Ax Shu Wang, Yixiang Fang, Yingli Zhou, Xilin Liu, Yuchi Ma 5/12/2026

ArchRAG: Attributed Community-based Hierarchical Retrieval-Augmented Generation

Hierarchical retrieval-augmented generation system using attributed community graphs for improved LLM question-answering.

Ax Lior Cohen, Kaixin Wang, Bingyi Kang, Uri Gadot, Shie Mannor 5/12/2026

Simulus: Combining Improvements in Sample-Efficient World Model Agents

Framework combining multiple improvements to world model agents for sample-efficient reinforcement learning.

Ax Shwai He, Weilin Cai, Jiayi Huang, Ang Li 5/12/2026

Capacity-Aware Inference: Mitigating the Straggler Effect in Mixture of Experts

Capacity-aware inference method addressing straggler effects in Mixture of Experts architectures during distributed inference.

Ax Nikhil Abhyankar, Parshin Shojaee, Chandan K. Reddy 5/12/2026

LLM-FE: Automated Feature Engineering for Tabular Data with LLMs as Evolutionary Optimizers

Automated feature engineering system using LLMs as evolutionary optimizers for tabular data machine learning tasks.

Ax Ming Li, Yanhong Li, Ziyue Li, Tianyi Zhou 5/12/2026

How Instruction and Reasoning Data shape Post-Training: Data Quality through the Lens of Layer-wise Gradients

Spectral analysis of layer-wise gradients to evaluate data quality in LLM post-training on instruction and reasoning tasks.

Ax Rui Wang, Junda Wu, Yu Xia, Tong Yu, Ruiyi Zhang, Ryan Rossi, Subrata Mitra, Lina Yao, Julian McAuley 5/12/2026

CachePrune: Teaching LLMs What Not to Follow via KV-Cache Editing

Defense method against indirect prompt injection attacks via KV-cache pruning to prevent unintended instruction-following in LLMs.

Ax Chenxi Jiang, Chuhao Zhou, Jianfei Yang 5/12/2026

REI-Bench: Can Embodied Agents Understand Vague Human Instructions in Task Planning?

Benchmark evaluating LLM-based robot task planners' ability to handle vague human instructions through semantic parsing.

Ax Yue Fan, Xuehai He, Diji Yang, Kaizhi Zheng, Ching-Chen Kuo, Yuting Zheng, Sravana Jyothi Narayanaraju, Xinze Guan, Xin Eric Wang 5/12/2026

GRIT: Teaching MLLMs to Think with Images

Method enabling multimodal LLMs to integrate visual reasoning with image information using reinforcement learning.

Ax Xinran Gu, Kaifeng Lyu, Jiazheng Li, Jingzhao Zhang 5/12/2026

Data Mixing Can Induce Phase Transitions in Knowledge Acquisition

Analysis of phase transitions in LLM knowledge acquisition when training on mixed data sources with varying quality levels.

Ax Liangwei Nathan Zheng, Wei Emma Zhang, Mingyu Guo, Olaf Maennel, Weitong Chen 5/12/2026

Rethinking Gating Mechanism in Sparse MoE: Handling Arbitrary Modality Inputs with Confidence-Guided Gate

Confidence-guided gating mechanism for Sparse MoE architectures to handle missing modalities in multimodal learning.

Ax Chenhao Zheng, Jieyu Zhang, Mohammadreza Salehi, Ziqi Gao, Vishnu Iyengar, Norimasa Kobori, Quan Kong, Ranjay Krishna 5/12/2026

One Trajectory, One Token: Grounded Video Tokenization via Panoptic Sub-object Trajectory

Novel video tokenization approach using panoptic sub-object trajectories to reduce tokens for transformer-based long video processing.

Ax Xenia Heilmann, Luca Corbucci, Mattia Cerrato, Anna Monreale 5/12/2026

FeDa4Fair: Client-Level Federated Datasets for Fairness Evaluation

arXiv paper on federated learning fairness evaluation, addressing discrimination at client level despite global model fairness.

Ax Weihong Li, Baohong Li, Anpeng Wu, Zhihan Li, Ming Ma, Keting Yin, Kun Kuang 5/12/2026

Causal Discovery for Irregularly Time Series with Consistency Guarantees

Causal discovery method for irregularly sampled time series with consistency guarantees, addressing missing data challenges.

Ax Qiyang Li, Zhiyuan Zhou, Sergey Levine 5/12/2026

Reinforcement Learning with Action Chunking

Q-chunking algorithm for improving reinforcement learning on long-horizon sparse-reward tasks in offline-to-online settings.

Ax Paulo Salem, Robert Sim, Christopher Olsen, Prerit Saxena, Rafael Barcelos, Yi Ding 5/12/2026

TinyTroupe: An LLM-powered Multiagent Persona Simulation Toolkit

TinyTroupe toolkit for LLM-powered multiagent persona simulation with fine-grained specifications for realistic human behavior simulation.

Ax Conrad Borchers, Bahar Shahrokhian, Francesco Balzan, Elham Tajik, Sreecharan Sankaranarayanan, Sebastian Simon 5/12/2026

Temperature and Persona Shape LLM Agent Consensus With Minimal Accuracy Gains in Qualitative Coding

Experimental study of multi-agent LLM systems for qualitative coding, examining how persona and temperature affect consensus-building vs single agents.

Ax Tom S\"uhr, Florian E. Dorner, Olawale Salaudeen, Augustin Kelava, Samira Samadi 5/12/2026

Position: Stop Evaluating AI with Human Tests, Develop Principled, AI-specific Tests instead

Argues LLM evaluations on human-designed tests are ontologically flawed; proposes AI-specific evaluation frameworks instead of anthropomorphizing model performance.

Ax Yadong Niu, Tianzi Wang, Heinrich Dinkel, Xingwei Sun, Jiahao Zhou, Gang Li, Jizhong Liu, Xunying Liu, Junbo Zhang, Jian Luan 5/12/2026

MECAT: A Multi-Experts Constructed Benchmark for Fine-Grained Audio Understanding Tasks

MECAT benchmark for fine-grained audio understanding using multi-expert annotations to distinguish detailed model outputs.

Ax Yunfeng Ge, Ming Jin, Yiji Zhao, Hongyan Li, Bo Du, Chang Xu, Shirui Pan 5/12/2026

EventTSF: Event-Aware Non-Stationary Time Series Forecasting

EventTSF integrates natural language events with time series data for improved non-stationary forecasting.

Ax Koffi Ismael Ouattara, Ioannis Krontiris, Theo Dimitrakos, Frank Kargl 5/12/2026

Assessing Trustworthiness of AI Training Dataset using Subjective Logic -- A Use Case on Bias

Subjective logic framework for assessing trustworthiness of AI training datasets regarding bias and fairness.

Ax Yang Qiao, Yuntong Hu, Bowen Zhu, Hasibul Haque, Liang Zhao 5/12/2026

Multimodal Representation Learning Conditioned on Semantic Relations

Multimodal representation learning conditioned on semantic relations beyond single shared embeddings like CLIP.

Ax Haibo Jin, Ruoxi Chen, Peiyan Zhang, Andy Zhou, Zelei Cheng, Haohan Wang 5/12/2026

GUARD: Guideline Upholding Test through Adaptive Role-play and Jailbreak Diagnostics for LLMs

GUARD testing framework translating ethics guidelines into actionable jailbreak diagnostics for LLM safety evaluation.

Ax Zhiyang Chen, Tara Saba, Xun Deng, Xujie Si, Fan Long 5/12/2026

Scam2Prompt: A Scalable Framework for Auditing Malicious Scam Endpoints in Production LLMs

Scam2Prompt framework auditing malicious endpoints in production LLMs for security vulnerabilities from training data.

Ax Erfan Baghaei Potraghloo, Seyedarmin Azizi, Souvik Kundu, Massoud Pedram 5/12/2026

Top-H Decoding: Adapting the Creativity and Coherence with Bounded Entropy in Text Generation

Top-H decoding method balancing creativity and coherence in LLM text generation via bounded entropy.

Ax Hoang Nguyen, Sidharth Surapaneni, Akshay Kalkunte, Jash Mehta, Aman Tiwari, Oluwanifemi Bamgbose, Khyati Mahajan, Jash Shah, Shruthan Radhakrishna, Sathwik Tejaswi Madhusudhan, Vikas Yadav, Sai Rajeswar 5/12/2026

AU-Harness: An Open-Source Toolkit for Holistic Evaluation of Audio LLMs

AU-Harness open-source toolkit for standardized evaluation of audio language models with multi-turn dialogue support.

Ax Hangzhan Jin, Sitao Luan, Tianwei Ni, Sicheng Lyu, Guillaume Rabusseau, Reihaneh Rabbany, Doina Precup, Mohammad Hamdaqa 5/12/2026

RL Fine-Tuning Heals OOD Forgetting in SFT

Analysis of why SFT+RL post-training works: SFT peaks on OOD early then declines; RL recovers generalization.

Ax Yu Wang, Sheng Shen, R\'emi Munos, Hongyuan Zhan, Yuandong Tian 5/12/2026

Positional Encoding via Token-Aware Phase Attention

Token-Aware Phase Attention (TAPA) positional encoding improving long-context modeling beyond RoPE limitations.

Ax Naibin Gu, Zhenyu Zhang, Yuchen Feng, Yilong Chen, Peng Fu, Zheng Lin, Shuohuan Wang, Yu Sun, Hua Wu, Weiping Wang, Haifeng Wang 5/12/2026

Elastic MoE: Unlocking the Inference-Time Scalability of Mixture-of-Experts

Elastic MoE method enabling runtime scaling of activated experts for heterogeneous hardware and varying workloads.

Ax Juangui Xu, Zikun Guo, Jingwei Lv, Xinyue Xu, Hongbin Lin, Shu Yang, Jun Wen, Di Wang, Lijie Hu 5/12/2026

Benchmarking and Mitigating Sycophancy in Medical Vision Language Models

Benchmark and mitigation techniques for sycophancy in medical vision language models for visual QA tasks.

Ax Jose I. Mestre, Alberto Fern\'andez-Hern\'andez, Cristian P\'erez-Corral, Manuel F. Dolz, Jose Duato, Enrique S. Quintana-Ort\'i 5/12/2026

GLAI: GreenLightningAI for Accelerated Training through Knowledge Decoupling

GLAI architectural block separating structural and quantitative knowledge for accelerated neural network training.

Ax Richard A. Dubniczky, Bertalan Borsos, Tamas Bisztray, Norbert Tihanyi 5/12/2026

You Have Been LaTeXpOsEd: A Systematic Analysis of Information Leakage in Preprint Archives Using Large Language Models

Security analysis of information leakage from LaTeX sources and metadata in arXiv preprints using LLMs.

Ax Xiaoyu Yang, En Yu, Wei Duan, Jie Lu 5/12/2026

Turning Drift into Constraint: Robust Reasoning Alignment in Non-Stationary Multi-Stream Environments

Multi-source reasoning alignment for MLLMs addressing concept drift in non-stationary environments via constraint satisfaction.

Ax Kai Yang, Yuqi Huang, Junheng Tao, Wanyu Wang, Qitian Wu 5/12/2026

PAINET: A Principled Efficient Transformer for 3D Dynamics Modeling

PAINET transformer for 3D dynamics modeling in multi-body systems using geometric symmetries for trajectory prediction.

Ax Firoj Alam, Ali Ezzat Shahroor, Md. Arid Hasan, Zien Sheikh Ali, Hunzalah Hassan Bhatti, Mohamed Bayan Kmainasi, Shammur Absar Chowdhury, Basel Mousi, Fahim Dalvi, Nadir Durrani, Natasa Milic-Frayling 5/12/2026

EverydayMMQA: A Multilingual and Multimodal Framework for Culturally Grounded Spoken Visual QA

EverydayMMQA dataset for multilingual/multimodal visual QA with cultural grounding using OASIS framework for low-resource languages.

Ax Prakhar Srivastava, Farrin Marouf Sofian, Francesco Immorlano, Kushagra Pandey, Stephan Mandt 5/12/2026

Control-Augmented Autoregressive Diffusion for Data Assimilation

Framework augmenting auto-regressive diffusion models with offline-trained controllers for guided generation via stepwise corrections.

Ax Shahriar Kabir Nahin, Hadi Askari, Muhao Chen, Anshuman Chhabra 5/12/2026

Less Diverse, Less Safe: The Indirect But Pervasive Risk of Test-Time Scaling in Large Language Models

Test-time scaling in LLMs reduces safety and reliability when candidate diversity decreases, revealing failure mode.

Ax Xinhu Li, Ayush Jain, Zhaojing Yang, Yigit Korkmaz, Erdem B{\i}y{\i}k 5/12/2026

When a Robot is More Capable than a Human: Learning from Constrained Demonstrators

Learning from constrained demonstrations for robots with limited control interfaces via policy learning.

Ax Kangping Hu, Stephen Mussmann 5/12/2026

Batch Bayesian Active Learning with Partial Batch Label Sampling

Batch Bayesian active learning with partial label sampling addressing scalability of acquisition functions.

Ax Zhongze Cai, Xiaocheng Li 5/12/2026

Users as Annotators: LLM Preference Learning from Comparison Mode

User-in-the-loop pairwise preference collection for LLM alignment training without professional annotators.

Ax Xi Yu, Yang Yang, Qun Liu, Yonghua Du, Sean McSweeney, Yuewei Lin 5/12/2026

GenCellAgent: Generalizable, Training-Free Cellular Image Segmentation via Large Language Model Agents

GenCellAgent multi-agent framework for training-free cellular image segmentation using vision-language models with planner-executor-evaluator loop.

Ax Katarzyna Kobalczyk, Zhiyuan Jerry Lin, Benjamin Letham, Zhuokai Zhao, Maximilian Balandat, Eytan Bakshy 5/12/2026

LILO: Bayesian Optimization with Natural Language Feedback

LILO framework uses LLMs to translate natural language feedback into structured preference signals for Bayesian optimization.

Ax Lukas Helff, Ruben H\"arle, Wolfgang Stammer, Felix Friedrich, Manuel Brack, Antonia W\"ust, Hikaru Shindo, Patrick Schramowski, Kristian Kersting 5/12/2026

ActivationReasoning: Logical Reasoning in Latent Activation Spaces

ActivationReasoning framework enables systematic logical reasoning in LLM latent spaces using sparse autoencoders for interpretability.

Ax Yulong Chen, Qi Zhang, Jiawen Zhang, Yadong Liu, Mu Li, Jie Wen, Yong Xu 5/12/2026

SAID: Safety-Aware Intent Defense via Prefix Probing for Large Language Models

SAID defense mechanism against jailbreaks via prefix probing for safety-aligned LLMs without external filtering.

Ax Yair Feldman, Yoav Artzi 5/12/2026

No Mean Feat: Simple, Strong Baselines for Context Compression

BenchPress evaluation suite and strong simple baselines for context compression in retrieval-augmented generation systems.

Ax Yuxiang Chen, Yifan Liu, Xiaoming Xu, Pengle Zhang, Michael Beyer, Martin Rapp, Jun Zhu, Jianfei Chen 5/12/2026

TetraJet-v2: Accurate NVFP4 Training for Large Language Models with Oscillation Suppression and Outlier Control

TetraJet-v2 enables 4-bit fully-quantized LLM training using NVFP4 format with oscillation suppression and outlier control.

Ax Bo Bai 5/12/2026

Forget BIT, It is All about TOKEN: Towards Semantic Information Theory for LLMs

Semantic Information Theory for LLMs replacing classical information theory bits with tokens, developing first-principles theory.

Ax Hanchen Li, Runyuan He, Qiuyang Mang, Qizheng Zhang, Huanzhi Mao, Xiaokun Chen, Hangrui Zhou, Alvin Cheung, Joseph Gonzalez, Ion Stoica 5/12/2026

Continuum: Efficient and Robust Multi-Turn LLM Agent Scheduling with KV Cache Time-to-Live

Continuum improves multi-turn LLM agent inference efficiency via KV cache time-to-live management across tool-interleaved calls.

Ax Wenlong Shang, Shihao Tian, Xutong Wan, Peng Chang 5/12/2026

COGNOS: Universal Enhancement for Time Series Anomaly Detection via Constrained Gaussian-Noise Optimization and Smoothing

COGNOS addresses MSE loss limitations in time series anomaly detection via constrained Gaussian-noise optimization and smoothing.

Ax Chanakya Ekbote, Vijay Lingam, Sujay Sanghavi, Jun Huan, Behrooz Omidvar-Tehrani, Anoop Deoras, Stefano Soatto 5/12/2026

MURPHY: Feedback-Aware GRPO with Retrospective Credit Assignment for Multi-Turn Code Generation

MURPHY improves GRPO training with feedback-aware retrospective credit assignment for multi-turn agentic code generation tasks.

Ax Punit Kumar, Asif Imran, Tevfik Kosar 5/12/2026

Energy Consumption of Dataframe Libraries for End-to-End Deep Learning Pipelines:A Comparative Analysis

Comparative analysis of Pandas, Polars, and Dask dataframe libraries in deep learning pipelines focusing on energy consumption and GPU interaction.