our agent through experimentation within a high-fidelity simulator for realistic military scenarios.
Authors: Venetia Pliatsika, Joao Fonseca, Kateryna Akhynko, Ivan Shevchenko, Julia Stoyanovich
Abstract: Algorithmic decisions in critical domains such as hiring, college admissions, and lending are often based on rankings. Given the impact of these decisions on individuals, organizations, and population groups, it is essential to understand them-to help individuals improve their ranking position, design better ranking procedures, and ensure legal compliance. In this paper, we argue that explainability methods for classification and regression, such as SHAP, are insufficient for ranking tasks, and present ShaRP-Shapley Values for Rankings and Preferences-a framework that explains the contributions of features to various aspects of a ranked outcome. ShaRP computes feature contributions for various ranking-specific profit functions, such as rank and top-k, and also includes a novel Shapley value-based method for explaining pairwise preference outcomes. We provide a flexible implementation of ShaRP, capable of efficiently and comprehensively explaining ranked and pairwise outcomes over tabular data, in score-based ranking and learning-to-rank tasks. Finally, to evaluate ShaRP and compare it with other explainability methods, we define ranking-specific explanation metrics and conduct an extensive experimental analysis, demonstrating the framework's flexibility and efficiency.
Authors: Xuhan Huang, Qingning Shen, Yan Hu, Anningzhe Gao, Benyou Wang
Abstract: Large Language Models (LLMs) have demonstrated strong performance across various natural language processing tasks, yet their proficiency in mathematical reasoning remains a key challenge. Addressing the gap between natural and mathematical language requires advanced reasoning capabilities, approaching those of Artificial General Intelligence (AGI). However, the evaluation remains challenging, as perfectly representing reality is inherently elusive, and traditional methods like manual or direct comparison of mathematical statements (Ramamonjison et al., 2023) are insufficient for assessing true modeling ability. We propose a process-oriented framework to evaluate LLMs' ability to construct mathematical models, using solvers to compare outputs with ground truth. Introducing Mamo, a benchmark with 1,209 questions covering ordinary differential equations, linear programming, and mixed-integer linear programming, we enable automatic evaluation of modeling accuracy. The results show that existing LLMs struggle with complex mathematical modeling tasks, with larger models demonstrating superior performance, while open-source models remain competitive in simpler cases but still fall short of proprietary models in more challenging problems.
Authors: Devon Graham, Kevin Leyton-Brown
Abstract: Utilitarian algorithm configuration is a general-purpose technique for automatically searching the parameter space of a given algorithm to optimize its performance, as measured by a given utility function, on a given set of inputs. Recently introduced utilitarian configuration procedures offer optimality guarantees about the returned parameterization while provably adapting to the hardness of the underlying problem. However, the applicability of these approaches is severely limited by the fact that they only search a finite, relatively small set of parameters. They cannot effectively search the configuration space of algorithms with continuous or uncountable parameters. In this paper we introduce a new procedure, which we dub COUP (Continuous, Optimistic Utilitarian Procrastination). COUP is designed to search infinite parameter spaces efficiently to find good configurations quickly. Furthermore, COUP maintains the theoretical benefits of previous utilitarian configuration procedures when applied to finite parameter spaces but is significantly faster, both provably and experimentally.
Authors: Vivek Farias, Joren Gijsbrechts, Aryan Khojandi, Tianyi Peng, Andrew Zheng
Abstract: Simulating a single trajectory of a dynamical system under some state-dependent policy is a core bottleneck in policy optimization (PO) algorithms. The many inherently serial policy evaluations that must be performed in a single simulation constitute the bulk of this bottleneck. In applying PO to supply chain optimization (SCO) problems, simulating a single sample path corresponding to one month of a supply chain can take several hours. We present an iterative algorithm to accelerate policy simulation, dubbed Picard Iteration. This scheme carefully assigns policy evaluation tasks to independent processes. Within an iteration, any given process evaluates the policy only on its assigned tasks while assuming a certain "cached" evaluation for other tasks; the cache is updated at the end of the iteration. Implemented on GPUs, this scheme admits batched evaluation of the policy across a single trajectory. We prove that the structure afforded by many SCO problems allows convergence in a small number of iterations independent of the horizon. We demonstrate practical speedups of 400x on large-scale SCO problems even with a single GPU, and also demonstrate practical efficacy in other RL environments.
Authors: Eleonora Mancini, Ana Tanevska, Andrea Galassi, Alessio Galatolo, Federico Ruggeri, Paolo Torroni
Abstract: Current research in machine learning and artificial intelligence is largely centered on modeling and performance evaluation, less so on data collection. However, recent research demonstrated that limitations and biases in data may negatively impact trustworthiness and reliability. These aspects are particularly impactful on sensitive domains such as mental health and neurological disorders, where speech data are used to develop AI applications for patients and healthcare providers. In this paper, we chart the landscape of available speech datasets for this domain, to highlight possible pitfalls and opportunities for improvement and promote fairness and diversity. We present a comprehensive list of desiderata for building speech datasets for mental health and neurological disorders and distill it into an actionable checklist focused on ethical concerns to foster more responsible research.
Authors: Siyin Wang, Xingsong Ye, Qinyuan Cheng, Junwen Duan, Shimin Li, Jinlan Fu, Xipeng Qiu, Xuanjing Huang
Abstract: As Artificial General Intelligence (AGI) becomes increasingly integrated into various facets of human life, ensuring the safety and ethical alignment of such systems is paramount. Previous studies primarily focus on single-modality threats, which may not suffice given the integrated and complex nature of cross-modality interactions. We introduce a novel safety alignment challenge called Safe Inputs but Unsafe Output (SIUO) to evaluate cross-modality safety alignment. Specifically, it considers cases where single modalities are safe independently but could potentially lead to unsafe or unethical outputs when combined. To empirically investigate this problem, we developed the SIUO, a cross-modality benchmark encompassing 9 critical safety domains, such as self-harm, illegal activities, and privacy violations. Our findings reveal substantial safety vulnerabilities in both closed- and open-source LVLMs, such as GPT-4V and LLaVA, underscoring the inadequacy of current models to reliably interpret and respond to complex, real-world scenarios.
Authors: Qiming Wu, Zichen Chen, Will Corcoran, Misha Sra, Ambuj K. Singh
Abstract: Large language models (LLMs) have achieved remarkable success in natural language processing (NLP), demonstrating significant capabilities in processing and understanding text data. However, recent studies have identified limitations in LLMs' ability to manipulate, program, and reason about structured data, especially graphs. We introduce GraphEval36K, the first comprehensive graph dataset, comprising 40 graph coding problems and 36,900 test cases to evaluate the ability of LLMs on graph problem-solving. Our dataset is categorized into eight primary and four sub-categories to ensure a thorough evaluation across different types of graphs. We benchmark ten LLMs, finding that private models outperform open-source ones, though the gap is narrowing. We also analyze the performance of LLMs across directed vs undirected graphs, different kinds of graph concepts, and network models. Furthermore, to improve the usability of our evaluation framework, we propose Structured Symbolic Decomposition (SSD), an instruction-based method designed to enhance LLM performance on complex graph tasks. Results show that SSD improves the average passing rate of GPT-4, GPT-4o, Gemini-Pro and Claude-3-Sonnet by 8.38%, 6.78%, 29.28% and 25.28%, respectively.
Authors: Jianheng Tang, Qifan Zhang, Yuhan Li, Nuo Chen, Jia Li
Abstract: The ``arms race'' of Large Language Models (LLMs) demands new benchmarks to examine their progresses. In this paper, we introduce GraphArena, a benchmarking tool designed to evaluate LLMs on real-world graph computational problems. It offers a suite of four polynomial-time tasks (e.g., Shortest Distance) and six NP-complete challenges (e.g., Traveling Salesman Problem). GraphArena features a rigorous evaluation framework that classifies LLM outputs as correct, suboptimal (feasible but not optimal), hallucinatory (properly formatted but infeasible), or missing. Evaluation of over 10 LLMs reveals that even top-performing LLMs struggle with larger, more complex graph problems and exhibit hallucination issues. We further explore four potential solutions to address this issue and improve LLMs on graph computation, including chain-of-thought prompting, instruction tuning, code writing, and scaling test-time compute, each demonstrating unique strengths and limitations. GraphArena complements the existing LLM benchmarks and is open-sourced at https://github.com/squareRoot3/GraphArena.
Authors: Ali AhmadiTeshnizi, Wenzhi Gao, Herman Brunborg, Shayan Talaei, Connor Lawless, Madeleine Udell
Abstract: Optimization problems are pervasive in sectors from manufacturing and distribution to healthcare. However, most such problems are still solved heuristically by hand rather than optimally by state-of-the art solvers because the expertise required to formulate and solve these problems limits the widespread adoption of optimization tools and techniques. We introduce a Large Language Model (LLM)-based system designed to formulate and solve (mixed integer) linear programming problems from their natural language descriptions. Our system is capable of developing mathematical models, writing and debugging solver code, evaluating the generated solutions, and improving efficiency and correctness of its model and code based on these evaluations. OptiMUS-0.3 utilizes a modular structure to process problems, allowing it to handle problems with long descriptions and complex data without long prompts. Experiments demonstrate that OptiMUS-0.3 outperforms existing state-of-the-art methods on easy datasets by more than 12% and on hard datasets (including a new dataset, NLP4LP, released with this paper that features long and complex problems) by more than 8%.
Authors: Yanqi Dai, Huanran Hu, Lei Wang, Shengjie Jin, Xu Chen, Zhiwu Lu
Abstract: Recently, Role-Playing Agents (RPAs) have garnered increasing attention for their potential to deliver emotional value and facilitate sociological research. However, existing studies are primarily confined to the textual modality, unable to simulate humans' multimodal perceptual capabilities. To bridge this gap, we introduce the concept of Multimodal Role-Playing Agents (MRPAs), and propose a comprehensive framework, MMRole, for their development and evaluation, which comprises a personalized multimodal dataset and a robust evaluation approach. Specifically, we construct a large-scale, high-quality dataset, MMRole-Data, consisting of 85 characters, 11K images, and 14K single or multi-turn dialogues. Additionally, we present a robust evaluation approach, MMRole-Eval, encompassing eight metrics across three dimensions, where a reward model is designed to score MRPAs with the constructed ground-truth data for comparison. Moreover, we develop the first specialized MRPA, MMRole-Agent. Extensive evaluation results demonstrate the improved performance of MMRole-Agent and highlight the primary challenges in developing MRPAs, emphasizing the need for enhanced multimodal understanding and role-playing consistency. The data, code, and models are all available at https://github.com/YanqiDai/MMRole.
Authors: Linjie Xu, Diego Perez-Liebana, Alexander Dockhorn
Abstract: Playing strategy games is a challenging problem for artificial intelligence (AI). One of the major challenges is the large search space due to a diverse set of game components. In recent works, state abstraction has been applied to search-based game AI and has brought significant performance improvements. State abstraction techniques rely on reducing the search space, e.g., by aggregating similar states. However, the application of these abstractions is hindered because the quality of an abstraction is difficult to evaluate. Previous works hence abandon the abstraction in the middle of the search to not bias the search to a local optimum. This mechanism introduces a hyper-parameter to decide the time to abandon the current state abstraction. In this work, we propose a size-constrained state abstraction (SCSA), an approach that limits the maximum number of nodes being grouped together. We found that with SCSA, the abstraction is not required to be abandoned. Our empirical results on $3$ strategy games show that the SCSA agent outperforms the previous methods and yields robust performance over different games. Codes are open-sourced at https://github.com/GAIGResearch/Stratega.
Authors: Antonio Rago, Bence Palfi, Purin Sukpanichnant, Hannibal Nabli, Kavyesh Vivek, Olga Kostopoulou, James Kinross, Francesca Toni
Abstract: AI-driven tools for healthcare are widely acknowledged as potentially beneficial to health practitioners and patients, e.g. the QCancer regression tool for cancer risk prediction. However, for these tools to be trusted, they need to be supplemented with explanations. We examine how explanations' content and format affect user comprehension and trust when explaining QCancer's predictions. Regarding content, we deploy SHAP and Occlusion-1. Regarding format, we present SHAP explanations, conventionally, as charts (SC) and Occlusion-1 explanations as charts (OC) as well as text (OT), to which their simpler nature lends itself. We conduct experiments with two sets of stakeholders: the general public (representing patients) and medical students (representing healthcare practitioners). Our experiments showed higher subjective comprehension and trust for Occlusion-1 over SHAP explanations based on content. However, when controlling for format, only OT outperformed SC, suggesting this trend is driven by preferences for text. Other findings corroborated that explanation format, rather than content, is often the critical factor.
Authors: Dezhi Luo, Haiyun Lyu, Qingying Gao, Haoran Sun, Yijiang Li, Hokin Deng
Abstract: Conservation is a critical milestone of cognitive development considered to be supported by both the understanding of quantitative concepts and the reversibility of operations. To assess whether this critical component of human intelligence has emerged in Vision Language Models, we have curated the ConserveBench, a battery of 365 cognitive experiments across four dimensions of physical quantities: volume, solid quantity, length, and number. The former two involve transformational tasks which require reversibility understanding. The latter two involve non-transformational tasks which assess quantity understanding. Surprisingly, we find that while Vision Language Models are generally good at transformational tasks, they tend to fail at non-transformational tasks. There is a dissociation between understanding the reversibility of operations and understanding of quantity, which both are believed to be the cornerstones of the understanding of law of conservation in humans. $\href{https://growing-ai-like-a-child.github.io/pages/Conservation/}{Website}$
URLs: https://growing-ai-like-a-child.github.io/pages/Conservation/
Authors: Nikolai Rozanov, Marek Rei
Abstract: Planning and acting to solve `real' tasks using large language models (LLMs) in interactive environments has become a new frontier for AI methods. While recent advances allowed LLMs to interact with online tools, solve robotics tasks and many more, long range reasoning tasks remain a problem for LLMs. Existing methods to address this issue are very resource intensive and require additional data or human crafted rules, instead, we propose a simple method based on few-shot in-context learning alone to enhance `chain-of-thought' with state-tracking for planning and acting with LLMs. We show that our method establishes the new state-of-the-art on Alfworld for in-context learning methods (+14\% over the previous best few-shot in-context learning method) and performs on par with methods that use additional training data and additional tools such as code-execution. We also demonstrate that our enhanced `chain-of-states' allows the agent to both solve longer horizon problems and to be more efficient in number of steps required to solve a task. We show that our method works across a variety of LLMs for both API-based and open source ones. Finally, we also conduct ablation studies and show that `chain-of-thoughts' helps state-tracking accuracy, while a json-structure harms overall performance. We open-source our code and annotations at https://github.com/ai-nikolai/StateAct.
Authors: Chengkun Cai, Xu Zhao, Haoliang Liu, Zhongyu Jiang, Tianfang Zhang, Zongkai Wu, Jenq-Neng Hwang, Serge Belongie, Lei Li
Abstract: Large Language Models (LLMs) have demonstrated impressive capabilities in reasoning tasks, yet their reliance on static prompt structures and limited adaptability to complex scenarios remains a significant challenge. In this paper, we propose the Deductive and InDuctive(DID) method, a novel framework that enhances LLM reasoning by dynamically integrating both deductive and inductive reasoning approaches. Drawing from cognitive science principles, DID implements a dual-metric complexity evaluation system that combines Littlestone dimension and information entropy to precisely assess task difficulty and guide decomposition strategies. DID enables the model to progressively adapt its reasoning pathways based on problem complexity, mirroring human cognitive processes. We evaluate DID's effectiveness across multiple benchmarks, including the AIW and MR-GSM8K, as well as our custom Holiday Puzzle dataset for temporal reasoning. Our results demonstrate significant improvements in reasoning quality and solution accuracy - achieving 70.3% accuracy on AIW (compared to 62.2% for Tree of Thought) while maintaining lower computational costs. The success of DID in improving LLM performance while preserving computational efficiency suggests promising directions for developing more cognitively aligned and capable language models. Our work contributes a theoretically grounded, input-centric approach to enhancing LLM reasoning capabilities, offering an efficient alternative to traditional output-exploration methods.
Authors: Ze Wang, Zekun Wu, Jeremy Zhang, Xin Guan, Navya Jain, Skylar Lu, Saloni Gupta, Adriano Koshiyama
Abstract: Model collapse, a phenomenon where models degrade in performance due to indiscriminate use of synthetic data is well studied. However, its role in bias amplification, the progressive reinforcement of preexisting social biases in Large Language Models (LLMs) remains underexplored. In this paper, we formally define the conditions for bias amplification and demonstrate through statistical simulations that bias can intensify even in the absence of sampling errors, the primary driver of model collapse. Empirically, we investigate political bias amplification in GPT2 using a custom built benchmark for sentence continuation tasks. Our findings reveal a progressively increasing right-leaning bias. Furthermore, we evaluate three mitigation strategies, Overfitting, Preservation, and Accumulation, and show that bias amplification persists even when model collapse is mitigated. Finally, a mechanistic interpretation identifies distinct sets of neurons responsible for model collapse and bias amplification, suggesting they arise from different underlying mechanisms.
Authors: Lang Cao, Chao Peng, Renhong Chen, Wu Ning, Yingtian Zou, Yitong Li
Abstract: Mathematical reasoning has been challenging for large language models (LLMs). However, the introduction of step-by-step Chain-of-Thought (CoT) inference has significantly advanced the mathematical capabilities of LLMs. Despite this progress, current approaches either necessitate extensive inference datasets for training or depend on few-shot methods that frequently compromise computational accuracy. To address these bottlenecks in mathematical reasoning, we propose a novel method called Step Guidied Reasoning, which is more stable and generalizable than few-shot methods and does not involve further fine-tuning of the model. In this approach, LLMs reflect on small reasoning steps, similar to how humans deliberate and focus attention on what to do next. By incorporating this reflective process into the inference stage, LLMs can effectively guide their reasoning from one step to the next. Through extensive experiments, we demonstrate the significant effect of Step Guidied Reasoning in augmenting mathematical performance in state-of-the-art language models. Qwen2-72B-Instruct outperforms its math-specific counterpart, Qwen2.5-72B-Math-Instruct, on MMLU- STEM with a score of 90.9%, compared to 87.3%. The average scores of Qwen2-7B-Instruct and Qwen2-72B-Instruct increase from 27.1% to 36.3% and from 36.5% to 47.4% on the mathematics domain, respectively.
Authors: Charles K. Assaad
Abstract: Understanding causal mechanisms across different populations is essential for designing effective public health interventions. Recently, difference graphs have been introduced as a tool to visually represent causal variations between two distinct populations. While there has been progress in inferring these graphs from data through causal discovery methods, there remains a gap in systematically leveraging their potential to enhance causal reasoning. This paper addresses that gap by establishing conditions for identifying causal changes and effects using difference graphs. It specifically focuses on identifying total causal changes and total effects in a nonparametric setting, as well as direct causal changes and direct effects in a linear setting. In doing so, it provides a novel approach to causal reasoning that holds potential for various public health applications.
Authors: Jiahao Wang, Mingyue Cheng, Qingyang Mao, Yitong Zhou, Feiyang Xu, Xin Li
Abstract: Large language models (LLMs) have demonstrated their effectiveness in multivariate time series classification (MTSC). Effective adaptation of LLMs for MTSC necessitates informative data representations. Existing LLM-based methods directly encode embeddings for time series within the latent space of LLMs from scratch to align with semantic space of LLMs. Despite their effectiveness, we reveal that these methods conceal three inherent bottlenecks: (1) they struggle to encode temporal and channel-specific information in a lossless manner, both of which are critical components of multivariate time series; (2) it is much difficult to align the learned representation space with the semantic space of the LLMs; (3) they require task-specific retraining, which is both computationally expensive and labor-intensive. To bridge these gaps, we propose TableTime, which reformulates MTSC as a table understanding task. Specifically, TableTime introduces the following strategies: (1) convert multivariate time series into a tabular form, thus minimizing information loss to the greatest extent; (2) represent tabular time series in text format to achieve natural alignment with the semantic space of LLMs; (3) design a reasoning framework that integrates contextual text information, neighborhood assistance, multi-path inference and problem decomposition to enhance the reasoning ability of LLMs and realize zero-shot classification. Extensive experiments performed on 10 publicly representative datasets from UEA archive verify the superiorities of the TableTime.
Authors: Lanxiang Hu, Qiyu Li, Anze Xie, Nan Jiang, Ion Stoica, Haojian Jin, Hao Zhang
Abstract: Evaluating the reasoning abilities of large language models (LLMs) is challenging. Existing benchmarks often depend on static datasets, which are vulnerable to data contamination and may get saturated over time, or on binary live human feedback that conflates reasoning with other abilities. As the most prominent dynamic benchmark, Chatbot Arena evaluates open-ended questions in real-world settings, but lacks the granularity in assessing specific reasoning capabilities. We introduce GameArena, a dynamic benchmark designed to evaluate LLM reasoning capabilities through interactive gameplay with humans. GameArena consists of three games designed to test specific reasoning capabilities (e.g., deductive and inductive reasoning), while keeping participants entertained and engaged. We analyze the gaming data retrospectively to uncover the underlying reasoning processes of LLMs and measure their fine-grained reasoning capabilities. We collect over 2000 game sessions and provide detailed assessments of various reasoning capabilities for five state-of-the-art LLMs. Our user study with 100 participants suggests that GameArena improves user engagement compared to Chatbot Arena. For the first time, GameArena enables the collection of step-by-step LLM reasoning data in the wild.
Authors: Shixin Jiang, Jiafeng Liang, Jiyuan Wang, Xuan Dong, Heng Chang, Weijiang Yu, Jinhua Du, Ming Liu, Bing Qin
Abstract: To tackle complex tasks in real-world scenarios, more researchers are focusing on Omni-MLLMs, which aim to achieve omni-modal understanding and generation. Beyond the constraints of any specific non-linguistic modality, Omni-MLLMs map various non-linguistic modalities into the embedding space of LLMs and enable the interaction and understanding of arbitrary combinations of modalities within a single model. In this paper, we systematically investigate relevant research and provide a comprehensive survey of Omni-MLLMs. Specifically, we first explain the four core components of Omni-MLLMs for unified multi-modal modeling with a meticulous taxonomy that offers novel perspectives. Then, we introduce the effective integration achieved through two-stage training and discuss the corresponding datasets as well as evaluation. Furthermore, we summarize the main challenges of current Omni-MLLMs and outline future directions. We hope this paper serves as an introduction for beginners and promotes the advancement of related research. Resources will be made public.
Authors: Eitan Wagner, Nitay Alon, Joseph M. Barnby, Omri Abend
Abstract: Theory of Mind (ToM) capabilities in LLMs have recently become a central object of investigation. Cognitive science distinguishes between two steps required for ToM tasks: 1) determine whether to invoke ToM, which includes the appropriate Depth of Mentalizing (DoM), or level of recursion required to complete a task; and 2) applying the correct inference given the DoM. In this position paper, we first identify several lines of work in different communities in AI, including LLM benchmarking, ToM add-ons, ToM probing, and formal models for ToM. We argue that recent work in AI tends to focus exclusively on the second step which are typically framed as static logic problems. We conclude with suggestions for improved evaluation of ToM capabilities inspired by dynamic environments used in cognitive tasks.
Authors: Tianyang Gu, Jingjin Wang, Zhihao Zhang, HaoHong Li
Abstract: Scientific idea generation has been extensively studied in creativity theory and computational creativity research, providing valuable frameworks for understanding and implementing creative processes. However, recent work using Large Language Models (LLMs) for research idea generation often overlooks these theoretical foundations. We present a framework that explicitly implements combinatorial creativity theory using LLMs, featuring a generalization-level retrieval system for cross-domain knowledge discovery and a structured combinatorial process for idea generation. The retrieval system maps concepts across different abstraction levels to enable meaningful connections between disparate domains, while the combinatorial process systematically analyzes and recombines components to generate novel solutions. Experiments on the OAG-Bench dataset demonstrate our framework's effectiveness, consistently outperforming baseline approaches in generating ideas that align with real research developments (improving similarity scores by 7\%-10\% across multiple metrics). Our results provide strong evidence that LLMs can effectively realize combinatorial creativity when guided by appropriate theoretical frameworks, contributing both to practical advancement of AI-assisted research and theoretical understanding of machine creativity.
Authors: Carl Qi, Dan Haramati, Tal Daniel, Aviv Tamar, Amy Zhang
Abstract: Object manipulation is a common component of everyday tasks, but learning to manipulate objects from high-dimensional observations presents significant challenges. These challenges are heightened in multi-object environments due to the combinatorial complexity of the state space as well as of the desired behaviors. While recent approaches have utilized large-scale offline data to train models from pixel observations, achieving performance gains through scaling, these methods struggle with compositional generalization in unseen object configurations with constrained network and dataset sizes. To address these issues, we propose a novel behavioral cloning (BC) approach that leverages object-centric representations and an entity-centric Transformer with diffusion-based optimization, enabling efficient learning from offline image data. Our method first decomposes observations into an object-centric representation, which is then processed by our entity-centric Transformer that computes attention at the object level, simultaneously predicting object dynamics and the agent's actions. Combined with the ability of diffusion models to capture multi-modal behavior distributions, this results in substantial performance improvements in multi-object tasks and, more importantly, enables compositional generalization. We present BC agents capable of zero-shot generalization to tasks with novel compositions of objects and goals, including larger numbers of objects than seen during training. We provide video rollouts on our webpage: https://sites.google.com/view/ec-diffuser.
Authors: Jiang Liu, Bolin Li, Haoyuan Li, Tianwei Lin, Wenqiao Zhang, Tao Zhong, Zhelun Yu, Jinghao Wei, Hao Cheng, Wanggui He, Fangxun Shu, Hao Jiang, Zheqi Lv, Juncheng Li, Siliang Tang, Yueting Zhuang
Abstract: Efficient multimodal large language models (EMLLMs), in contrast to multimodal large language models (MLLMs), reduce model size and computational costs and are often deployed on resource-constrained devices. However, due to data privacy concerns, existing open-source EMLLMs rarely have access to private domain-specific data during the pre-training process, making them difficult to directly apply in device-specific domains, such as certain business scenarios. To address this weakness, this paper focuses on the efficient adaptation of EMLLMs to private domains, specifically in two areas: 1) how to reduce data requirements, and 2) how to avoid parameter fine-tuning. Specifically, we propose a tun\textbf{\underline{I}}ng-free, a\textbf{\underline{D}}aptiv\textbf{\underline{E}}, univers\textbf{\underline{AL}} \textbf{\underline{Prompt}} Optimization Framework, abbreviated as \textit{\textbf{\ourmethod{}}} which consists of two stages: 1) Predefined Prompt, based on the reinforcement searching strategy, generate a prompt optimization strategy tree to acquire optimization priors; 2) Prompt Reflection initializes the prompt based on optimization priors, followed by self-reflection to further search and refine the prompt. By doing so, \ourmethod{} elegantly generates the ``ideal prompts'' for processing private domain-specific data. Note that our method requires no parameter fine-tuning and only a small amount of data to quickly adapt to the data distribution of private data. Extensive experiments across multiple tasks demonstrate that our proposed \ourmethod{} significantly improves both efficiency and performance compared to baselines.
Authors: Pritam Sil, Bhaskaran Raman, Pushpak Bhattacharyya
Abstract: Assessments play a vital role in a student's learning process by providing feedback on a student's proficiency level in a subject. While assessments often make use of short answer questions, it is often difficult to grade such questions at a large scale. Moreover, such questions often involve students drawing supporting diagrams along with their textual explanations. Such questions often promote multimodal literacy and are aligned with competency-based questions, which demand a deeper cognitive processing ability from students. However, existing literature does not deal with the automatic grading of such answers. Thus, to bridge this gap, we propose the Multimodal Short Answer Grading with Feedback (MMSAF) problem along with a dataset of 2197 data points. Additionally, we provide an automated framework for generating such datasets. Our evaluations on existing Large Language Models (LLMs) over this dataset achieved an overall accuracy of 55% on the Level of Correctness labels and 75% on Image Relevance labels. As per human experts, Pixtral was more aligned towards human judgement and values for biology and ChatGPT for physics and chemistry and achieved a score of 4 or more out of 5 in most parameters.
Authors: Leo Ardon, Daniel Furelos-Blanco, Roko Parac, Alessandra Russo
Abstract: Reward machines (RMs) are an effective approach for addressing non-Markovian rewards in reinforcement learning (RL) through finite-state machines. Traditional RMs, which label edges with propositional logic formulae, inherit the limited expressivity of propositional logic. This limitation hinders the learnability and transferability of RMs since complex tasks will require numerous states and edges. To overcome these challenges, we propose First-Order Reward Machines ($\texttt{FORM}$s), which use first-order logic to label edges, resulting in more compact and transferable RMs. We introduce a novel method for $\textbf{learning}$ $\texttt{FORM}$s and a multi-agent formulation for $\textbf{exploiting}$ them and facilitate their transferability, where multiple agents collaboratively learn policies for a shared $\texttt{FORM}$. Our experimental results demonstrate the scalability of $\texttt{FORM}$s with respect to traditional RMs. Specifically, we show that $\texttt{FORM}$s can be effectively learnt for tasks where traditional RM learning approaches fail. We also show significant improvements in learning speed and task transferability thanks to the multi-agent learning framework and the abstraction provided by the first-order language.
Authors: Nantheera Anantrasirichai, Fan Zhang, David Bull
Abstract: The rapid advancements in artificial intelligence (AI), particularly in generative AI and large language models (LLMs), have profoundly impacted the creative industries by enabling innovative content creation, enhancing workflows, and democratizing access to creative tools. This paper explores the significant technological shifts since our previous review in 2022, highlighting how these developments have expanded creative opportunities and efficiency. These technological advancements have enhanced the capabilities of text-to-image, text-to-video, and multimodal generation technologies. In particular, key breakthroughs in LLMs have established new benchmarks in conversational AI, while advancements in image generators have revolutionized content creation. We also discuss AI integration into post-production workflows, which has significantly accelerated and refined traditional processes. Despite these innovations, challenges remain, particularly for the media industry, due to the demands on communication traffic from creative content. We therefore include data compression and quality assessment in this paper. Furthermore, we highlight the trend toward unified AI frameworks capable of addressing multiple creative tasks and underscore the importance of human oversight to mitigate AI-generated inaccuracies. Finally, we explore AI's future potential in the creative sector, stressing the need to navigate emerging challenges to maximize its benefits while addressing associated risks.
Authors: Muhammet Alkan, Idris Zakariyya, Samuel Leighton, Kaushik Bhargav Sivangi, Christos Anagnostopoulos, Fani Deligianni
Abstract: As artificial intelligence (AI) becomes increasingly embedded in healthcare delivery, this chapter explores the critical aspects of developing reliable and ethical Clinical Decision Support Systems (CDSS). Beginning with the fundamental transition from traditional statistical models to sophisticated machine learning approaches, this work examines rigorous validation strategies and performance assessment methods, including the crucial role of model calibration and decision curve analysis. The chapter emphasizes that creating trustworthy AI systems in healthcare requires more than just technical accuracy; it demands careful consideration of fairness, explainability, and privacy. The challenge of ensuring equitable healthcare delivery through AI is stressed, discussing methods to identify and mitigate bias in clinical predictive models. The chapter then delves into explainability as a cornerstone of human-centered CDSS. This focus reflects the understanding that healthcare professionals must not only trust AI recommendations but also comprehend their underlying reasoning. The discussion advances in an analysis of privacy vulnerabilities in medical AI systems, from data leakage in deep learning models to sophisticated attacks against model explanations. The text explores privacy-preservation strategies such as differential privacy and federated learning, while acknowledging the inherent trade-offs between privacy protection and model performance. This progression, from technical validation to ethical considerations, reflects the multifaceted challenges of developing AI systems that can be seamlessly and reliably integrated into daily clinical practice while maintaining the highest standards of patient care and data protection.
Authors: Muhammet Alkan, Hester Huijsdens, Yola Jones, Fani Deligianni
Abstract: The pivotal shift from traditional paper-based records to sophisticated Electronic Health Records (EHR), enabled systematic collection and analysis of patient data through descriptive statistics, providing insight into patterns and trends across patient populations. This evolution continued toward predictive analytics, allowing healthcare providers to anticipate patient outcomes and potential complications before they occur. This progression from basic digital record-keeping to sophisticated predictive modelling and digital twins reflects healthcare's broader evolution toward more integrated, patient-centred approaches that combine data-driven insights with personalized care delivery. This chapter explores the evolution and significance of healthcare information systems, beginning with an examination of the implementation of EHR in the UK and the USA. It provides a comprehensive overview of the International Classification of Diseases (ICD) system, tracing its development from ICD-9 to ICD-10. Central to this discussion is the MIMIC-III database, a landmark achievement in healthcare data sharing and arguably the most comprehensive critical care database freely available to researchers worldwide. MIMIC-III has democratized access to high-quality healthcare data, enabling unprecedented opportunities for research and analysis. The chapter examines its structure, clinical outcome analysis capabilities, and practical applications through case studies, with a particular focus on mortality and length of stay metrics, vital signs extraction, and ICD coding. Through detailed entity-relationship diagrams and practical examples, the text illustrates MIMIC's complex data structure and demonstrates how different querying approaches can lead to subtly different results, emphasizing the critical importance of understanding the database's architecture for accurate data extraction.
Authors: Joao Marques-Silva, Xuanxiang Huang, Olivier Letoffe
Abstract: Recent work demonstrated the existence of critical flaws in the current use of Shapley values in explainable AI (XAI), i.e. the so-called SHAP scores. These flaws are significant in that the scores provided to a human decision-maker can be misleading. Although these negative results might appear to indicate that Shapley values ought not be used in XAI, this paper argues otherwise. Concretely, this paper proposes a novel definition of SHAP scores that overcomes existing flaws. Furthermore, the paper outlines a practically efficient solution for the rigorous estimation of the novel SHAP scores. Preliminary experimental results confirm our claims, and further underscore the flaws of the current SHAP scores.
Authors: Yulan Hu, Sheng Ouyang, Yong Liu
Abstract: The Process Reward Model (PRM) plays a crucial role in mathematical reasoning tasks, requiring high-quality supervised process data. However, we observe that reasoning steps generated by Large Language Models (LLMs) often fail to exhibit strictly incremental information, leading to redundancy that can hinder effective reasoning. To address this issue, we propose \model, a simple yet effective coarse-to-fine strategy. Instead of focusing on the detection of redundant steps, our approach first establishes a coarse-grained window to merge adjacent reasoning steps into unified, holistic steps. The window size is then progressively reduced to extract fine-grained reasoning steps, enabling data collection at multiple granularities for training. By leveraging this hierarchical refinement process, \model mitigates redundancy while preserving essential fine-grained knowledge. Extensive experiments on two reasoning datasets across three loss criteria validate the \model's effectiveness and versatility.
Authors: Yutong Yin, Zhaoran Wang
Abstract: Humans exhibit remarkable compositional reasoning by integrating knowledge from various sources. For example, if someone learns ( B = f(A) ) from one source and ( C = g(B) ) from another, they can deduce ( C=g(B)=g(f(A)) ) even without encountering ( ABC ) together, showcasing the generalization ability of human intelligence. In this paper, we introduce a synthetic learning task, "FTCT" (Fragmented at Training, Chained at Testing), to validate the potential of Transformers in replicating this skill and interpret its inner mechanism. In the training phase, data consist of separated knowledge fragments from an overall causal graph. During testing, Transformers must infer complete causal graph traces by integrating these fragments. Our findings demonstrate that few-shot Chain-of-Thought prompting enables Transformers to perform compositional reasoning on FTCT by revealing correct combinations of fragments, even if such combinations were absent in the training data. Furthermore, the emergence of compositional reasoning ability is strongly correlated with the model complexity and training-testing data similarity. We propose, both theoretically and empirically, that Transformers learn an underlying generalizable program from training, enabling effective compositional reasoning during testing.
Authors: Jialun Cao, Yaojie Lu, Meiziniu Li, Haoyang Ma, Haokun Li, Mengda He, Cheng Wen, Le Sun, Hongyu Zhang, Shengchao Qin, Shing-Chi Cheung, Cong Tian
Abstract: The research in AI-based formal mathematical reasoning has shown an unstop- pable growth trend. These studies have excelled in mathematical competitions like IMO and have made significant progress. This paper focuses on formal verification, an immediate application scenario of formal reasoning, and breaks it down into sub-tasks. We constructed 18k high-quality instruction-response pairs across five formal specification languages (Coq, Lean4, Dafny, ACSL, and TLA+) by distilling gpt-4o and evaluated against ten open-sourced LLMs, including recent popular DeepSeek-R1. We also fine-tuned several 7~8B small models to achieve comparable performance with Deepseek-R1-671B. Interestingly, we observed that fine-tuning with formal data also enhances mathematics, reasoning, and coding capabilities. Fine-tuned models are released at https: //huggingface.co/fm-universe.
Authors: Yingxuan Yang, Bo Huang, Siyuan Qi, Chao Feng, Haoyi Hu, Yuxuan Zhu, Jinbo Hu, Haoran Zhao, Ziyi He, Xiao Liu, Zongyu Wang, Lin Qiu, Xuezhi Cao, Xunliang Cai, Yong Yu, Weinan Zhang
Abstract: Large Language Model (LLM) agents frameworks often employ modular architectures, incorporating components such as planning, reasoning, action execution, and reflection to tackle complex tasks. However, quantifying the contribution of each module to overall system performance remains a significant challenge, impeding optimization and interpretability. To address this, we introduce CapaBench (Capability-level Assessment Benchmark), an evaluation framework grounded in cooperative game theory's Shapley Value, which systematically measures the marginal impact of individual modules and their interactions within an agent's architecture. By replacing default modules with test variants across all possible combinations, CapaBench provides a principle method for attributing performance contributions. Key contributions include: (1) We are the first to propose a Shapley Value-based methodology for quantifying the contributions of capabilities in LLM agents; (2) Modules with high Shapley Values consistently lead to predictable performance gains when combined, enabling targeted optimization; and (3) We build a multi-round dataset of over 1,500 entries spanning diverse domains and practical task scenarios, enabling comprehensive evaluation of agent capabilities. CapaBench bridges the gap between component-level evaluation and holistic system assessment, providing actionable insights for optimizing modular LLM agents and advancing their deployment in complex, real-world scenarios.
Authors: Haozhe Wang, Long Li, Chao Qu, Fengming Zhu, Weidi Xu, Wei Chu, Fangzhen Lin
Abstract: Recent advances in mathematical problem-solving with language models (LMs) integrate chain-of-thought (CoT) reasoning and code execution to harness their complementary strengths. However, existing hybrid frameworks exhibit a critical limitation: they depend on externally dictated instructions or rigid code-integration templates, lacking metacognitive awareness -- the capacity to dynamically evaluate intrinsic capabilities and autonomously determine when and how to integrate tools. This rigidity motivates our study of autonomous code integration, enabling models to adapt tool-usage strategies as their reasoning abilities evolve during training. While reinforcement learning (RL) shows promise for boosting LLM reasoning at scale (e.g., DeepSeek-R1), we demonstrate its inefficiency in learning autonomous code integration due to inadequate exploration of the vast combinatorial space of CoT-code interleaving patterns. To address this challenge, we propose a novel Expectation-Maximization (EM) framework that synergizes structured exploration (E-step) with off-policy RL optimization (M-step), creating a self-reinforcing cycle between metacognitive tool-use decisions and evolving capabilities. Experiments reveal our method achieves superior results through improved exploration. Notably, our 7B model improves over 11% on MATH500 and 9.4% on AIME without o1-like CoT.
Authors: Amitayush Thakur, George Tsoukalas, Greg Durrett, Swarat Chaudhuri
Abstract: Neural networks have shown substantial promise at automatic theorem-proving in interactive proof assistants (ITPs) like Lean and Coq. However, most neural theorem-proving models are restricted to specific ITPs, leaving out opportunities for cross-lingual $\textit{transfer}$ between ITPs. We address this weakness with a multilingual proof framework, ${\rm P{\small ROOF}W{\small ALA}}$, that allows a standardized form of interaction between neural theorem-provers and two established ITPs (Coq and Lean). It enables the collection of multilingual proof step data -- data recording the result of proof actions on ITP states -- for training neural provers. ${\rm P{\small ROOF}W{\small ALA}}$ allows the systematic evaluation of a model's performance across different ITPs and problem domains via efficient parallel proof search algorithms. We show that multilingual training enabled by ${\rm P{\small ROOF}W{\small ALA}}$ can lead to successful transfer across ITPs. Specifically, a model trained on a mix of ${\rm P{\small ROOF}W{\small ALA}}$-generated Coq and Lean data outperforms Lean-only and Coq-only models on the standard prove-at-$k$ metric. We open source all code including code for the ${\rm P{\small ROOF}W{\small ALA}}$ Framework (https://github.com/trishullab/proof-wala), and the Multilingual ITP interaction framework (https://github.com/trishullab/itp-interface).
URLs: https://github.com/trishullab/proof-wala),, https://github.com/trishullab/itp-interface).
Authors: Fan Liu, Wenshuo Chao, Naiqiang Tan, Hao Liu
Abstract: With the advancement of large language models (LLMs), solving complex reasoning tasks has gained increasing attention. Inference-time computation methods (e.g., Best-of-N, beam search, et al.) are particularly valuable as they can enhance reasoning performance without modifying model parameters or requiring additional training. However, these techniques come with implementation challenges, and most existing methods remain at the proof-of-concept stage with limited practical adoption due to their computational complexity and varying effectiveness across different tasks. In this paper, we investigate and benchmark diverse inference-time computation strategies across reasoning tasks of varying complexity. Since most current methods rely on a proposer-verifier pipeline that first generates candidate solutions (e.g., reasoning solutions) and then selects the best one based on reward signals (e.g., RLHF rewards, process rewards), our research focuses on optimizing both candidate solution generation (e.g., instructing prompts, hyperparameters such as temperature and top-p) and reward mechanisms (e.g., self-evaluation, reward types). Through extensive experiments (more than 20,000 A100-80G GPU hours with over 1,000 experiments) across a variety of models (e.g., Llama, Qwen, and Mistral families) of various sizes, our ablation studies reveal that previously overlooked strategies can significantly enhance performance (e.g., tuning temperature can improve reasoning task performance by up to 5%). Furthermore, we establish a standardized benchmark for inference-time computation by systematically evaluating six representative methods across eight reasoning tasks. These findings provide a stronger foundation for future research. The code is available at https://github.com/usail-hkust/benchmark_inference_time_computation_LLM
URLs: https://github.com/usail-hkust/benchmark_inference_time_computation_LLM
Authors: Josh Rozner, Christopher Potts, Kyle Mahowald
Abstract: Cryptic crosswords, the dominant crossword variety in the UK, are a promising target for advancing NLP systems that seek to process semantically complex, highly compositional language. Cryptic clues read like fluent natural language but are adversarially composed of two parts: a definition and a wordplay cipher requiring character-level manipulations. Expert humans use creative intelligence to solve cryptics, flexibly combining linguistic, world, and domain knowledge. In this paper, we make two main contributions. First, we present a dataset of cryptic clues as a challenging new benchmark for NLP systems that seek to process compositional language in more creative, human-like ways. After showing that three non-neural approaches and T5, a state-of-the-art neural language model, do not achieve good performance, we make our second main contribution: a novel curriculum approach, in which the model is first fine-tuned on related tasks such as unscrambling words.We also introduce a challenging data split, examine the meta-linguistic capabilities of subword-tokenized models, and investigate model systematicity by perturbing the wordplay part of clues, showing that T5 exhibits behavior partially consistent with human solving strategies. Although our curricular approach considerably improves on the T5 baseline, our best-performing model still fails to generalize to the extent that humans can. Thus, cryptic crosswords remain an unsolved challenge for NLP systems and a potential source of future innovation.
Authors: Tianya Zhang
Abstract: Traffic state data, such as speed, volume and travel time collected from ubiquitous traffic monitoring sensors require advanced network level analytics for forecasting and identifying significant traffic patterns. This paper leverages diverse traffic state datasets from the Caltrans Performance Measurement System (PeMS) hosted on the open benchmark and achieved promising performance compared to well recognized spatial-temporal models. Drawing inspiration from the success of hierarchical architectures in various Artificial Intelligence (AI) tasks, we integrate cell and hidden states from low-level to high-level Long Short-Term Memory (LSTM) networks with an attention pooling mechanism, similar to human perception systems. The developed hierarchical structure is designed to account for dependencies across different time scales, capturing the spatial-temporal correlations of network-level traffic states, enabling the prediction of traffic states for all corridors rather than a single link or route. The efficiency of designed attention-based LSTM is analyzed by ablation study. Comparative results with baseline LSTM models demonstrate that the Hierarchical Attention LSTM (HierAttnLSTM) model not only provides higher prediction accuracy but also effectively forecasts unusual congestion patterns. Data and code are made publicly available to support reproducible scientific research.
Authors: Lukas Muttenthaler, Jonas Dippel, Lorenz Linhardt, Robert A. Vandermeulen, Simon Kornblith
Abstract: Today's computer vision models achieve human or near-human level performance across a wide variety of vision tasks. However, their architectures, data, and learning algorithms differ in numerous ways from those that give rise to human vision. In this paper, we investigate the factors that affect the alignment between the representations learned by neural networks and human mental representations inferred from behavioral responses. We find that model scale and architecture have essentially no effect on the alignment with human behavioral responses, whereas the training dataset and objective function both have a much larger impact. These findings are consistent across three datasets of human similarity judgments collected using two different tasks. Linear transformations of neural network representations learned from behavioral responses from one dataset substantially improve alignment with human similarity judgments on the other two datasets. In addition, we find that some human concepts such as food and animals are well-represented by neural networks whereas others such as royal or sports-related objects are not. Overall, although models trained on larger, more diverse datasets achieve better alignment with humans than models trained on ImageNet alone, our results indicate that scaling alone is unlikely to be sufficient to train neural networks with conceptual representations that match those used by humans.
Authors: Di Wang, Yao Wang, Shao-Bo Lin
Abstract: In recent years, large amounts of electronic health records (EHRs) concerning chronic diseases have been collected to facilitate medical diagnosis. Modeling the dynamic properties of EHRs related to chronic diseases can be efficiently done using dynamic treatment regimes (DTRs). While reinforcement learning (RL) is a widely used method for creating DTRs, there is ongoing research in developing RL algorithms that can effectively handle large amounts of data. In this paper, we present a scalable kernel-based distributed Q-learning algorithm for generating DTRs. We perform both theoretical assessments and numerical analysis for the proposed approach. The results demonstrate that our algorithm significantly reduces the computational complexity associated with the state-of-the-art deep reinforcement learning methods, while maintaining comparable generalization performance in terms of accumulated rewards across stages, such as survival time or cumulative survival probability.
Authors: Saviz Mowlavi, Ken Kamrin
Abstract: Most noninvasive imaging techniques utilize electromagnetic or acoustic waves originating from multiple locations and directions to identify hidden geometrical structures. Surprisingly, it is also possible to image hidden voids and inclusions buried within an object using a single static thermal or mechanical loading experiment by observing the response of the exposed surface of the body, but this problem is challenging to invert. Although physics-informed neural networks (PINNs) have shown promise as a simple-yet-powerful tool for problem inversion, they have not yet been applied to imaging problems with a priori unknown topology. Here, we introduce a topology optimization framework based on PINNs that identifies concealed geometries using exposed surface data from a single loading experiment, without prior knowledge of the number or types of shapes. We allow for arbitrary solution topology by representing the geometry using a material density field combined with a novel eikonal regularization technique. We validate our framework by detecting the number, locations, and shapes of hidden voids and inclusions in many example cases, in both 2D and 3D, and we demonstrate the method's robustness to noise and sparsity in the data. Our methodology opens a pathway for PINNs to solve geometry optimization problems in engineering.
Authors: Zixuan Hu, Li Shen, Zhenyi Wang, Tongliang Liu, Chun Yuan, Dacheng Tao
Abstract: The goal of data-free meta-learning is to learn useful prior knowledge from a collection of pre-trained models without accessing their training data. However, existing works only solve the problem in parameter space, which (i) ignore the fruitful data knowledge contained in the pre-trained models; (ii) can not scale to large-scale pre-trained models; (iii) can only meta-learn pre-trained models with the same network architecture. To address those issues, we propose a unified framework, dubbed PURER, which contains: (1) ePisode cUrriculum inveRsion (ECI) during data-free meta training; and (2) invErsion calibRation following inner loop (ICFIL) during meta testing. During meta training, we propose ECI to perform pseudo episode training for learning to adapt fast to new unseen tasks. Specifically, we progressively synthesize a sequence of pseudo episodes by distilling the training data from each pre-trained model. The ECI adaptively increases the difficulty level of pseudo episodes according to the real-time feedback of the meta model. We formulate the optimization process of meta training with ECI as an adversarial form in an end-to-end manner. During meta testing, we further propose a simple plug-and-play supplement-ICFIL-only used during meta testing to narrow the gap between meta training and meta testing task distribution. Extensive experiments in various real-world scenarios show the superior performance of ours.
Authors: Tianya Zhang, Ph. D., Peter J. Jin, Ph. D., Sean T. McQuade, Ph. D., Alexandre Bayen, Ph. D., Benedetto Piccoli
Abstract: Car-following (CF) algorithms are crucial components of traffic simulations and have been integrated into many production vehicles equipped with Advanced Driving Assistance Systems (ADAS). Insights from the model of car-following behavior help us understand the causes of various macro phenomena that arise from interactions between pairs of vehicles. Car-following models encompass multiple disciplines, including traffic engineering, physics, dynamic system control, cognitive science, machine learning, and reinforcement learning. This paper presents an extensive survey that highlights the differences, complementarities, and overlaps among microscopic traffic flow and control models based on their underlying principles and design logic. It reviews representative algorithms, ranging from theory-based kinematic models, Psycho-Physical Models, and Adaptive cruise control models to data-driven algorithms like Reinforcement Learning (RL) and Imitation Learning (IL). The manuscript discusses the strengths and limitations of these models and explores their applications in different contexts. This review synthesizes existing researches across different domains to fill knowledge gaps and offer guidance for future research by identifying the latest trends in car following models and their applications.
Authors: Caiyang Yu, Xianggen Liu, Yifan Wang, Yun Liu, Wentao Feng, Deng Xiong, Chenwei Tang, Jiancheng Lv
Abstract: Neural Architecture Search (NAS) has emerged as one of the effective methods to design the optimal neural network architecture automatically. Although neural architectures have achieved human-level performances in several tasks, few of them are obtained from the NAS method. The main reason is the huge search space of neural architectures, making NAS algorithms inefficient. This work presents a novel architecture search algorithm, called GPT-NAS, that optimizes neural architectures by Generative Pre-Trained (GPT) model with an evolutionary algorithm (EA) as the search strategy. In GPT-NAS, we assume that a generative model pre-trained on a large-scale corpus could learn the fundamental law of building neural architectures. Therefore, GPT-NAS leverages the GPT model to propose reasonable architecture components given the basic one and then utilizes EAs to search for the optimal solution. Such an approach can largely reduce the search space by introducing prior knowledge in the search process. Extensive experimental results show that our GPT-NAS method significantly outperforms seven manually designed neural architectures and thirteen architectures provided by competing NAS methods. In addition, our experiments also indicate that the proposed algorithm improves the performance of finely tuned neural architectures by up to about 12% compared to those without GPT, further demonstrating its effectiveness in searching neural architectures.
Authors: Zixuan Hu, Li Shen, Zhenyi Wang, Baoyuan Wu, Chun Yuan, Dacheng Tao
Abstract: Data-free meta-learning (DFML) aims to enable efficient learning of new tasks by meta-learning from a collection of pre-trained models without access to the training data. Existing DFML work can only meta-learn from (i) white-box and (ii) small-scale pre-trained models (iii) with the same architecture, neglecting the more practical setting where the users only have inference access to the APIs with arbitrary model architectures and model scale inside. To solve this issue, we propose a Bi-level Data-free Meta Knowledge Distillation (BiDf-MKD) framework to transfer more general meta knowledge from a collection of black-box APIs to one single meta model. Specifically, by just querying APIs, we inverse each API to recover its training data via a zero-order gradient estimator and then perform meta-learning via a novel bi-level meta knowledge distillation structure, in which we design a boundary query set recovery technique to recover a more informative query set near the decision boundary. In addition, to encourage better generalization within the setting of limited API budgets, we propose task memory replay to diversify the underlying task distribution by covering more interpolated tasks. Extensive experiments in various real-world scenarios show the superior performance of our BiDf-MKD framework.
Authors: Chao Bian, Yawen Zhou, Miqing Li, Chao Qian
Abstract: Evolutionary algorithms (EAs) have been widely and successfully applied to solve multi-objective optimization problems, due to their nature of population-based search. Population update, a key component in multi-objective EAs (MOEAs), is usually performed in a greedy, deterministic manner. That is, the next-generation population is formed by selecting the best solutions from the current population and newly-generated solutions (irrespective of the selection criteria used such as Pareto dominance, crowdedness and indicators). In this paper, we analytically present that stochastic population update can be beneficial for the search of MOEAs. Specifically, we prove that the expected running time of two well-established MOEAs, SMS-EMOA and NSGA-II, for solving two bi-objective problems, OneJumpZeroJump and bi-objective RealRoyalRoad, can be exponentially decreased if replacing its deterministic population update mechanism by a stochastic one. Empirical studies also verify the effectiveness of the proposed population update method. This work is an attempt to show the benefit of introducing randomness into the population update of MOEAs. Its positive results, which might hold more generally, should encourage the exploration of developing new MOEAs in the area.
Authors: Yijun Li, Cheuk Hang Leung, Qi Wu
Abstract: Probabilistic forecasting of multivariate time series is essential for various downstream tasks. Most existing approaches rely on the sequences being uniformly spaced and aligned across all variables. However, real-world multivariate time series often suffer from temporal irregularities, including nonuniform intervals and misaligned variables, which pose significant challenges for accurate forecasting. To address these challenges, we propose an end-to-end framework that models temporal irregularities while capturing the joint distribution of variables at arbitrary continuous-time points. Specifically, we introduce a dynamic conditional continuous normalizing flow to model data distributions in a non-parametric manner, accommodating the complex, non-Gaussian characteristics commonly found in real-world datasets. Then, by leveraging a carefully factorized log-likelihood objective, our approach captures both temporal and cross-sectional dependencies efficiently. Extensive experiments on a range of real-world datasets demonstrate the superiority and adaptability of our method compared to existing approaches.
Authors: Yidong Zhu, Shao-Hsien Liu, Mohammad Arif Ul Alam
Abstract: The integration of diverse health data, such as IoT (Internet of Things), EHR (Electronic Health Record), and clinical surveys, with scalable AI(Artificial Intelligence) has enabled the identification of physical, behavioral, and psycho-social indicators of pain. However, the adoption of AI in clinical pain evaluation is hindered by challenges like personalization and fairness. Many AI models, including machine and deep learning, exhibit biases, discriminating against specific groups based on gender or ethnicity, causing skepticism among medical professionals about their reliability. This paper proposes a Multi-attribute Fairness Loss (MAFL) based Convolutional Neural Network (CNN) model designed to account for protected attributes in data, ensuring fair pain status predictions while minimizing disparities between privileged and unprivileged groups. We evaluate whether a balance between accuracy and fairness is achievable by comparing the proposed model with existing mitigation methods. Our findings indicate that the model performs favorably against state-of-the-art techniques. Using the NIH All-Of-US dataset, comprising data from 868 individuals over 1500 days, we demonstrate our model's effectiveness, achieving accuracy rates between 75% and 85%.
Authors: Weiqiu You, Helen Qu, Marco Gatti, Bhuvnesh Jain, Eric Wong
Abstract: Self-attributing neural networks (SANNs) present a potential path towards interpretable models for high-dimensional problems, but often face significant trade-offs in performance. In this work, we formally prove a lower bound on errors of per-feature SANNs, whereas group-based SANNs can achieve zero error and thus high performance. Motivated by these insights, we propose Sum-of-Parts (SOP), a framework that transforms any differentiable model into a group-based SANN, where feature groups are learned end-to-end without group supervision. SOP achieves state-of-the-art performance for SANNs on vision and language tasks, and we validate that the groups are interpretable on a range of quantitative and semantic metrics. We further validate the utility of SOP explanations in model debugging and cosmological scientific discovery. Code is available at https://github.com/BrachioLab/sop.
Authors: Huaman Sun, Jiaxin Pei, Minje Choi, David Jurgens
Abstract: Human judgments are inherently subjective and are actively affected by personal traits such as gender and ethnicity. While Large Language Models (LLMs) are widely used to simulate human responses across diverse contexts, their ability to account for demographic differences in subjective tasks remains uncertain. In this study, leveraging the POPQUORN dataset, we evaluate nine popular LLMs on their ability to understand demographic differences in two subjective judgment tasks: politeness and offensiveness. We find that in zero-shot settings, most models' predictions for both tasks align more closely with labels from White participants than those from Asian or Black participants, while only a minor gender bias favoring women appears in the politeness task. Furthermore, sociodemographic prompting does not consistently improve and, in some cases, worsens LLMs' ability to perceive language from specific sub-populations. These findings highlight potential demographic biases in LLMs when performing subjective judgment tasks and underscore the limitations of sociodemographic prompting as a strategy to achieve pluralistic alignment. Code and data are available at: https://github.com/Jiaxin-Pei/LLM-as-Subjective-Judge.
URLs: https://github.com/Jiaxin-Pei/LLM-as-Subjective-Judge.
Authors: Lei Guan, Dongsheng Li, Yongle Chen, Jiye Liang, Wenjian Wang, Xicheng Lu
Abstract: Asynchronous pipeline model parallelism with a "1F1B" (one forward, one backward) schedule generates little bubble overhead and always provides quite a high throughput. However, the "1F1B" schedule inevitably leads to weight inconsistency and weight staleness issues due to the cross-training of different mini-batches across GPUs. To simultaneously address these two problems, in this paper, we propose an optimizer-dependent weight prediction strategy (a.k.a PipeOptim) for asynchronous pipeline training. The key insight of our proposal is that we employ a weight prediction strategy in the forward pass to ensure that each mini-batch uses consistent and staleness-free weights to compute the forward pass. To be concrete, we first construct the weight prediction scheme based on the update rule of the used optimizer when training the deep neural network models. Then throughout the "1F1B" pipelined training, each mini-batch is mandated to execute weight prediction ahead of the forward pass, subsequently employing the predicted weights to perform the forward pass. As a result, PipeOptim 1) inherits the advantage of the "1F1B" schedule and generates pretty high throughput, and 2) can ensure effective parameter learning regardless of the type of the used optimizer. To verify the effectiveness of our proposal, we conducted extensive experimental evaluations using eight different deep-learning models spanning three machine-learning tasks including image classification, sentiment analysis, and machine translation. The experiment results demonstrate that PipeOptim outperforms the popular pipelined approaches including GPipe, PipeDream, PipeDream-2BW, and SpecTrain. The code of PipeOptim can be accessible at https://github.com/guanleics/PipeOptim.
Authors: Ling Li, Shaohua Li, Winda Marantika, Alex C. Kot, Huijing Zhan
Abstract: Denoising Diffusion Probabilistic Model (DDPM) has shown great competence in image and audio generation tasks. However, there exist few attempts to employ DDPM in the text generation, especially review generation under recommendation systems. Fueled by the predicted reviews explainability that justifies recommendations could assist users better understand the recommended items and increase the transparency of recommendation system, we propose a Diffusion Model-based Review Generation towards EXplainable Recommendation named Diffusion-EXR. Diffusion-EXR corrupts the sequence of review embeddings by incrementally introducing varied levels of Gaussian noise to the sequence of word embeddings and learns to reconstruct the original word representations in the reverse process. The nature of DDPM enables our lightweight Transformer backbone to perform excellently in the recommendation review generation task. Extensive experimental results have demonstrated that Diffusion-EXR can achieve state-of-the-art review generation for recommendation on two publicly available benchmark datasets.
Authors: Kun Wu, Ning Liu, Zhen Zhao, Di Qiu, Jinming Li, Zhengping Che, Zhiyuan Xu, Qinru Qiu, Jian Tang
Abstract: Improving data utilization, especially for imperfect data from task failures, is crucial for robotic manipulation due to the challenging, time-consuming, and expensive data collection process in the real world. Current imitation learning (IL) typically discards imperfect data, focusing solely on successful expert data. While reinforcement learning (RL) can learn from explorations and failures, the sim2real gap and its reliance on dense reward and online exploration make it difficult to apply effectively in real-world scenarios. In this work, we aim to conquer the challenge of leveraging imperfect data without the need for reward information to improve the model performance for robotic manipulation in an offline manner. Specifically, we introduce a Self-Supervised Data Filtering framework (SSDF) that combines expert and imperfect data to compute quality scores for failed trajectory segments. High-quality segments from the failed data are used to expand the training dataset. Then, the enhanced dataset can be used with any downstream policy learning method for robotic manipulation tasks. Extensive experiments on the ManiSkill2 benchmark built on the high-fidelity Sapien simulator and real-world robotic manipulation tasks using the Franka robot arm demonstrated that the SSDF can accurately expand the training dataset with high-quality imperfect data and improve the success rates for all robotic manipulation tasks.
Authors: Xiaopeng Li, Shasha Li, Shezheng Song, Huijun Liu, Bin Ji, Xi Wang, Jun Ma, Jie Yu, Xiaodong Liu, Jing Wang, Weimin Zhang
Abstract: The general capabilities of large language models (LLMs) make them the infrastructure for various AI applications, but updating their inner knowledge requires significant resources. Recent model editing is a promising technique for efficiently updating a small amount of knowledge of LLMs and has attracted much attention. In particular, local editing methods, which directly update model parameters, are proven suitable for updating small amounts of knowledge. Local editing methods update weights by computing least squares closed-form solutions and identify edited knowledge by vector-level matching in inference, which achieve promising results. However, these methods still require a lot of time and resources to complete the computation. Moreover, vector-level matching lacks reliability, and such updates disrupt the original organization of the model's parameters. To address these issues, we propose a detachable and expandable Subject Word Embedding Altering (SWEA) framework, which finds the editing embeddings through token-level matching and adds them to the subject word embeddings in Transformer input. To get these editing embeddings, we propose optimizing then suppressing fusion method, which first optimizes learnable embedding vectors for the editing target and then suppresses the Knowledge Embedding Dimensions (KEDs) to obtain final editing embeddings. We thus propose SWEA$\oplus$OS method for editing factual knowledge in LLMs. We demonstrate the overall state-of-the-art (SOTA) performance of SWEA$\oplus$OS on the CounterFact and zsRE datasets. To further validate the reasoning ability of SWEA$\oplus$OS in editing knowledge, we evaluate it on the more complex RippleEdits benchmark. The results demonstrate that SWEA$\oplus$OS possesses SOTA reasoning ability.
Authors: Yuncheng Hua, Lizhen Qu, Gholamreza Haffari
Abstract: We develop assistive agents based on Large Language Models (LLMs) that aid interlocutors in business negotiations. Specifically, we simulate business negotiations by letting two LLM-based agents engage in role play. A third LLM acts as a remediator agent to rewrite utterances violating norms for improving negotiation outcomes. We introduce a simple tuning-free and label-free In-Context Learning (ICL) method to identify high-quality ICL exemplars for the remediator, where we propose a novel select criteria, called value impact, to measure the quality of the negotiation outcomes. We provide rich empirical evidence to demonstrate its effectiveness in negotiations across three different negotiation topics. We have released our source code and the generated dataset at: https://github.com/tk1363704/SADAS.
Authors: Guangya Wan, Yunsheng Lu, Yuqi Wu, Mengxuan Hu, Sheng Li
Abstract: Causal discovery (CD) and Large Language Models (LLMs) have emerged as transformative fields in artificial intelligence that have evolved largely independently. While CD specializes in uncovering cause-effect relationships from data, and LLMs excel at natural language processing and generation, their integration presents unique opportunities for advancing causal understanding. This survey examines how LLMs are transforming CD across three key dimensions: direct causal extraction from text, integration of domain knowledge into statistical methods, and refinement of causal structures. We systematically analyze approaches that leverage LLMs for CD tasks, highlighting their innovative use of metadata and natural language for causal inference. Our analysis reveals both LLMs' potential to enhance traditional CD methods and their current limitations as imperfect expert systems. We identify key research gaps, outline evaluation frameworks and benchmarks for LLM-based causal discovery, and advocate future research efforts for leveraging LLMs in causality research. As the first comprehensive examination of the synergy between LLMs and CD, this work lays the groundwork for future advances in the field.
Authors: Peyman Hosseini, Mehran Hosseini, Ignacio Castro, Matthew Purver
Abstract: From natural language processing to vision, Scaled Dot Product Attention (SDPA) is the backbone of most modern deep learning applications. Unfortunately, its memory and computational requirements can be prohibitive in low-resource settings. In this paper, we improve its efficiency without sacrificing its versatility. We propose three attention variants where we remove consecutive linear transformations or add a novel one, and evaluate them on a range of standard NLP and vision tasks. Our proposed models are substantially lighter than standard SDPA (and have 25-50% fewer parameters). We show that the performance cost of these changes is negligible relative to size reduction and that in one case (Super Attention) we succeed in outperforming SDPA by up to 10% while improving its speed and reducing its parameters by 25%.
Authors: Bhishma Dedhia, Niraj K. Jha
Abstract: Several accounts of human cognition posit that our intelligence is rooted in our ability to form abstract composable concepts, ground them in our environment, and reason over these grounded entities. This trifecta of human thought has remained elusive in modern intelligent machines. In this work, we investigate whether slot representations extracted from visual scenes serve as appropriate compositional abstractions for grounding and reasoning. We present the Neural Slot Interpreter (NSI), which learns to ground object semantics in slots. At the core of NSI is an XML-like schema that uses simple syntax rules to organize the object semantics of a scene into object-centric schema primitives. Then, the NSI metric learns to ground primitives into slots through a structured contrastive learning objective that reasons over the intermodal alignment. Experiments with a bi-modal object-property and scene retrieval task demonstrate the grounding efficacy and interpretability of correspondences learned by NSI. From a scene representation standpoint, we find that emergent NSI slots that move beyond the image grid by binding to spatial objects facilitate improved visual grounding compared to conventional bounding-box-based approaches. From a data efficiency standpoint, we empirically validate that NSI learns more generalizable representations from a fixed amount of annotation data than the traditional approach. We also show that the grounded slots surpass unsupervised slots in real-world object discovery and scale with scene complexity. Finally, we investigate the reasoning abilities of the grounded slots. Vision Transformers trained on grounding-aware NSI tokenizers using as few as ten tokens outperform patch-based tokens on challenging few-shot classification tasks.
Authors: Liang Yao
Abstract: Prompting methods play a crucial role in enhancing the capabilities of pre-trained large language models (LLMs). We explore how contrastive prompting (CP) significantly improves the ability of large language models to perform complex reasoning. We demonstrate that LLMs are decent contrastive reasoners by simply adding "Let's give a correct and a wrong answer." before LLMs provide answers. Experiments on various large language models show that zero-shot contrastive prompting improves the performance of standard zero-shot prompting on a range of arithmetic, commonsense, and symbolic reasoning tasks without any hand-crafted few-shot examples, such as increasing the accuracy on GSM8K from 35.9% to 88.8% and AQUA-RAT from 41.3% to 62.2% with the state-of-the-art GPT-4 model. Our method not only surpasses zero-shot CoT and few-shot CoT in most arithmetic and commonsense reasoning tasks but also can seamlessly integrate with existing prompting methods, resulting in improved or comparable results when compared to state-of-the-art methods. Our code is available at https://github.com/yao8839836/cp
Authors: Xiaoyu Liu, Paiheng Xu, Junda Wu, Jiaxin Yuan, Yifan Yang, Yuhang Zhou, Fuxiao Liu, Tianrui Guan, Haoliang Wang, Tong Yu, Julian McAuley, Wei Ai, Furong Huang
Abstract: Causal inference has shown potential in enhancing the predictive accuracy, fairness, robustness, and explainability of Natural Language Processing (NLP) models by capturing causal relationships among variables. The emergence of generative Large Language Models (LLMs) has significantly impacted various NLP domains, particularly through their advanced reasoning capabilities. This survey focuses on evaluating and improving LLMs from a causal view in the following areas: understanding and improving the LLMs' reasoning capacity, addressing fairness and safety issues in LLMs, complementing LLMs with explanations, and handling multimodality. Meanwhile, LLMs' strong reasoning capacities can in turn contribute to the field of causal inference by aiding causal relationship discovery and causal effect estimations. This review explores the interplay between causal inference frameworks and LLMs from both perspectives, emphasizing their collective potential to further the development of more advanced and equitable artificial intelligence systems.
Authors: Ziyao Liu, Huanyi Ye, Chen Chen, Yongsen Zheng, Kwok-Yan Lam
Abstract: Machine Unlearning (MU) has recently gained considerable attention due to its potential to achieve Safe AI by removing the influence of specific data from trained Machine Learning (ML) models. This process, known as knowledge removal, addresses AI governance concerns of training data such as quality, sensitivity, copyright restrictions, and obsolescence. This capability is also crucial for ensuring compliance with privacy regulations such as the Right To Be Forgotten (RTBF). Furthermore, effective knowledge removal mitigates the risk of harmful outcomes, safeguarding against biases, misinformation, and unauthorized data exploitation, thereby enhancing the safe and responsible use of AI systems. Efforts have been made to design efficient unlearning approaches, with MU services being examined for integration with existing machine learning as a service (MLaaS), allowing users to submit requests to remove specific data from the training corpus. However, recent research highlights vulnerabilities in machine unlearning systems, such as information leakage and malicious unlearning, that can lead to significant security and privacy concerns. Moreover, extensive research indicates that unlearning methods and prevalent attacks fulfill diverse roles within MU systems. This underscores the intricate relationship and complex interplay among these mechanisms in maintaining system functionality and safety. This survey aims to fill the gap between the extensive number of studies on threats, attacks, and defenses in machine unlearning and the absence of a comprehensive review that categorizes their taxonomy, methods, and solutions, thus offering valuable insights for future research directions and practical implementations.
Authors: Jian Li, Pu Ren, Yang Liu, Hao Sun
Abstract: Object-centric learning aims to break down complex visual scenes into more manageable object representations, enhancing the understanding and reasoning abilities of machine learning systems toward the physical world. Recently, slot-based video models have demonstrated remarkable proficiency in segmenting and tracking objects, but they overlook the importance of the effective reasoning module. In the real world, reasoning and predictive abilities play a crucial role in human perception and object tracking; in particular, these abilities are closely related to human intuitive physics. Inspired by this, we designed a novel reasoning module called the Slot-based Time-Space Transformer with Memory buffer (STATM) to enhance the model's perception ability in complex scenes. The memory buffer primarily serves as storage for slot information from upstream modules, the Slot-based Time-Space Transformer makes predictions through slot-based spatiotemporal attention computations and fusion. Our experimental results on various datasets indicate that the STATM module can significantly enhance the capabilities of multiple state-of-the-art object-centric learning models for video. Moreover, as a predictive model, the STATM module also performs well in downstream prediction and Visual Question Answering (VQA) tasks. We will release our codes and data at https://github.com/intell-sci-comput/STATM.
Authors: Shuyu Chang, Rui Wang, Peng Ren, Qi Wang, Haiping Huang
Abstract: Modeling topics effectively in short texts, such as tweets and news snippets, is crucial to capturing rapidly evolving social trends. Existing topic models often struggle to accurately capture the underlying semantic patterns of short texts, primarily due to the sparse nature of such data. This nature of texts leads to an unavoidable lack of co-occurrence information, which hinders the coherence and granularity of mined topics. This paper introduces a novel model-agnostic mechanism, termed Topic Refinement, which leverages the advanced text comprehension capabilities of Large Language Models (LLMs) for short-text topic modeling. Unlike traditional methods, this post-processing mechanism enhances the quality of topics extracted by various topic modeling methods through prompt engineering. We guide LLMs in identifying semantically intruder words within the extracted topics and suggesting coherent alternatives to replace these words. This process mimics human-like identification, evaluation, and refinement of the extracted topics. Extensive experiments on four diverse datasets demonstrate that Topic Refinement boosts the topic quality and improves the performance in topic-related text classification tasks.
Authors: Shahzeb Naeem, Muhammad Riyyan Khan, Usman Tariq, Abhinav Dhall, Carlos Ivan Colon, Hasan Al-Nashash
Abstract: This research explores the positive application of deepfake technology for upper body generation, specifically sign language for the Deaf and Hard of Hearing (DHoH) community. Given the complexity of sign language and the scarcity of experts, the generated videos are vetted by a sign language expert for accuracy. We construct a reliable deepfake dataset, evaluating its technical and visual credibility using computer vision and natural language processing models. The dataset, consisting of over 1200 videos featuring both seen and unseen individuals, is also used to detect deepfake videos targeting vulnerable individuals. Expert annotations confirm that the generated videos are comparable to real sign language content. Linguistic analysis, using textual similarity scores and interpreter evaluations, shows that the interpretation of generated videos is at least 90% similar to authentic sign language. Visual analysis demonstrates that convincingly realistic deepfakes can be produced, even for new subjects. Using a pose/style transfer model, we pay close attention to detail, ensuring hand movements are accurate and align with the driving video. We also apply machine learning algorithms to establish a baseline for deepfake detection on this dataset, contributing to the detection of fraudulent sign language videos.
Authors: Yizhou Chi, Kevin Yang, Dan Klein
Abstract: We present THOUGHTSCULPT, a general reasoning and search method for tasks with outputs that can be decomposed into components. THOUGHTSCULPT explores a search tree of potential solutions using Monte Carlo Tree Search (MCTS), building solutions one action at a time and evaluating according to any domain-specific heuristic, which in practice is often simply an LLM evaluator. Critically, our action space includes revision actions: THOUGHTSCULPT may choose to revise part of its previous output rather than continuing to build the rest of its output. Empirically, THOUGHTSCULPT outperforms state-of-the-art reasoning methods across three challenging tasks: Story Outline Improvement (up to +30% interestingness), Mini-Crosswords Solving (up to +16% word success rate), and Constrained Generation (up to +10% concept coverage).
Authors: Zukang Yang, Zixuan Zhu, Xuan Zhu
Abstract: Large Language Models (LLMs) have achieved significant success in open-domain question answering. However, they continue to face challenges such as hallucinations and knowledge cutoffs. These issues can be mitigated through in-context learning by providing LLMs with relevant context before generating answers. Recent literature proposes Knowledge Graph Prompting (KGP) which integrates knowledge graphs with an LLM-based traversal agent to substantially enhance document retrieval quality. However, KGP requires costly fine-tuning with large datasets and remains prone to hallucination. In this paper, we propose CuriousLLM, an enhancement that integrates a curiosity-driven reasoning mechanism into an LLM agent. This mechanism enables the agent to generate relevant follow-up questions, thereby guiding the information retrieval process more efficiently. Central to our approach is the development of the new Follow-upQA dataset, which includes questions and supporting evidence as input, with follow-up questions serving as ground truths. These follow-up questions either inquire about what is still missing to fully answer the user's query or use special tokens to signify that the retrieved evidence is sufficient. Our experiments show that CuriousLLM significantly boosts LLM performance in multi-document question answering (MD-QA), circumventing the substantial computational costs and latency from the original KGP framework.
Authors: Qiao Deng, Zhongzhen Huang, Yunqi Wang, Zhichuan Wang, Zhao Wang, Xiaofan Zhang, Qi Dou, Yeung Yu Hui, Edward S. Hui
Abstract: Medical foundation models have the potential to revolutionize healthcare by providing robust and generalized representations of medical data. Medical vision-language pre-training has emerged as a promising approach for learning domain-general representations of medical image and text. Current algorithms that exploit global and local alignment between medical image and text could however be marred by redundant information in medical data. To address this issue, we propose a grounded knowledge-enhanced medical vision-language pre-training (GK-MVLP) framework for chest X-ray. In this framework, medical knowledge was grounded to the appropriate anatomical regions by using a transformer-based grounded knowledge-enhanced module for fine-grained alignment between textural features of medical knowledge and the corresponding anatomical region-level visual features. The performance of GK-MVLP was competitive with or exceeded the state of the art on downstream image understanding tasks (chest X-ray disease classification, disease localization), generative task (report generation), and vision-language understanding task (medical visual question-answering). Our results demonstrate the advantage of incorporating grounding mechanism to remove biases and improve the alignment between chest X-ray image and radiology report.
Authors: Wenwen Qiang, Jingyao Wang, Lingyu Si, Chuxiong Sun, Fuchun Sun, Hui Xiong
Abstract: In this paper, we investigate the characteristics that define a good representation or model. We propose that such a representation or model should possess universality, characterized by: (i) discriminability: performing well on training samples; (ii) generalization: performing well on unseen datasets; and (iii) transferability: performing well on unseen tasks with distribution shifts. Despite its importance, current self-supervised learning (SSL) methods lack explicit modeling of universality, and theoretical analysis remains underexplored. To address these issues, we aim to explore and incorporate universality into SSL. Specifically, we first revisit SSL from a task perspective and find that each mini-batch can be viewed as a multi-class classification task. We then propose that a universal SSL model should achieve: (i) learning universality by minimizing loss across all training samples, and (ii) evaluation universality by learning causally invariant representations that generalize well to unseen tasks. To quantify this, we introduce a $\sigma$-measurement that assesses the gap between the performance of SSL model and optimal task-specific models. Furthermore, to model universality, we propose the GeSSL framework. It first learns task-specific models by minimizing SSL loss, then incorporates future updates to enhance discriminability, and finally integrates these models to learn from multiple tasks. Theoretical and empirical evidence supports the effectiveness of GeSSL.
Authors: Arkadiy Saakyan, Shreyas Kulkarni, Tuhin Chakrabarty, Smaranda Muresan
Abstract: Large Vision-Language Models (VLMs) have demonstrated strong capabilities in tasks requiring a fine-grained understanding of literal meaning in images and text, such as visual question-answering or visual entailment. However, there has been little exploration of the capabilities of these models when presented with images and captions containing figurative meaning, such as metaphors or humor. To close this gap, we propose a new task framing the figurative meaning understanding problem as an explainable visual entailment task, where the model has to predict whether the image (premise) entails a caption (hypothesis) and justify the predicted label with a textual explanation. The figurative phenomena can be present in the image, in the caption, or both. Using a human-AI collaboration approach, we build the accompanying expert-verified dataset V-FLUTE, containing 6,027 {image, caption, label, explanation} instances spanning five diverse figurative phenomena: metaphors, similes, idioms, sarcasm, and humor. Through automatic evaluation, we find that VLMs struggle to generalize from literal to figurative meaning, particularly when it is present in images. Further, we identify common types of errors in VLM reasoning (hallucination and incomplete or unsound reasoning) across classes of models via human evaluation.
Authors: Gabriel Freedman, Adam Dejl, Deniz Gorur, Xiang Yin, Antonio Rago, Francesca Toni
Abstract: The profusion of knowledge encoded in large language models (LLMs) and their ability to apply this knowledge zero-shot in a range of settings makes them promising candidates for use in decision-making. However, they are currently limited by their inability to provide outputs which can be faithfully explained and effectively contested to correct mistakes. In this paper, we attempt to reconcile these strengths and weaknesses by introducing \emph{argumentative LLMs (ArgLLMs)}, a method for augmenting LLMs with argumentative reasoning. Concretely, ArgLLMs construct argumentation frameworks, which then serve as the basis for formal reasoning in support of decision-making. The interpretable nature of these argumentation frameworks and formal reasoning means that any decision made by ArgLLMs may be explained and contested. We evaluate ArgLLMs' performance experimentally in comparison with state-of-the-art techniques, in the context of the decision-making task of claim verification. We also define novel properties to characterise contestability and assess ArgLLMs formally in terms of these properties.
Authors: Yashuai Yan, Esteve Valls Mascaro, Tobias Egle, Dongheui Lee
Abstract: Humanoid robots have the potential to mimic human motions with high visual fidelity, yet translating these motions into practical, physical execution remains a significant challenge. Existing techniques in the graphics community often prioritize visual fidelity over physics-based feasibility, posing a significant challenge for deploying bipedal systems in practical applications. This paper addresses these issues through bounded residual reinforcement learning to produce physics-based high-quality motion imitation onto legged humanoid robots that enhance motion resemblance while successfully following the reference human trajectory. Our framework, Imitation to Control Humanoid Robots Through Bounded Residual Reinforcement Learning (I-CTRL), reformulates motion imitation as a constrained refinement over non-physics-based retargeted motions. I-CTRL excels in motion imitation with simple and unique rewards that generalize across five robots. Moreover, our framework introduces an automatic priority scheduler to manage large-scale motion datasets when efficiently training a unified RL policy across diverse motions. The proposed approach signifies a crucial step forward in advancing the control of bipedal robots, emphasizing the importance of aligning visual and physical realism for successful motion imitation.
Authors: Chengkun Cai, Xu Zhao, Yucheng Du, Haoliang Liu, Lei Li
Abstract: Large Language Models (LLMs) have emerged as powerful tools in artificial intelligence, especially in complex decision-making scenarios, but their static problem-solving strategies often limit their adaptability to dynamic environments. We explore the enhancement of reasoning capabilities in LLMs through Temperature Tree ($T^2$) prompting via a heuristic algorithm, termed as $T^2$ of Thoughts ($T^2oT$). The primary focus is on enhancing decision-making processes by dynamically adjusting search parameters, especially temperature, to improve accuracy without increasing computational demands. We empirically validate that our hybrid $T^2oT$ approach yields enhancements in, single-solution accuracy, multi-solution generation and text generation quality. Our findings suggest that while dynamic search depth adjustments based on temperature can yield mixed results, a fixed search depth, when coupled with adaptive capabilities of $T^2oT$, provides a more reliable and versatile problem-solving strategy. This work highlights the potential for future explorations in optimizing algorithmic interactions with foundational language models, particularly illustrated by our development for the Game of 24 and Creative Writing tasks.
Authors: Xingchen Zou, Jiani Huang, Xixuan Hao, Yuhao Yang, Haomin Wen, Yibo Yan, Chao Huang, Chao Chen, Yuxuan Liang
Abstract: Regional socioeconomic indicators are critical across various domains, yet their acquisition can be costly. Inferring global socioeconomic indicators from a limited number of regional samples is essential for enhancing management and sustainability in urban areas and human settlements. Current inference methods typically rely on spatial interpolation based on the assumption of spatial continuity, which does not adequately address the complex variations present within regional spaces. In this paper, we present GeoHG, the first space-aware socioeconomic indicator inference method that utilizes a heterogeneous graph-based structure to represent geospace for non-continuous inference. Extensive experiments demonstrate the effectiveness of GeoHG in comparison to existing methods, achieving an $R^2$ score exceeding 0.8 under extreme data scarcity with a masked ratio of 95\%.
Authors: Shashata Sawmya, Linghao Kong, Ilia Markov, Dan Alistarh, Nir Shavit
Abstract: Disentangling polysemantic neurons is at the core of many current approaches to interpretability of large language models. Here we attempt to study how disentanglement can be used to understand performance, particularly under weight sparsity, a leading post-training optimization technique. We suggest a novel measure for estimating neuronal entanglement: the Wasserstein distance of a neuron's output distribution to a Gaussian. Moreover, we show the existence of a small number of highly entangled "Wasserstein Neurons" in each linear layer of an LLM, characterized by their highly non-Gaussian output distributions, their role in mapping similar inputs to dissimilar outputs, and their significant impact on model accuracy. To study these phenomena, we propose a new experimental framework for disentangling polysemantic neurons. Our framework separates each layer's inputs to create a mixture of experts where each neuron's output is computed by a mixture of neurons of lower Wasserstein distance, each better at maintaining accuracy when sparsified without retraining. We provide strong evidence that this is because the mixture of sparse experts is effectively disentangling the input-output relationship of individual neurons, in particular the difficult Wasserstein neurons.
Authors: Cristian Rodriguez-Opazo, Ehsan Abbasnejad, Damien Teney, Hamed Damirchi, Edison Marrese-Taylor, Anton van den Hengel
Abstract: Contrastive Language-Image Pretraining (CLIP) stands out as a prominent method for image representation learning. Various architectures, from vision transformers (ViTs) to convolutional networks (ResNets) have been trained with CLIP to serve as general solutions to diverse vision tasks. This paper explores the differences across various CLIP-trained vision backbones. Despite using the same data and training objective, we find that these architectures have notably different representations, different classification performance across datasets, and different robustness properties to certain types of image perturbations. Our findings indicate a remarkable possible synergy across backbones by leveraging their respective strengths. In principle, classification accuracy could be improved by over 40 percentage with an informed selection of the optimal backbone per test example.Using this insight, we develop a straightforward yet powerful approach to adaptively ensemble multiple backbones. The approach uses as few as one labeled example per class to tune the adaptive combination of backbones. On a large collection of datasets, the method achieves a remarkable increase in accuracy of up to 39.1% over the best single backbone, well beyond traditional ensembles
Authors: Alissa A. Valentine, Lauren A. Lepow, Lili Chan, Alexander W. Charney, Isotta Landi
Abstract: Negative patient descriptions and stigmatizing language can contribute to generating healthcare disparities in two ways: (1) read by patients, they can harm their trust and engagement with the medical center; (2) read by physicians, they may negatively influence their perspective of a future patient. In psychiatry, the patient-clinician therapeutic alliance is a major determinant of clinical outcomes. Therefore, language usage in psychiatric clinical notes may not only create healthcare disparities, but also perpetuate them. Recent advances in NLP systems have facilitated the efforts to detect discriminatory language in healthcare. However, such attempts have only focused on the perspectives of the medical center and its physicians. Considering both physicians and non-physicians' point of view is a more translatable approach to identifying potentially harmful language in clinical notes. By leveraging pre-trained and large language models (PLMs and LLMs), this work aims to characterize potentially harmful language usage in psychiatric notes by identifying the sentiment expressed in sentences describing patients based on the reader's point of view. Extracting 39 sentences from the Mount Sinai Health System containing psychiatric lexicon, we fine-tuned three PLMs (RoBERTa, GatorTron, and GatorTron + Task Adaptation) and implemented zero-shot and few-shot ICL approaches for three LLMs (GPT-3.5, Llama-3.1, and Mistral) to classify the sentiment of the sentences according to the physician or non-physician point of view. Results showed that GPT-3.5 aligned best to physician point of view and Mistral aligned best to non-physician point of view. These results underline the importance of recognizing the reader's point of view, not only for improving the note writing process, but also for the quantification, identification, and reduction of bias in computational systems for downstream analyses.
Authors: Ziyang Wang, Jianzhou You, Haining Wang, Tianwei Yuan, Shichao Lv, Yang Wang, Limin Sun
Abstract: Honeypots, as a strategic cyber-deception mechanism designed to emulate authentic interactions and bait unauthorized entities, often struggle with balancing flexibility, interaction depth, and deception. They typically fail to adapt to evolving attacker tactics, with limited engagement and information gathering. Fortunately, the emergent capabilities of large language models and innovative prompt-based engineering offer a transformative shift in honeypot technologies. This paper introduces HoneyGPT, a pioneering shell honeypot architecture based on ChatGPT, characterized by its cost-effectiveness and proactive engagement. In particular, we propose a structured prompt engineering framework that incorporates chain-of-thought tactics to improve long-term memory and robust security analytics, enhancing deception and engagement. Our evaluation of HoneyGPT comprises a baseline comparison based on a collected dataset and a three-month field evaluation. The baseline comparison demonstrates HoneyGPT's remarkable ability to strike a balance among flexibility, interaction depth, and deceptive capability. The field evaluation further validates HoneyGPT's superior performance in engaging attackers more deeply and capturing a wider array of novel attack vectors.
Authors: Sungho Jeon, Xinyue Ma, Kwang In Kim, Myeongjae Jeon
Abstract: Recent rehearsal-free methods, guided by prompts, excel in vision-related continual learning (CL) with drifting data but lack resource efficiency, making real-world deployment challenging. In this paper, we introduce Resource-Efficient Prompting (REP), which improves the computational and memory efficiency of prompt-based rehearsal-free methods while minimizing accuracy trade-offs. Our approach employs swift prompt selection to refine input data using a carefully provisioned model and introduces adaptive token merging (AToM) and layer dropping (ALD) for efficient prompt updates. AToM and ALD selectively skip data and model layers while preserving task-specific features during new-task learning. Extensive experiments on multiple image classification datasets demonstrates REP's superior resource efficiency over state-of-the-art ViT- and CNN-based methods.
Authors: Zahraa Al Sahili, Ioannis Patras, Matthew Purver
Abstract: In the domain of text-to-image generative models, biases inherent in training datasets often propagate into generated content, posing significant ethical challenges, particularly in socially sensitive contexts. We introduce FairCoT, a novel framework that enhances fairness in text to image models through Chain of Thought (CoT) reasoning within multimodal generative large language models. FairCoT employs iterative CoT refinement to systematically mitigate biases, and dynamically adjusts textual prompts in real time, ensuring diverse and equitable representation in generated images. By integrating iterative reasoning processes, FairCoT addresses the limitations of zero shot CoT in sensitive scenarios, balancing creativity with ethical responsibility. Experimental evaluations across popular text-to-image systems including DALLE and various Stable Diffusion variants, demonstrate that FairCoT significantly enhances fairness and diversity without sacrificing image quality or semantic fidelity. By combining robust reasoning, lightweight deployment, and extensibility to multiple models, FairCoT represents a promising step toward more socially responsible and transparent AI driven content generation.
Authors: Samar Khanna, Medhanie Irgau, David B. Lobell, Stefano Ermon
Abstract: Parameter-efficient fine-tuning (PEFT) techniques such as low-rank adaptation (LoRA) can effectively adapt large pre-trained foundation models to downstream tasks using only a small fraction (0.1%-10%) of the original trainable weights. An under-explored question of PEFT is in extending the pre-training phase without supervised labels; that is, can we adapt a pre-trained foundation model to a new domain via efficient self-supervised pre-training on this new domain? In this work, we introduce ExPLoRA, a highly effective technique to improve transfer learning of pre-trained vision transformers (ViTs) under domain shifts. Initializing a ViT with pre-trained weights on large, natural-image datasets such as from DinoV2 or MAE, ExPLoRA continues the unsupervised pre-training objective on a new domain, unfreezing 1-2 pre-trained ViT blocks and tuning all other layers with LoRA. We then fine-tune the resulting model only with LoRA on this new domain for supervised learning. Our experiments demonstrate state-of-the-art results on satellite imagery, even outperforming fully pre-training and fine-tuning ViTs. Using the DinoV2 training objective, we demonstrate up to 8% improvement in linear probing top-1 accuracy on downstream tasks while using <10% of the number of parameters that are used in prior fully-tuned state-of-the art approaches. Our ablation studies confirm the efficacy of our approach over other baselines, including PEFT and unfreezing more ViT blocks. Code is available on the project website: https://samar-khanna.github.io/ExPLoRA/
Authors: Yanming Liu, Xinyue Peng, Yuwei Zhang, Xiaolan Ke, Songhang Deng, Jiannan Cao, Chen Ma, Mengchen Fu, Sheng Cheng, Xun Wang, Jianwei Yin, Tianyu Du, Xuhong Zhang
Abstract: Large language models have repeatedly shown outstanding performance across diverse applications. However, deploying these models can inadvertently risk user privacy. The significant memory demands during training pose a major challenge in terms of resource consumption. This substantial size places a heavy load on memory resources, raising considerable practical concerns. In this paper, we introduce DP-MemArc, a novel training framework aimed at reducing the memory costs of large language models while emphasizing the protection of user data privacy. DP-MemArc incorporates side network or reversible network designs to support a variety of differential privacy memory-efficient fine-tuning schemes. Our approach not only achieves in memory optimization but also ensures robust privacy protection, keeping user data secure and confidential. Extensive experiments have demonstrated that DP-MemArc effectively provides differential privacy-efficient fine-tuning across different task scenarios.
Authors: Keon Lee, Dong Won Kim, Jaehyeon Kim, Seungjun Chung, Jaewoong Cho
Abstract: Large-scale latent diffusion models (LDMs) excel in content generation across various modalities, but their reliance on phonemes and durations in text-to-speech (TTS) limits scalability and access from other fields. While recent studies show potential in removing these domain-specific factors, performance remains suboptimal. In this work, we introduce DiTTo-TTS, a Diffusion Transformer (DiT)-based TTS model, to investigate whether LDM-based TTS can achieve state-of-the-art performance without domain-specific factors. Through rigorous analysis and empirical exploration, we find that (1) DiT with minimal modifications outperforms U-Net, (2) variable-length modeling with a speech length predictor significantly improves results over fixed-length approaches, and (3) conditions like semantic alignment in speech latent representations are key to further enhancement. By scaling our training data to 82K hours and the model size to 790M parameters, we achieve superior or comparable zero-shot performance to state-of-the-art TTS models in naturalness, intelligibility, and speaker similarity, all without relying on domain-specific factors. Speech samples are available at https://ditto-tts.github.io.
Authors: Boxuan Lyu, Hidetaka Kamigaito, Kotaro Funakoshi, Manabu Okumura
Abstract: Maximum a posteriori decoding, a commonly used method for neural machine translation (NMT), aims to maximize the estimated posterior probability. However, high estimated probability does not always lead to high translation quality. Minimum Bayes Risk (MBR) decoding (\citealp{kumar2004minimum}) offers an alternative by seeking hypotheses with the highest expected utility. Inspired by Quality Estimation (QE) reranking which uses the QE model as a ranker (\citealp{fernandes-etal-2022-quality}), we propose source-based MBR (sMBR) decoding, a novel approach that utilizes quasi-sources (generated via paraphrasing or back-translation) as ``support hypotheses'' and a reference-free quality estimation metric as the utility function, marking the first work to solely use sources in MBR decoding. Experiments show that sMBR outperforms QE reranking and the standard MBR decoding. Our findings suggest that sMBR is a promising approach for NMT decoding.
Authors: Ronny Luss, Erik Miehling, Amit Dhurandhar
Abstract: The advent of black-box deep neural network classification models has sparked the need to explain their decisions. However, in the case of generative AI, such as large language models (LLMs), there is no class prediction to explain. Rather, one can ask why an LLM output a particular response to a given prompt. In this paper, we answer this question by proposing a contrastive explanation method requiring simply black-box/query access. Our explanations suggest that an LLM outputs a reply to a given prompt because if the prompt was slightly modified, the LLM would have given a different response that is either less preferable or contradicts the original response. The key insight is that contrastive explanations simply require a scoring function that has meaning to the user and not necessarily a specific real valued quantity (viz. class label). To this end, we offer a novel budgeted algorithm, our main algorithmic contribution, which intelligently creates contrasts based on such a scoring function while adhering to a query budget, necessary for longer contexts. We show the efficacy of our method on important natural language tasks such as open-text generation and chatbot conversations.
Authors: Tong Ye, Tengfei Ma, Xuhong Zhang, Hang Yu, Jianwei Yin, Wenhai Wang
Abstract: Large language models (LLMs) have shown remarkable capabilities in solving various programming tasks, such as code generation. However, their potential for code optimization, particularly in performance enhancement, remains largely unexplored. This paper investigates the capabilities of LLMs in optimizing code for minimal execution time, addressing a critical gap in current research. The recently proposed code optimization dataset constructs program optimization pairs based on iterative submissions from the same programmer for the same problem. However, this approach limits LLMs to local performance improvements, neglecting global algorithmic innovation. To overcome this limitation, we adopt a completely different perspective by reconstructing the optimization pairs into a problem-oriented approach. This allows for the integration of various ideas from multiple programmers tackling the same problem. Experimental results demonstrate that adapting LLMs to problem-oriented optimization pairs significantly enhances their optimization capabilities. Furthermore, recognizing the inherent trade-offs in code optimization, we introduce an anchor verification mechanism to mitigate the "optimization tax". Ultimately, our approach elevates both the optimization ratio and speedup to new levels.
Authors: Jiho Kim, Woosog Chay, Hyeonji Hwang, Daeun Kyung, Hyunseung Chung, Eunbyeol Cho, Yohan Jo, Edward Choi
Abstract: Recent advancements in Large Language Models (LLMs) have significantly enhanced the capabilities of conversation systems, making them applicable to various fields (e.g., education). Despite their progress, the evaluation of the systems often overlooks the complexities of real-world conversations, such as real-time interactions, multi-party dialogues, and extended contextual dependencies. To bridge this gap, we introduce DialSim, a real-time dialogue simulator. In this simulator, a conversation system is assigned the role of a character from popular TV shows, requiring it to respond to spontaneous questions using past dialogue information and to distinguish between known and unknown information. Key features of DialSim include assessing the system's ability to respond within a reasonable time limit, handling long-term multi-party dialogues, and evaluating performance under randomized questioning with LongDialQA, a novel, high-quality question-answering dataset. Our experiments using DialSim reveal the strengths and weaknesses of the latest conversation systems, offering valuable insights for future advancements in conversational AI. DialSim is available at https://dialsim.github.io/.
Authors: Yuncheng Hua, Yujin Huang, Shuo Huang, Tao Feng, Lizhen Qu, Chris Bain, Richard Bassed, Gholamreza Haffari
Abstract: This paper tackles the task of emotion-cause pair extraction in the unsupervised domain adaptation setting. The problem is challenging as the distributions of the events causing emotions in target domains are dramatically different than those in source domains, despite the distributions of emotional expressions between domains are overlapped. Inspired by causal discovery, we propose a novel deep latent model in the variational autoencoder (VAE) framework, which not only captures the underlying latent structures of data but also utilizes the easily transferable knowledge of emotions as the bridge to link the distributions of events in different domains. To facilitate knowledge transfer across domains, we also propose a novel variational posterior regularization technique to disentangle the latent representations of emotions from those of events in order to mitigate the damage caused by the spurious correlations related to the events in source domains. Through extensive experiments, we demonstrate that our model outperforms the strongest baseline by approximately 11.05\% on a Chinese benchmark and 2.45\% on a English benchmark in terms of weighted-average F1 score. We have released our source code and the generated dataset publicly at: https://github.com/tk1363704/CAREL-VAE.
Authors: Ming Li, Han Chen, Chenguang Wang, Dang Nguyen, Dianqi Li, Tianyi Zhou
Abstract: Large language models (LLMs) still lack delicate controllability over their responses, which is critical to enhancing their performance and the user experience. However, curating supervised fine-tuning (SFT) datasets to improve LLM controllability usually relies on human experts or proprietary LLMs, which requires additional costs. To bridge this gap, we propose Rule-based Data Recycling (RuleR), a data augmentation method incorporating multiple constraints into the original data samples according to predefined rules, which creates new training tasks to consolidate the controllability of LLMs. Instead of creating new data from scratch, RuleR "recycles" existing data by simply applying rule-based edits to their responses and appending the rule-instructions in their original instructions. Experimental results demonstrate RuleR's effectiveness in improving LLM controllability while maintaining general instruction-following capabilities.
Authors: Shayne Longpre, Stella Biderman, Alon Albalak, Hailey Schoelkopf, Daniel McDuff, Sayash Kapoor, Kevin Klyman, Kyle Lo, Gabriel Ilharco, Nay San, Maribeth Rauh, Aviya Skowron, Bertie Vidgen, Laura Weidinger, Arvind Narayanan, Victor Sanh, David Adelani, Percy Liang, Rishi Bommasani, Peter Henderson, Sasha Luccioni, Yacine Jernite, Luca Soldaini
Abstract: Foundation model development attracts a rapidly expanding body of contributors, scientists, and applications. To help shape responsible development practices, we introduce the Foundation Model Development Cheatsheet: a growing collection of 250+ tools and resources spanning text, vision, and speech modalities. We draw on a large body of prior work to survey resources (e.g. software, documentation, frameworks, guides, and practical tools) that support informed data selection, processing, and understanding, precise and limitation-aware artifact documentation, efficient model training, advance awareness of the environmental impact from training, careful model evaluation of capabilities, risks, and claims, as well as responsible model release, licensing and deployment practices. We hope this curated collection of resources helps guide more responsible development. The process of curating this list, enabled us to review the AI development ecosystem, revealing what tools are critically missing, misused, or over-used in existing practices. We find that (i) tools for data sourcing, model evaluation, and monitoring are critically under-serving ethical and real-world needs, (ii) evaluations for model safety, capabilities, and environmental impact all lack reproducibility and transparency, (iii) text and particularly English-centric analyses continue to dominate over multilingual and multi-modal analyses, and (iv) evaluation of systems, rather than just models, is needed so that capabilities and impact are assessed in context.
Authors: Zifan Liu, Xinran Li, Shibo Chen, Gen Li, Jiashuo Jiang, Jun Zhang
Abstract: Reinforcement learning (RL) has proven to be well-performed and general-purpose in the inventory control (IC). However, further improvement of RL algorithms in the IC domain is impeded due to two limitations of online experience. First, online experience is expensive to acquire in real-world applications. With the low sample efficiency nature of RL algorithms, it would take extensive time to train the RL policy to convergence. Second, online experience may not reflect the true demand due to the lost sales phenomenon typical in IC, which makes the learning process more challenging. To address the above challenges, we propose a decision framework that combines reinforcement learning with feedback graph (RLFG) and intrinsically motivated exploration (IME) to boost sample efficiency. In particular, we first take advantage of the inherent properties of lost-sales IC problems and design the feedback graph (FG) specially for lost-sales IC problems to generate abundant side experiences aid RL updates. Then we conduct a rigorous theoretical analysis of how the designed FG reduces the sample complexity of RL methods. Based on the theoretical insights, we design an intrinsic reward to direct the RL agent to explore to the state-action space with more side experiences, further exploiting FG's power. Experimental results demonstrate that our method greatly improves the sample efficiency of applying RL in IC. Our code is available at https://anonymous.4open.science/r/RLIMFG4IC-811D/
Authors: Yonatan Sverdlov, Nadav Dym
Abstract: Motivated by applications in chemistry and other sciences, we study the expressive power of message-passing neural networks for geometric graphs, whose node features correspond to 3-dimensional positions. Recent work has shown that such models can separate generic pairs of non-isomorphic geometric graphs, though they may fail to separate some rare and complicated instances. However, these results assume a fully connected graph, where each node possesses complete knowledge of all other nodes. In contrast, often, in application, every node only possesses knowledge of a small number of nearest neighbors. This paper shows that generic pairs of non-isomorphic geometric graphs can be separated by message-passing networks with rotation equivariant features as long as the underlying graph is connected. When only invariant intermediate features are allowed, generic separation is guaranteed for generically globally rigid graphs. We introduce a simple architecture, EGENNET, which achieves our theoretical guarantees and compares favorably with alternative architecture on synthetic and chemical benchmarks. Our code is available at https://github.com/yonatansverdlov/E-GenNet.
Authors: Yunxiang Zhang, Nan Wu, Connor Z. Lin, Gordon Wetzstein, Qi Sun
Abstract: Diffusion models offer unprecedented image generation power given just a text prompt. While emerging approaches for controlling diffusion models have enabled users to specify the desired spatial layouts of the generated content, they cannot predict or control where viewers will pay more attention due to the complexity of human vision. Recognizing the significance of attention-controllable image generation in practical applications, we present a saliency-guided framework to incorporate the data priors of human visual attention mechanisms into the generation process. Given a user-specified viewer attention distribution, our control module conditions a diffusion model to generate images that attract viewers' attention toward the desired regions. To assess the efficacy of our approach, we performed an eye-tracked user study and a large-scale model-based saliency analysis. The results evidence that both the cross-user eye gaze distributions and the saliency models' predictions align with the desired attention distributions. Lastly, we outline several applications, including interactive design of saliency guidance, attention suppression in unwanted regions, and adaptive generation for varied display/viewing conditions.
Authors: C\'ecile Decker, Julian Sester
Abstract: In this paper we propose a novel $Q$-learning algorithm allowing to solve distributionally robust Markov decision problems for which the ambiguity set of probability measures can be chosen arbitrarily as long as it comprises only a finite amount of measures. Therefore, our approach goes beyond the well-studied cases involving ambiguity sets of balls around some reference measure with the distance to reference measure being measured with respect to the Wasserstein distance or the Kullback--Leibler divergence. Hence, our approach allows the applicant to create ambiguity sets better tailored to her needs and to solve the associated robust Markov decision problem via a $Q$-learning algorithm whose convergence is guaranteed by our main result. Moreover, we showcase in several numerical experiments the tractability of our approach.
Authors: Ye Liu, Jiajun Zhu, Xukai Liu, Haoyu Tang, Yanghai Zhang, Kai Zhang, Xiaofang Zhou, Enhong Chen
Abstract: Few-Shot Fake News Detection (FS-FND) aims to distinguish inaccurate news from real ones in extremely low-resource scenarios. This task has garnered increased attention due to the widespread dissemination and harmful impact of fake news on social media. Large Language Models (LLMs) have demonstrated competitive performance with the help of their rich prior knowledge and excellent in-context learning abilities. However, existing methods face significant limitations, such as the Understanding Ambiguity and Information Scarcity, which significantly undermine the potential of LLMs. To address these shortcomings, we propose a Dual-perspective Knowledge-guided Fake News Detection (DKFND) model, designed to enhance LLMs from both inside and outside perspectives. Specifically, DKFND first identifies the knowledge concepts of each news article through a Detection Module. Subsequently, DKFND creatively designs an Investigation Module to retrieve inside and outside valuable information concerning to the current news, followed by another Judge Module to evaluate the relevance and confidence of them. Finally, a Determination Module further derives two respective predictions and obtain the final result. Extensive experiments on two public datasets show the efficacy of our proposed method, particularly in low-resource settings.
Authors: Aditi Khandelwal, Harman Singh, Hengrui Gu, Tianlong Chen, Kaixiong Zhou
Abstract: Large language models are often expected to constantly adapt to new sources of knowledge and knowledge editing techniques aim to efficiently patch the outdated model knowledge, with minimal modification. Most prior works focus on monolingual knowledge editing in English, even though new information can emerge in any language from any part of the world. We propose the Cross-Lingual Multi-Hop Knowledge Editing paradigm, for measuring and analyzing the performance of various SoTA knowledge editing techniques in a cross-lingual setup. Specifically, we create a parallel cross-lingual benchmark, CROLIN-MQUAKE for measuring the knowledge editing capabilities. Our extensive analysis over various knowledge editing techniques uncover significant gaps in performance between the cross-lingual and English-centric setting. Following this, we propose a significantly improved system for cross-lingual multi-hop knowledge editing, CLEVER-CKE. CLEVER-CKE is based on a retrieve, verify and generate knowledge editing framework, where a retriever is formulated to recall edited facts and support an LLM to adhere to knowledge edits. We develop language-aware and hard-negative based contrastive objectives for improving the cross-lingual and fine-grained fact retrieval and verification process used in this framework. Extensive experiments on three LLMs, eight languages, and two datasets show CLEVER-CKE's significant gains of up to 30% over prior methods.
Authors: Yuqi Jiang, Xudong Lu, Qian Jin, Qi Sun, Hanming Wu, Cheng Zhuo
Abstract: Intelligence is key to advancing integrated circuit (IC) fabrication. Recent breakthroughs in Large Multimodal Models (LMMs) have unlocked extraditionary abilities in understanding images and text, fostering intelligent fabrication. Leveraging the power of LMMs, we introduce FabGPT, a customized IC fabrication large multimodal model for wafer defect knowledge query. FabGPT manifests expertise in conducting defect detection in Scanning Electron Microscope (SEM) images, performing root cause analysis, and providing expert Q&A on fabrication processes. FabGPT matches enhanced multimodal features to automatically detect minute defects under complex wafer backgrounds and reduce the subjectivity of manual threshold settings. Besides, the proposed modulation module and interactive corpus training strategy embed wafer defect knowledge into the pre-trained model, effectively balancing Q&A queries related to defect knowledge and original knowledge and mitigating the modality bias issues. Experiments on in-house fab data show that FabGPT achieves significant performance improvement in wafer defect detection and knowledge querying.
Authors: Lei Ren, Haiteng Wang, Jinwang Li, Yang Tang, Chunhua Yang
Abstract: With the remarkable success of generative models like ChatGPT, Artificial Intelligence Generated Content (AIGC) is undergoing explosive development. Not limited to text and images, generative models can generate industrial time series data, addressing challenges such as the difficulty of data collection and data annotation. Due to their outstanding generation ability, they have been widely used in Internet of Things, metaverse, and cyber-physical-social systems to enhance the efficiency of industrial production. In this paper, we present a comprehensive overview of generative models for industrial time series from deep generative models (DGMs) to large generative models (LGMs). First, a DGM-based AIGC framework is proposed for industrial time series generation. Within this framework, we survey advanced industrial DGMs and present a multi-perspective categorization. Furthermore, we systematically analyze the critical technologies required to construct industrial LGMs from four aspects: large-scale industrial dataset, LGMs architecture for complex industrial characteristics, self-supervised training for industrial time series, and fine-tuning of industrial downstream tasks. Finally, we conclude the challenges and future directions to enable the development of generative models in industry.
Authors: Shaomang Huang, Jianfeng Pan, Min Peng, Hanzhong Zheng
Abstract: Large Language Models (LLMs) have achieved exceptional performance across diverse domains through training on massive datasets. However, scaling LLMs to support multiple downstream domain applications remains a significant challenge, especially under resource constraints. Existing approaches often struggle to balance performance across multiple domains with resource efficiency, limiting their broader applicability. To address this, we introduce the CCoE architecture, a modular framework that seamlessly integrates domain-specific experts into a unified LLM. By leveraging independently trained expert subnetworks on a shared backbone partition, CCoE achieves state-of-the-art performance while significantly reducing the resource requirements for multi-expert deployments. Furthermore, rule-based gating and expert planning in CCoE enable flexible task allocation, promoting expert collaboration to handle complex reasoning tasks. CCoE not only reduces inference costs but also provides a flexible and scalable solution for integrating domain expertise across diverse applications. Experiments on five domains demonstrate that CCoE achieves comparable performance to current domain-specific LLMs. Moreover, compared to existing multi-domain model ensemble methods, CCoE reduces memory usage by 61.3%, while improving inference efficiency by 0.76x over parameter-efficient multi-expert integration approaches.
Authors: Peng Xu, Wei Ping, Xianchao Wu, Chejian Xu, Zihan Liu, Mohammad Shoeybi, Bryan Catanzaro
Abstract: In this work, we introduce ChatQA 2, an Llama 3.0-based model with a 128K context window, designed to bridge the gap between open-source LLMs and leading proprietary models (e.g., GPT-4-Turbo-2024-04-09) in long context understanding and retrieval-augmented generation (RAG) capabilities. These two capabilities are complementary to each other and essential for LLMs to process large volumes of information that cannot fit into a single prompt. We present a detailed continued training recipe to extend the context window of Llama3-70B-base from 8K to 128K tokens, along with a three-stage instruction tuning process to enhance the model's instruction-following, RAG performance, and long-context understanding capabilities. Our results demonstrate that the Llama3-ChatQA-2-70B model outperforms most existing state-of-the-art models, including GPT-4-Turbo-2024-04-09, Qwen2-72B-Instruct, and Llama3.1-70B-Instruct, on ultra-long tasks beyond 100K tokens, as well as on the RAG benchmark using only a 4K context window, showing the strong long context capability across varying sequence lengths. We further provide extensive comparisons between direct long-context and RAG solutions using the same state-of-the-art long-context LLMs. Interestingly, we find that the performance of strong long-context LLMs using RAG improves when retrieving a larger number of chunks. With a large set of top-k chunks, RAG consistently outperforms direct long-context solution using the same state-of-the-art long-context models (e.g., Llama3-ChatQA-2-70B and Qwen2-72B-Instruct) on both 32K and 128K benchmarks. We open-source the model weights, training data, and the evaluation setup for the for the community: https://chatqa2-project.github.io/
Authors: Zheng Chong, Xiao Dong, Haoxiang Li, Shiyue Zhang, Wenqing Zhang, Xujie Zhang, Hanqing Zhao, Dongmei Jiang, Xiaodan Liang
Abstract: Virtual try-on methods based on diffusion models achieve realistic effects but often require additional encoding modules, a large number of training parameters, and complex preprocessing, which increases the burden on training and inference. In this work, we re-evaluate the necessity of additional modules and analyze how to improve training efficiency and reduce redundant steps in the inference process. Based on these insights, we propose CatVTON, a simple and efficient virtual try-on diffusion model that transfers in-shop or worn garments of arbitrary categories to target individuals by concatenating them along spatial dimensions as inputs of the diffusion model. The efficiency of CatVTON is reflected in three aspects: (1) Lightweight network. CatVTON consists only of a VAE and a simplified denoising UNet, removing redundant image and text encoders as well as cross-attentions, and includes just 899.06M parameters. (2) Parameter-efficient training. Through experimental analysis, we identify self-attention modules as crucial for adapting pre-trained diffusion models to the virtual try-on task, enabling high-quality results with only 49.57M training parameters. (3) Simplified inference. CatVTON eliminates unnecessary preprocessing, such as pose estimation, human parsing, and captioning, requiring only a person image and garment reference to guide the virtual try-on process, reducing over 49% memory usage compared to other diffusion-based methods. Extensive experiments demonstrate that CatVTON achieves superior qualitative and quantitative results compared to baseline methods and demonstrates strong generalization performance in in-the-wild scenarios, despite being trained solely on public datasets with 73K samples.
Authors: Shi Lin, Hongming Yang, Rongchang Li, Xun Wang, Changting Lin, Wenpeng Xing, Meng Han
Abstract: The rapid development of Large Language Models (LLMs) has brought significant advancements across various tasks. However, despite these achievements, LLMs still exhibit inherent safety vulnerabilities, especially when confronted with jailbreak attacks. Existing jailbreak methods suffer from two main limitations: reliance on complicated prompt engineering and iterative optimization, which lead to low attack success rate (ASR) and attack efficiency (AE). In this work, we propose an efficient jailbreak attack method, Analyzing-based Jailbreak (ABJ), which leverages the advanced reasoning capability of LLMs to autonomously generate harmful content, revealing their underlying safety vulnerabilities during complex reasoning process. We conduct comprehensive experiments on ABJ across various open-source and closed-source LLMs. In particular, ABJ achieves high ASR (82.1% on GPT-4o-2024-11-20) with exceptional AE among all target LLMs, showcasing its remarkable attack effectiveness, transferability, and efficiency. Our findings underscore the urgent need to prioritize and improve the safety of LLMs to mitigate the risks of misuse.
Authors: Yuqi Zhou, Joseph Severino, Sanjana Vijayshankar, Juliette Ugirumurera, Jibo Sanyal
Abstract: Timely and effective load shedding in power systems is critical for maintaining supply-demand balance and preventing cascading blackouts. To eliminate load shedding bias against specific regions in the system, optimization-based methods are uniquely positioned to help balance between economical and equity considerations. However, the resulting optimization problem involves complex constraints, which can be time-consuming to solve and thus cannot meet the real-time requirements of load shedding. To tackle this challenge, in this paper we present an efficient machine learning algorithm to enable millisecond-level computation for the optimization-based load shedding problem. Numerical studies on both a 3-bus toy example and a realistic RTS-GMLC system have demonstrated the validity and efficiency of the proposed algorithm for delivering equitable and real-time load shedding decisions.
Authors: Luca Mouchel, Debjit Paul, Shaobo Cui, Robert West, Antoine Bosselut, Boi Faltings
Abstract: Despite the remarkable performance of Large Language Models (LLMs) in natural language processing tasks, they still struggle with generating logically sound arguments, resulting in potential risks such as spreading misinformation. To address this issue, we introduce FIPO, a fallacy-informed framework that leverages preference optimization methods to steer LLMs toward logically sound arguments. FIPO includes a classification loss, to capture the fine-grained information on fallacy types. Our results on argumentation datasets show that our method reduces the fallacy errors by up to 17.5%. Furthermore, our human evaluation results indicate that the quality of the generated arguments by our method significantly outperforms the fine-tuned baselines, as well as other preference optimization methods, such as DPO. These findings highlight the importance of ensuring models are aware of logical fallacies for effective argument generation. Our code is available at github.com/lucamouchel/Logical-Fallacies.
Authors: Jonas Oppenlaender, Hannah Johnston, Johanna Silvennoinen, Helena Barranha
Abstract: Image generation using generative artificial intelligence has become a popular activity. However, text-to-image generation - where images are produced from typed prompts - can be less engaging in public settings since the act of typing tends to limit interactive audience participation, thereby reducing its suitability for designing dynamic public installations. In this article, we explore body prompting as input modality for image generation in the context of installations at public event settings. Body prompting extends interaction with generative AI beyond textual inputs to reconnect the creative act of image generation with the physical act of creating artworks. We implement this concept in an interactive art installation, Artworks Reimagined, designed to transform existing artworks via body prompting. We deployed the installation at an event with hundreds of visitors in a public and private setting. Our semi-structured interviews with a sample of visitors (N = 79) show that body prompting was well-received and provides an engaging and fun experience to the installation's visitors. We present insights into participants' experience of body prompting and AI co-creation and identify three distinct strategies of embodied interaction focused on re-creating, reimagining, or casual interaction. We provide valuable recommendations for practitioners seeking to design interactive generative AI experiences in museums, galleries, and public event spaces.
Authors: Geonhee Kim, Marco Valentino, Andr\'e Freitas
Abstract: Recent studies on logical reasoning in Language Models (LMs) have sparked a debate on whether they can learn systematic reasoning principles during pre-training or merely exploit superficial patterns in the training data. This paper presents a mechanistic interpretation of syllogistic reasoning in LMs to advance the understanding of internal dynamics. Specifically, we present a methodology for circuit discovery aimed at interpreting content-independent reasoning mechanisms. Through two distinct intervention methods, we uncover a sufficient and necessary circuit involving middle-term suppression that elucidates how LMs transfer information to derive valid conclusions from premises. Furthermore, we investigate how belief biases manifest in syllogistic reasoning, finding evidence of partial contamination from additional attention heads responsible for encoding commonsense and contextualized knowledge. Finally, we explore the generalization of the discovered mechanisms across various syllogistic schemes, model sizes and architectures, finding that the identified circuit is sufficient and necessary for the schemes on which the models achieve high downstream accuracy (> 60%), and that the activation patterns apply to models of different families. Overall, our findings suggest that LMs indeed learn transferable content-independent reasoning mechanisms, but that, at the same time, such mechanisms do not involve generalizable and abstract logical primitives, being susceptible to contamination by the same world knowledge acquired during pre-training.
Authors: Niloofar Azizi, Nils Kriege, Horst Bischof
Abstract: Graph Neural Networks (GNNs) excel in handling graph-structured data but often underperform in link prediction tasks compared to classical methods, mainly due to the limitations of the commonly used message-passing principle. Notably, their ability to distinguish non-isomorphic graphs is limited by the 1-dimensional Weisfeiler-Lehman test. Our study presents a novel method to enhance the expressivity of GNNs by embedding induced subgraphs into the graph Laplacian matrix's eigenbasis. We introduce a Learnable Lanczos algorithm with Linear Constraints (LLwLC), proposing two novel subgraph extraction strategies: encoding vertex-deleted subgraphs and applying Neumann eigenvalue constraints. For the former, we demonstrate the ability to distinguish graphs that are indistinguishable by 2-WL, while maintaining efficient time complexity. The latter focuses on link representations enabling differentiation between $k$-regular graphs and node automorphism, a vital aspect for link prediction tasks. Our approach results in an extremely lightweight architecture, reducing the need for extensive training datasets. Empirically, our method improves performance in challenging link prediction tasks across benchmark datasets, establishing its practical utility and supporting our theoretical findings. Notably, LLwLC achieves 20x and 10x speedup by only requiring 5% and 10% data from the PubMed and OGBL-Vessel datasets while comparing to the state-of-the-art.
Authors: Yuxia Wu, Shujie Li, Yuan Fang, Chuan Shi
Abstract: Large language models (LLMs) have presented significant opportunities to enhance various machine learning applications, including graph neural networks (GNNs). By leveraging the vast open-world knowledge within LLMs, we can more effectively interpret and utilize textual data to better characterize heterophilic graphs, where neighboring nodes often have different labels. However, existing approaches for heterophilic graphs overlook the rich textual data associated with nodes, which could unlock deeper insights into their heterophilic contexts. In this work, we explore the potential of LLMs for modeling heterophilic graphs and propose a novel two-stage framework: LLM-enhanced edge discriminator and LLM-guided edge reweighting. In the first stage, we fine-tune the LLM to better identify homophilic and heterophilic edges based on the textual content of their nodes. In the second stage, we adaptively manage message propagation in GNNs for different edge types based on node features, structures, and heterophilic or homophilic characteristics. To cope with the computational demands when deploying LLMs in practical scenarios, we further explore model distillation techniques to fine-tune smaller, more efficient models that maintain competitive performance. Extensive experiments validate the effectiveness of our framework, demonstrating the feasibility of using LLMs to enhance node classification on heterophilic graphs.
Authors: Yuhao Du, Zhuo Li, Pengyu Cheng, Xiang Wan, Anningzhe Gao
Abstract: Despite the substantial advancements in artificial intelligence, large language models (LLMs) remain being challenged by generation safety. With adversarial jailbreaking prompts, one can effortlessly induce LLMs to output harmful content, causing unexpected negative social impacts. This vulnerability highlights the necessity for robust LLM red-teaming strategies to identify and mitigate such risks before large-scale application. To detect specific types of risks, we propose a novel red-teaming method that $\textbf{A}$ttacks LLMs with $\textbf{T}$arget $\textbf{Toxi}$c $\textbf{A}$nswers ($\textbf{Atoxia}$). Given a particular harmful answer, Atoxia generates a corresponding user query and a misleading answer opening to examine the internal defects of a given LLM. The proposed attacker is trained within a reinforcement learning scheme with the LLM outputting probability of the target answer as the reward. We verify the effectiveness of our method on various red-teaming benchmarks, such as AdvBench and HH-Harmless. The empirical results demonstrate that Atoxia can successfully detect safety risks in not only open-source models but also state-of-the-art black-box models such as GPT-4o.
Authors: Wenxuan Wang, Juluan Shi, Zixuan Ling, Yuk-Kit Chan, Chaozheng Wang, Cheryl Lee, Youliang Yuan, Jen-tse Huang, Wenxiang Jiao, Michael R. Lyu
Abstract: Equipped with the capability to call functions, modern large language models (LLMs) can leverage external tools for addressing a range of tasks unattainable through language skills alone. However, the effective execution of these tools relies heavily not just on the advanced capabilities of LLMs but also on precise user instructions, which often cannot be ensured in the real world. To evaluate the performance of LLMs tool-use under imperfect instructions, we meticulously examine the real-world instructions queried from users, analyze the error patterns, and build a challenging tool-use benchmark called Noisy ToolBench (NoisyToolBench). We find that due to the next-token prediction training objective, LLMs tend to arbitrarily generate the missed argument, which may lead to hallucinations and risks. To address this issue, we propose a novel framework, Ask-when-Needed (AwN), which prompts LLMs to ask questions to users whenever they encounter obstacles due to unclear instructions. Moreover, to reduce the manual labor involved in user-LLM interaction and assess LLMs performance in tool utilization from both accuracy and efficiency perspectives, we design an automated evaluation tool named ToolEvaluator. Our experiments demonstrate that the AwN significantly outperforms existing frameworks for tool learning in the NoisyToolBench. We will release all related code and datasets to support future research.
Authors: Chanjun Park, Hyeonwoo Kim
Abstract: This paper conducts a longitudinal study over eleven months to address the limitations of prior research on the Open Ko-LLM Leaderboard, which have relied on empirical studies with restricted observation periods of only five months. By extending the analysis duration, we aim to provide a more comprehensive understanding of the progression in developing Korean large language models (LLMs). Our study is guided by three primary research questions: (1) What are the specific challenges in improving LLM performance across diverse tasks on the Open Ko-LLM Leaderboard over time? (2) How does model size impact task performance correlations across various benchmarks? (3) How have the patterns in leaderboard rankings shifted over time on the Open Ko-LLM Leaderboard?. By analyzing 1,769 models over this period, our research offers a comprehensive examination of the ongoing advancements in LLMs and the evolving nature of evaluation frameworks.
Authors: Wenyu Zhang, Shuo Sun, Bin Wang, Xunlong Zou, Zhuohan Liu, Yingxu He, Geyu Lin, Nancy F. Chen, Ai Ti Aw
Abstract: The rapid advancements in large language models (LLMs) have significantly enhanced natural language processing capabilities, facilitating the development of AudioLLMs that process and understand speech and audio inputs alongside text. Existing AudioLLMs typically combine a pre-trained audio encoder with a pre-trained LLM, which are subsequently finetuned on specific audio tasks. However, the pre-trained audio encoder has constrained capacity to capture features for new tasks and datasets. To address this, we propose to incorporate mixtures of `weak' encoders (MoWE) into the AudioLLM framework. MoWE supplements a base encoder with a pool of relatively light weight encoders, selectively activated based on the audio input to enhance feature extraction without significantly increasing model size. Our empirical results demonstrate that MoWE effectively improves multi-task performance, broadening the applicability of AudioLLMs to more diverse audio tasks.
Authors: Sichun Wu, Kazi Injamamul Haque, Zerrin Yumak
Abstract: Audio-driven 3D facial animation synthesis has been an active field of research with attention from both academia and industry. While there are promising results in this area, recent approaches largely focus on lip-sync and identity control, neglecting the role of emotions and emotion control in the generative process. That is mainly due to the lack of emotionally rich facial animation data and algorithms that can synthesize speech animations with emotional expressions at the same time. In addition, majority of the models are deterministic, meaning given the same audio input, they produce the same output motion. We argue that emotions and non-determinism are crucial to generate diverse and emotionally-rich facial animations. In this paper, we propose ProbTalk3D a non-deterministic neural network approach for emotion controllable speech-driven 3D facial animation synthesis using a two-stage VQ-VAE model and an emotionally rich facial animation dataset 3DMEAD. We provide an extensive comparative analysis of our model against the recent 3D facial animation synthesis approaches, by evaluating the results objectively, qualitatively, and with a perceptual user study. We highlight several objective metrics that are more suitable for evaluating stochastic outputs and use both in-the-wild and ground truth data for subjective evaluation. To our knowledge, that is the first non-deterministic 3D facial animation synthesis method incorporating a rich emotion dataset and emotion control with emotion labels and intensity levels. Our evaluation demonstrates that the proposed model achieves superior performance compared to state-of-the-art emotion-controlled, deterministic and non-deterministic models. We recommend watching the supplementary video for quality judgement. The entire codebase is publicly available (https://github.com/uuembodiedsocialai/ProbTalk3D/).
Authors: Hyunjong Ok, Jaeho Lee
Abstract: Singing voices contain much richer information than common voices, such as diverse vocal and acoustic characteristics. However, existing open-source audio-text datasets for singing voices capture only a limited set of attributes and lacks acoustic features, leading to limited utility towards downstream tasks, such as style captioning. To fill this gap, we formally consider the task of singing style captioning and introduce S2Cap, a singing voice dataset with comprehensive descriptions of diverse vocal, acoustic and demographic attributes. Based on this dataset, we develop a simple yet effective baseline algorithm for the singing style captioning. The algorithm utilizes two novel technical components: CRESCENDO for mitigating misalignment between pretrained unimodal models, and demixing supervision to regularize the model to focus on the singing voice. Despite its simplicity, the proposed method outperforms state-of-the-art baselines.
Authors: Theodoros Sotirou, Vassilis Lyberatos, Orfeas Menis Mastromichalakis, Giorgos Stamou
Abstract: Multimodal models are critical for music understanding tasks, as they capture the complex interplay between audio and lyrics. However, as these models become more prevalent, the need for explainability grows-understanding how these systems make decisions is vital for ensuring fairness, reducing bias, and fostering trust. In this paper, we introduce MusicLIME, a model-agnostic feature importance explanation method designed for multimodal music models. Unlike traditional unimodal methods, which analyze each modality separately without considering the interaction between them, often leading to incomplete or misleading explanations, MusicLIME reveals how audio and lyrical features interact and contribute to predictions, providing a holistic view of the model's decision-making. Additionally, we enhance local explanations by aggregating them into global explanations, giving users a broader perspective of model behavior. Through this work, we contribute to improving the interpretability of multimodal music models, empowering users to make informed choices, and fostering more equitable, fair, and transparent music understanding systems.
Authors: David Herel, Vojtech Bartek, Jiri Jirak, Tomas Mikolov
Abstract: Who is the US President? The answer changes depending on when the question is asked. While large language models (LLMs) are evaluated on various reasoning tasks, they often miss a crucial dimension: time. In real-world scenarios, the correctness of answers is frequently tied to temporal context. To address this gap, we present a novel framework and dataset spanning over 8,000 events from 2018 to 2024, annotated with day-level granularity and sourced globally across domains such as politics, science, and business. Our TimeShift evaluation method systematically probes LLMs for temporal reasoning, revealing that base models often outperform instruction-tuned and synthetic-trained counterparts on time-sensitive recall. Additionally, we find that even large-scale models exhibit brittleness in handling paraphrased facts, highlighting unresolved challenges in temporal consistency. By identifying these limitations, our work provides a significant step toward advancing time-aware language models capable of adapting to the dynamic nature of real-world knowledge.
Authors: Shuo Yu (State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China, Hefei, China), Mingyue Cheng (State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China, Hefei, China), Jiqian Yang (State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China, Hefei, China), Jie Ouyang (State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China, Hefei, China), Yucong Luo (State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China, Hefei, China), Chenyi Lei (Kuaishou Technology, Beijing, China), Qi Liu (State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China, Hefei, China), Enhong Chen (State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China, Hefei, China)
Abstract: Retrieval-augmented generation (RAG) is increasingly recognized as an effective approach to mitigating the hallucination of large language models (LLMs) through the integration of external knowledge. While numerous efforts, most studies focus on a single type of external knowledge source. In contrast, most real-world applications involve diverse knowledge from various sources, a scenario that has been relatively underexplored. The main dilemma is the lack of a suitable dataset incorporating multiple knowledge sources and pre-exploration of the associated issues. To address these challenges, we standardize a benchmark dataset that combines structured and unstructured knowledge across diverse and complementary domains. Building upon the dataset, we identify the limitations of existing methods under such conditions. Therefore, we develop PruningRAG, a plug-and-play RAG framework that uses multi-granularity pruning strategies to more effectively incorporate relevant context and mitigate the negative impact of misleading information. Extensive experimental results demonstrate superior performance of PruningRAG and our insightful findings are also reported. Our dataset and code are publicly available\footnote{https://github.com/USTCAGI/PruningRAG}.
Authors: Shilong Yang, Chulong Zhang, Qi Zang, Juan Yu, Liang Zeng, Xiao Luo, Yexuan Xing, Xin Pan, Qi Li, Xiaokun Liang, Yaoqin Xie
Abstract: Breast cancer is a significant global health issue, and the diagnosis of breast imaging has always been challenging. Mammography images typically have extremely high resolution, with lesions occupying only a very small area. Down-sampling in neural networks can easily lead to the loss of microcalcifications or subtle structures, making it difficult for traditional neural network architectures to address these issues. To tackle these challenges, we propose a Context Clustering Network with triple information fusion. Firstly, compared to CNNs or transformers, we find that Context clustering methods (1) are more computationally efficient and (2) can more easily associate structural or pathological features, making them suitable for the clinical tasks of mammography. Secondly, we propose a triple information fusion mechanism that integrates global information, feature-based local information, and patch-based local information. The proposed approach is rigorously evaluated on two public datasets, Vindr-Mammo and CBIS-DDSM, using five independent splits to ensure statistical robustness. Our method achieves an AUC of 0.828 on Vindr-Mammo and 0.805 on CBIS-DDSM, outperforming the next best method by 3.1% and 2.4%, respectively. These improvements are statistically significant (p<0.05), underscoring the benefits of Context Clustering Network with triple information fusion. Overall, our Context Clustering framework demonstrates strong potential as a scalable and cost-effective solution for large-scale mammography screening, enabling more efficient and accurate breast cancer detection. Access to our method is available at https://github.com/Sohyu1/Mammo_Clustering.
Authors: Xin Hong, Yuan Gong, Vidhyasaharan Sethu, Ting Dang
Abstract: Recent advancements in Large Language Models (LLMs) have demonstrated great success in many Natural Language Processing (NLP) tasks. In addition to their cognitive intelligence, exploring their capabilities in emotional intelligence is also crucial, as it enables more natural and empathetic conversational AI. Recent studies have shown LLMs' capability in recognizing emotions, but they often focus on single emotion labels and overlook the complex and ambiguous nature of human emotions. This study is the first to address this gap by exploring the potential of LLMs in recognizing ambiguous emotions, leveraging their strong generalization capabilities and in-context learning. We design zero-shot and few-shot prompting and incorporate past dialogue as context information for ambiguous emotion recognition. Experiments conducted using three datasets indicate significant potential for LLMs in recognizing ambiguous emotions, and highlight the substantial benefits of including context information. Furthermore, our findings indicate that LLMs demonstrate a high degree of effectiveness in recognizing less ambiguous emotions and exhibit potential for identifying more ambiguous emotions, paralleling human perceptual capabilities.
Authors: Haobo Li, Zhaowei Wang, Jiachen Wang, Yueya Wang, Alexis Kai Hon Lau, Huamin Qu
Abstract: Forecasting weather and climate events is crucial for making appropriate measures to mitigate environmental hazards and minimize losses. However, existing environmental forecasting research focuses narrowly on predicting numerical meteorological variables (e.g., temperature), neglecting the translation of these variables into actionable textual narratives of events and their consequences. To bridge this gap, we proposed Weather and Climate Event Forecasting (WCEF), a new task that leverages numerical meteorological raster data and textual event data to predict weather and climate events. This task is challenging to accomplish due to difficulties in aligning multimodal data and the lack of supervised datasets. To address these challenges, we present CLLMate, the first multimodal dataset for WCEF, using 26,156 environmental news articles aligned with ERA5 reanalysis data. We systematically benchmark 23 existing MLLMs on CLLMate, including closed-source, open-source, and our fine-tuned models. Our experiments reveal the advantages and limitations of existing MLLMs and the value of CLLMate for the training and benchmarking of the WCEF task.
Authors: Guanyu Cui, Zhewei Wei, Hsin-Hao Su
Abstract: The success of graph neural networks (GNNs) has spurred theoretical explorations into their expressive power. In the graph machine learning community, researchers often equate GNNs with the Weisfeiler-Lehman (WL) tests as a foundation for theoretical analysis. However, we identify two major limitations of this approach: (1) the semantics of WL tests involve verifying purely structural equivalences through a set of logical sentences. As a result, they do not align well with the concept of expressive power, which is typically defined as the class of functions that GNNs can express, and they are not well-suited for handling graphs with features; (2) by leveraging communication complexity, we show that the lower bound on a GNN's capacity (depth multiplied by width) to simulate one iteration of the WL test grows almost linearly with the graph size. This finding indicates that the WL test is not locally computable and is misaligned with the message-passing GNNs. Furthermore, we show that allowing unlimited precomputation or directly integrating features computed by external models, while claiming that these precomputations enhance the expressiveness of GNNs, can sometimes lead to issues. Such problems can even be observed in an influential paper published in a top-tier machine learning conference. We argue that using well-defined computational models, such as the CONGEST model from distributed computing, is a reasonable approach to characterizing and exploring GNNs' expressive power. Following this approach, we present some results on the effects of virtual nodes and edges. Finally, we highlight several open problems regarding GNN expressive power for further exploration.
Authors: Klaus-Rudolf Kladny, Bernhard Sch\"olkopf, Michael Muehlebach
Abstract: Generative models lack rigorous statistical guarantees for their outputs and are therefore unreliable in safety-critical applications. In this work, we propose Sequential Conformal Prediction for Generative Models (SCOPE-Gen), a sequential conformal prediction method producing prediction sets that satisfy a rigorous statistical guarantee called conformal admissibility control. This guarantee states that with high probability, the prediction sets contain at least one admissible (or valid) example. To this end, our method first samples an initial set of i.i.d. examples from a black box generative model. Then, this set is iteratively pruned via so-called greedy filters. As a consequence of the iterative generation procedure, admissibility of the final prediction set factorizes as a Markov chain. This factorization is crucial, because it allows to control each factor separately, using conformal prediction. In comparison to prior work, our method demonstrates a large reduction in the number of admissibility evaluations during calibration. This reduction is important in safety-critical applications, where these evaluations must be conducted manually by domain experts and are therefore costly and time consuming. We highlight the advantages of our method in terms of admissibility evaluations and cardinality of the prediction sets through experiments in natural language generation and molecular graph extension tasks.
Authors: Seokhyun An, Minji Kim, Hyounghun Kim
Abstract: Instruction tuning -- supervised fine-tuning using instruction-response pairs -- is a key step in making pre-trained large language models (LLMs) instructable. Meanwhile, LLMs perform multitask learning during their pre-training, acquiring extensive knowledge and capabilities. We hypothesize that the pre-training stage can enable them to develop the ability to comprehend and address instructions. To verify this, we propose Response Tuning (RT), which removes the instruction and its corresponding mapping to the response from instruction tuning. Instead, it focuses solely on establishing the response distribution. Our experiments demonstrate that RT models, trained only on responses, can effectively respond to a wide range of instructions and exhibit helpfulness approaching that of their instruction-tuned counterparts. In addition, we observe that the models can recognize and reject unsafe queries after learning the refusal conditions from training responses. Furthermore, we demonstrate that these observations also hold in an in-context learning setting. These findings support our hypothesis, highlighting the extensive inherent capabilities of pre-trained LLMs.
Authors: Tiexin Qin, Mengxu Zhu, Chunyang Li, Terry Lyons, Hong Yan, Haoliang Li
Abstract: Understanding protein dynamics are essential for deciphering protein functional mechanisms and developing molecular therapies. However, the complex high-dimensional dynamics and interatomic interactions of biological processes pose significant challenge for existing computational techniques. In this paper, we approach this problem for the first time by introducing Deep Signature, a novel computationally tractable framework that characterizes complex dynamics and interatomic interactions based on their evolving trajectories. Specifically, our approach incorporates soft spectral clustering that locally aggregates cooperative dynamics to reduce the size of the system, as well as signature transform that collects iterated integrals to provide a global characterization of the non-smooth interactive dynamics. Theoretical analysis demonstrates that Deep Signature exhibits several desirable properties, including invariance to translation, near invariance to rotation, equivariance to permutation of atomic coordinates, and invariance under time reparameterization. Furthermore, experimental results on three benchmarks of biological processes verify that our approach can achieve superior performance compared to baseline methods.
Authors: Yan Scholten, Stephan G\"unnemann, Leo Schwinn
Abstract: Comprehensive evaluation of Large Language Models (LLMs) is an open research problem. Existing evaluations rely on deterministic point estimates generated via greedy decoding. However, we find that deterministic evaluations fail to capture the whole output distribution of a model, yielding inaccurate estimations of model capabilities. This is particularly problematic in critical contexts such as unlearning and alignment, where precise model evaluations are crucial. To remedy this, we introduce the first formal probabilistic evaluation framework for LLMs. Namely, we propose novel metrics with high probability guarantees concerning the output distribution of a model. Our metrics are application-independent and allow practitioners to make more reliable estimates about model capabilities before deployment. Our experimental analysis reveals that deterministic evaluations falsely indicate successful unlearning and alignment, whereas our probabilistic evaluations better capture model capabilities. We show how to overcome challenges associated with probabilistic outputs in a case study on unlearning by introducing (1) a novel loss based on entropy optimization, and (2) adaptive temperature scaling. We demonstrate that our approach significantly enhances unlearning in probabilistic settings on recent benchmarks. Overall, our proposed shift from point estimates to probabilistic evaluations of output distributions represents an important step toward comprehensive evaluations of LLMs. Code available at https://www.cs.cit.tum.de/daml/probabilistic-unlearning/.
URLs: https://www.cs.cit.tum.de/daml/probabilistic-unlearning/.
Authors: Fatemeh Pesaran Zadeh, Juyeon Kim, Jin-Hwa Kim, Gunhee Kim
Abstract: Large language models (LLMs) have demonstrated strong capabilities across various language tasks, notably through instruction-tuning methods. However, LLMs face challenges in visualizing complex, real-world data through charts and plots. Firstly, existing datasets rarely cover a full range of chart types, such as 3D, volumetric, and gridded charts. Secondly, supervised fine-tuning methods do not fully leverage the intricate relationships within rich datasets, including text, code, and figures. To address these challenges, we propose a hierarchical pipeline and a new dataset for chart generation. Our dataset, Text2Chart31, includes 31 unique plot types referring to the Matplotlib library, with 11.1K tuples of descriptions, code, data tables, and plots. Moreover, we introduce a reinforcement learning-based instruction tuning technique for chart generation tasks without requiring human feedback. Our experiments show that this approach significantly enhances the model performance, enabling smaller models to outperform larger open-source models and be comparable to state-of-the-art proprietary models in data visualization tasks. We make the code and dataset available at https://github.com/fatemehpesaran310/Text2Chart31.
Authors: Zi Wang, Divyam Anshumaan, Ashish Hooda, Yudong Chen, Somesh Jha
Abstract: Optimization methods are widely employed in deep learning to identify and mitigate undesired model responses. While gradient-based techniques have proven effective for image models, their application to language models is hindered by the discrete nature of the input space. This study introduces a novel optimization approach, termed the \emph{functional homotopy} method, which leverages the functional duality between model training and input generation. By constructing a series of easy-to-hard optimization problems, we iteratively solve these problems using principles derived from established homotopy methods. We apply this approach to jailbreak attack synthesis for large language models (LLMs), achieving a $20\%-30\%$ improvement in success rate over existing methods in circumventing established safe open-source models such as Llama-2 and Llama-3.
Authors: Manuel Brenner, Elias Weber, Georgia Koppe, Daniel Durstewitz
Abstract: In science, we are often interested in obtaining a generative model of the underlying system dynamics from observed time series. While powerful methods for dynamical systems reconstruction (DSR) exist when data come from a single domain, how to best integrate data from multiple dynamical regimes and leverage it for generalization is still an open question. This becomes particularly important when individual time series are short, and group-level information may help to fill in for gaps in single-domain data. Here we introduce a hierarchical framework that enables to harvest group-level (multi-domain) information while retaining all single-domain characteristics, and showcase it on popular DSR benchmarks, as well as on neuroscience and medical data. In addition to faithful reconstruction of all individual dynamical regimes, our unsupervised methodology discovers common low-dimensional feature spaces in which datasets with similar dynamics cluster. The features spanning these spaces were further dynamically highly interpretable, surprisingly in often linear relation to control parameters that govern the dynamics of the underlying system. Finally, we illustrate transfer learning and generalization to new parameter regimes, paving the way toward DSR foundation models.
Authors: Zelin Zang, Yongjie Xu, Chenrui Duan, Yue Yuan, Jinlin Wu, Zhen Lei, Stan Z. Li
Abstract: Biological tree (BioTree) analysis is a foundational tool in biology, enabling the exploration of evolutionary and differentiation relationships among organisms, genes, and cells. Traditional tree construction methods, while instrumental in early research, face significant challenges in handling the growing complexity and scale of modern biological data, particularly in integrating multimodal datasets. Advances in deep learning (DL) offer transformative opportunities by enabling the fusion of biological prior knowledge with data-driven models. These approaches address key limitations of traditional methods, facilitating the construction of more accurate and interpretable BioTrees. This review highlights critical biological priors essential for phylogenetic and differentiation tree analyses and explores strategies for integrating these priors into DL models to enhance accuracy and interpretability. Additionally, the review systematically examines commonly used data modalities and databases, offering a valuable resource for developing and evaluating multimodal fusion models. Traditional tree construction methods are critically assessed, focusing on their biological assumptions, technical limitations, and scalability issues. Recent advancements in DL-based tree generation methods are reviewed, emphasizing their innovative approaches to multimodal integration and prior knowledge incorporation. Finally, the review discusses diverse applications of BioTrees in various biological disciplines, from phylogenetics to developmental biology, and outlines future trends in leveraging DL to advance BioTree research. By addressing the challenges of data complexity and prior knowledge integration, this review aims to inspire interdisciplinary innovation at the intersection of biology and DL.
Authors: Feng Tian, Yixuan Li, Yichao Yan, Shanyan Guan, Yanhao Ge, Xiaokang Yang
Abstract: In the field of image editing, three core challenges persist: controllability, background preservation, and efficiency. Inversion-based methods rely on time-consuming optimization to preserve the features of the initial images, which results in low efficiency due to the requirement for extensive network inference. Conversely, inversion-free methods lack theoretical support for background similarity, as they circumvent the issue of maintaining initial features to achieve efficiency. As a consequence, none of these methods can achieve both high efficiency and background consistency. To tackle the challenges and the aforementioned disadvantages, we introduce PostEdit, a method that incorporates a posterior scheme to govern the diffusion sampling process. Specifically, a corresponding measurement term related to both the initial features and Langevin dynamics is introduced to optimize the estimated image generated by the given target prompt. Extensive experimental results indicate that the proposed PostEdit achieves state-of-the-art editing performance while accurately preserving unedited regions. Furthermore, the method is both inversion- and training-free, necessitating approximately 1.5 seconds and 18 GB of GPU memory to generate high-quality results.
Authors: Yuxin Xiao, Shujian Zhang, Wenxuan Zhou, Marzyeh Ghassemi, Sanqiang Zhao
Abstract: To acquire instruction-following capabilities, large language models (LLMs) undergo instruction tuning, where they are trained on instruction-response pairs using next-token prediction (NTP). Efforts to improve instruction tuning often focus on higher-quality supervised fine-tuning (SFT) datasets, typically requiring data filtering with proprietary LLMs or human annotation. In this paper, we take a different approach by proposing SFTMix, a novel Mixup-based recipe that elevates LLM instruction tuning beyond the conventional NTP paradigm, without relying on well-curated datasets. Observing that LLMs exhibit uneven confidence across the semantic representation space, we argue that examples with different confidence levels should play distinct roles in instruction tuning--confident data is prone to overfitting, while unconfident data is harder to generalize. Based on this insight, SFTMix leverages training dynamics to identify examples with varying confidence levels, interpolates them to bridge the confidence gap, and applies a Mixup-based regularization to support learning on these additional, interpolated examples. By propagating supervision signals across confidence regions and encouraging linear behavior between them, SFTMix mitigates overfitting in confident examples while enhancing generalization in unconfident ones. We demonstrate the effectiveness of SFTMix in both instruction-following and healthcare-specific SFT tasks, with consistent improvements across LLM families and SFT datasets of varying sizes and qualities. Extensive analyses across six directions highlight SFTMix's compatibility with data selection, adaptability to compute-constrained scenarios, and scalability to broader applications.
Authors: Yonatan Sverdlov, Ido Springer, Nadav Dym
Abstract: This paper explores the characterization of equivariant linear layers for representations of permutations and related groups. Unlike traditional approaches, which address these problems using parameter-sharing, we consider an alternative methodology based on irreducible representations and Schur's lemma. Using this methodology, we obtain an alternative derivation for existing models like DeepSets, 2-IGN graph equivariant networks, and Deep Weight Space (DWS) networks. The derivation for DWS networks is significantly simpler than that of previous results. Next, we extend our approach to unaligned symmetric sets, where equivariance to the wreath product of groups is required. Previous works have addressed this problem in a rather restrictive setting, in which almost all wreath equivariant layers are Siamese. In contrast, we give a full characterization of layers in this case and show that there is a vast number of additional non-Siamese layers in some settings. We also show empirically that these additional non-Siamese layers can improve performance in tasks like graph anomaly detection, weight space alignment, and learning Wasserstein distances. Our code is available at \href{https://github.com/yonatansverdlov/Irreducible-Representations-of-Deep-Weight-Spaces}{GitHub}.
URLs: https://github.com/yonatansverdlov/Irreducible-Representations-of-Deep-Weight-Spaces
Authors: Jona Ruthardt, Gertjan J. Burghouts, Serge Belongie, Yuki M. Asano
Abstract: How well do text-only Large Language Models (LLMs) naturally align with the visual world? We provide the first direct analysis by utilizing frozen text representations in a discriminative vision-language model framework and measuring zero-shot generalization on unseen classes. We find decoder-based LLMs exhibit high intrinsic visual alignment. In particular, more capable LLMs reliably demonstrate stronger generalization. Moreover, utilizing frozen LLMs leads to strong gains in cross-lingual settings, where our approach surpasses CLIP's accuracy of 1.4% with 38.7% for Chinese. Our proposed method improves both robustness and generalization and also significantly reduces the need for paired data and compute, making vision-language models more accessible and adaptable.
Authors: Youquan Li, Miao Zheng, Fan Yang, Guosheng Dong, Bin Cui, Weipeng Chen, Zenan Zhou, Wentao Zhang
Abstract: Human feedback is crucial in the interactions between humans and Large Language Models (LLMs). However, existing research primarily focuses on benchmarking LLMs in single-turn dialogues. Even in benchmarks designed for multi-turn dialogues, the user inputs are often independent, neglecting the nuanced and complex nature of human feedback within real-world usage scenarios. To fill this research gap, we introduce FB-Bench, a fine-grained, multi-task benchmark designed to evaluate LLMs' responsiveness to human feedback under real-world usage scenarios in Chinese. Drawing from the two main interaction scenarios, FB-Bench comprises 591 meticulously curated samples, encompassing eight task types, five deficiency types of response, and nine feedback types. We extensively evaluate a broad array of popular LLMs, revealing significant variations in their performance across different interaction scenarios. Further analysis indicates that task, human feedback, and deficiencies of previous responses can also significantly impact LLMs' responsiveness. Our findings underscore both the strengths and limitations of current models, providing valuable insights and directions for future research. Code and datasets are available at https://github.com/PKU-Baichuan-MLSystemLab/FB-Bench.
Authors: Shu Yang, Shenzhe Zhu, Liang Liu, Lijie Hu, Mengdi Li, Di Wang
Abstract: Large language models (LLMs) have significantly advanced dialogue systems and role-playing agents through their ability to generate human-like text. While prior studies have shown that LLMs can exhibit distinct and consistent personalities, the mechanisms through which these models encode and express specific personality traits remain poorly understood. To address this, we investigate how various factors, such as cultural norms and environmental stressors, encoded within LLMs, shape their personality traits, guided by the theoretical framework of social determinism. Inspired by related work on LLM interpretability, we propose a training-free approach to modify the model's behavior by extracting and steering latent features corresponding to factors within the model, thereby eliminating the need for retraining. Furthermore, we analyze the implications of these factors for model safety, focusing on their impact through the lens of personality.
Authors: Luyi Ma, Xiaohan Li, Zezhong Fan, Kai Zhao, Jianpeng Xu, Jason Cho, Praveen Kanumala, Kaushiki Nag, Sushant Kumar, Kannan Achan
Abstract: Integrating diverse data modalities is crucial for enhancing the performance of personalized recommendation systems. Traditional models, which often rely on singular data sources, lack the depth needed to accurately capture the multifaceted nature of item features and user behaviors. This paper introduces a novel framework for multi-behavior recommendations, leveraging the fusion of triple-modality, which is visual, textual, and graph data through alignment with large language models (LLMs). By incorporating visual information, we capture contextual and aesthetic item characteristics; textual data provides insights into user interests and item features in detail; and graph data elucidates relationships within the item-behavior heterogeneous graphs. Our proposed model called Triple Modality Fusion (TMF) utilizes the power of LLMs to align and integrate these three modalities, achieving a comprehensive representation of user behaviors. The LLM models the user's interactions including behaviors and item features in natural languages. Initially, the LLM is warmed up using only natural language-based prompts. We then devise the modality fusion module based on cross-attention and self-attention mechanisms to integrate different modalities from other models into the same embedding space and incorporate them into an LLM. Extensive experiments demonstrate the effectiveness of our approach in improving recommendation accuracy. Further ablation studies validate the effectiveness of our model design and benefits of the TMF.
Authors: Jiahao Yuan, Dehui Du, Hao Zhang, Zixiang Di, Usman Naseem
Abstract: Large language models (LLMs) have shown remarkable performance in reasoning tasks but face limitations in mathematical and complex logical reasoning. Existing methods to improve LLMs' logical capabilities either involve traceable or verifiable logical sequences that generate more reliable responses by constructing logical structures yet increase computational costs, or introduces rigid logic template rules, reducing flexibility. In this paper, we propose Reversal of Thought (RoT), a plug-and-play and cost-effective reasoning framework designed to enhance the logical reasoning abilities of LLMs during the warm-up phase prior to batch inference. RoT utilizes a Preference-Guided Reverse Reasoning warm-up strategy, which integrates logical symbols for pseudocode planning through meta-cognitive mechanisms and pairwise preference self-evaluation to generate task-specific prompts solely through demonstrations, aligning with LLMs' cognitive preferences shaped by RLHF. Through reverse reasoning, we utilize a Cognitive Preference Manager to assess knowledge boundaries and further expand LLMs' reasoning capabilities by aggregating solution logic for known tasks and stylistic templates for unknown tasks. Experiments across various tasks demonstrate that RoT surpasses existing baselines in both reasoning accuracy and efficiency.
Authors: Qingren Yao, Chao-Han Huck Yang, Renhe Jiang, Yuxuan Liang, Ming Jin, Shirui Pan
Abstract: Scaling laws offer valuable insights into the design of time series foundation models (TSFMs). However, previous research has largely focused on the scaling laws of TSFMs for in-distribution (ID) data, leaving their out-of-distribution (OOD) scaling behavior and the influence of model architectures less explored. In this work, we examine two common TSFM architectures, encoder-only and decoder-only Transformers, and investigate their scaling behavior on both ID and OOD data. These models are trained and evaluated across varying parameter counts, compute budgets, and dataset sizes. Our experiments reveal that the log-likelihood loss of TSFMs exhibits similar scaling behavior in both OOD and ID settings. We further compare the scaling properties across different architectures, incorporating two state-of-the-art TSFMs as case studies, showing that model architecture plays a significant role in scaling. The encoder-only Transformers demonstrate better scalability than the decoder-only Transformers, while the architectural enhancements in the two advanced TSFMs primarily improve ID performance but reduce OOD scalability. While scaling up TSFMs is expected to drive performance breakthroughs, the lack of a comprehensive understanding of TSFM scaling laws has hindered the development of a robust framework to guide model scaling. We fill this gap in this work by synthesizing our findings and providing practical guidelines for designing and scaling larger TSFMs with enhanced model capabilities.
Authors: Hyeonwoo Kim, Dahyun Kim, Jihoo Kim, Sukyung Lee, Yungi Kim, Chanjun Park
Abstract: The Open Ko-LLM Leaderboard has been instrumental in benchmarking Korean Large Language Models (LLMs), yet it has certain limitations. Notably, the disconnect between quantitative improvements on the overly academic leaderboard benchmarks and the qualitative impact of the models should be addressed. Furthermore, the benchmark suite is largely composed of translated versions of their English counterparts, which may not fully capture the intricacies of the Korean language. To address these issues, we propose Open Ko-LLM Leaderboard2, an improved version of the earlier Open Ko-LLM Leaderboard. The original benchmarks are entirely replaced with new tasks that are more closely aligned with real-world capabilities. Additionally, four new native Korean benchmarks are introduced to better reflect the distinct characteristics of the Korean language. Through these refinements, Open Ko-LLM Leaderboard2 seeks to provide a more meaningful evaluation for advancing Korean LLMs.
Authors: Yongqin Xu, Huan Li, Ke Chen, Lidan Shou
Abstract: Schema matching (SM) and entity matching (EM) tasks are crucial for data integration. While large language models (LLMs) have shown promising results in these tasks, they suffer from hallucinations and confusion about task instructions. This study presents the Knowledge-Compliant Matching Framework (KcMF), an LLM-based approach that addresses these issues without the need for domain-specific fine-tuning. KcMF employs a once-and-for-all pseudo-code-based task decomposition strategy to adopt natural language statements that guide LLM reasoning and reduce confusion across various task types. We also propose two mechanisms, Dataset as Knowledge (DaK) and Example as Knowledge (EaK), to build domain knowledge sets when unstructured domain knowledge is lacking. Moreover, we introduce a result-ensemble strategy to leverage multiple knowledge sources and suppress badly formatted outputs. Extensive evaluations confirm that KcMF clearly enhances five LLM backbones in both SM and EM tasks while outperforming the non-LLM competitors by an average F1-score of 17.93%.
Authors: So Kuroki, Taishi Nakamura, Takuya Akiba, Yujin Tang
Abstract: Training large language models to acquire specific skills remains a challenging endeavor. Conventional training approaches often struggle with data distribution imbalances and inadequacies in objective functions that do not align well with task-specific performance. To address these challenges, we introduce CycleQD, a novel approach that leverages the Quality Diversity framework through a cyclic adaptation of the algorithm, along with a model merging based crossover and an SVD-based mutation. In CycleQD, each task's performance metric is alternated as the quality measure while the others serve as the behavioral characteristics. This cyclic focus on individual tasks allows for concentrated effort on one task at a time, eliminating the need for data ratio tuning and simplifying the design of the objective function. Empirical results from AgentBench indicate that applying CycleQD to LLAMA3-8B-INSTRUCT based models not only enables them to surpass traditional fine-tuning methods in coding, operating systems, and database tasks, but also achieves performance on par with GPT-3.5-TURBO, which potentially contains much more parameters, across these domains. Crucially, this enhanced performance is achieved while retaining robust language capabilities, as evidenced by its performance on widely adopted language benchmark tasks. We highlight the key design choices in CycleQD, detailing how these contribute to its effectiveness. Furthermore, our method is general and can be applied to image segmentation models, highlighting its applicability across different domains.
Authors: Can Cui, Yunsheng Ma, Zichong Yang, Yupeng Zhou, Peiran Liu, Juanwu Lu, Lingxi Li, Yaobin Chen, Jitesh H. Panchal, Amr Abdelraouf, Rohit Gupta, Kyungtae Han, Ziran Wang
Abstract: With the broader usage and highly successful development of Large Language Models (LLMs), there has been a growth of interest and demand for applying LLMs to autonomous driving technology. Driven by their natural language understanding and reasoning ability, LLMs have the potential to enhance various aspects of autonomous driving systems, from perception and scene understanding to language interaction and decision-making. In this paper, we first introduce the novel concept of designing LLMs for autonomous driving (LLM4AD). Then, we propose a comprehensive benchmark for evaluating the instruction-following abilities of LLM4AD in simulation. Furthermore, we conduct a series of experiments on real-world vehicle platforms, thoroughly evaluating the performance and potential of our LLM4AD systems. Finally, we envision the main challenges of LLM4AD, including latency, deployment, security and privacy, safety, trust and transparency, and personalization. Our research highlights the significant potential of LLMs to enhance various aspects of autonomous vehicle technology, from perception and scene understanding to language interaction and decision-making.
Authors: Lawrence Jang, Yinheng Li, Dan Zhao, Charles Ding, Justin Lin, Paul Pu Liang, Rogerio Bonatti, Kazuhito Koishida
Abstract: Videos are often used to learn or extract the necessary information to complete tasks in ways different than what text and static imagery alone can provide. However, many existing agent benchmarks neglect long-context video understanding, instead focusing on text or static image inputs. To bridge this gap, we introduce VideoWebArena (VideoWA), a benchmark for evaluating the capabilities of long-context multimodal agents for video understanding. VideoWA consists of 2,021 web agent tasks based on manually crafted video tutorials, which total almost four hours of content. For our benchmark, we define a taxonomy of long-context video-based agent tasks with two main areas of focus: skill retention and factual retention. While skill retention tasks evaluate whether an agent can use a given human demonstration to complete a task efficiently, the factual retention task evaluates whether an agent can retrieve instruction-relevant information from a video to complete a task. We find that the best model achieves 13.3% success on factual retention tasks and 45.8% on factual retention QA pairs, far below human performance at 73.9% and 79.3%, respectively. On skill retention tasks, long-context models perform worse with tutorials than without, exhibiting a 5% performance decrease in WebArena tasks and a 10.3% decrease in VisualWebArena tasks. Our work highlights the need to improve the agentic abilities of long-context multimodal models and provides a testbed for future development with long-context video agents.
Authors: Zheyuan Liu, Guangyao Dou, Mengzhao Jia, Zhaoxuan Tan, Qingkai Zeng, Yongle Yuan, Meng Jiang
Abstract: Generative models such as Large Language Models (LLM) and Multimodal Large Language models (MLLMs) trained on massive web corpora can memorize and disclose individuals' confidential and private data, raising legal and ethical concerns. While many previous works have addressed this issue in LLM via machine unlearning, it remains largely unexplored for MLLMs. To tackle this challenge, we introduce Multimodal Large Language Model Unlearning Benchmark (MLLMU-Bench), a novel benchmark aimed at advancing the understanding of multimodal machine unlearning. MLLMU-Bench consists of 500 fictitious profiles and 153 profiles for public celebrities, each profile feature over 14 customized question-answer pairs, evaluated from both multimodal (image+text) and unimodal (text) perspectives. The benchmark is divided into four sets to assess unlearning algorithms in terms of efficacy, generalizability, and model utility. Finally, we provide baseline results using existing generative model unlearning algorithms. Surprisingly, our experiments show that unimodal unlearning algorithms excel in generation and cloze tasks, while multimodal unlearning approaches perform better in classification tasks with multimodal inputs.
Authors: Zhongwu Chen, Chengjin Xu, Dingmin Wang, Zhen Huang, Yong Dou, Xuhui Jiang, Jian Guo
Abstract: Retrieval-augmented generation (RAG) has shown promising potential in knowledge intensive question answering (QA). However, existing approaches only consider the query itself, neither specifying the retrieval preferences for the retrievers nor informing the generators of how to refer to the retrieved documents for the answers, which poses a significant challenge to the QA performance. To address these issues, we propose Rule-guided Retrieval-Augmented Generation with LMs, which explicitly introduces rules for in-context learning (RuleRAG-ICL) to guide retrievers to recall related documents in the directions of rules and uniformly guide generators to reason attributed by the same rules. Moreover, most existing RAG datasets were constructed without considering rules and Knowledge Graphs (KGs) are recognized as providing high-quality rules. Therefore, we construct five rule-aware RAG benchmarks for QA, RuleQA, based on KGs to stress the significance of retrieval and reasoning with rules. Experiments on RuleQA demonstrate RuleRAG-ICL improves the retrieval quality of +89.2% in Recall@10 and answer accuracy of +103.1% in Exact Match, and RuleRAG-FT yields more enhancement. In addition, experiments on four existing RAG datasets show RuleRAG is also effective by offering rules in RuleQA to them, further proving the generalization of rule guidance in RuleRAG.
Authors: Junting Chen, Checheng Yu, Xunzhe Zhou, Tianqi Xu, Yao Mu, Mengkang Hu, Wenqi Shao, Yikai Wang, Guohao Li, Lin Shao
Abstract: Heterogeneous multi-robot systems (HMRS) have emerged as a powerful approach for tackling complex tasks that single robots cannot manage alone. Current large-language-model-based multi-agent systems (LLM-based MAS) have shown success in areas like software development and operating systems, but applying these systems to robot control presents unique challenges. In particular, the capabilities of each agent in a multi-robot system are inherently tied to the physical composition of the robots, rather than predefined roles. To address this issue, we introduce a novel multi-agent framework designed to enable effective collaboration among heterogeneous robots with varying embodiments and capabilities, along with a new benchmark named Habitat-MAS. One of our key designs is $\textit{Robot Resume}$: Instead of adopting human-designed role play, we propose a self-prompted approach, where agents comprehend robot URDF files and call robot kinematics tools to generate descriptions of their physics capabilities to guide their behavior in task planning and action execution. The Habitat-MAS benchmark is designed to assess how a multi-agent framework handles tasks that require embodiment-aware reasoning, which includes 1) manipulation, 2) perception, 3) navigation, and 4) comprehensive multi-floor object rearrangement. The experimental results indicate that the robot's resume and the hierarchical design of our multi-agent system are essential for the effective operation of the heterogeneous multi-robot system within this intricate problem context.
Authors: Wenxiao Wang, Lihui Gu, Liye Zhang, Yunxiang Luo, Yi Dai, Chen Shen, Liang Xie, Binbin Lin, Xiaofei He, Jieping Ye
Abstract: The rapid advancement of large language models (LLMs) has opened new possibilities for automating the proposal of innovative scientific ideas. This process involves two key phases: literature retrieval and idea generation. However, existing approaches often fall short due to their reliance on keyword-based search tools during the retrieval phase, which neglects crucial semantic information and frequently results in incomplete retrieval outcomes. Similarly, in the idea generation phase, current methodologies tend to depend solely on the internal knowledge of LLMs or metadata from retrieved papers, thereby overlooking significant valuable insights contained within the full texts. To address these limitations, we introduce SciPIP, an innovative framework designed to enhance the LLM-based proposal of scientific ideas through improvements in both literature retrieval and idea generation. Our approach begins with the construction of a comprehensive literature database that supports advanced retrieval based not only on keywords but also on semantics and citation relationships. This is complemented by the introduction of a multi-granularity retrieval algorithm aimed at ensuring more thorough and exhaustive retrieval results. For the idea generation phase, we propose a dual-path framework that effectively integrates both the content of retrieved papers and the extensive internal knowledge of LLMs. This integration significantly boosts the novelty, feasibility, and practical value of proposed ideas. Our experiments, conducted across various domains such as natural language processing and computer vision, demonstrate SciPIP's capability to generate a multitude of innovative and useful ideas. These findings underscore SciPIP's potential as a valuable tool for researchers seeking to advance their fields with groundbreaking concepts.
Authors: Xinghao Wang, Pengyu Wang, Bo Wang, Dong Zhang, Yunhua Zhou, Xipeng Qiu
Abstract: Large language models (LLMs) have revolutionized numerous applications, yet their deployment remains challenged by memory constraints on local devices. While scaling laws have enhanced LLM capabilities, the primary bottleneck has shifted from \textit{capability} to \textit{availability}, emphasizing the need for efficient memory management. Traditional compression methods, such as quantization, often require predefined compression ratios and separate compression processes for each setting, complicating deployment in variable memory environments. In this paper, we introduce \textbf{BitStack}, a novel, training-free weight compression approach that enables megabyte-level trade-offs between memory usage and model performance. By leveraging weight decomposition, BitStack can dynamically adjust the model size with minimal transmission between running memory and storage devices. Our approach iteratively decomposes weight matrices while considering the significance of each parameter, resulting in an approximately 1-bit per parameter residual block in each decomposition iteration. These blocks are sorted and stacked in storage as basic transmission units, with different quantities loaded based on current memory availability. Extensive experiments across a wide range of tasks demonstrate that, despite offering fine-grained size control, BitStack consistently matches or surpasses strong quantization baselines, particularly at extreme compression ratios. To the best of our knowledge, this is the first decomposition-based method that effectively bridges the gap to practical compression techniques like quantization. Code is available at https://github.com/xinghaow99/BitStack.
Authors: Kung-Hsiang Huang, Akshara Prabhakar, Sidharth Dhawan, Yixin Mao, Huan Wang, Silvio Savarese, Caiming Xiong, Philippe Laban, Chien-Sheng Wu
Abstract: Customer Relationship Management (CRM) systems are vital for modern enterprises, providing a foundation for managing customer interactions and data. Integrating AI agents into CRM systems can automate routine processes and enhance personalized service. However, deploying and evaluating these agents is challenging due to the lack of realistic benchmarks that reflect the complexity of real-world CRM tasks. To address this issue, we introduce CRMArena, a novel benchmark designed to evaluate AI agents on realistic tasks grounded in professional work environments. Following guidance from CRM experts and industry best practices, we designed CRMArena with nine customer service tasks distributed across three personas: service agent, analyst, and manager. The benchmark includes 16 commonly used industrial objects (e.g., account, order, knowledge article, case) with high interconnectivity, along with latent variables (e.g., complaint habits, policy violations) to simulate realistic data distributions. Experimental results reveal that state-of-the-art LLM agents succeed in less than 40% of the tasks with ReAct prompting, and less than 55% even with function-calling abilities. Our findings highlight the need for enhanced agent capabilities in function-calling and rule-following to be deployed in real-world work environments. CRMArena is an open challenge to the community: systems that can reliably complete tasks showcase direct business value in a popular work environment.
Authors: Qisheng Hu, Quanyu Long, Wenya Wang
Abstract: Fact-checking pipelines increasingly adopt the Decompose-Then-Verify paradigm, where texts are broken down into smaller claims for individual verification and subsequently combined for a veracity decision. While decomposition is widely-adopted in such pipelines, its effects on final fact-checking performance remain underexplored. Some studies have reported improvements from decompostition, while others have observed performance declines, indicating its inconsistent impact. To date, no comprehensive analysis has been conducted to understand this variability. To address this gap, we present an in-depth analysis that explicitly examines the impact of decomposition on downstream verification performance. Through error case inspection and experiments, we introduce a categorization of decomposition errors and reveal a trade-off between accuracy gains and the noise introduced through decomposition. Our analysis provides new insights into understanding current system's instability and offers guidance for future studies toward improving claim decomposition in fact-checking pipelines.
Authors: Dingjie Song, Sicheng Lai, Shunian Chen, Lichao Sun, Benyou Wang
Abstract: The rapid progression of multimodal large language models (MLLMs) has demonstrated superior performance on various multimodal benchmarks. However, the issue of data contamination during training creates challenges in performance evaluation and comparison. While numerous methods exist for detecting models' contamination in large language models (LLMs), they are less effective for MLLMs due to their various modalities and multiple training phases. In this study, we introduce a multimodal data contamination detection framework, MM-Detect, designed for MLLMs. Our experimental results indicate that MM-Detect is quite effective and sensitive in identifying varying degrees of contamination, and can highlight significant performance improvements due to the leakage of multimodal benchmark training sets. Furthermore, we explore whether the contamination originates from the base LLMs used by MLLMs or the multimodal training phase, providing new insights into the stages at which contamination may be introduced.
Authors: Zhanwei Zhang, Shizhao Sun, Wenxiao Wang, Deng Cai, Jiang Bian
Abstract: Recently, there is a growing interest in creating computer-aided design (CAD) models based on user intent, known as controllable CAD generation. Existing work offers limited controllability and needs separate models for different types of control, reducing efficiency and practicality. To achieve controllable generation across all CAD construction hierarchies, such as sketch-extrusion, extrusion, sketch, face, loop and curve, we propose FlexCAD, a unified model by fine-tuning large language models (LLMs). First, to enhance comprehension by LLMs, we represent a CAD model as a structured text by abstracting each hierarchy as a sequence of text tokens. Second, to address various controllable generation tasks in a unified model, we introduce a hierarchy-aware masking strategy. Specifically, during training, we mask a hierarchy-aware field in the CAD text with a mask token. This field, composed of a sequence of tokens, can be set flexibly to represent various hierarchies. Subsequently, we ask LLMs to predict this masked field. During inference, the user intent is converted into a CAD text with a mask token replacing the part the user wants to modify, which is then fed into FlexCAD to generate new CAD models. Comprehensive experiments on public dataset demonstrate the effectiveness of FlexCAD in both generation quality and controllability. Code will be available at https://github.com/microsoft/FlexCAD.
Authors: Hyukhun Koh, Minha Jhang, Dohyung Kim, Sangmook Lee, Kyomin Jung
Abstract: Although auto-regressive models excel in natural language processing, they often struggle to generate diverse text and provide limited controllability. Non-auto-regressive methods could be an alternative but often produce degenerate outputs and exhibit shortcomings in conditional generation. To address these challenges, we propose Diffusion-EAGS, a novel framework that integrates conditional masked language models into diffusion language models through the theoretical lens of a conditional Markov Random Field. In doing so, we propose entropy-adaptive Gibbs sampling and entropy-based noise scheduling to counterbalance each model's shortcomings. Experimental results show that Diffusion-EAGS outperforms baselines and achieves the best quality-diversity tradeoff, demonstrating its effectiveness in non-autoregressive text generation.
Authors: YeongHyeon Park, Myung Jin Kim, Hyeong Seok Kim
Abstract: A pre-trained visual-language model, contrastive language-image pre-training (CLIP), successfully accomplishes various downstream tasks with text prompts, such as finding images or localizing regions within the image. Despite CLIP's strong multi-modal data capabilities, it remains limited in specialized environments, such as medical applications. For this purpose, many CLIP variants-i.e., BioMedCLIP, and MedCLIP-SAMv2-have emerged, but false positives related to normal regions persist. Thus, we aim to present a simple yet important goal of reducing false positives in medical anomaly detection. We introduce a Contrastive LAnguage Prompting (CLAP) method that leverages both positive and negative text prompts. This straightforward approach identifies potential lesion regions by visual attention to the positive prompts in the given image. To reduce false positives, we attenuate attention on normal regions using negative prompts. Extensive experiments with the BMAD dataset, including six biomedical benchmarks, demonstrate that CLAP method enhances anomaly detection performance. Our future plans include developing an automated fine prompting method for more practical usage.
Authors: Yeong-Joon Ju, Ho-Joong Kim, Seong-Whan Lee
Abstract: Recent multimodal retrieval methods have endowed text-based retrievers with multimodal capabilities by utilizing pre-training strategies for visual-text alignment. They often directly fuse the two modalities for cross-reference during the alignment to understand multimodal queries. However, existing methods often overlook crucial visual information due to a text-dominant issue, which overly depends on text-driven signals. In this paper, we introduce MIRe, a retrieval framework that achieves modality interaction without fusing textual features during the alignment. Our method allows the textual query to attend to visual embeddings while not feeding text-driven signals back into the visual representations. Additionally, we construct a pre-training dataset for multimodal query retrieval by transforming concise question-answer pairs into extended passages. Our experiments demonstrate that our pre-training strategy significantly enhances the understanding of multimodal queries, resulting in strong performance across four multimodal retrieval benchmarks under zero-shot settings. Our code is publicly available: https://github.com/yeongjoonJu/MIRe.
Authors: Zekun Fei, Biao Yi, Jianing Geng, Ruiqi He, Lihai Nie, Zheli Liu
Abstract: Embedding-as-a-Service (EaaS) has emerged as a successful business pattern but faces significant challenges related to various forms of copyright infringement, particularly, the API misuse and model extraction attacks. Various studies have proposed backdoor-based watermarking schemes to protect the copyright of EaaS services. In this paper, we reveal that previous watermarking schemes possess semantic-independent characteristics and propose the Semantic Perturbation Attack (SPA). Our theoretical and experimental analysis demonstrate that this semantic-independent nature makes current watermarking schemes vulnerable to adaptive attacks that exploit semantic perturbations tests to bypass watermark verification. Extensive experimental results across multiple datasets demonstrate that the True Positive Rate (TPR) for identifying watermarked samples under SPA can reach up to more than 95\%, rendering watermarks ineffective while maintaining the high utility of embeddings. Furthermore, we discuss potential defense strategies to mitigate SPA. Our code is available at https://github.com/Zk4-ps/EaaS-Embedding-Watermark.
Authors: Erik Miehling, Michael Desmond, Karthikeyan Natesan Ramamurthy, Elizabeth M. Daly, Pierre Dognin, Jesus Rios, Djallel Bouneffouf, Miao Liu
Abstract: Building pluralistic AI requires designing models that are able to be shaped to represent a wide range of value systems and cultures. Achieving this requires first being able to evaluate the degree to which a given model is capable of reflecting various personas. To this end, we propose a benchmark for evaluating the steerability of model personas as a function of prompting. Our design is based on a formal definition of prompt steerability, which analyzes the degree to which a model's joint behavioral distribution can be shifted from its baseline. By defining steerability indices and inspecting how these indices change as a function of steering effort, we can estimate the steerability of a model across various persona dimensions and directions. Our benchmark reveals that the steerability of many current models is limited -- due to both a skew in their baseline behavior and an asymmetry in their steerability across many persona dimensions. We release an implementation of our benchmark at https://github.com/IBM/prompt-steering.
Authors: Huaizhi Ge, Yiming Li, Qifan Wang, Yongfeng Zhang, Ruixiang Tang
Abstract: Large Language Models (LLMs) are known to be vulnerable to backdoor attacks, where triggers embedded in poisoned samples can maliciously alter LLMs' behaviors. In this paper, we move beyond attacking LLMs and instead examine backdoor attacks through the novel lens of natural language explanations. Specifically, we leverage LLMs' generative capabilities to produce human-readable explanations for their decisions, enabling direct comparisons between explanations for clean and poisoned samples. Our results show that backdoored models produce coherent explanations for clean inputs but diverse and logically flawed explanations for poisoned data, a pattern consistent across classification and generation tasks for different backdoor attacks. Further analysis reveals key insights into the explanation generation process. At the token level, explanation tokens associated with poisoned samples only appear in the final few transformer layers. At the sentence level, attention dynamics indicate that poisoned inputs shift attention away from the original input context during explanation generation. These findings enhance our understanding of backdoor mechanisms in LLMs and present a promising framework for detecting vulnerabilities through explainability.
Authors: Roland Daynauth, Christopher Clarke, Krisztian Flautner, Lingjia Tang, Jason Mars
Abstract: Deciding which large language model (LLM) to use is a complex challenge. Pairwise ranking has emerged as a new method for evaluating human preferences for LLMs. This approach entails humans evaluating pairs of model outputs based on a predefined criterion. By collecting these comparisons, a ranking can be constructed using methods such as Elo. However, applying these algorithms as constructed in the context of LLM evaluation introduces several challenges. In this paper, we explore the effectiveness of ranking systems for head-to-head comparisons of LLMs. We formally define a set of fundamental principles for effective ranking and conduct a series of extensive evaluations on the robustness of several ranking algorithms in the context of LLMs. Our analysis uncovers key insights into the factors that affect ranking accuracy and efficiency, offering guidelines for selecting the most appropriate methods based on specific evaluation contexts and resource constraints.
Authors: Jinglei Cheng, Ruilin Zhou, Yuhang Gan, Chen Qian, Junyu Liu
Abstract: We introduce Quantum Hamiltonian Descent as a novel approach to solve the graph partition problem. By reformulating graph partition as a Quadratic Unconstrained Binary Optimization (QUBO) problem, we leverage QHD's quantum-inspired dynamics to identify optimal community structures. Our method implements a multi-level refinement strategy that alternates between QUBO formulation and QHD optimization to iteratively improve partition quality. Experimental results demonstrate that our QHD-based approach achieves superior modularity scores (up to 5.49\%) improvement with reduced computational overhead compared to traditional optimization methods. This work establishes QHD as an effective quantum-inspired framework for tackling graph partition challenges in large-scale networks.
Authors: Eric Tang, Bangding Yang, Xingyou Song
Abstract: With the rise of large language models (LLMs) for flexibly processing information as strings, a natural application is regression, specifically by preprocessing string representations into LLM embeddings as downstream features for metric prediction. In this paper, we provide one of the first comprehensive investigations into embedding-based regression and demonstrate that LLM embeddings as features can be better for high-dimensional regression tasks than using traditional feature engineering. This regression performance can be explained in part due to LLM embeddings over numeric data inherently preserving Lipschitz continuity over the feature space. Furthermore, we quantify the contribution of different model effects, most notably model size and language understanding, which we find surprisingly do not always improve regression performance.
Authors: Yiqun Zhang, Mingjie Zhao, Hong Jia, Yang Lu, Mengke Li, Yiu-ming Cheung
Abstract: Categorical data composed of qualitative valued attributes are ubiquitous in machine learning tasks. Due to the lack of well-defined metric space, categorical data distributions are difficult to be intuitively understood. Clustering is a popular data analysis technique suitable for data distribution understanding. However, the success of clustering often relies on reasonable distance metrics, which happens to be what categorical data naturally lack. This paper therefore introduces a new finding that the order relation among attribute values is the decisive factor in clustering accuracy, and is also the key to understanding categorical data clusters, because the essence of clustering is to order the clusters in terms of their admission to samples. To obtain the orders, we propose a new learning paradigm that allows joint learning of clusters and the orders. It alternatively partitions the data into clusters based on the distance metric built upon the orders and estimates the most likely orders according to the clusters. The algorithm achieves superior clustering accuracy with a convergence guarantee, and the learned orders facilitate the understanding of the non-intuitive cluster distribution of categorical data. Extensive experiments with ablation studies, statistical evidence, and case studies have validated the new insight into the importance of value order and the method proposition. The source code is temporarily opened in https://anonymous.4open.science/r/OCL-demo.
Authors: Matthew Niedoba, Berend Zwartsenberg, Kevin Murphy, Frank Wood
Abstract: We propose a simple, training-free mechanism which explains the generalization behaviour of diffusion models. By comparing pre-trained diffusion models to their theoretically optimal empirical counterparts, we identify a shared local inductive bias across a variety of network architectures. From this observation, we hypothesize that network denoisers generalize through localized denoising operations, as these operations approximate the training objective well over much of the training distribution. To validate our hypothesis, we introduce novel denoising algorithms which aggregate local empirical denoisers to replicate network behaviour. Comparing these algorithms to network denoisers across forward and reverse diffusion processes, our approach exhibits consistent visual similarity to neural network outputs, with lower mean squared error than previously proposed methods.
Authors: Xi Zhang, Zaiqiao Meng, Jake Lever, Edmond S. L. Ho
Abstract: Radiology report generation (RRG) requires advanced medical image analysis, effective temporal reasoning, and accurate text generation. While multimodal large language models (MLLMs) align with pre-trained vision encoders to enhance visual-language understanding, most existing methods rely on single-image analysis or rule-based heuristics to process multiple images, failing to fully leverage temporal information in multi-modal medical datasets. In this paper, we introduce Libra, a temporal-aware MLLM tailored for chest X-ray report generation. Libra combines a radiology-specific image encoder with a novel Temporal Alignment Connector (TAC), designed to accurately capture and integrate temporal differences between paired current and prior images. Extensive experiments on the MIMIC-CXR dataset demonstrate that Libra establishes a new state-of-the-art benchmark among similarly scaled MLLMs, setting new standards in both clinical relevance and lexical accuracy.
Authors: Hisashi Oshima, Tsuyoshi Ishizone, Tomoyuki Higuchi
Abstract: Recent developments in the unsupervised domain adaptation (UDA) enable the unsupervised machine learning (ML) prediction for target data, thus this will accelerate real world applications with ML models such as image recognition tasks in self-driving. Researchers have reported the UDA techniques are not working well under large co-variate shift problems where e.g. supervised source data consists of handwritten digits data in monotone color and unsupervised target data colored digits data from the street view. Thus there is a need for a method to resolve co-variate shift and transfer source labelling rules under this dynamics. We perform two stages domain invariant representation learning to bridge the gap between source and target with semantic intermediate data (unsupervised). The proposed method can learn domain invariant features simultaneously between source and intermediate also intermediate and target. Finally this achieves good domain invariant representation between source and target plus task discriminability owing to source labels. This induction for the gradient descent search greatly eases learning convergence in terms of classification performance for target data even when large co-variate shift. We also derive a theorem for measuring the gap between trained models and unsupervised target labelling rules, which is necessary for the free parameters optimization. Finally we demonstrate that proposing method is superiority to previous UDA methods using 4 representative ML classification datasets including 38 UDA tasks. Our experiment will be a basis for challenging UDA problems with large co-variate shift.
Authors: Hanqing Zhu, Zhenyu Zhang, Wenyan Cong, Xi Liu, Sem Park, Vikas Chandra, Bo Long, David Z. Pan, Zhangyang Wang, Jinwon Lee
Abstract: Large language models (LLMs) are notoriously memory-intensive during training, particularly with the popular AdamW optimizer. This memory burden necessitates using more or higher-end GPUs or reducing batch sizes, limiting training scalability and throughput. To address this, various memory-efficient optimizers have been proposed to reduce optimizer memory usage. However, they face critical challenges: (i) reliance on costly SVD operations; (ii) significant performance trade-offs compared to AdamW; and (iii) still substantial optimizer memory overhead to maintain competitive performance. In this work, we identify that AdamW's learning rate adaptation rule can be effectively coarsened as a structured learning rate update. Based on this insight, we propose Approximated Gradient Scaling for Memory-Efficient LLM Optimization (APOLLO), which approximates learning rate scaling using an auxiliary low-rank optimizer state based on pure random projection. This structured learning rate update rule makes APOLLO highly tolerant to further memory reductions while delivering comparable pre-training performance. Even its rank-1 variant, APOLLO-Mini, achieves superior pre-training performance compared to AdamW with SGD-level memory costs. Extensive experiments demonstrate that the APOLLO series performs on-par with or better than AdamW, while achieving greater memory savings by nearly eliminating the optimization states of AdamW. These savings provide significant system-level benefits: (1) Enhanced Throughput: 3x throughput on an 8xA100-80GB setup compared to AdamW by supporting 4x larger batch sizes. (2) Improved Model Scalability: Pre-training LLaMA-13B with naive DDP on A100-80GB GPUs without system-level optimizations. (3) Low-End GPU Friendly Pre-training: Pre-training LLaMA-7B on a single GPU using less than 12 GB of memory with weight quantization.
Authors: Shahran Rahman Alve
Abstract: This project aims to develop a robust video surveillance system, which can segment videos into smaller clips based on the detection of activities. It uses CCTV footage, for example, to record only major events-like the appearance of a person or a thief-so that storage is optimized and digital searches are easier. It utilizes the latest techniques in object detection and tracking, including Convolutional Neural Networks (CNNs) like YOLO, SSD, and Faster R-CNN, as well as Recurrent Neural Networks (RNNs) and Long Short-Term Memory networks (LSTMs), to achieve high accuracy in detection and capture temporal dependencies. The approach incorporates adaptive background modeling through Gaussian Mixture Models (GMM) and optical flow methods like Lucas-Kanade to detect motions. Multi-scale and contextual analysis are used to improve detection across different object sizes and environments. A hybrid motion segmentation strategy combines statistical and deep learning models to manage complex movements, while optimizations for real-time processing ensure efficient computation. Tracking methods, such as Kalman Filters and Siamese networks, are employed to maintain smooth tracking even in cases of occlusion. Detection is improved on various-sized objects for multiple scenarios by multi-scale and contextual analysis. Results demonstrate high precision and recall in detecting and tracking objects, with significant improvements in processing times and accuracy due to real-time optimizations and illumination-invariant features. The impact of this research lies in its potential to transform video surveillance, reducing storage requirements and enhancing security through reliable and efficient object detection and tracking.
Authors: Harvey McGuinness, Tianyu Wang, Carey E. Priebe, Hayden Helm
Abstract: Alignment is a social phenomenon wherein individuals share a common goal or perspective. Mirroring, or mimicking the behaviors and opinions of another individual, is one mechanism by which individuals can become aligned. Large scale investigations of the effect of mirroring on alignment have been limited due to the scalability of traditional experimental designs in sociology. In this paper, we introduce a simple computational framework that enables studying the effect of mirroring behavior on alignment in multi-agent systems. We simulate systems of interacting large language models in this framework and characterize overall system behavior and alignment with quantitative measures of agent dynamics. We find that system behavior is strongly influenced by the range of communication of each agent and that these effects are exacerbated by increased rates of mirroring. We discuss the observed simulated system behavior in the context of known human social dynamics.
Authors: Kayne Uriel K. Rodrigo, Jerriane Hillary Heart S. Marcial, Samuel C. Brillo, Khatalyn E. Mata, Jonathan C. Morano
Abstract: This study focuses on enhancing rice leaf disease image classification algorithms, which have traditionally relied on Convolutional Neural Network (CNN) models. We employed transfer learning with MobileViTV2_050 using ImageNet-1k weights, a lightweight model that integrates CNN's local feature extraction with Vision Transformers' global context learning through a separable self-attention mechanism. Our approach resulted in a significant 15.66% improvement in classification accuracy for MobileViTV2_050-A, our first enhanced model trained on the baseline dataset, achieving 93.14%. Furthermore, MobileViTV2_050-B, our second enhanced model trained on a broader rice leaf dataset, demonstrated a 22.12% improvement, reaching 99.6% test accuracy. Additionally, MobileViTV2-A attained an F1-score of 93% across four rice labels and a Receiver Operating Characteristic (ROC) curve ranging from 87% to 97%. In terms of resource consumption, our enhanced models reduced the total parameters of the baseline CNN model by up to 92.50%, from 14 million to 1.1 million. These results indicate that MobileViTV2_050 not only improves computational efficiency through its separable self-attention mechanism but also enhances global context learning. Consequently, it offers a lightweight and robust solution suitable for mobile deployment, advancing the interpretability and practicality of models in precision agriculture.
Authors: Jialiang Cheng, Ning Gao, Yun Yue, Zhiling Ye, Jiadi Jiang, Jian Sha
Abstract: Distributed training methods are crucial for large language models (LLMs). However, existing distributed training methods often suffer from communication bottlenecks, stragglers, and limited elasticity, particularly in heterogeneous or large-scale environments. Local SGD methods have been proposed to address these issues, but their effectiveness remains limited to small-scale training due to additional memory overhead and lack of concerns on efficiency and stability. To tackle these issues, we propose EDiT, an innovative Efficient Distributed Training method that combines a tailored Local SGD approach with model sharding techniques to enhance large-scale training efficiency. EDiT performs layer-wise parameter synchronization during forward pass, reducing communication and memory overhead and enabling overlap. Besides, EDiT employs a pseudo gradient penalty strategy to suppress loss spikes, which ensures training stability and improves performance. Additionally, we introduce A-EDiT, a fully asynchronous variant of EDiT that accommodates heterogeneous clusters. Building on EDiT/A-EDiT, we conduct a series of experiments to validate large-scale asynchronous training for LLMs, accompanied by comprehensive analyses. Experimental results demonstrate the superior performance of EDiT/A-EDiT, establishing them as robust solutions for distributed LLM training in diverse computational ecosystems. The code is available at Atorch codebase: https://github.com/intelligent-machine-learning/atorch/tree/main/atorch/local_sgd.
URLs: https://github.com/intelligent-machine-learning/atorch/tree/main/atorch/local_sgd.
Authors: Jiquan Wang, Sha Zhao, Zhiling Luo, Yangxuan Zhou, Haiteng Jiang, Shijian Li, Tao Li, Gang Pan
Abstract: Electroencephalography (EEG) is a non-invasive technique to measure and record brain electrical activity, widely used in various BCI and healthcare applications. Early EEG decoding methods rely on supervised learning, limited by specific tasks and datasets, hindering model performance and generalizability. With the success of large language models, there is a growing body of studies focusing on EEG foundation models. However, these studies still leave challenges: Firstly, most of existing EEG foundation models employ full EEG modeling strategy. It models the spatial and temporal dependencies between all EEG patches together, but ignores that the spatial and temporal dependencies are heterogeneous due to the unique structural characteristics of EEG signals. Secondly, existing EEG foundation models have limited generalizability on a wide range of downstream BCI tasks due to varying formats of EEG data, making it challenging to adapt to. To address these challenges, we propose a novel foundation model called CBraMod. Specifically, we devise a criss-cross transformer as the backbone to thoroughly leverage the structural characteristics of EEG signals, which can model spatial and temporal dependencies separately through two parallel attention mechanisms. And we utilize an asymmetric conditional positional encoding scheme which can encode positional information of EEG patches and be easily adapted to the EEG with diverse formats. CBraMod is pre-trained on a very large corpus of EEG through patch-based masked EEG reconstruction. We evaluate CBraMod on up to 10 downstream BCI tasks (12 public datasets). CBraMod achieves the state-of-the-art performance across the wide range of tasks, proving its strong capability and generalizability. The source code is publicly available at https://github.com/wjq-learning/CBraMod.
Authors: Chongkai Gao, Haozhuo Zhang, Zhixuan Xu, Zhehao Cai, Lin Shao
Abstract: We aim to develop a model-based planning framework for world models that can be scaled with increasing model and data budgets for general-purpose manipulation tasks with only language and vision inputs. To this end, we present FLow-centric generative Planning (FLIP), a model-based planning algorithm on visual space that features three key modules: 1. a multi-modal flow generation model as the general-purpose action proposal module; 2. a flow-conditioned video generation model as the dynamics module; and 3. a vision-language representation learning model as the value module. Given an initial image and language instruction as the goal, FLIP can progressively search for long-horizon flow and video plans that maximize the discounted return to accomplish the task. FLIP is able to synthesize long-horizon plans across objects, robots, and tasks with image flows as the general action representation, and the dense flow information also provides rich guidance for long-horizon video generation. In addition, the synthesized flow and video plans can guide the training of low-level control policies for robot execution. Experiments on diverse benchmarks demonstrate that FLIP can improve both the success rates and quality of long-horizon video plan synthesis and has the interactive world model property, opening up wider applications for future works.Video demos are on our website: https://nus-lins-lab.github.io/flipweb/.
Authors: Chenghui Yu, Peiyi Li, Haoze Wu, Yiri Wen, Bingfeng Deng, Hongyu Xiong
Abstract: Reducing negative user experiences is essential for the success of recommendation platforms. Exposing users to inappropriate content could not only adversely affect users' psychological well-beings, but also potentially drive users away from the platform, sabotaging the platform's long-term success. However, recommendation algorithms tend to weigh more heavily on positive feedback signals due to the scarcity of negative ones, which may result in the neglect of valuable negative user feedback. In this paper, we propose an approach aimed at limiting negative user experiences. Our method primarily relies on distributing in-feed surveys to the users, modeling the users' feedback collected from the survey, and integrating the model predictions into the recommendation system. We further enhance the baseline survey model by integrating the Learning Hidden Unit Contributions module and the Squeeze-and-Excitation module. In addition, we strive to resolve the problem of response Bias by applying a survey-submit model; The A/B testing results indicate a reduction in survey sexual rate and survey inappropriate rate, ranging from -1.44\% to -3.9\%. Additionally, we compared our methods against an online baseline that does not incorporate our approach. The results indicate that our approach significantly reduces the report rate and dislike rate by 1\% to 2.27\% compared to the baseline, confirming the effectiveness of our methods in enhancing user experience. After we launched the survey model based our approach on our platform, the model is able to bring reductions of 1.75\%, 2.57\%, 2.06\% on reports, dislikes, survey inappropriate rate, respectively.
Authors: Nianze Tao
Abstract: Generating novel molecules with higher properties than the training space, namely the out-of-distribution generation, is important for ${de~novo}$ drug design. However, it is not easy for distribution learning-based models, for example diffusion models, to solve this challenge as these methods are designed to fit the distribution of training data as close as possible. In this paper, we show that Bayesian flow network is capable of effortlessly generating high quality out-of-distribution samples that meet several scenarios. We introduce a semi-autoregressive training/sampling method that helps to enhance the model performance and surpass the state-of-the-art models.
Authors: Yu Yan, Sheng Sun, Junqi Tong, Min Liu, Qi Li
Abstract: Metaphor serves as an implicit approach to convey information, while enabling the generalized comprehension of complex subjects. However, metaphor can potentially be exploited to bypass the safety alignment mechanisms of Large Language Models (LLMs), leading to the theft of harmful knowledge. In our study, we introduce a novel attack framework that exploits the imaginative capacity of LLMs to achieve jailbreaking, the J\underline{\textbf{A}}ilbreak \underline{\textbf{V}}ia \underline{\textbf{A}}dversarial Me\underline{\textbf{TA}} -pho\underline{\textbf{R}} (\textit{AVATAR}). Specifically, to elicit the harmful response, AVATAR extracts harmful entities from a given harmful target and maps them to innocuous adversarial entities based on LLM's imagination. Then, according to these metaphors, the harmful target is nested within human-like interaction for jailbreaking adaptively. Experimental results demonstrate that AVATAR can effectively and transferablly jailbreak LLMs and achieve a state-of-the-art attack success rate across multiple advanced LLMs. Our study exposes a security risk in LLMs from their endogenous imaginative capabilities. Furthermore, the analytical study reveals the vulnerability of LLM to adversarial metaphors and the necessity of developing defense methods against jailbreaking caused by the adversarial metaphor. \textcolor{orange}{ \textbf{Warning: This paper contains potentially harmful content from LLMs.}}
Authors: Hongbin Zhang, Kehai Chen, Xuefeng Bai, Yang Xiang, Min Zhang
Abstract: Large language models (LLMs) have exhibited impressive multilingual reasoning capabilities, driven by extensive multilingual pre-training corpora and instruction fine-tuning data. However, a performance gap exists between high- and low-resource language reasoning tasks due to the language imbalance in the pre-training corpus, which is exacerbated by evaluation bias in existing reasoning benchmarks lacking low-resource language coverage. To alleviate this issue, we propose LinguaLIFT, a two-stage instruction tuning framework for advancing low-resource language reasoning. LinguaLIFT employs a language alignment layer to capture multilingual alignment in a code-switched tuning way without requiring multilingual instruction or parallel data, thereby transferring the cross-lingual reasoning capabilities to low-resource languages through English-only instruction tuning data. To comprehensively evaluate the multilingual reasoning capabilities, we introduce the Multilingual Math World Problem (MMWP) benchmark, which spans 21 low-resource, 17 medium-resource, and 10 high-resource languages. Experimental results show that LinguaLIFT outperforms several competitive baselines across MMWP and four widely used benchmarks.
Authors: Wenyu Zhang, Wei En Ng, Lixin Ma, Yuwen Wang, Jungqi Zhao, Allison Koenecke, Boyang Li, Lu Wang
Abstract: Current vision-language models may grasp basic spatial cues and simple directions (e.g. left, right, front, back), but struggle with the multi-dimensional spatial reasoning necessary for human-like understanding and real-world applications. To address this gap, we develop SPHERE (Spatial Perception and Hierarchical Evaluation of REasoning), a hierarchical evaluation framework supported by a new human-annotated dataset. SPHERE systematically probes models across increasing levels of complexity, from fundamental skills to multi-skill integration and high-level reasoning that combines spatial, visual, and logical understanding. Benchmark evaluation of state-of-the-art models reveals significant deficiencies, especially in reasoning about distance and proximity, understanding both egocentric and allocentric perspectives, and applying spatial logic in physical contexts. These findings expose critical blind spots in existing models and underscore the need for more advanced spatial reasoning techniques, driving the development of vision-language models that align more closely with human spatial cognition. The dataset will be open-sourced upon publication.
Authors: Zihui Cheng, Qiguang Chen, Jin Zhang, Hao Fei, Xiaocheng Feng, Wanxiang Che, Min Li, Libo Qin
Abstract: Large Vision-Language Models (LVLMs) have recently demonstrated amazing success in multi-modal tasks, including advancements in Multi-modal Chain-of-Thought (MCoT) reasoning. Despite these successes, current benchmarks still follow a traditional paradigm with multi-modal input and text-modal output, which leads to significant drawbacks such as missing visual operations and vague expressions. Motivated by this, we introduce a novel Chain of Multi-modal Thought (CoMT) benchmark to address these limitations. Different from the traditional MCoT benchmark, CoMT requires both multi-modal input and multi-modal reasoning output, aiming to mimic human-like reasoning that inherently integrates visual operation. Specifically, CoMT consists of four categories: (1) Visual Creation, (2) Visual Deletion, (3) Visual Update, and (4) Visual Selection to comprehensively explore complex visual operations and concise expression in real scenarios. We evaluate various LVLMs and strategies on CoMT, revealing some key insights into the capabilities and limitations of the current approaches. We hope that CoMT can inspire more research on introducing multi-modal generation into the reasoning process.
Authors: Xi Ding, Lei Wang
Abstract: Large language models (LLMs) have revolutionized video-based computer vision applications, including action recognition, anomaly detection, and video summarization. Videos inherently pose unique challenges, combining spatial complexity with temporal dynamics that are absent in static images or textual data. Current approaches to video understanding with LLMs often rely on pretrained video encoders to extract spatiotemporal features and text encoders to capture semantic meaning. These representations are integrated within LLM frameworks, enabling multimodal reasoning across diverse video tasks. However, the critical question persists: Can LLMs truly understand the concept of time, and how effectively can they reason about temporal relationships in videos? This work critically examines the role of LLMs in video processing, with a specific focus on their temporal reasoning capabilities. We identify key limitations in the interaction between LLMs and pretrained encoders, revealing gaps in their ability to model long-term dependencies and abstract temporal concepts such as causality and event progression. Furthermore, we analyze challenges posed by existing video datasets, including biases, lack of temporal annotations, and domain-specific limitations that constrain the temporal understanding of LLMs. To address these gaps, we explore promising future directions, including the co-evolution of LLMs and encoders, the development of enriched datasets with explicit temporal labels, and innovative architectures for integrating spatial, temporal, and semantic reasoning. By addressing these challenges, we aim to advance the temporal comprehension of LLMs, unlocking their full potential in video analysis and beyond. Our paper's GitHub repository can be found at https://github.com/Darcyddx/Video-LLM.
Authors: Lo Pang-Yun Ting, Ali \c{S}enol, Huan-Yang Wang, Hsu-Chao Lai, Kun-Ta Chuang, Huan Liu
Abstract: The advanced bidirectional EV charging and discharging technology, aimed at supporting grid stability and emergency operations, has driven a growing interest in workplace applications. It not only reduces electricity expenses but also enhances the resilience in handling practical matters, such as peak power limitation, fluctuating energy prices, and unpredictable EV departures. Considering these factors systematically can benefit energy efficiency in office buildings and for EV users simultaneously. To employ AI to address these issues, we propose HUCA, a novel real-time charging control for regulating energy demands for both the building and EVs. HUCA employs hierarchical actor-critic networks to dynamically reduce electricity costs in buildings, accounting for the needs of EV charging in the dynamic pricing scenario. To tackle the uncertain EV departures, we introduce a new critic augmentation to account for departure uncertainties in evaluating the charging decisions, while maintaining the robustness of the charging control. Experiments on real-world electricity datasets under both simulated certain and uncertain departure scenarios demonstrate that HUCA outperforms baselines in terms of total electricity costs while maintaining competitive performance in fulfilling EV charging requirements. A case study also manifests that HUCA effectively balances energy supply between the building and EVs based on real-time information, showcasing its potential as a key AI-driven solution for vehicle charging control.
Authors: Xin Zhao, Zehui Jiang, Naoki Yoshinaga
Abstract: Although feed-forward neurons in pre-trained language models (PLMs) can store knowledge and their importance in influencing model outputs has been studied, existing work focuses on finding a limited set of neurons and analyzing their relative importance. However, the global quantitative role of activation values in shaping outputs remains unclear, hindering further advancements in applications like knowledge editing. Our study first investigates the numerical relationship between neuron activations and model output and discovers the global linear relationship between them through neuron interventions on a knowledge probing dataset. We refer to the gradient of this linear relationship as neuron empirical gradient (NEG), and introduce NeurGrad, an accurate and efficient method for computing NEG. NeurGrad enables quantitative analysis of all neurons in PLMs, advancing our understanding of neurons' controllability. Furthermore, we explore NEG's ability to represent language skills across diverse prompts via skill neuron probing. Experiments on MCEval8k, a multi-choice knowledge benchmark spanning various genres, validate NEG's representational ability. The data and code are released.
Authors: Tingxu Han, Zhenting Wang, Chunrong Fang, Shiyu Zhao, Shiqing Ma, Zhenyu Chen
Abstract: Reasoning is critical for large language models (LLMs) to excel in a wide range of tasks. While methods like Chain-of-Thought (CoT) reasoning enhance LLM performance by decomposing problems into intermediate steps, they also incur significant overhead in token usage, leading to increased costs. We find that the reasoning process of current LLMs is unnecessarily lengthy and it can be compressed by including a reasonable token budget in the prompt, but the choice of token budget plays a crucial role in the actual compression effectiveness. We then propose a token-budget-aware LLM reasoning framework, which dynamically estimates token budgets for different problems based on reasoning complexity and uses the estimated token budgets to guide the reasoning process. Experiments show that our method effectively reduces token costs in CoT reasoning with only a slight performance reduction, offering a practical solution to balance efficiency and accuracy in LLM reasoning. Code: https://github.com/GeniusHTX/TALE.
Authors: Dongwei Sun, Jing Yao, Changsheng Zhou, Xiangyong Cao, Pedram Ghamisi
Abstract: Remote sensing image change description represents an innovative multimodal task within the realm of remote sensing processing. This task not only facilitates the detection of alterations in surface conditions, but also provides comprehensive descriptions of these changes, thereby improving human interpretability and interactivity.Generally, existing deep-learning-based methods predominantly utilized a three-stage framework that successively perform feature extraction, feature fusion, and localization from bitemporal images before text generation. However, this reliance often leads to an excessive focus on the design of specific network architectures and restricts the feature distributions to the dataset at hand, which in turn results in limited generalizability and robustness during application.To address these limitations, this paper proposes a novel approach for remote sensing image change detection and description that incorporates diffusion models, aiming to transition the emphasis of modeling paradigms from conventional feature learning to data distribution learning. The proposed method primarily includes a simple multi-scale change detection module, whose output features are subsequently refined by an well-designed diffusion model. Furthermore, we introduce a frequency-guided complex filter module to boost the model performance by managing high-frequency noise throughout the diffusion process. We validate the effectiveness of our proposed method across several datasets for remote sensing change detection and description, showcasing its superior performance compared to existing techniques. The code will be available at \href{https://github.com/sundongwei}{MaskApproxNet} after a possible publication.
Authors: Huan-Hsin Tseng, Hsin-Yi Lin, Samuel Yen-Chi Chen, Shinjae Yoo
Abstract: This work analyzes transfer learning of the Variational Quantum Circuit (VQC). Our framework begins with a pretrained VQC configured in one domain and calculates the transition of 1-parameter unitary subgroups required for a new domain. A formalism is established to investigate the adaptability and capability of a VQC under the analysis of loss bounds. Our theory observes knowledge transfer in VQCs and provides a heuristic interpretation for the mechanism. An analytical fine-tuning method is derived to attain the optimal transition for adaptations of similar domains.
Authors: Kangcheng Luo, Quzhe Huang, Cong Jiang, Yansong Feng
Abstract: Legal articles often include vague concepts for adapting to the ever-changing society. Providing detailed interpretations of these concepts is a critical and challenging task even for legal practitioners. It requires meticulous and professional annotations and summarizations by legal experts, which are admittedly time-consuming and expensive to collect at scale. By emulating legal experts' doctrinal method, we introduce a novel framework, ATRIE, using large language models (LLMs) to AuTomatically Retrieve concept-related information, Interpret legal concepts, and Evaluate generated interpretations, eliminating dependence on legal experts. ATRIE comprises a legal concept interpreter and a legal concept interpretation evaluator. The interpreter uses LLMs to retrieve relevant information from judicial precedents and interpret legal concepts. The evaluator uses performance changes on legal concept entailment, a downstream task we propose, as a proxy of interpretation quality. Automatic and multifaceted human evaluations indicate that the quality of our interpretations is comparable to those written by legal experts, with superior comprehensiveness and readability. Although there remains a slight gap in accuracy, it can already assist legal practitioners in improving the efficiency of concept interpretation.
Authors: Jushang Qiu, Lei Wang
Abstract: Skeleton-based action recognition has gained significant attention for its ability to efficiently represent spatiotemporal information in a lightweight format. Most existing approaches use graph-based models to process skeleton sequences, where each pose is represented as a skeletal graph structured around human physical connectivity. Among these, the Spatiotemporal Graph Convolutional Network (ST-GCN) has become a widely used framework. Alternatively, hypergraph-based models, such as the Hyperformer, capture higher-order correlations, offering a more expressive representation of complex joint interactions. A recent advancement, termed Taylor Videos, introduces motion-enhanced skeleton sequences by embedding motion concepts, providing a fresh perspective on interpreting human actions in skeleton-based action recognition. In this paper, we conduct a comprehensive evaluation of both traditional skeleton sequences and Taylor-transformed skeletons using ST-GCN and Hyperformer models on the NTU-60 and NTU-120 datasets. We compare skeletal graph and hypergraph representations, analyzing static poses against motion-injected poses. Our findings highlight the strengths and limitations of Taylor-transformed skeletons, demonstrating their potential to enhance motion dynamics while exposing current challenges in fully using their benefits. This study underscores the need for innovative skeletal modelling techniques to effectively handle motion-rich data and advance the field of action recognition.
Authors: Zhen Li, Yupeng Su, Runming Yang, Congkai Xie, Zheng Wang, Zhongwei Xie, Ngai Wong, Hongxia Yang
Abstract: Large language models have achieved significant advancements in complex mathematical reasoning benchmarks, such as MATH. However, their substantial computational requirements present challenges for practical deployment. Model quantization has emerged as an effective strategy to reduce memory usage and computational costs by employing lower precision and bit-width representations. In this study, we systematically evaluate the impact of quantization on mathematical reasoning tasks. Our results demonstrate that aggressive quantization methods like AWQ and GPTQ introduce up to 32.39% accuracy degradation (average 11.31%) on Llama-3 models, particularly in numerical computation and reasoning planning. To address this, we introduce a multidimensional evaluation framework combining qualitative capability analysis and quantitative error assessment. We further develop targeted recovery strategies, showing that fine-tuning quantized models on only 545 task-specific examples for 3 minutes on 4 GPUs effectively restores reasoning capabilities to near full-precision levels. Additionally, our error assessment pipeline achieves 98.9% accuracy in diagnosing and localizing errors across 3,366 failure cases, providing actionable insights for mitigating quantization-induced degradation.
Authors: Beichen Zhang, Yuhong Liu, Xiaoyi Dong, Yuhang Zang, Pan Zhang, Haodong Duan, Yuhang Cao, Dahua Lin, Jiaqi Wang
Abstract: Large language models (LLMs) have demonstrated impressive ability in solving complex mathematical problems with multi-step reasoning and can be further enhanced with well-designed in-context learning (ICL) examples. However, this potential is often constrained by two major challenges in ICL: granularity mismatch and irrelevant information. We observe that while LLMs excel at decomposing mathematical problems, they often struggle with reasoning errors in fine-grained steps. Moreover, ICL examples retrieved at the question level may omit critical steps or even mislead the model with irrelevant details. To address this issue, we propose BoostStep, a method that enhances reasoning accuracy through step-aligned ICL, a novel mechanism that carefully aligns retrieved reference steps with the corresponding reasoning steps. Additionally, BoostStep incorporates an effective "first-try" strategy to deliver exemplars highly relevant to the current state of reasoning. BoostStep is a flexible and powerful method that integrates seamlessly with chain-of-thought (CoT) and tree search algorithms, refining both candidate selection and decision-making. Empirical results show that BoostStep improves GPT-4o's CoT performance by 4.6% across mathematical benchmarks, significantly surpassing traditional few-shot learning's 1.2%. Moreover, it can achieve an additional 7.5\% gain combined with tree search. Surprisingly, it enhances state-of-the-art LLMs to solve challenging math problems using simpler examples. It improves DeepSeek-R1-671B's performance on AIME by 2.2%, leveraging simple examples only from the MATH dataset.
Authors: Bahman Torkamandi
Abstract: In the realm of fractal geometry, intricate structures emerge from simple iterative processes that partition parameter spaces into regions of stability and instability. Likewise, training large language models involves iteratively applying update functions, such as Adam, where even slight hyperparameter adjustments can shift the training process from convergence to divergence. Recent evidence from miniature neural networks suggests that the boundary separating these outcomes displays fractal characteristics. Building on these insights, this study extends them to medium-sized, decoder-only transformer architectures by employing a more consistent convergence measure and examining the learning rate hyperparameter landscape for attention and fully connected layers. The results show that the trainability frontier is not a simple threshold; rather, it forms a self-similar yet seemingly random structure at multiple scales, with statistically consistent and repeating patterns. Within this landscape, a region of stable convergence is surrounded by a complex chaotic border, illustrating the sensitive nature of the underlying training dynamics.
Authors: Qingyu Ren, Jie Zeng, Qianyu He, Jiaqing Liang, Yanghua Xiao, Weikang Zhou, Zeye Sun, Fei Yu
Abstract: It is crucial for large language models (LLMs) to follow instructions that involve multiple constraints. However, it is an unexplored area to enhance LLMs' ability to follow soft constraints. To bridge the gap, we initially design a pipeline to construct datasets with high-quality outputs automatically. Additionally, to fully utilize the positive and negative samples generated during the data construction process, we choose Direct Preference Optimization (DPO) as the training method. Furthermore, taking into account the difficulty of soft constraints indicated by the number of constraints, we design a curriculum learning training paradigm based on the constraint quantity. We experimentally evaluate the effectiveness of our methods in improving LLMs' soft constraint following ability and analyze the factors driving the improvements.The datasets and code are publicly available at https://github.com/Rainier-rq/FollowSoftConstraint.
Authors: Joonho Ko, Jinheon Baek, Sung Ju Hwang
Abstract: Large language models (LLMs) have shown remarkable performance across a wide range of natural language tasks. However, a critical challenge remains in that they sometimes generate factually incorrect answers. To address this, while many previous work has focused on identifying errors in their generation and further refining them, they are slow in deployment since they are designed to verify the response from LLMs only after their entire generation (from the first to last tokens) is done. Further, we observe that once LLMs generate incorrect tokens early on, there is a higher likelihood that subsequent tokens will also be factually incorrect. To this end, in this work, we propose Streaming-VR (Streaming Verification and Refinement), a novel approach designed to enhance the efficiency of verification and refinement of LLM outputs. Specifically, the proposed Streaming-VR enables on-the-fly verification and correction of tokens as they are being generated, similar to a streaming process, ensuring that each subset of tokens is checked and refined in real-time by another LLM as the LLM constructs its response. Through comprehensive evaluations on multiple datasets, we demonstrate that our approach not only enhances the factual accuracy of LLMs, but also offers a more efficient solution compared to prior refinement methods.
Authors: Aishwarya Jadhav, Jeffery Cao, Abhishree Shetty, Urvashi Priyam Kumar, Aditi Sharma, Ben Sukboontip, Jayant Sravan Tamarapalli, Jingyi Zhang, Anirudh Koul
Abstract: This paper presents AI Guide Dog (AIGD), a lightweight egocentric (first-person) navigation system for visually impaired users, designed for real-time deployment on smartphones. AIGD employs a vision-only multi-label classification approach to predict directional commands, ensuring safe navigation across diverse environments. We introduce a novel technique for goal-based outdoor navigation by integrating GPS signals and high-level directions, while also handling uncertain multi-path predictions for destination-free indoor navigation. As the first navigation assistance system to handle both goal-oriented and exploratory navigation across indoor and outdoor settings, AIGD establishes a new benchmark in blind navigation. We present methods, datasets, evaluations, and deployment insights to encourage further innovations in assistive navigation systems.
Authors: Yirong Zeng, Xiao Ding, Yuxian Wang, Weiwen Liu, Wu Ning, Yutai Hou, Xu Huang, Bing Qin, Ting Liu
Abstract: Augmenting large language models (LLMs) with external tools is known as a promising approach to enhancing their capabilities, especially for complex tasks. Synthesizing tool-use data through real-world simulations is an effective way to achieve it. Nevertheless, our investigation reveals that (1) training gains significantly decay as synthetic data increases. The model struggles to benefit from more synthetic data due to potential data diversity issues, resulting in poor performance in complex scenarios. Moreover, we find that (2) this challenge primarily manifests as minor discrepancies between the model's output and the ground truth response (termed as deficiency), such as errors in parameter values that require complex reasoning from the context to resolve. To this end, we propose an iterative reinforced fine-tuning strategy designed to alleviate these challenges. This strategy involves: (1) enhancing the diversity of synthetic data through path exploration of Monte Carlo Tree Search. (2) iteratively identifying deficiency-related data, constructing fine-grained preference pairs to pinpoint deficiencies, and then applying preference optimization to optimize these deficiencies. Our experiments show that models trained using our method achieve about 3\% better performance than same-size models, outperforming larger open-source and closed-source models.
Authors: Jialun Cao, Yuk-Kit Chan, Zixuan Ling, Wenxuan Wang, Shuqing Li, Mingwei Liu, Ruixi Qiao, Yuting Han, Chaozheng Wang, Boxi Yu, Pinjia He, Shuai Wang, Zibin Zheng, Michael R. Lyu, Shing-Chi Cheung
Abstract: Various benchmarks have been proposed to assess the performance of large language models (LLMs) in different coding scenarios. We refer to them as code-related benchmarks. However, there are no systematic guidelines by which such a benchmark should be developed to ensure its quality, reliability, and reproducibility. We propose How2Bench, which is comprised of a 55-criteria checklist as a set of guidelines to govern the development of code-related benchmarks comprehensively. Using HOW2BENCH, we profiled 274 benchmarks released within the past decade and found concerning issues. Nearly 70% of the benchmarks did not take measures for data quality assurance; over 10% did not even open source or only partially open source. Many highly cited benchmarks have loopholes, including duplicated samples, incorrect reference codes/tests/prompts, and unremoved sensitive/confidential information. Finally, we conducted a human study involving 49 participants, which revealed significant gaps in awareness of the importance of data quality, reproducibility, and transparency.
Authors: Jing Liu, Zhenchao Ma, Zepu Wang, Chenxuanyin Zou, Jiayang Ren, Zehua Wang, Liang Song, Bo Hu, Yang Liu, Victor C. M. Leung
Abstract: Diffusion models (DMs) have emerged as a powerful class of generative AI models, showing remarkable potential in anomaly detection (AD) tasks across various domains, such as cybersecurity, fraud detection, healthcare, and manufacturing. The intersection of these two fields, termed diffusion models for anomaly detection (DMAD), offers promising solutions for identifying deviations in increasingly complex and high-dimensional data. In this survey, we review recent advances in DMAD research. We begin by presenting the fundamental concepts of AD and DMs, followed by a comprehensive analysis of classic DM architectures including DDPMs, DDIMs, and Score SDEs. We further categorize existing DMAD methods into reconstruction-based, density-based, and hybrid approaches, providing detailed examinations of their methodological innovations. We also explore the diverse tasks across different data modalities, encompassing image, time series, video, and multimodal data analysis. Furthermore, we discuss critical challenges and emerging research directions, including computational efficiency, model interpretability, robustness enhancement, edge-cloud collaboration, and integration with large language models. The collection of DMAD research papers and resources is available at https://github.com/fdjingliu/DMAD.
Authors: Zijin Hong, Hao Wu, Su Dong, Junnan Dong, Yilin Xiao, Yujing Zhang, Zhu Wang, Feiran Huang, Linyi Li, Hongxia Yang, Xiao Huang
Abstract: Recent studies have raised concerns about the reliability of current mathematical benchmarks, highlighting issues such as simplistic design and potential data contamination. Therefore, creating a reliable benchmark that effectively evaluates the genuine capabilities of large language models (LLMs) in mathematical reasoning remains a significant challenge. To address this, we propose RV-Bench, a framework for Benchmarking LLMs via Random Variables in mathematical reasoning. Specifically, the background content of a random variable question (RV question) mirrors the original problem in existing benchmarks, but the variable combinations are randomized, making it "unseen" by the LLMs. Models must completely understand the question pattern of the original problem to correctly answer RV questions with various variable values. As a result, the LLM's genuine capability in mathematical reasoning is reflected by its accuracy and robustness on RV-Bench. We conducted extensive experiments on over 30 representative LLMs across more than 1000 RV questions. Our findings suggest that LLMs exhibit an imbalance in proficiency between encountered and "unseen" data domains. Proficiency generalization across similar mathematical reasoning tasks is verified to be limited by accuracy and robustness, but it can still be enhanced through test-time scaling.
Authors: Xurui Song, Zhixin Xie, Shuo Huai, Jiayi Kong, Jun Luo
Abstract: The wide adoption of Large Language Models (LLMs) has attracted significant attention from $\textit{jailbreak}$ attacks, where adversarial prompts crafted through optimization or manual design exploit LLMs to generate malicious contents. However, optimization-based attacks have limited efficiency and transferability, while existing manual designs are either easily detectable or demand intricate interactions with LLMs. In this paper, we first point out a novel perspective for jailbreak attacks: LLMs are more responsive to $\textit{positive}$ prompts. Based on this, we deploy Happy Ending Attack (HEA) to wrap up a malicious request in a scenario template involving a positive prompt formed mainly via a $\textit{happy ending}$, it thus fools LLMs into jailbreaking either immediately or at a follow-up malicious request.This has made HEA both efficient and effective, as it requires only up to two turns to fully jailbreak LLMs. Extensive experiments show that our HEA can successfully jailbreak on state-of-the-art LLMs, including GPT-4o, Llama3-70b, Gemini-pro, and achieves 88.79\% attack success rate on average. We also provide quantitative explanations for the success of HEA.
Authors: Xuemiao Zhang, Liangyu Xu, Feiyu Duan, Yongwei Zhou, Sirui Wang, Rongxiang Weng, Jingang Wang, Xunliang Cai
Abstract: Large language models (LLMs) generally utilize a consistent data distribution throughout the pretraining process. However, as the model's capability improves, it is intuitive that its data preferences dynamically change, indicating the need for pretraining with different data at various training stages. To achieve it, we propose the Perplexity Difference (PD) based Preference Curriculum learning (PDPC) framework, which always perceives and uses the data preferred by LLMs to train and boost them. First, we introduce the PD metric to quantify the difference in how challenging a sample is for weak versus strong models. Samples with high PD are more challenging for weak models to learn and are more suitable to be arranged in the later stage of pretraining. Second, we propose the preference function to approximate and predict the data preference of the LLM at any training step, so as to complete the arrangement of the dataset offline and ensure continuous training without interruption. Experimental results on 1.3B and 3B models demonstrate that PDPC significantly surpasses baselines. Notably, the 3B model trained on 1T tokens achieves an increased average accuracy of over 8.1% across MMLU and CMMLU.
Authors: Shezheng Song, Hao Xu, Jun Ma, Shasha Li, Long Peng, Qian Wan, Xiaodong Liu, Jie Yu
Abstract: Large Language Models (LLMs) exhibit strong general language capabilities. However, fine-tuning these models on domain-specific tasks often leads to catastrophic forgetting, where the model overwrites or loses essential knowledge acquired during pretraining. This phenomenon significantly limits the broader applicability of LLMs. To address this challenge, we propose a novel approach to compute the element-wise importance of model parameters crucial for preserving general knowledge during fine-tuning. Our method utilizes a dual-objective optimization strategy: (1) regularization loss based on element-wise parameter importance, which constrains the updates to parameters crucial for general knowledge; (2) cross-entropy loss to adapt to domain-specific tasks. Additionally, we introduce layer-wise coefficients to account for the varying contributions of different layers, dynamically balancing the dual-objective optimization. Extensive experiments on scientific, medical, and physical tasks using GPT-J and LLaMA-3 demonstrate that our approach mitigates catastrophic forgetting while enhancing model adaptability. Compared to previous methods, our solution is approximately 20 times faster and requires only 10-15% of the storage, highlighting the practical efficiency. The code will be released.
Authors: Libo Wang
Abstract: In view of the gap in the current large language model in sharing memory across dialogues, this research proposes a wormhole memory module (WMM) to realize memory as a Rubik's cube that can be arbitrarily retrieved between different dialogues. Through simulation experiments, the researcher built an experimental framework based on the Python environment and used setting memory barriers to simulate the current situation where memories between LLMs dialogues are difficult to share. The CoQA development data set was imported into the experiment, and the feasibility of its cross-dialogue memory retrieval function was verified for WMM's nonlinear indexing and dynamic retrieval, and a comparative analysis was conducted with the capabilities of Titans and MemGPT memory modules. Experimental results show that WMM demonstrated the ability to retrieve memory across dialogues and the stability of quantitative indicators in eight experiments. It contributes new technical approaches to the optimization of memory management of LLMs and provides experience for the practical application in the future.
Authors: Zhenhao Zhu, Bulou Liu, Qingyao Ai, Yiqun Liu
Abstract: Multiple choice questions (MCQs) are a popular and important task for evaluating large language models (LLMs). Based on common strategies people use when answering MCQs, the process of elimination (PoE) has been proposed as an effective problem-solving method. Existing methods to the PoE generally fall into two categories: one involves having the LLM directly select the incorrect options, while the other involves scoring the options. However, both methods incur high computational costs and often perform worse than methods that directly answer the MCQs with the option IDs. To address this issue, this paper proposes a PoE based on option ID. Specifically, our method eliminates option by selecting the option ID with the lowest probability. We conduct experiments with 10 different LLMs in zero-shot settings on 7 publicly available datasets. The experimental results demonstrate that our method significantly improves the LLM's performance. Further analysis reveals that the sequential elimination strategy can effectively enhance the LLM's reasoning ability. Additionally, we find that sequential elimination is also applicable to few-shot settings and can be combined with debias methods to further improve LLM's performance.
Authors: Chuanyang Zheng
Abstract: We present a new family of mobile hybrid vision networks, called iFormer, with a focus on optimizing latency and accuracy on mobile applications. iFormer effectively integrates the fast local representation capacity of convolution with the efficient global modeling ability of self-attention. The local interactions are derived from transforming a standard convolutional network, \textit{i.e.}, ConvNeXt, to design a more lightweight mobile network. Our newly introduced mobile modulation attention removes memory-intensive operations in MHA and employs an efficient modulation mechanism to boost dynamic global representational capacity. We conduct comprehensive experiments demonstrating that iFormer outperforms existing lightweight networks across various tasks. Notably, iFormer achieves an impressive Top-1 accuracy of 80.4\% on ImageNet-1k with a latency of only 1.10 ms on an iPhone 13, surpassing the recently proposed MobileNetV4 under similar latency constraints. Additionally, our method shows significant improvements in downstream tasks, including COCO object detection, instance segmentation, and ADE20k semantic segmentation, while still maintaining low latency on mobile devices for high-resolution inputs in these scenarios.
Authors: Rajath Rao, Adithya Ganesan, Oscar Kjell, Jonah Luby, Akshay Raghavan, Scott Feltman, Whitney Ringwald, Ryan L. Boyd, Benjamin Luft, Camilo Ruggero, Neville Ryant, Roman Kotov, H. Andrew Schwartz
Abstract: Current speech encoding pipelines often rely on an additional text-based LM to get robust representations of human communication, even though SotA speech-to-text models often have a LM within. This work proposes an approach to improve the LM within an audio model such that the subsequent text-LM is unnecessary. We introduce WhiSPA (Whisper with Semantic and Psychological Alignment), which leverages a novel audio training objective: contrastive loss with a language model embedding as a teacher. Using over 500k speech segments from mental health audio interviews, we evaluate the utility of aligning Whisper's latent space with semantic representations from a text autoencoder (SBERT) and lexically derived embeddings of basic psychological dimensions: emotion and personality. Over self-supervised affective tasks and downstream psychological tasks, WhiSPA surpasses current speech encoders, achieving an average error reduction of 73.4% and 83.8%, respectively. WhiSPA demonstrates that it is not always necessary to run a subsequent text LM on speech-to-text output in order to get a rich psychological representation of human communication.
Authors: Hao Dong, Moru Liu, Kaiyang Zhou, Eleni Chatzi, Juho Kannala, Cyrill Stachniss, Olga Fink
Abstract: In real-world scenarios, achieving domain adaptation and generalization poses significant challenges, as models must adapt to or generalize across unknown target distributions. Extending these capabilities to unseen multimodal distributions, i.e., multimodal domain adaptation and generalization, is even more challenging due to the distinct characteristics of different modalities. Significant progress has been made over the years, with applications ranging from action recognition to semantic segmentation. Besides, the recent advent of large-scale pre-trained multimodal foundation models, such as CLIP, has inspired works leveraging these models to enhance adaptation and generalization performances or adapting them to downstream tasks. This survey provides the first comprehensive review of recent advances from traditional approaches to foundation models, covering: (1) Multimodal domain adaptation; (2) Multimodal test-time adaptation; (3) Multimodal domain generalization; (4) Domain adaptation and generalization with the help of multimodal foundation models; and (5) Adaptation of multimodal foundation models. For each topic, we formally define the problem and thoroughly review existing methods. Additionally, we analyze relevant datasets and applications, highlighting open challenges and potential future research directions. We maintain an active repository that contains up-to-date literature at https://github.com/donghao51/Awesome-Multimodal-Adaptation.
URLs: https://github.com/donghao51/Awesome-Multimodal-Adaptation.
Authors: Pratik Shrestha, Sujan Kapali, Swikar Gautam, Vishal Pokharel, Santosh Giri
Abstract: This paper introduces a mobile-based solution that enhances online shoe shopping through 3D modeling and Augmented Reality (AR), leveraging the efficiency of 3D Gaussian Splatting. Addressing the limitations of static 2D images, the framework generates realistic 3D shoe models from 2D images, achieving an average Peak Signal-to-Noise Ratio (PSNR) of 32, and enables immersive AR interactions via smartphones. A custom shoe segmentation dataset of 3120 images was created, with the best-performing segmentation model achieving an Intersection over Union (IoU) score of 0.95. This paper demonstrates the potential of 3D modeling and AR to revolutionize online shopping by offering realistic virtual interactions, with applicability across broader fashion categories.
Authors: Ting-Yao E. Hsu, Yi-Li Hsu, Shaurya Rohatgi, Chieh-Yang Huang, Ho Yin Sam Ng, Ryan Rossi, Sungchul Kim, Tong Yu, Lun-Wei Ku, C. Lee Giles, Ting-Hao K. Huang
Abstract: Since the SCICAP datasets launch in 2021, the research community has made significant progress in generating captions for scientific figures in scholarly articles. In 2023, the first SCICAP Challenge took place, inviting global teams to use an expanded SCICAP dataset to develop models for captioning diverse figure types across various academic fields. At the same time, text generation models advanced quickly, with many powerful pre-trained large multimodal models (LMMs) emerging that showed impressive capabilities in various vision-and-language tasks. This paper presents an overview of the first SCICAP Challenge and details the performance of various models on its data, capturing a snapshot of the fields state. We found that professional editors overwhelmingly preferred figure captions generated by GPT-4V over those from all other models and even the original captions written by authors. Following this key finding, we conducted detailed analyses to answer this question: Have advanced LMMs solved the task of generating captions for scientific figures?
Authors: Harshith Padigela, Chintan Shah, Dinkar Juyal
Abstract: In this report, we present ML-Dev-Bench, a benchmark aimed at testing agentic capabilities on applied Machine Learning development tasks. While existing benchmarks focus on isolated coding tasks or Kaggle-style competitions, ML-Dev-Bench tests agents' ability to handle the full complexity of ML development workflows. The benchmark assesses performance across critical aspects including dataset handling, model training, improving existing models, debugging, and API integration with popular ML tools. We evaluate three agents -- ReAct, Openhands, and AIDE -- on a diverse set of 30 tasks, providing insights into their strengths and limitations in handling practical ML development challenges.
Authors: Yuhang Zhou, Giannis Karamanolakis, Victor Soto, Anna Rumshisky, Mayank Kulkarni, Furong Huang, Wei Ai, Jianhua Lu
Abstract: The recent success of specialized Large Language Models (LLMs) in domains such as mathematical reasoning and coding has led to growing interest in methods for merging these expert LLMs into a unified Mixture-of-Experts (MoE) model, with the goal of enhancing performance in each domain while retaining effectiveness on general tasks. However, the effective merging of expert models remains an open challenge, especially for models with highly divergent weight parameters or different architectures. State-of-the-art MoE merging methods only work with homogeneous model architectures and rely on simple unweighted averaging to merge expert layers, which does not address parameter interference and requires extensive fine-tuning of the merged MoE to restore performance. To address these limitations, this paper introduces new MoE merging techniques, including strategies to mitigate parameter interference, routing heuristics to reduce the need for MoE fine-tuning, and a novel method for merging experts with different architectures. Extensive experiments across multiple domains demonstrate the effectiveness of our proposed methods, reducing fine-tuning costs, improving performance over state-of-the-art methods, and expanding the applicability of MoE merging.
Authors: Lifan Jiang, Shuang Chen, Boxi Wu, Xiaotong Guan, Jiahui Zhang
Abstract: With the advancement of generative artificial intelligence, previous studies have achieved the task of generating aesthetic images from hand-drawn sketches, fulfilling the public's needs for drawing. However, these methods are limited to static images and lack the ability to control video animation generation using hand-drawn sketches. To address this gap, we propose VidSketch, the first method capable of generating high-quality video animations directly from any number of hand-drawn sketches and simple text prompts, bridging the divide between ordinary users and professional artists. Specifically, our method introduces a Level-Based Sketch Control Strategy to automatically adjust the guidance strength of sketches during the generation process, accommodating users with varying drawing skills. Furthermore, a TempSpatial Attention mechanism is designed to enhance the spatiotemporal consistency of generated video animations, significantly improving the coherence across frames. You can find more detailed cases on our official website.
Authors: Anneketh Vij, Changhao Liu, Rahul Anil Nair, Theodore Eugene Ho, Edward Shi, Ayan Bhowmick
Abstract: This research presents an exploration and study of the recipe generation task by fine-tuning various very small language models, with a focus on developing robust evaluation metrics and comparing across different language models the open-ended task of recipe generation. This study presents extensive experiments with multiple model architectures, ranging from T5-small (Raffel et al., 2023) and SmolLM-135M(Allal et al., 2024) to Phi-2 (Research, 2023), implementing both traditional NLP metrics and custom domain-specific evaluation metrics. Our novel evaluation framework incorporates recipe-specific metrics for assessing content quality and introduces approaches to allergen substitution. The results indicate that, while larger models generally perform better on standard metrics, the relationship between model size and recipe quality is more nuanced when considering domain-specific metrics. SmolLM-360M and SmolLM-1.7B demonstrate comparable performance despite their size difference before and after fine-tuning, while fine-tuning Phi-2 shows notable limitations in recipe generation despite its larger parameter count. The comprehensive evaluation framework and allergen substitution systems provide valuable insights for future work in recipe generation and broader NLG tasks that require domain expertise and safety considerations.
Authors: Christine Konicki, Mithun Chakraborty, Michael P. Wellman
Abstract: Policy Space Response Oracles (PSRO) interleaves empirical game-theoretic analysis with deep reinforcement learning (DRL) to solve games too complex for traditional analytic methods. Tree-exploiting PSRO (TE-PSRO) is a variant of this approach that iteratively builds a coarsened empirical game model in extensive form using data obtained from querying a simulator that represents a detailed description of the game. We make two main methodological advances to TE-PSRO that enhance its applicability to complex games of imperfect information. First, we introduce a scalable representation for the empirical game tree where edges correspond to implicit policies learned through DRL. These policies cover conditions in the underlying game abstracted in the game model, supporting sustainable growth of the tree over epochs. Second, we leverage extensive form in the empirical model by employing refined Nash equilibria to direct strategy exploration. To enable this, we give a modular and scalable algorithm based on generalized backward induction for computing a subgame perfect equilibrium (SPE) in an imperfect-information game. We experimentally evaluate our approach on a suite of games including an alternating-offer bargaining game with outside offers; our results demonstrate that TE-PSRO converges toward equilibrium faster when new strategies are generated based on SPE rather than Nash equilibrium, and with reasonable time/memory requirements for the growing empirical model.
Authors: Sherif Eissa, Sander Stuijk, Floran De Putter, Andrea Nardi-Dei, Federico Corradi, Henk Corporaal
Abstract: Spiking Neural Networks (SNNs) are promising bio-inspired third-generation neural networks. Recent research has trained deep SNN models with accuracy on par with Artificial Neural Networks (ANNs). Although the event-driven and sparse nature of SNNs show potential for more energy efficient computation than ANNs, SNN neurons have internal states which evolve over time. Keeping track of SNN states can significantly increase data movement and storage requirements, potentially losing its advantages with respect to ANNs. This paper investigates the energy effects of having neuron states, and how it is influenced by the chosen mapping to realistic hardware architectures with advanced memory hierarchies. Therefore, we develop STEMS, a mapping design space exploration tool for SNNs. STEMS models SNN's stateful behavior and explores intra-layer and inter-layer mapping optimizations to minimize data movement, considering both spatial and temporal SNN dimensions. Using STEMS, we show up to 12x reduction in off-chip data movement and 5x reduction in energy (on top of intra-layer optimizations), on two event-based vision SNN benchmarks. Finally, neuron states may not be needed for all SNN layers. By optimizing neuron states for one of our benchmarks, we show 20x reduction in neuron states and 1.4x better performance without accuracy loss.
Authors: Fred Zhangzhi Peng, Zachary Bezemek, Sawan Patel, Jarrid Rector-Brooks, Sherwood Yao, Alexander Tong, Pranam Chatterjee
Abstract: In this paper, we explore how token unmasking order influences generative quality in masked diffusion models (MDMs). We derive an expanded evidence lower bound (ELBO) that introduces a planner to select which tokens to unmask at each step. Our analysis reveals that alternative unmasking strategies can enhance generation performance. Building on this, we propose Path Planning (P2), a sampling framework that uses a pre-trained BERT model or the denoiser itself to guide unmasking decisions. P2 generalizes all known MDM sampling strategies and significantly improves performance across diverse domains, including language generation (in-context learning, code generation, story infilling, mathematical reasoning, reverse curse correction) and biological sequence generation (protein and RNA sequences).
Authors: Emanuele La Malfa, Christoph Weinhuber, Orazio Torre, Fangru Lin, X. Angelo Huang, Samuele Marro, Anthony Cohn, Nigel Shadbolt, Michael Wooldridge
Abstract: Many reasoning, planning, and problem-solving tasks share an intrinsic algorithmic nature: correctly simulating each step is a sufficient condition to solve them correctly. We collect pairs of naturalistic and synthetic reasoning tasks to assess the capabilities of Large Language Models (LLM). While naturalistic tasks often require careful human handcrafting, we show that synthetic data is, in many cases, a good proxy that is much easier to collect at scale. We leverage common constructs in programming as the counterpart of the building blocks of naturalistic reasoning tasks, such as straight-line programs, code that contains critical paths, and approximate and redundant instructions. We further assess the capabilities of LLMs on sorting problems and repeated operations via sorting algorithms and nested loops. Our synthetic datasets further reveal that while the most powerful LLMs exhibit relatively strong execution capabilities, the process is fragile: it is negatively affected by memorisation and seems to rely heavily on pattern recognition. Our contribution builds upon synthetically testing the reasoning capabilities of LLMs as a scalable complement to handcrafted human-annotated problems.
Authors: Kaicheng Zhang, Piero Deidda, Desmond Higham, Francesco Tudisco
Abstract: Oversmoothing is a fundamental challenge in graph neural networks (GNNs): as the number of layers increases, node embeddings become increasingly similar, and model performance drops sharply. Traditionally, oversmoothing has been quantified using metrics that measure the similarity of neighbouring node features, such as the Dirichlet energy. While these metrics are related to oversmoothing, we argue they have critical limitations and fail to reliably capture oversmoothing in realistic scenarios. For instance, they provide meaningful insights only for very deep networks and under somewhat strict conditions on the norm of network weights and feature representations. As an alternative, we propose measuring oversmoothing by examining the numerical or effective rank of the feature representations. We provide theoretical support for this approach, demonstrating that the numerical rank of feature representations converges to one for a broad family of nonlinear activation functions under the assumption of nonnegative trained weights. To the best of our knowledge, this is the first result that proves the occurrence of oversmoothing without assumptions on the boundedness of the weight matrices. Along with the theoretical findings, we provide extensive numerical evaluation across diverse graph architectures. Our results show that rank-based metrics consistently capture oversmoothing, whereas energy-based metrics often fail. Notably, we reveal that a significant drop in the rank aligns closely with performance degradation, even in scenarios where energy metrics remain unchanged.
Authors: Yi Li, Yuquan Deng, Jesse Zhang, Joel Jang, Marius Memmel, Raymond Yu, Caelan Reed Garrett, Fabio Ramos, Dieter Fox, Anqi Li, Abhishek Gupta, Ankit Goyal
Abstract: Large foundation models have shown strong open-world generalization to complex problems in vision and language, but similar levels of generalization have yet to be achieved in robotics. One fundamental challenge is the lack of robotic data, which are typically obtained through expensive on-robot operation. A promising remedy is to leverage cheaper, off-domain data such as action-free videos, hand-drawn sketches or simulation data. In this work, we posit that hierarchical vision-language-action (VLA) models can be more effective in utilizing off-domain data than standard monolithic VLA models that directly finetune vision-language models (VLMs) to predict actions. In particular, we study a class of hierarchical VLA models, where the high-level VLM is finetuned to produce a coarse 2D path indicating the desired robot end-effector trajectory given an RGB image and a task description. The intermediate 2D path prediction is then served as guidance to the low-level, 3D-aware control policy capable of precise manipulation. Doing so alleviates the high-level VLM from fine-grained action prediction, while reducing the low-level policy's burden on complex task-level reasoning. We show that, with the hierarchical design, the high-level VLM can transfer across significant domain gaps between the off-domain finetuning data and real-robot testing scenarios, including differences on embodiments, dynamics, visual appearances and task semantics, etc. In the real-robot experiments, we observe an average of 20% improvement in success rate across seven different axes of generalization over OpenVLA, representing a 50% relative gain. Visual results are provided at: https://hamster-robot.github.io/
Authors: Ada Defne Tur, Gaurav Kamath, Siva Reddy
Abstract: Though English sentences are typically inflexible vis-\`a-vis word order, constituents often show far more variability in ordering. One prominent theory presents the notion that constituent ordering is directly correlated with constituent weight: a measure of the constituent's length or complexity. Such theories are interesting in the context of natural language processing (NLP), because while recent advances in NLP have led to significant gains in the performance of large language models (LLMs), much remains unclear about how these models process language, and how this compares to human language processing. In particular, the question remains whether LLMs display the same patterns with constituent movement, and may provide insights into existing theories on when and how the shift occurs in human language. We compare a variety of LLMs with diverse properties to evaluate broad LLM performance on four types of constituent movement: heavy NP shift, particle movement, dative alternation, and multiple PPs. Despite performing unexpectedly around particle movement, LLMs generally align with human preferences around constituent ordering.
Authors: Hongye Liu, Ricardo Henao
Abstract: Smart word substitution aims to enhance sentence quality by improving word choices; however current benchmarks rely on human-labeled data. Since word choices are inherently subjective, ground-truth word substitutions generated by a small group of annotators are often incomplete and likely not generalizable. To circumvent this issue, we instead employ a model-based score (BARTScore) to quantify sentence quality, thus forgoing the need for human annotations. Specifically, we use this score to define a distribution for each word substitution, allowing one to test whether a substitution is statistically superior relative to others. In addition, we propose a loss function that directly optimizes the alignment between model predictions and sentence scores, while also enhancing the overall quality score of a substitution. Crucially, model learning no longer requires human labels, thus avoiding the cost of annotation while maintaining the quality of the text modified with substitutions. Experimental results show that the proposed approach outperforms both masked language models (BERT, BART) and large language models (GPT-4, LLaMA). The source code is available at https://github.com/Hyfred/Substitute-Words-with-Ranking.
URLs: https://github.com/Hyfred/Substitute-Words-with-Ranking.
Authors: Junyu Lu, Kai Ma, Kaichun Wang, Kelaiti Xiao, Roy Ka-Wei Lee, Bo Xu, Liang Yang, Hongfei Lin
Abstract: Large Language Models (LLMs) have become essential for offensive language detection, yet their ability to handle annotation disagreement remains underexplored. Disagreement samples, which arise from subjective interpretations, pose a unique challenge due to their ambiguous nature. Understanding how LLMs process these cases, particularly their confidence levels, can offer insight into their alignment with human annotators. This study systematically evaluates the performance of multiple LLMs in detecting offensive language at varying levels of annotation agreement. We analyze binary classification accuracy, examine the relationship between model confidence and human disagreement, and explore how disagreement samples influence model decision-making during few-shot learning and instruction fine-tuning. Our findings reveal that LLMs struggle with low-agreement samples, often exhibiting overconfidence in these ambiguous cases. However, utilizing disagreement samples in training improves both detection accuracy and model alignment with human judgment. These insights provide a foundation for enhancing LLM-based offensive language detection in real-world moderation tasks.
Authors: Yiwei Guo, Zhihan Li, Hankun Wang, Bohan Li, Chongtian Shao, Hanglei Zhang, Chenpeng Du, Xie Chen, Shujie Liu, Kai Yu
Abstract: The rapid advancement of speech generation technologies in the era of large language models (LLMs) has established discrete speech tokens as a foundational paradigm for speech representation. These tokens, characterized by their discrete, compact, and concise nature, are not only advantageous for efficient transmission and storage, but also inherently compatible with the language modeling framework, enabling seamless integration of speech into text-dominated LLM architectures. Current research categorizes discrete speech tokens into two principal classes: acoustic tokens and semantic tokens, each of which has evolved into a rich research domain characterized by unique design philosophies and methodological approaches. This survey systematically synthesizes the existing taxonomy and recent innovations in discrete speech tokenization, conducts a critical examination of the strengths and limitations of each paradigm, and presents systematic experimental comparisons across token types. Furthermore, we identify persistent challenges in the field and propose potential research directions, aiming to offer actionable insights to inspire future advancements in the development and application of discrete speech tokens.
Authors: Gon\c{c}alo Gomes, Chrysoula Zerva, Bruno Martins
Abstract: The evaluation of image captions, looking at both linguistic fluency and semantic correspondence to visual contents, has witnessed a significant effort. Still, despite advancements such as the CLIPScore metric, multilingual captioning evaluation has remained relatively unexplored. This work presents several strategies, and extensive experiments, related to evaluating CLIPScore variants in multilingual settings. To address the lack of multilingual test data, we consider two different strategies: (1) using quality aware machine-translated datasets with human judgements, and (2) re-purposing multilingual datasets that target semantic inference and reasoning. Our results highlight the potential of finetuned multilingual models to generalize across languages and to handle complex linguistic challenges. Tests with machine-translated data show that multilingual CLIPScore models can maintain a high correlation with human judgements across different languages, and additional tests with natively multilingual and multicultural data further attest to the high-quality assessments.
Authors: Adilzhan Adilkhanov, Amir Yelenov, Assylkhan Seitzhanov, Ayan Mazhitov, Azamat Abdikarimov, Danissa Sandykbayeva, Daryn Kenzhebek, Dinmukhammed Mukashev, Ilyas Umurbekov, Jabrail Chumakov, Kamila Spanova, Karina Burunchina, Madina Yergibay, Margulan Issa, Moldir Zabirova, Nurdaulet Zhuzbay, Nurlan Kabdyshev, Nurlan Zhaniyar, Rasul Yermagambet, Rustam Chibar, Saltanat Seitzhan, Soibkhon Khajikhanov, Tasbolat Taunyazov, Temirlan Galimzhanov, Temirlan Kaiyrbay, Tleukhan Mussin, Togzhan Syrymova, Valeriya Kostyukova, Yerkebulan Massalim, Yermakhan Kassym, Zerde Nurbayeva, Zhanat Kappassov
Abstract: This paper presents an AI-generated review of Vision-Language-Action (VLA) models, summarizing key methodologies, findings, and future directions. The content is produced using large language models (LLMs) and is intended only for demonstration purposes. This work does not represent original research, but highlights how AI can help automate literature reviews. As AI-generated content becomes more prevalent, ensuring accuracy, reliability, and proper synthesis remains a challenge. Future research will focus on developing a structured framework for AI-assisted literature reviews, exploring techniques to enhance citation accuracy, source credibility, and contextual understanding. By examining the potential and limitations of LLM in academic writing, this study aims to contribute to the broader discussion of integrating AI into research workflows. This work serves as a preliminary step toward establishing systematic approaches for leveraging AI in literature review generation, making academic knowledge synthesis more efficient and scalable.
Authors: Jinyu Xiang, Jiayi Zhang, Zhaoyang Yu, Fengwei Teng, Jinhao Tu, Xinbing Liang, Sirui Hong, Chenglin Wu, Yuyu Luo
Abstract: Well-designed prompts are crucial for enhancing Large language models' (LLMs) reasoning capabilities while aligning their outputs with task requirements across diverse domains. However, manually designed prompts require expertise and iterative experimentation. While existing prompt optimization methods aim to automate this process, they rely heavily on external references such as ground truth or by humans, limiting their applicability in real-world scenarios where such data is unavailable or costly to obtain. To address this, we propose Self-Supervised Prompt Optimization (SPO), a cost-efficient framework that discovers effective prompts for both closed and open-ended tasks without requiring external reference. Motivated by the observations that prompt quality manifests directly in LLM outputs and LLMs can effectively assess adherence to task requirements, we derive evaluation and optimization signals purely from output comparisons. Specifically, SPO selects superior prompts through pairwise output comparisons evaluated by an LLM evaluator, followed by an LLM optimizer that aligns outputs with task requirements. Extensive experiments demonstrate that SPO outperforms state-of-the-art prompt optimization methods, achieving comparable or superior results with significantly lower costs (e.g., 1.1% to 5.6% of existing methods) and fewer samples (e.g., three samples). The code is available at https://github.com/geekan/MetaGPT/blob/main/examples/spo
URLs: https://github.com/geekan/MetaGPT/blob/main/examples/spo
Authors: Daniel Rodriguez-Cardenas, Alejandro Velasco, Denys Poshyvanyk
Abstract: Language Models (LLMs), such as transformer-based neural networks trained on billions of parameters, have become increasingly prevalent in software engineering (SE). These models, trained on extensive datasets that include code repositories, exhibit remarkable capabilities for SE tasks. However, evaluating their effectiveness poses significant challenges, primarily due to the potential overlap between the datasets used for training and those employed for evaluation. To address this issue, we introduce SnipGen, a comprehensive repository mining framework designed to leverage prompt engineering across various downstream tasks for code generation. SnipGen aims to mitigate data contamination by generating robust testbeds and crafting tailored data points to assist researchers and practitioners in evaluating LLMs for code-related tasks. In our exploratory study, SnipGen mined approximately 227K data points from 338K recent code changes in GitHub commits, focusing on method-level granularity. SnipGen features a collection of prompt templates that can be combined to create a Chain-of-Thought-like sequence of prompts, enabling a nuanced assessment of LLMs' code generation quality. By providing the mining tool, the methodology, and the dataset, SnipGen empowers researchers and practitioners to rigorously evaluate and interpret LLMs' performance in software engineering contexts.
Authors: Junlong Li, Daya Guo, Dejian Yang, Runxin Xu, Yu Wu, Junxian He
Abstract: Reasoning is a fundamental capability of Large Language Models. While prior research predominantly focuses on enhancing narrow skills like math or code generation, improving performance on many other reasoning tasks remains challenging due to sparse and fragmented training data. To address this issue, we propose CodeI/O, a novel approach that systematically condenses diverse reasoning patterns inherently embedded in contextually-grounded codes, through transforming the original code into a code input-output prediction format. By training models to predict inputs/outputs given code and test cases entirely in natural language as Chain-of-Thought (CoT) rationales, we expose them to universal reasoning primitives -- like logic flow planning, state-space searching, decision tree traversal, and modular decomposition -- while decoupling structured reasoning from code-specific syntax and preserving procedural rigor. Experimental results demonstrate CodeI/O leads to consistent improvements across symbolic, scientific, logic, math & numerical, and commonsense reasoning tasks. By matching the existing ground-truth outputs or re-executing the code with predicted inputs, we can verify each prediction and further enhance the CoTs through multi-turn revision, resulting in CodeI/O++ and achieving higher performance. Our data and models are available at https://github.com/hkust-nlp/CodeIO.
Authors: Shuzheng Si, Haozhe Zhao, Gang Chen, Cheng Gao, Yuzhuo Bai, Zhitong Wang, Kaikai An, Kangyang Luo, Chen Qian, Fanchao Qi, Baobao Chang, Maosong Sun
Abstract: Training LLMs on data containing unfamiliar knowledge during the instruction tuning stage can encourage hallucinations. To address this challenge, we introduce NOVA, a novel framework designed to identify high-quality data that aligns well with the LLM's learned knowledge to reduce hallucinations. NOVA includes Internal Consistency Probing (ICP) and Semantic Equivalence Identification (SEI) to measure how familiar the LLM is with instruction data. Specifically, ICP evaluates the LLM's understanding of the given instruction by calculating the tailored consistency among multiple self-generated responses. SEI further assesses the familiarity of the LLM with the target response by comparing it to the generated responses, using the proposed semantic clustering and well-designed voting strategy. Finally, to ensure the quality of selected samples, we introduce an expert-aligned reward model, considering characteristics beyond just familiarity. By considering data quality and avoiding unfamiliar data, we can utilize the selected data to effectively align LLMs to follow instructions and hallucinate less.
Authors: Alan Saji, Jaavid Aktar Husain, Thanmay Jayakumar, Raj Dabre, Anoop Kunchukuttan, Ratish Puduppully
Abstract: Large Language Models (LLMs) exhibit remarkable multilingual generalization despite being predominantly trained on English-centric corpora. A fundamental question arises: how do LLMs achieve such robust multilingual capabilities? We take the case of non-Roman script languages, we investigate the role of Romanization - the representation of non-Roman scripts using Roman characters - as a bridge in multilingual processing. Using mechanistic interpretability techniques, we analyze next-token generation and find that intermediate layers frequently represent target words in Romanized form before transitioning to native script, a phenomenon we term Latent Romanization. Further, through activation patching experiments, we demonstrate that LLMs encode semantic concepts similarly across native and Romanized scripts, suggesting a shared underlying representation. Additionally, for translation into non-Roman script languages, our findings reveal that when the target language is in Romanized form, its representations emerge earlier in the model's layers compared to native script. These insights contribute to a deeper understanding of multilingual representation in LLMs and highlight the implicit role of Romanization in facilitating language transfer.
Authors: Mohsinul Kabir, Ajwad Abrar, Sophia Ananiadou
Abstract: A large number of studies rely on closed-style multiple-choice surveys to evaluate cultural alignment in Large Language Models (LLMs). In this work, we challenge this constrained evaluation paradigm and explore more realistic, unconstrained approaches. Using the World Values Survey (WVS) and Hofstede Cultural Dimensions as case studies, we demonstrate that LLMs exhibit stronger cultural alignment in less constrained settings, where responses are not forced. Additionally, we show that even minor changes, such as reordering survey choices, lead to inconsistent outputs, exposing the limitations of closed-style evaluations. Our findings advocate for more robust and flexible evaluation frameworks that focus on specific cultural proxies, encouraging more nuanced and accurate assessments of cultural alignment in LLMs.
Authors: Dongqi Liu, Chenxi Whitehouse, Xi Yu, Louis Mahon, Rohit Saxena, Zheng Zhao, Yifu Qiu, Mirella Lapata, Vera Demberg
Abstract: Transforming recorded videos into concise and accurate textual summaries is a growing challenge in multimodal learning. This paper introduces VISTA, a dataset specifically designed for video-to-text summarization in scientific domains. VISTA contains 18,599 recorded AI conference presentations paired with their corresponding paper abstracts. We benchmark the performance of state-of-the-art large models and apply a plan-based framework to better capture the structured nature of abstracts. Both human and automated evaluations confirm that explicit planning enhances summary quality and factual consistency. However, a considerable gap remains between models and human performance, highlighting the challenges of scientific video summarization.
Authors: Bin Wu, Yihang Wang, Yuanhao Zeng, Jiawei Liu, Jiashu Zhao, Cheng Yang, Yawen Li, Long Xia, Dawei Yin, Chuan Shi
Abstract: Recommender systems (RS) serve as a fundamental tool for navigating the vast expanse of online information, with deep learning advancements playing an increasingly important role in improving ranking accuracy. Among these, graph neural networks (GNNs) excel at extracting higher-order structural information, while large language models (LLMs) are designed to process and comprehend natural language, making both approaches highly effective and widely adopted. Recent research has focused on graph foundation models (GFMs), which integrate the strengths of GNNs and LLMs to model complex RS problems more efficiently by leveraging the graph-based structure of user-item relationships alongside textual understanding. In this survey, we provide a comprehensive overview of GFM-based RS technologies by introducing a clear taxonomy of current approaches, diving into methodological details, and highlighting key challenges and future directions. By synthesizing recent advancements, we aim to offer valuable insights into the evolving landscape of GFM-based recommender systems.
Authors: Jiyuan Ren, Zhaocheng Du, Zhihao Wen, Qinglin Jia, Sunhao Dai, Chuhan Wu, Zhenhua Dong
Abstract: As large language models (LLMs) advance, their ability to perform in-context learning and few-shot language generation has improved significantly. This has spurred using LLMs to produce high-quality synthetic data to enhance the performance of smaller models like online retrievers or weak LLMs. However, LLM-generated synthetic data often differs from the real data in key language attributes (e.g., styles, tones, content proportions, etc.). As a result, mixing these synthetic data directly with real data may distort the original data distribution, potentially hindering performance improvements. To solve this, we introduce SynAlign: a synthetic data generation and filtering framework based on key attribute distribution matching. Before generation, SynAlign employs an uncertainty tracker surrogated by the Gaussian Process model to iteratively select data clusters distinct from selected ones as demonstrations for new data synthesis, facilitating the efficient exploration diversity of the real data. Then, a latent attribute reasoning method is employed: the LLM summarizes linguistic attributes of demonstrations and then synthesizes new data based on them. This approach facilitates synthesizing diverse data with linguistic attributes that appear in real data.After generation, the Maximum Mean Discrepancy is used as the objective function to learn the sampling weight of each synthetic data, ensuring distribution matching with the real data. Our experiments on multiple text prediction tasks show significant performance improvements. We also conducted an online A/B test on an online retriever to demonstrate SynAlign's effectiveness.
Authors: Sebastian Sanokowski, Wilhelm Berghammer, Martin Ennemoser, Haoyu Peter Wang, Sepp Hochreiter, Sebastian Lehner
Abstract: Learning to sample from complex unnormalized distributions over discrete domains emerged as a promising research direction with applications in statistical physics, variational inference, and combinatorial optimization. Recent work has demonstrated the potential of diffusion models in this domain. However, existing methods face limitations in memory scaling and thus the number of attainable diffusion steps since they require backpropagation through the entire generative process. To overcome these limitations we introduce two novel training methods for discrete diffusion samplers, one grounded in the policy gradient theorem and the other one leveraging Self-Normalized Neural Importance Sampling (SN-NIS). These methods yield memory-efficient training and achieve state-of-the-art results in unsupervised combinatorial optimization. Numerous scientific applications additionally require the ability of unbiased sampling. We introduce adaptations of SN-NIS and Neural Markov Chain Monte Carlo that enable for the first time the application of discrete diffusion models to this problem. We validate our methods on Ising model benchmarks and find that they outperform popular autoregressive approaches. Our work opens new avenues for applying diffusion models to a wide range of scientific applications in discrete domains that were hitherto restricted to exact likelihood models.
Authors: Timoth\'ee Darcet, Federico Baldassarre, Maxime Oquab, Julien Mairal, Piotr Bojanowski
Abstract: Masked Image Modeling (MIM) offers a promising approach to self-supervised representation learning, however existing MIM models still lag behind the state-of-the-art. In this paper, we systematically analyze target representations, loss functions, and architectures, to introduce CAPI - a novel pure-MIM framework that relies on the prediction of latent clusterings. Our approach leverages a clustering-based loss, which is stable to train, and exhibits promising scaling properties. Our ViT-L backbone, CAPI, achieves 83.8% accuracy on ImageNet and 32.1% mIoU on ADE20K with simple linear probes, substantially outperforming previous MIM methods and approaching the performance of the current state-of-the-art, DINOv2. We release all our code and models.
Authors: Wangyang Ying, Cong Wei, Nanxu Gong, Xinyuan Wang, Haoyue Bai, Arun Vignesh Malarkkan, Sixun Dong, Dongjie Wang, Denghui Zhang, Yanjie Fu
Abstract: Tabular data is one of the most widely used data formats across various domains such as bioinformatics, healthcare, and marketing. As artificial intelligence moves towards a data-centric perspective, improving data quality is essential for enhancing model performance in tabular data-driven applications. This survey focuses on data-driven tabular data optimization, specifically exploring reinforcement learning (RL) and generative approaches for feature selection and feature generation as fundamental techniques for refining data spaces. Feature selection aims to identify and retain the most informative attributes, while feature generation constructs new features to better capture complex data patterns. We systematically review existing generative methods for tabular data engineering, analyzing their latest advancements, real-world applications, and respective strengths and limitations. This survey emphasizes how RL-based and generative techniques contribute to the automation and intelligence of feature engineering. Finally, we summarize the existing challenges and discuss future research directions, aiming to provide insights that drive continued innovation in this field.
Authors: Kunat Pipatanakul, Pittawat Taveekitworachai, Potsawee Manakul, Kasima Tharnpipitchai
Abstract: This paper investigates data selection and model merging methodologies aimed at incorporating advanced reasoning capabilities such as those of DeepSeek R1 into language-specific large language models (LLMs), with a particular focus on the Thai LLM. Our goal is to enhance the reasoning capabilities of language-specific LLMs while maintaining their target language abilities. DeepSeek R1 excels in reasoning but primarily benefits high-resource languages such as English and Chinese. However, low-resource languages remain underserved due to the dominance of English-centric training data and model optimizations, which limit performance in these languages. This limitation results in unreliable code-switching and diminished effectiveness on tasks in low-resource languages. Meanwhile, local and regional LLM initiatives have attempted to bridge this gap by developing language-specific LLMs that focus on improving local linguistic fidelity. We demonstrate that, with only publicly available datasets and a computational budget of $120, it is possible to enhance the reasoning capabilities of language-specific LLMs to match the level of DeepSeek R1, without compromising their performance on target language tasks.
Authors: Quantao Yang, Michael C. Welle, Danica Kragic, Olov Andersson
Abstract: Recent advances in skill learning has propelled robot manipulation to new heights by enabling it to learn complex manipulation tasks from a practical number of demonstrations. However, these skills are often limited to the particular action, object, and environment \textit{instances} that are shown in the training data, and have trouble transferring to other instances of the same category. In this work we present an open-vocabulary Spatial-Semantic Diffusion policy (S$^2$-Diffusion) which enables generalization from instance-level training data to category-level, enabling skills to be transferable between instances of the same category. We show that functional aspects of skills can be captured via a promptable semantic module combined with a spatial representation. We further propose leveraging depth estimation networks to allow the use of only a single RGB camera. Our approach is evaluated and compared on a diverse number of robot manipulation tasks, both in simulation and in the real world. Our results show that S$^2$-Diffusion is invariant to changes in category-irrelevant factors as well as enables satisfying performance on other instances within the same category, even if it was not trained on that specific instance. Full videos of all real-world experiments are available in the supplementary material.
Authors: Caleb Cranney, Jesse G. Meyer
Abstract: Transformer architectures have transformed AI applications but remain complex to customize for domain experts lacking low-level implementation expertise. We introduce AttentionSmithy, a modular software package that simplifies transformer innovation by breaking down key components into reusable building blocks: attention modules, feed-forward networks, normalization layers, and positional encodings. Users can rapidly prototype and evaluate transformer variants without extensive coding. Our framework supports four positional encoding strategies and integrates with neural architecture search for automated design. We validate AttentionSmithy by replicating the original transformer under resource constraints and optimizing translation performance by combining positional encodings. Additionally, we demonstrate its adaptability in gene-specific modeling, achieving over 95% accuracy in cell type classification. These case studies highlight AttentionSmithy's potential to accelerate research across diverse fields by removing framework implementation barriers.
Authors: Mingmeng Geng, Roberto Trotta
Abstract: With a statistical analysis of arXiv paper abstracts, we report a marked drop in the frequency of several words previously identified as overused by ChatGPT, such as "delve", starting soon after they were pointed out in early 2024. The frequency of certain other words favored by ChatGPT, such as "significant", has instead kept increasing. These phenomena suggest that some authors of academic papers have adapted their use of large language models (LLMs), for example, by selecting outputs or applying modifications to the LLM-generated content. Such coevolution and cooperation of humans and LLMs thus introduce additional challenges to the detection of machine-generated text in real-world scenarios. Estimating the impact of LLMs on academic writing by examining word frequency remains feasible, and more attention should be paid to words that were already frequently employed, including those that have decreased in frequency due to LLMs' disfavor.
Authors: Jin Hyun Park, Seyyed Ali Ayati, Yichen Cai
Abstract: The increasing prevalence of microphones in everyday devices and the growing reliance on online services have amplified the risk of acoustic side-channel attacks (ASCAs) targeting keyboards. This study explores deep learning techniques, specifically vision transformers (VTs) and large language models (LLMs), to enhance the effectiveness and applicability of such attacks. We present substantial improvements over prior research, with the CoAtNet model achieving state-of-the-art performance. Our CoAtNet shows a 5.0% improvement for keystrokes recorded via smartphone (Phone) and 5.9% for those recorded via Zoom compared to previous benchmarks. We also evaluate transformer architectures and language models, with the best VT model matching CoAtNet's performance. A key advancement is the introduction of a noise mitigation method for real-world scenarios. By using LLMs for contextual understanding, we detect and correct erroneous keystrokes in noisy environments, enhancing ASCA performance. Additionally, fine-tuned lightweight language models with Low-Rank Adaptation (LoRA) deliver comparable performance to heavyweight models with 67X more parameters. This integration of VTs and LLMs improves the practical applicability of ASCA mitigation, marking the first use of these technologies to address ASCAs and error correction in real-world scenarios.
Authors: Tianwei Lin, Wenqiao Zhang, Sijing Li, Yuqian Yuan, Binhe Yu, Haoyuan Li, Wanggui He, Hao Jiang, Mengze Li, Xiaohui Song, Siliang Tang, Jun Xiao, Hui Lin, Yueting Zhuang, Beng Chin Ooi
Abstract: We present HealthGPT, a powerful Medical Large Vision-Language Model (Med-LVLM) that integrates medical visual comprehension and generation capabilities within a unified autoregressive paradigm. Our bootstrapping philosophy is to progressively adapt heterogeneous comprehension and generation knowledge to pre-trained large language models (LLMs). This is achieved through a novel heterogeneous low-rank adaptation (H-LoRA) technique, which is complemented by a tailored hierarchical visual perception approach and a three-stage learning strategy. To effectively learn the HealthGPT, we devise a comprehensive medical domain-specific comprehension and generation dataset called VL-Health. Experimental results demonstrate exceptional performance and scalability of HealthGPT in medical visual unified tasks. Our project can be accessed at https://github.com/DCDmllm/HealthGPT.
Authors: Ishika Agarwal, Dilek Hakkani-T\"ur
Abstract: Influence functions provide crucial insights into model training, but existing methods suffer from large computational costs and limited generalization. Particularly, recent works have proposed various metrics and algorithms to calculate the influence of data using language models, which do not scale well with large models and datasets. This is because of the expensive forward and backward passes required for computation, substantial memory requirements to store large models, and poor generalization of influence estimates to new data. In this paper, we explore the use of small neural networks -- which we refer to as the InfluenceNetwork -- to estimate influence values, achieving up to 99% cost reduction. Our evaluation demonstrates that influence values can be estimated with models just 0.0027% the size of full language models (we use 7B and 8B versions). We apply our algorithm of estimating influence values (called NN-CIFT: Neural Networks for effiCient Instruction Fine-Tuning) to the downstream task of subset selection for general instruction fine-tuning. In our study, we include four state-of-the-art influence functions and show no compromise in performance, despite large speedups, between NN-CIFT and the original influence functions. We provide an in-depth hyperparameter analyses of NN-CIFT. The code for our method can be found here: https://github.com/agarwalishika/NN-CIFT.