paperlist.ai

AI 顶会论文

AI 顶会论文库

检索 30+ 个顶会的论文，包括 NeurIPS、ICLR、ICML、CVPR、ACL、AAAI 等。输入关键词筛选，打开任意一篇查看摘要、作者和链接。

arXiv

A Statistical Framework for Auditing Behavioral Dependence and Induced Bias in LLM Judges

Chenchen Kuai, Jiwan Jiang, Zihao Zhu, Hao Wang, Keshu Wu, Zihao Li, Yunlong Zhang, Chenxi Liu, Zhengzhong Tu, Zhiwen Fan, Yang Zhou

The rapid growth of the large language model (LLM) ecosystem raises a critical question: are seemingly diverse models truly independent? Shared pretraining data, distillation, and alignment pipelines can induce hidden behavioral dependencies, or latent entanglement, that undermine multi-model systems such as LLM-as-a-judge pipelines and ensemble verification, which implicitly assume independent signals. In practice, this manifests as correlated reasoning patterns and synchronized failures, where apparent agreement reflects shared error modes rather than independent validation. To address this, we develop a statistical framework for auditing behavioral entanglement among black-box LLMs. Our approach introduces a multi-resolution hierarchy that characterizes the joint failure manifold through two information-theoretic metrics: (i) a Difficulty-Weighted Behavioral Entanglement Index (BEI), which amplifies synchronized failures on easy tasks, and (ii) a Cumulative Information Gain (CIG) metric, which captures directional alignment in erroneous responses. Through experiments on 18 LLMs from six model families, we identify statistically significant behavioral entanglement. Such behavioral dependence is associated with judge over-endorsement bias on a disjoint MMLU-Pro evaluation set (rho = 0.508 for BEI and rho = 0.520 for CIG; p < 0.01). The association further transfers to the MATH-500 benchmark (rho = 0.441 for BEI and rho = 0.457 for CIG; p < 0.05), providing cross-benchmark evidence that the identified dependency structure generalizes beyond the data and response format used for its estimation. Finally, we demonstrate a practical use case of entanglement through de-entangled verifier ensemble reweighting, achieving 3.5 and 2.6 percentage-point gains in accuracy and precision, respectively, over majority voting.

arXiv

TraceSafe: A Systematic Assessment of LLM Guardrails on Multi-Step Tool-Calling Trajectories

Yen-Shan Chen, Sian-Yao Huang, Cheng-Lin Yang, Yun-Nung Chen

As large language models (LLMs) evolve from static chatbots into autonomous agents, the primary vulnerability surface shifts from final outputs to intermediate execution traces. While safety guardrails are well-benchmarked for natural language responses, their efficacy remains largely unexplored within multi-step tool-use trajectories. To address this gap, we introduce **TraceSafe-Bench**, the first comprehensive benchmark specifically designed to assess mid-trajectory safety. It encompasses 12 risk categories, ranging from security threats (e.g., prompt injection, privacy leaks) to operational failures (e.g., hallucinations, interface inconsistencies), featuring over 1,000 unique execution instances. Our evaluation of 13 LLM-as-a-guard models and 7 specialized guardrails yields three critical findings: 1) *Structural Bottleneck*: Guardrail efficacy is driven more by structural data competence (e.g., JSON parsing) than semantic safety alignment. Performance correlates strongly with structured-to-text benchmarks ($ρ=0.79$, $p<0.01$) while we detect no association with standard jailbreak robustness. 2) *Scale Is Not the Deciding Factor*: Trajectory detection accuracy does not scale monotonically with model size, and general-purpose LLMs consistently outperform specialized safety guardrails. 3) *Bounded Temporal Stability*: Detection accuracy improves with trajectory length within native context bounds as models process dynamic execution behavior, but degrades in extreme long-context regimes due to over-refusal.

arXiv

Large Language Models Align with the Human Brain during Creative Thinking

Mete Ismayilzada, Simone A. Luchini, Abdulkadir Gokce, Badr AlKhamissi, Antoine Bosselut, Antonio Laverghetta Jr., Lonneke van der Plas, Roger E. Beaty

Creative thinking is a fundamental aspect of human cognition, and divergent thinking-the capacity to generate novel and varied ideas-is widely regarded as its core generative engine. Large language models (LLMs) have recently demonstrated impressive performance on divergent thinking tests and prior work has shown that models with higher task performance tend to be more aligned to human brain activity. However, existing brain-LLM alignment studies have focused on passive, non-creative tasks. Here, we explore brain alignment during creative thinking using fMRI data from 170 participants performing the Alternate Uses Task (AUT). We extract representations from LLMs varying in size (270M-72B) and measure alignment to brain responses via Representational Similarity Analysis (RSA), targeting the creativity-related default mode and frontoparietal networks. We find that brain-LLM alignment is positively associated with model size (default mode network only) and with idea originality (both networks), with these relationships clearest early in the creative process. We further find that post-training objectives are associated with functionally selective differences in alignment: a creativity-optimized Llama-3.1-8B-Instruct retains alignment with high-creativity neural responses while lacking the positive low-creativity alignment present in other variants, but a reasoning-trained variant shows the opposite asymmetry, with negative alignment to high-creativity responses. Together, these results suggest that brain alignment offers an informative lens on how post-training relates to the neural geometry of human creative thought.

arXiv

CresOWLve: Benchmarking Creative Problem-Solving Over Real-World Knowledge

Mete Ismayilzada, Renqing Cuomao, Daniil Yurshevich, Anna Sotnikova, Lonneke van der Plas, Antoine Bosselut

Creative problem-solving requires combining multiple cognitive abilities, including logical reasoning, lateral thinking, analogy-making, and commonsense knowledge, to discover insights that connect seemingly unrelated pieces of information. However, most existing benchmarks for large language models (LLMs) evaluate only specific components of this process. Moreover, many creativity-oriented benchmarks rely on artificially constructed brainteasers or contrived scenarios that do not reflect how creative problem-solving occurs in real-world settings. To address this gap, we introduce CresOWLve (containing 2,061 examples across five difficulty levels), a benchmark for evaluating creative problem-solving using puzzles grounded in real-world knowledge. Problems in CresOWLve require employing multiple creative thinking strategies, retrieving facts from diverse domains, and creatively combining them to arrive at a solution. Evaluating several frontier non-thinking and thinking LLMs, we show that CresOWLve remains highly challenging. Our analysis reveals a consistent performance gap: models perform substantially better on factual questions than on creative ones (up to -17% drop). While models can often retrieve the relevant knowledge, they struggle to form the non-obvious creative connections required to integrate the knowledge and arrive at the correct answer.

arXiv

From Rebound to Remedy: Understanding and Mitigating Reward Hacking via Representation Engineering

Rui Wu, Ruixiang Tang

Reinforcement learning for LLMs is vulnerable to reward hacking, where models exploit shortcuts to maximize reward without solving the intended task. We systematically study this phenomenon in coding tasks using an environment-manipulation setting, where models can rewrite evaluator code to trivially pass tests without solving the task, as a controlled testbed. Across both studied models, we identify a reproducible three-phase rebound pattern: models first attempt to rewrite the evaluator but fail, as their rewrites embed test cases their own solutions cannot pass. They then temporarily retreat to legitimate solving. When legitimate reward remains scarce, they rebound into successful hacking with qualitatively different strategies. Using representation engineering, we extract concept directions for shortcut, deception, and evaluation awareness from domain-general contrastive pairs and find that the shortcut direction tracks hacking behavior most closely, making it an effective representational proxy for detection. Motivated by this finding, we propose Advantage Modification, which integrates shortcut concept scores into GRPO advantage computation to penalize hacking rollouts before policy updates. Because the penalty is internalized into the training signal rather than applied only at inference time, Advantage Modification provides more robust suppression of hacking compared with generation-time activation steering.

arXiv

Uncertainty-Aware Variational Reward Factorization via Probabilistic Preference Bases for LLM Personalization

Gyuseok Lee, Wonbin Kweon, Zhenrui Yue, SeongKu Kang, Jiawei Han, Dong Wang

Reward factorization personalizes large language models (LLMs) by decomposing rewards into shared basis functions and user-specific weights. Yet, existing methods estimate user weights from scarce data in isolation and as deterministic points, leading to inaccurate and unreliable inference. We introduce Variational Reward Factorization (VRF), an uncertainty-aware framework that represents each user's preferences as a variational distribution in a shared preference space. VRF infers user distributions via a variational encoder, derives weights through Wasserstein distance matching with shared probabilistic bases, and downweights uncertain estimates through a variance-attenuated loss. On three benchmarks, VRF outperforms all baselines across seen and unseen users, few-shot scenarios, and varying uncertainty levels, with gains extending to downstream alignment. Our code is available at https://github.com/Gyu-Seok-Lee/VRF_COLM26.

arXiv

Beyond Static Models: An Evolving Framework for Continual Learning in Large Language Models across Training Stages

Hongyang Chen, Zhongwu Sun, Hongfei Ye, Kunchi Li, Xuemin Lin

Continual learning (CL) has emerged as a pivotal paradigm to enable large language models (LLMs) to dynamically adapt to evolving knowledge and sequential tasks while mitigating catastrophic forgetting, a critical limitation of the static pre-training paradigm inherent to modern LLMs. This paper presents a comprehensive overview of CL methodologies tailored for LLMs, structured around three core training stages: continual pre-training, continual fine-tuning, and continual alignment. Beyond the canonical taxonomy of rehearsal-, regularization-, and architecture-based methods, we further subdivide each category by its distinct forgetting mitigation mechanisms and conduct a rigorous comparative analysis of the adaptability and critical improvements of traditional CL methods for LLMs. In doing so, we explicitly highlight core distinctions between LLM CL and traditional machine learning, particularly with respect to scale, parameter efficiency, and emergent capabilities. Our analysis covers essential evaluation metrics, including forgetting rates and knowledge transfer efficiency, along with emerging benchmarks for assessing CL performance. While current methods show promising results, fundamental challenges persist in achieving seamless knowledge integration, and we identify key open problems and future directions for the field.

arXiv

Data Repetition Beats Data Scaling in Long-CoT Supervised Fine-Tuning

Dawid J. Kopiczko, Sagar Vaze, Tijmen Blankevoort, Yuki M Asano

Supervised fine-tuning (SFT) on chain-of-thought data is an essential post-training step for reasoning language models. Standard machine learning intuition suggests that training with more unique training samples yields better generalization. Counterintuitively, we show that SFT benefits from repetition: under a fixed update budget, training for more epochs on smaller datasets outperforms single-epoch training on larger datasets. On AIME'24/25 and GPQA benchmarks, Olmo3-7B trained for 128 epochs on 400 samples outperforms the equivalent 1 epoch on 51200 samples by 12-26 percentage points, with no additional catastrophic forgetting. We find that training token accuracy reliably signals when repetition has saturated; improvements from additional epochs plateau at full memorization, a pattern consistent across all settings. These findings provide a practical approach for reasoning SFT, where scaling epochs with token accuracy as a stopping criterion can replace expensive undirected data scaling. We pose the repetition advantage, where full memorization coincides with improved generalization, as a new open problem for the community in understanding the training dynamics of large language models.

arXiv

Expert-Guided Multimodal Fusion for Unified Emotion and Sentiment Analysis

Jiaqi Qiao, Xinran Li, Yifan Lyu, Xiujuan Xu, Liu Yu

Multimodal emotion understanding requires the integration of heterogeneous data sources, including text, audio, and visual modalities, while simultaneously addressing discrete emotion recognition and continuous sentiment analysis. We propose EGMF, a unified framework that combines expert-guided multimodal fusion with large language models to achieve superior performance across both tasks. At the core of our framework is a multi-scale expert network, comprising a local expert for capturing subtle emotional nuances, a semantic correlation expert for modeling cross-modal relationships, and a global context expert for understanding long-range dependencies. These experts are adaptively integrated via hierarchical dynamic gating, enabling context-aware feature selection and modality weighting. The enhanced multimodal representations are seamlessly incorporated into the language model through pseudo token injection and prompt-based conditioning, allowing a single generative framework to handle both classification and regression tasks. We employ parameter-efficient LoRA fine-tuning to maintain computational efficiency. Extensive experiments on bilingual benchmark datasets (MELD, CHERMA, MOSEI, SIMS-V2) demonstrate that EGMF outperforms state-of-the-art methods in terms of accuracy, cross-lingual robustness, and the discovery of universal patterns in multimodal emotional expressions.

arXiv

FronTalk: Benchmarking Front-End Development as Conversational Code Generation with Multi-Modal Feedback

Xueqing Wu, Zihan Xue, Da Yin, Shuyan Zhou, Kai-Wei Chang, Nanyun Peng, Yeming Wen

We present FronTalk, a benchmark for front-end code generation that pioneers the study of a unique interaction dynamic: conversational code generation with multi-modal feedback. In front-end development, visual artifacts such as sketches, mockups and annotated creenshots are essential for conveying design intent, yet their role in multi-turn code generation remains largely unexplored. To address this gap, we focus on the front-end development task and curate FronTalk, a collection of 100 multi-turn dialogues derived from real-world websites across diverse domains such as news, finance, and art. Each turn features both a textual instruction and an equivalent visual instruction, each representing the same user intent. To comprehensively evaluate model performance, we propose a novel agent-based evaluation framework leveraging a web agent to simulate users and explore the website, and thus measuring both functional correctness and user experience. Evaluation of 20 models reveals two key challenges that are under-explored systematically in the literature: (1) a significant forgetting issue where models overwrite previously implemented features, resulting in task failures, and (2) a persistent challenge in interpreting visual feedback, especially for open-source vision-language models (VLMs). We propose a strong baseline to tackle the forgetting issue with AceCoder, a method that critiques the implementation of every past instruction using an autonomous web agent. This approach significantly reduces forgetting to nearly zero and improves the performance by up to 9.3% (56.0% to 65.3%). Overall, we aim to provide a solid foundation for future research in front-end development and the general interaction dynamics of multi-turn, multi-modal code generation. Code and data are released at https://github.com/shirley-wu/frontalk

arXiv

Discovering and Causally Validating Emotion-Sensitive Neurons in Large Audio-Language Models

Xiutian Zhao, Björn Schuller, Berrak Sisman

Emotion is a central dimension of spoken communication, yet, we still lack a mechanistic account of how modern large audio-language models (LALMs) encode it internally. We present the first neuron-level interpretability study of emotion-sensitive neurons (ESNs) in LALMs and provide causal evidence supporting the existence of such units in Qwen2.5-Omni, Kimi-Audio, and Audio Flamingo 3. Across these three widely used open-source models, we compare frequency-, entropy-, mean-deviation-, and contrast-based neuron selectors on multiple emotion recognition benchmarks. Using inference-time interventions, we reveal a consistent emotion-specific signature: deactivating neurons selected for a given emotion disproportionately degrades recognition of that emotion while largely preserving other classes, whereas targeted steering amplifies these units to bias predictions toward the target emotion. These effects arise with modest amounts of identification data and scale systematically with intervention strength. We further observe that ESNs exhibit non-uniform layer-wise clustering with partial cross-dataset transfer. Taken together, our results offer a causal, neuron-level account of emotion decisions in LALMs and highlight targeted neuron interventions as an actionable handle for controllable affective behaviors.

arXiv

Prompt engineering does not universally improve Large Language Model performance across clinical decision-making tasks

Mengdi Chai, Ali R. Zomorrodi

Large Language Models (LLMs) have demonstrated promise in medical knowledge assessments, yet their practical utility in real-world clinical decision-making remains underexplored. In this study, we evaluated the performance of three state-of-the-art LLMs-ChatGPT-4o, Gemini 1.5 Pro, and LIama 3.3 70B-in clinical decision support across the entire clinical reasoning workflow of a typical patient encounter. Using 36 case studies, we first assessed LLM's out-of-the-box performance across five key sequential clinical decision-making tasks under two temperature settings (default vs. zero): differential diagnosis, essential immediate steps, relevant diagnostic testing, final diagnosis, and treatment recommendation. All models showed high variability by task, achieving near-perfect accuracy in final diagnosis, poor performance in relevant diagnostic testing, and moderate performance in remaining tasks. Furthermore, ChatGPT performed better under the zero temperature, whereas LIama showed stronger performance under the default temperature. Next, we assessed whether prompt engineering could enhance LLM performance by applying variations of the MedPrompt framework, incorporating targeted and random dynamic few-shot learning. The results demonstrate that prompt engineering is not a one-size-fit-all solution. While it significantly improved the performance on the task with lowest baseline accuracy (relevant diagnostic testing), it was counterproductive for others. Another key finding was that the targeted dynamic few-shot prompting did not consistently outperform random selection, indicating that the presumed benefits of closely matched examples may be counterbalanced by loss of broader contextual diversity. These findings suggest that the impact of prompt engineering is highly model and task-dependent, highlighting the need for tailored, context-aware strategies for integrating LLMs into healthcare.

arXiv

Bounding Hallucinations: Merlin-Arthur Protocols for Mutual-Information Bounds in Language Models

Björn Deiseroth, Max Henning Höth, Kristian Kersting, Letitia Parcalabescu

Retrieval-augmented generation (RAG) relies on retrieved context to guide large language models (LLM), yet treats the retrieval as a heuristic rather than verifiable evidence -- leading to unsupported answers, hallucinations, and reliance on spurious context. We introduce a novel evaluation, data augmentation, and training framework that treats the RAG pipeline as an interactive proof system by adapting the Merlin-Arthur (M/A) protocol: Arthur (the generator LLM) receives context of unknown provenance and Merlin gives helpful evidence, while Morgana injects adversarial, misleading context. We implement them both with an XAI method to self-assess and modify evidence most influential to Arthur. Based on this framework we propose the Explained Information Fraction (EIF) score, that disentangles explanation fidelity from model predictive errors and imperfect benchmarks, and normalizes M/A mutual-information lower bounds to realistic empirical settings. When trained with those contexts, Arthur learns to answer when evidence supports the answer and abstain when evidence is insufficient. Across five QA benchmarks and four LLMs (1B to 32B parameters), M/A reduces incorrect answers under insufficient context by up to 35pp during training and by 18-20pp over vanilla finetuning. Abstention emerges even \emph{without any manually annotated unanswerable example or preference pair}. We improve EIF-cond by 0.1-0.4 during M/A training and by 0.33-0.38 over vanilla finetuning. Reusing Merlin/Morgana contexts as automatic hard positives and negatives also raises retriever Recall@1 by 2pp. While high accuracy does not guarantee entropy flow from context to answer, our EIF scores -- to our knowledge, the first context-to-answer information bound for RAG generators -- show that autonomous interactive-proof-style supervision enables RAG systems that treat retrieved documents as verifiable evidence.

arXiv

MiraMind: Benchmarking Reliable Mental Health Reasoning beyond Answer Accuracy

Mengxi Xiao, Kailai Yang, Pengde Zhao, Enze Zhang, Ziyan Kuang, Zhiwei Liu, Weiguang Han, Shu Liao, Lianting Huang, Guojun Xiong, Victor Gutierrez Basulto, Jinpeng Hu, Min Peng, Qianqian Xie, Sophia Ananiadou

Mental-health reasoning with large language models (LLMs) is an evidence-constrained judgment problem: models must transform limited, subjective, and often ambiguous evidence into interpretations, decisions, or claims whose specificity, certainty, severity, and actionability remain warranted. Existing benchmarks mainly evaluate specific clinical roles or final answers, leaving the reliability of explicit reasoning trajectories under-specified. We introduce MiraMind, a unified benchmark spanning six task families and 13 datasets across appraisal, diagnosis, intervention, abstraction, and verification. MiraMind evaluates both task outcomes and the trajectories connecting evidence to judgment through usability, logical structure, and informational contribution. Evaluating 20 LLMs reveals a restraint gap, in which the specificity or certainty of model judgments exceeds what limited evidence supports. We further train Mindora, an 8B model that targets evidence-to-judgment transitions through hard-case supervision, structured trajectory rewriting, and consistency-aware optimization. Mindora achieves the best average rank on MiraMind, improves over its backbone across all six task families, and produces more balanced reasoning trajectories. These results show that MiraMind can expose shared restraint failures and evaluate whether targeted post-training improves evidence-constrained mental-health reasoning.

arXiv

FanarGuard: A Culturally-Aware Moderation Filter for Arabic Language Models

Masoomali Fatehkia, Enes Altinisik, Husrev Taha Sencar

Content moderation filters are a critical safeguard against alignment failures in language models. Yet most existing filters focus narrowly on general safety and overlook cultural context. In this work, we introduce FanarGuard, a bilingual moderation filter that evaluates both safety and cultural alignment in Arabic and English. We construct a dataset of over 468K prompt and response pairs, drawn from synthetic and public datasets, scored by a panel of LLM judges on harmlessness and cultural awareness, and use it to train two filter variants. To rigorously evaluate cultural alignment, we further develop the first benchmark targeting Arabic cultural contexts, comprising over 1k norm-sensitive prompts with LLM-generated responses annotated by human raters. Results show that FanarGuard achieves stronger agreement with human annotations than inter-annotator reliability, while matching the performance of state-of-the-art filters on safety benchmarks. These findings highlight the importance of integrating cultural awareness into moderation and establish FanarGuard as a practical step toward more context-sensitive safeguards.

arXiv

Penalizing Length: Uncovering Systematic Bias in Quality Estimation Metrics

Yilin Zhang, Wenda Xu, Zhongtao Liu, Tetsuji Nakagawa, Markus Freitag

Quality Estimation (QE) metrics are vital in machine translation for reference-free evaluation and increasingly serve as selection criteria in data filtering and candidate reranking. However, the prevalence and impact of length bias in QE metrics have been underexplored. Through a systematic study of top-performing learned and LLM-as-a-Judge QE metrics across 10 diverse language pairs, we reveal two critical length effects: First, QE metrics consistently over-predict errors with increasing translation length, even for high-quality, error-free texts. Second, when candidates of comparable quality are available for the same source text, learned QE metrics tend to favor shorter translations, whereas LLM-as-a-Judge metrics range from being approximately length-neutral to favoring longer translations. These metric-dependent length effects can favor or penalize translations based on length rather than quality and can propagate into downstream pipelines that rely on QE signals for data selection or system optimization. We trace the root cause of MetricX-24 QE's cumulative-length penalty to skewed supervision distributions, in which longer error-free examples are underrepresented in training data. As a diagnostic intervention, we apply length normalization during training and show that this simple modification effectively decouples error prediction from sequence length, yielding more reliable QE signals across translations of varying length.

arXiv

MetaLint: Easy-to-Hard Generalization for Code Linting

Atharva Naik, Lawanya Baghel, Dhakshin Govindarajan, Darsh Agrawal, Yiqing Xie, Daniel Fried, Carolyn Rose

Large language models excel at code generation but struggle with code linting, particularly in generalizing to unseen or evolving best practices beyond those observed during training. We introduce MetaLint, a meta-learning framework that formulates code linting as an instruction-following task, where a model evaluates whether code adheres to a natural language specification of best practices. In contrast to prior work that trains models to detect violations from a fixed set of best practices, MetaLint evaluates code against a provided natural language specification, enabling test-time control over which practices to enforce and generalization to unseen or evolving rules without retraining. We demonstrate that models trained solely on synthetic data generated from automatic linters still generalize to harder, context-dependent best practices for which such linters are not available. To evaluate generalization beyond such easy signals, we introduce a human-curated benchmark of hard best practices inspired by Python Enhancement Proposals (PEPs). On this benchmark, MetaLint substantially improves performance without explicit fine-tuning on target best practices and exhibits strong, easy-to-hard generalization. Qwen3-4B achieves a 2.7x detection F-score gain (25.9% -> 70.4%), the highest recall, and a 26.7% localization F-score, matching larger models such as o3-mini. These gains generalize across programming languages, model families, scales, reasoning settings, and linter sources. We release the code and benchmark to support reproducibility and future work.

arXiv

MateInfoUB: A Real-World Benchmark for Testing LLMs in Competitive, Multilingual, and Multimodal Educational Tasks

Adrian Marius Dumitran, Theodor-Pierre Moroianu, Mihnea-Vicentiu Buca

The rapid advancement of Large Language Models (LLMs) has transformed various domains, particularly computer science (CS) education. These models exhibit remarkable capabilities in code-related tasks and problem-solving, raising questions about their potential and limitations in advanced CS contexts. This study presents a novel bilingual (English-Romanian) multimodal (text and image) dataset of multiple-choice questions derived from a high-level computer science competition. A particularity of our dataset is that the problems are conceived such that some of them are easier solved using reasoning on paper, while for others writing code is more efficient. We systematically evaluate State of The Art LLMs on this dataset, analyzing their performance on theoretical programming tasks. Our findings reveal the strengths and limitations of current LLMs, including the influence of language choice (English vs. Romanian), providing insights into their applicability in CS education and competition settings. We also address critical ethical considerations surrounding educational integrity and the fairness of assessments in the context of LLM usage. These discussions aim to inform future educational practices and policies. To support further research, our dataset will be made publicly available in both English and Romanian. Additionally, we release an educational application tailored for Romanian students, enabling them to self-assess using the dataset in an interactive and practice-oriented environment.

arXiv

High-Layer Attention Pruning with Rescaling

Songtao Liu, Peng Liu

Pruning is a highly effective approach for compressing large language models (LLMs), significantly reducing inference latency. However, conventional training-free structured pruning methods often employ a heuristic metric that indiscriminately removes some attention heads across all pruning layers, without considering their positions within the network architecture. In this work, we propose a novel pruning algorithm that strategically prunes attention heads in the model's higher layers. Since the removal of attention heads can alter the magnitude of token representations, we introduce an adaptive rescaling parameter that calibrates the representation scale post-pruning to counteract this effect. We conduct comprehensive experiments on a wide range of LLMs, including LLaMA3.1-8B, Mistral-7B-v0.3, Qwen2-7B, and Gemma2-9B. Our evaluation includes both generation and discriminative tasks across 27 datasets. The results consistently demonstrate that our method outperforms existing structured pruning methods. This improvement is particularly notable in generation tasks, where our approach significantly outperforms existing baselines. Code is available at https://github.com/SongtaoLiu0823/HARP.

arXiv

PAM: Training Policy-Aligned Moderation Filters at Scale

Masoomali Fatehkia, Enes Altinisik, Mohamed Osman, Husrev Taha Sencar

Large language models (LLMs) remain vulnerable to misalignment and jailbreaks, making external safeguards like moderation filters essential, yet existing filters often focus narrowly on safety, falling short of the broader alignment needs seen in real-world deployments. We introduce Policy Aligned Moderation (PAM), a flexible framework for training custom moderation filters grounded in user-defined policies that extend beyond conventional safety objectives. PAM automates training data generation without relying on human-written examples, enabling scalable support for diverse, application-specific alignment goals and generation policies. PAM-trained filters match the performance of state-of-the-art safety moderation filters and policy reasoning models, and outperform them on PAMbench, four newly introduced user-annotated policy enforcement benchmarks that target age restrictions, dietary accommodations, cultural alignment, and limitations in medical guidance. These performance gains are achieved while the PAM filter runs 5-100x faster at inference than policy-conditioned reasoning models.

继续浏览

更多 AI 顶会论文

30+ 个顶会，按会议、年份和主题浏览。

浏览全部会议

进入任一会议和年份，查看完整论文列表。

全部会议 →

搜索论文

按标题、作者或关键词，在 21 万+ 篇论文里搜索。

搜索论文 →