Ax Shir Rozenfeld, Rahul Pankajakshan, Itay Zloczower, Eyal Lenga, Gilad Gressel, Yisroel Mirsky 5/1/2026

GAVEL: Towards Rule-Based Safety Through Activation Monitoring

Rule-based activation monitoring approach for LLM safety detection with improved precision and interpretability over broad training.

Ax Niv Fono, Yftah Ziser, Omer Ben-Porat 5/1/2026

From Competition to Collaboration: Designing Sustainable Mechanisms Between LLMs and Online Forums

Framework proposing collaboration mechanisms between GenAI systems and Q&A forums to address data dependency and user displacement.

Ax Hyomin Kim, Sang-Yeon Hwang, Jaechang Lim, Yinhua Piao, Yunhak Oh, Woo Youn Kim, Chanyoung Park, Sungsoo Ahn, Junhyeok Jeon 5/1/2026

Progressive Multi-Agent Reasoning for Biological Perturbation Prediction

Multi-agent reasoning system using LLMs to predict gene regulation and chemical perturbation responses in drug discovery.

Ax Yuyu Guo, Wenjie Yang, Siyuan Yang, Ziyang Liu, Cheng Chen, Yuan Wei, Yun Hu, Yang Huang, Guoliang Hao, Dongsheng Yuan, Jianming Wang, Xin Chen, Hang Yu, Lei Lei, Peng Di 5/1/2026

OpAgent: Operator Agent for Web Navigation

Web navigation agent using online reinforcement learning to handle complex, volatile real-world websites beyond offline datasets.

Ax Hugh Xuechen Liu, K{\i}van\c{c} Tatar 5/1/2026

Grounding Machine Creativity in Game Design Knowledge Representations: Empirical Probing of LLM-Based Executable Synthesis of Goal Playable Patterns under Structural Constraints

LLM-based synthesis of executable game design patterns using gameplay knowledge representations and structural constraints.

Ax Mohamed Elfeki, Tu Trinh, Kelvin Luu, Guangze Luo, Nathan Hunt, Ernesto Montoya, Nandan Marwaha, Yannis He, Charles Wang, Fernando Crabedo, Alessa Castilo, Bing Liu 5/1/2026

HiL-Bench (Human-in-Loop Benchmark): Do Agents Know When to Ask for Help?

Benchmark evaluating whether coding agents know when to request human help instead of guessing on ambiguous specifications.

Ax Yadong Li, Guoxin Wu, Haiping Hou, Biye Li 5/1/2026

UAF: A Unified Audio Front-end LLM for Full-Duplex Speech Interaction

End-to-end audio LLM architecture for full-duplex speech interaction avoiding latency and error propagation of cascaded pipelines.

Ax Xue Xia, Chengkai Yao, Mingyu Tsoi, Xinjie Mao, Wenxuan Huang, Jiaqi Wei, Hao Wu, Cheng Tan, Lang Yu, Yuejin Yang, Mengdi Liu, Siqi Sun, Zhangyang Gao 5/1/2026

AblateCell: A Reproduce-then-Ablate Agent for Virtual Cell Repositories

Coding agent system that reproduces baselines and performs ablations on AI Virtual Cell repositories to identify critical components.

Ax Chenyang An, Qihao Ye, Minghao Pan, Jiayaun Zhang 5/1/2026

QED: An Open-Source Multi-Agent System for Generating Mathematical Proofs on Open Problems

Open-source multi-agent system using LLMs to generate original proofs for open mathematical problems, identifying seven failure modes in proof generation.

Ax Chengyang Huang, Siddhartha Srivastava, Kenneth K. Y. Ho, Kathy E. Luker, Gary D. Luker, Xun Huan, Krishna Garikipati 5/1/2026

FP-IRL: Fokker--Planck Inverse Reinforcement Learning -- A Physics-Constrained Approach to Markov Decision Processes

Fokker-Planck inverse reinforcement learning approach for inferring reward functions in MDPs without requiring known transition dynamics.

Ax Lu Yin, Ajay Jaiswal, Shiwei Liu, Souvik Kundu, Zhangyang Wang 5/1/2026

Junk DNA Hypothesis: Pruning Small Pre-Trained Weights Irreversibly and Monotonically Impairs "Difficult" Downstream Tasks in LLMs

Study showing small-magnitude weights in pre-trained LLMs are essential for difficult downstream tasks, contradicting pruning assumptions.

Ax Rahul Nair, Bhanu Tokas, Hannah Kerner 5/1/2026

A Woman with a Knife or A Knife with a Woman? Measuring Directional Bias Amplification in Image Captions

Metric for measuring directional bias amplification in image captioning models, extending bias analysis beyond classification tasks.

Ax Musfiqur Rahman, SayedHassan Khatoonabadi, Emad Shihab 5/1/2026

OpenClassGen: A Large-Scale Corpus of Real-World Python Classes for LLM Research

OpenClassGen: Large-scale corpus of 324,843 real-world Python classes from open-source projects for training and evaluating code generation models.

Ax Luoxi Tang, Tharunya Sundar, Yuqiao Meng, Shuai Yang, Ankita Patra, Lakshmi Manohar Chippada, Jiqian Zhao, Yi Li, Weicheng Ma, Zhaohan Xi 5/1/2026

From Test-taking to Cognitive Scaffolding: A Pedagogical Diagnostic Benchmark for LLMs on English Standardized Tests

Pedagogical diagnostic framework for evaluating LLMs on English standardized tests, measuring reasoning fidelity and misconception diagnosis beyond accuracy scores.

Ax Zhuohan Xie, Daniil Orel, Rushil Thareja, Dhruv Sahnan, Hachem Madmoun, Fan Zhang, Debopriyo Banerjee, Georgi Georgiev, Xueqing Peng, Lingfei Qian, Jimin Huang, Jinyan Su, Aaryamonvikram Singh, Rui Xing, Rania Elbadry, Chen Xu, Haonan Li, Fajri Koto, Ivan Koychev, Tanmoy Chakraborty, Yuxia Wang, Salem Lahlou, Veselin Stoyanov, Sophia Ananiadou, Preslav Nakov 5/1/2026

FinChain: A Symbolic Benchmark for Verifiable Chain-of-Thought Financial Reasoning

FinChain benchmark for evaluating chain-of-thought reasoning in financial analysis, addressing gap in existing datasets by focusing on intermediate reasoning steps rather than just final answers.

Ax Wenrui Zhou, Mohamed Hendy, Shu Yang, Qingsong Yang, Zikun Guo, Yuyu Luo, Lijie Hu, Di Wang 5/1/2026

Flattery in Motion: Benchmarking and Analyzing Sycophancy in Video-LLMs

Benchmark analyzing sycophancy in video-LLMs. Measures tendency to align with user input contradicting visual evidence. Evaluates factual consistency.

Ax Weiyi Zhao, Xiaoyu Tan, Liang Liu, Sijia Li, Youwei Song, Xihe Qiu 5/1/2026

OR-VSKC: Resolving Visual-Semantic Knowledge Conflicts in Operating Rooms with Synthetic Data-Guided Alignment

Addresses visual-semantic knowledge conflicts in medical MLLMs. Uses synthetic data to align safety knowledge activation in surgical contexts.

Ax Perry Dong, Qiyang Li, Dorsa Sadigh, Chelsea Finn 5/1/2026

EXPO: Stable Reinforcement Learning with Expressive Policies

EXPO: reinforcement learning approach for training expressive policies with diffusion/flow-matching. Addresses stability challenges in online RL fine-tuning.

Ax Santosh Vasa, Aditi Ramadwar, Jnana Rama Krishna Darabattula, Md Zafar Anwar, Stanislaw Antol, Andrei Vatavu, Thomas Monninger, Sihao Ding 5/1/2026

AutoVDC: Automated Vision Data Cleaning Using Vision-Language Models

Framework using vision-language models for automated dataset quality control. Reduces manual annotation effort for autonomous driving datasets.

Ax Joshua Ong Jun Leang, Zheng Zhao, Aryo Pradipta Gema, Sohee Yang, Wai-Chung Kwan, Xuanli He, Wenda Li, Pasquale Minervini, Eleonora Giunchiglia, Shay B. Cohen 5/1/2026

PiCSAR: Probabilistic Confidence Selection And Ranking for Reasoning Chains

PiCSAR: training-free scoring function for selecting correct reasoning chains. Improves LLM and reasoning model accuracy via best-of-n sampling.

Ax Deniz Bayazit, Aaron Mueller, Antoine Bosselut 5/1/2026

Crosscoding Through Time: Tracking Emergence & Consolidation Of Linguistic Representations Throughout LLM Pretraining

Sparse coding analysis of how linguistic representations emerge during LLM pretraining. Tracks when models acquire specific linguistic abilities.

Ax Ching-Chun Chang, Isao Echizen 5/1/2026

Tell-Tale Watermarks for Explanatory Reasoning in Synthetic Media Forensics

Forensics method using watermarks for detecting synthetic media edits. Provides explanatory reasoning for identifying semantic, photometric, and geometric alterations.

Ax Brahim Mahmoudi, Naouel Moha, Quentin Sti\'evenart, Florent Avellaneda 5/1/2026

ML Code Smells: From Specification to Detection

Framework to detect code smells in ML pipelines. Identifies implementation practices that compromise reproducibility, robustness, and maintainability.

Ax Lute Lillo, Nick Cheney 5/1/2026

Activation Function Design Sustains Plasticity in Continual Learning

Studies activation function design impact on continual learning. Shows role of non-linearity in preventing plasticity loss during task adaptation.

Ax Yucheng Wang, Yifan Hou, Aydin Javadov, Mubashara Akhtar, Mrinmaya Sachan 5/1/2026

Compose and Fuse: Revisiting the Foundational Bottlenecks in Multimodal Reasoning

Analysis of multimodal reasoning in MLLMs. Investigates when cross-modal interactions improve or undermine performance on reasoning tasks.

Ax Jinho Chang, Jaemin Kim, Jong Chul Ye 5/1/2026

Training-Free Reward-Guided Image Editing via Trajectory Optimal Control

Training-free reward-guided image editing using diffusion and flow-matching models. Steers generation via trajectory optimal control.

Ax Francesca Gomez 5/1/2026

From surveillance to signalling: escalation channels as environmental controls for agentic AI

Environmental controls for AI agent safety. Proposes escalation channels to handle task-rule conflicts during agentic operation with sensitive information.

Ax Lorenzo Nikiforos, Luciano Prono, Charalampos Antoniadis, Fabio Pareschi, Riccardo Rovatti, Gianluca Setti 5/1/2026

Vanishing Contributions: A Unified Framework for Smooth and Iterative Model Compression

Unified framework for model compression combining pruning, quantization, and low-rank decomposition with iterative gradual techniques.

Ax Lang Gao, Xuhui Li, Chenxi Wang, Mingzhe Li, Wei Liu, Zirui Song, Jinghui Zhang, Rui Yan, Preslav Nakov, Xiuying Chen 5/1/2026

When Personalization Tricks Detectors: The Feature-Inversion Trap in Machine-Generated Text Detection

Benchmark dataset for detecting personalized machine-generated text. Evaluates detector robustness when LLMs imitate personal writing styles.

Ax Jiaqi Leng, Xiang Hu, Junxiong Wang, Jianguo Li, Wei Wu, Yucheng Lu 5/1/2026

Understanding and Improving Length Generalization in Hierarchical Sparse Attention Models

Studies length generalization in sparse attention models for long context processing. Compares sliding window, state space models, and chunk-based sparse attention.

Ax Ming Li, Pei Chen, Zhenhao Zhang, Tao Yang, Xinyang Zhang, Han Li, Tianyu Cao, Ming Zeng, Zhuofeng Wu, Meng Jiang, Huasheng Li, Lihong Li, Bing Yin 5/1/2026

Mitigating Lost in Multi-turn Conversation via Curriculum RL with Verifiable Accuracy and Abstention Rewards

Curriculum RL framework (RLAAR) to improve LLM multi-turn conversation performance. Addresses Lost-in-Conversation degradation using verifiable rewards.

Ax Elena Celledoni, Brynjulf Owren, Lars Ruthotto, Tianjiao Nicole Yang 5/1/2026

Mixed Precision Training of Neural ODEs

Mixed precision training techniques for neural ODEs addressing computational costs and numerical stability.

Ax Meixiu Long, Duolin Sun, Dan Yang, Yihan Jiao, Lei Liu, Jiahai Wang, BinBin Hu, Yue Shen, Jie Feng, Zhehao Tan, Junjie Wang, Lianzhen Zhong, Jian Wang, Peng Wei, Jinjie Gu 5/1/2026

GroupRank: A Groupwise Paradigm for Effective and Efficient Passage Reranking with LLMs

GroupRank method for efficient passage reranking with LLMs using groupwise paradigm to balance efficiency-accuracy tradeoff in information retrieval.

Ax Daniel Agyei Asante, Md Mokarram Chowdhury, Yang Li 5/1/2026

Decomposed Trust: Privacy, Adversarial Robustness, Ethics, and Fairness in Low-Rank LLMs

Study of trustworthiness implications including privacy, robustness, ethics, and fairness in low-rank factorized LLMs for resource-constrained deployment.

Ax Yuji Takubo, Arpit Dwivedi, Sukeerth Ramkumar, Luis A. Pabon, Daniele Gammelli, Marco Pavone, Simone D'Amico 5/1/2026

Language-Conditioned Safe Trajectory Generation for Spacecraft Rendezvous

Language-conditioned trajectory optimization for autonomous spacecraft rendezvous using deep learning instead of manual expert inputs.

Ax Arijit Ray, Ahmed Abdelkader, Chengzhi Mao, Bryan A. Plummer, Kate Saenko, Ranjay Krishna, Leonidas Guibas, Wen-Sheng Chu 5/1/2026

Mull-Tokens: Modality-Agnostic Latent Thinking

Multimodal model using latent thinking tokens for reasoning across text, images, and other modalities without specialist tools or handcrafted data.

Ax Zhenguo Zhang, Haohan Zheng, Yishen Wang, Le Xu, Tianchen Deng, Xuefeng Chen, Qu Chen, Bo Zhang, Wuxiong Huang 5/1/2026

OmniDrive-R1: Reinforcement-driven Interleaved Multi-modal Chain-of-Thought for Trustworthy Vision-Language Autonomous Driving

Vision-Language Model for autonomous driving using reinforcement-driven multimodal chain-of-thought reasoning to mitigate hallucination and improve reliability.

Ax Yonggan Fu, Lexington Whalen, Zhifan Ye, Xin Dong, Shizhe Diao, Jingyu Liu, Chengyue Wu, Hao Zhang, Enze Xie, Song Han, Maksim Khadkevich, Jan Kautz, Yingyan Celine Lin, Pavlo Molchanov 5/1/2026

Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speed

Research on converting pretrained autoregressive language models into efficient diffusion language models for faster parallel generation while maintaining accuracy.

Ax Chao Hu, Wenhao Zeng, Yuling Shi, Beijun Shen, Xiaodong Gu 5/1/2026

In Line with Context: Repository-Level Code Generation via Context Inlining

Research on repository-level code generation using context inlining to improve model understanding of complex dependencies.

Ax Kai Li, Xuanqing Yu, Ziyi Ni, Yi Zeng, Yao Xu, Zheqing Zhang, Xin Li, Jitao Sang, Xiaogang Duan, Xuelei Wang, Chengbao Liu, Jie Tan 5/1/2026

TiMem: Temporal-Hierarchical Memory Consolidation for Long-Horizon Conversational Agents

TiMem temporal-hierarchical memory framework for long-horizon conversational agents, organizing interaction histories across multiple time scales.

Ax Ahmed Abdelkawy, Ahmed Elsayed, Asem Ali, Aly Farag, Thomas Tretter, Michael McIntyre 5/1/2026

Context Matters: Peer-Aware Student Behavioral Engagement Measurement via VLM Action Parsing and LLM Sequence Classification

Combines VLM action parsing and LLM sequence classification to measure student behavioral engagement in classrooms using peer context.

Ax Yixi Zhou, Fan Zhang, Yu Chen, Haipeng Zhang, Preslav Nakov, Zhuohan Xie 5/1/2026

FinCARDS: Card-Based Analyst Reranking for Financial Document Question Answering

FinCARDS structured reranking framework for financial document QA, optimizing constraint-based evidence selection over semantic relevance.

Ax Jihang Li, Qing Liu, Zulong Chen, Jing Wang, Wei Wang, Chuanfei Xu, Zeyi Wen 5/1/2026

Taxon: Hierarchical Tax Code Prediction with Semantically Aligned LLM Expert Guidance

Taxon framework uses LLM expert guidance for hierarchical tax code prediction in e-commerce compliance and invoicing automation.

Ax Mark Rothermel, Marcus Kornmann, Marcus Rohrbach, Anna Rohrbach 5/1/2026

VeriTaS: The First Dynamic Benchmark for Multimodal Automated Fact-Checking

VeriTaS dynamic benchmark for multimodal automated fact-checking, designed to prevent data leakage from LLM pretraining.

Ax Ruozhen Yang, Yucheng Jiang, Yueqi Jiang, Priyanka Kargupta, Yunyi Zhang, Jiawei Han 5/1/2026

Grounding Agent Memory in Contextual Intent

STITCH agentic memory system indexes trajectory steps with contextual intent for long-horizon goal-oriented LLM interactions.

Ax Yelin Chen, Fanjin Zhang, Suping Sun, Yunhe Pang, Yuanchun Wang, Jian Song, Xiaoyan Li, Lei Hou, Shu Zhao, Jie Tang, Juanzi Li 5/1/2026

RPC-Bench: A Fine-grained Benchmark for Research Paper Comprehension

RPC-Bench large-scale QA benchmark from research paper review exchanges evaluates LLMs on paper comprehension with fine-grained metrics.

Ax Ken Deng, Yifu Qiu, Yoni Kasten, Shay B. Cohen, Yftah Ziser 5/1/2026

Lost in Space? Vision-Language Models Struggle with Relative Camera Pose Estimation

Studies vision-language model performance on relative camera pose estimation, introducing benchmarks for multi-view spatial reasoning.

Ax Weiyu Sun, Liangliang Chen, Yongnuo Cai, Huiru Xie, Yi Zeng, Ying Zhang 5/1/2026

EDU-CIRCUIT-HW: Evaluating Multimodal Large Language Models on Real-World University-Level STEM Student Handwritten Solutions

EDU-CIRCUIT-HW benchmark evaluates multimodal LLMs on real-world university-level STEM student handwritten solutions with mixed mathematical content.

Ax Huiyang Yi, Xiaojian Shen, Yonggang Wu, Duxin Chen, He Wang, Wenwu Yu 5/1/2026

CausalCompass: Evaluating the Robustness of Time-Series Causal Discovery in Misspecified Scenarios

CausalCompass benchmark framework evaluates robustness of time-series causal discovery methods under violated causal assumptions.

Ax Nuno Fachada, Daniel Fernandes, Carlos M. Fernandes, Jo\~ao P. Matos-Carvalho 5/1/2026

Can Large Language Models Implement Agent-Based Models? An ODD-based Replication Study

Evaluates whether 17 contemporary LLMs can reliably implement agent-based models from ODD specifications, testing code synthesis capabilities.