Ax Yule Liu, Heyi Zhang, Jinyi Zheng, Zhen Sun, Zifan Peng, Jiaheng Wei, Tianshuo Cong, Yilong Yang, Xinlei He 5/12/2026

Auditing Data Membership in Reinforcement Learning With Verifiable Rewards

Membership inference attacks to audit unauthorized data use in RLVR training pipelines used in LLM post-training.

Ax Zhirui Liu, Kaiyang Ji, Ke Yang, Yahao Fan, Jingyi Yu, Ye Shi, Jingya Wang 5/12/2026

Commanding Humanoid by Free-form Language: A Large Language Action Model with Unified Motion Vocabulary

Large Language Action model (Humanoid-LLA) that translates free-form natural language commands to robotic humanoid control with diverse motions.

Ax Timo Stoll, Chendi Qian, Ben Finkelshtein, Ali Parviz, Darius Weber, Fabrizio Frasca, Hadar Shavit, Antoine Siraudin, Arman Mielke, Marie Anastacio, Erik M\"uller, Maya Bechler-Speicher, Michael Bronstein, Mikhail Galkin, Holger Hoos, Mathias Niepert, Bryan Perozzi, Jan T\"onshoff, Christopher Morris 5/12/2026

GraphBench: Next-generation graph learning benchmarking

GraphBench provides standardized benchmarking suite for graph machine learning with consistent evaluation protocols for foundation models.

Ax Leyang Shen, Yang Zhang, Chun Kai Ling, Xiaoyan Zhao, Tat-Seng Chua 5/12/2026

CARL: Criticality-Aware Agentic Reinforcement Learning

CARL improves multi-step reinforcement learning for agents by identifying and optimizing criticality-aware action choices rather than treating all steps equally.

Ax Joshua Ward, Bochao Gu, Chi-Hua Wang, Guang Cheng 5/12/2026

When Tables Leak: Attacking String Memorization in LLM-Based Tabular Data Generation

Security analysis showing LLM-based tabular data generation systems memorize and leak sensitive string information from training data.

Ax Nirhoshan Sivaroopan, Hansi Karunarathna, Chamara Madarasingha, Anura Jayasumana, Kanchana Thilakarathna 5/12/2026

RAG-HAR: Retrieval Augmented Generation-based Human Activity Recognition

RAG-HAR applies retrieval-augmented generation with LLMs for human activity recognition without dataset-specific training.

Ax Ming Li, Han Chen, Yunze Xiao, Jian Chen, Hong Jiao, Tianyi Zhou 5/12/2026

Can LLMs Estimate Student Struggles? Human-AI Difficulty Alignment with Proficiency Simulation for Item Difficulty Prediction

Analysis of whether LLMs can estimate question difficulty and perceive student learning struggles for educational assessment.

Ax Tiange Luo, Lajanugen Logeswaran, Jaekyeom Kim, Justin Johnson, Honglak Lee 5/12/2026

Selective LoRA for Visual Tokens and Attention Heads

Parameter-efficient fine-tuning method applying LoRA selectively to visual tokens and attention heads in vision-language models.

Ax Ming Li, Chenrui Fan, Yize Cheng, Soheil Feizi, Tianyi Zhou 5/12/2026

Schoenfeld's Anatomy of Mathematical Reasoning by Language Models

Framework (ThinkARM) for analyzing and abstracting LLM reasoning traces using episode theory into functional reasoning steps.

Ax Wenxuan Xu, Arvind Pillai, Subigya Nepal, Amanda C Collins, Daniel M Mackin, Michael V Heinz, Tess Z Griffin, Nicholas C Jacobson, Andrew Campbell 5/12/2026

LENS: LLM-Enabled Narrative Synthesis for Mental Health by Aligning Multimodal Sensing with Language Models

Framework aligning multimodal health sensor data with LLMs to generate clinical narratives for mental health assessment.

Ax Yiwen Chen, Yiqing Wu, Huishi Luo, Fuzhen Zhuang, Deqing Wang, Zhao Zhang 5/12/2026

TextBridgeGNN: Pre-training Graph Neural Network for Cross-Domain Recommendation via Text-Guided Transfer

Graph neural network approach for cross-domain recommendation using text-guided transfer learning without ID embeddings.

Ax Yuhuan You, Lai Wei, Xihong Wu, Tianshu Qu 5/12/2026

The World is Not Mono: Enabling Spatial Understanding in Large Audio-Language Models

Framework for spatial audio understanding in audio-language models including localization and scene-level plausibility reasoning.

Ax Hossein Hosseini Kasnavieh, Gholamreza Haffari, Chris Leckie, Adel N. Toosi 5/12/2026

IntroLM: Introspective Language Models via Prefilling-Time Self-Evaluation

Method enabling LLMs to self-evaluate output quality during inference without external classifiers, via prefilling-time introspection.

Ax Chenxiao Yu, Bowen Yi, Farzan Karimi-Malekabadi, Suhaib Abdurahman, Jinyi Ye, Shrikanth Narayanan, Yue Zhao, Morteza Dehghani 5/12/2026

Tracing Moral Foundations in Large Language Models

Study of how moral foundations are encoded and organized across 14 LLMs to determine if models have internal moral structure.

Ax Qiuyu Tian, Zequn Liu, Yiding Li, Fengyi Chen, Zequn Liu, Youyong Kong, Fan Guo, Yuyao Li, Jinjing Shen, Zhijing Xie, Yiyun Luo, Xin Zhang, Yingce Xia 5/12/2026

STAGE: A Full-Screenplay Benchmark for Reasoning over Evolving Storie

Benchmark of long-form narrative understanding using movie screenplays to evaluate story world reasoning and generation consistency.

Ax Chi Zhang, Mengqi Zhang, Xiaotian Ye, Runxi Cheng, Zisheng Zhou, Ying Zhou, Pengjie Ren, Zhumin Chen 5/12/2026

Spectral Characterization and Mitigation of Sequential Knowledge Editing Collapse

Spectral analysis of sequential knowledge editing in LLMs identifying mechanisms of catastrophic collapse and mitigation strategies.

Ax Pingzhi Tang, Yiding Wang, Muhan Zhang 5/12/2026

Knowledge is Not Enough: Injecting RL Skills for Continual Adaptation

Method combining supervised fine-tuning with reinforcement learning to help LLMs adapt to new knowledge beyond parametric memory.

Ax Chun-Yi Kuan, Hung-yi Lee 5/12/2026

AQUA-Bench: Beyond Finding Answers to Knowing When There Are None in Audio Question Answering

Benchmark for audio question answering including unanswerable questions to evaluate model robustness beyond correct answers.

Ax Akram Elbouanani, Aboubacar Tuo, Adrian Popescu 5/12/2026

A Scalable Entity-Based Framework for Auditing Bias in LLMs

Scalable framework using named entities as controlled probes to audit bias in LLMs with synthetic data generation.

Ax Qiyang Li, Sergey Levine 5/12/2026

Q-learning with Adjoint Matching

Novel reinforcement learning algorithm for continuous-action RL using adjoint matching with diffusion/flow-matching policies.

Ax Ren Zhuang, Ben Wang, Shuifa Sun 5/12/2026

The Geometric Reasoner: Manifold-Informed Latent Foresight Search for Long-Context Reasoning

Training-free framework for long-context reasoning using manifold-informed latent foresight search with strict memory bounds.

Ax Maxwell Crouse, Ibrahim Abdelaziz, Kshitij Fadnis, Siva Sankalp Patel, Kinjal Basu, Chulaka Gunasekara, Sadhana Kumaravel, Asim Munawar, Pavan Kapanipathi 5/12/2026

Simulating Complex Multi-Turn Tool Calling Interactions in Stateless Execution Environments

Framework for generating synthetic multi-turn tool-calling data for LLMs in stateless execution environments, enabling tuning of smaller models.

Ax Jim Maar, Denis Paperno, Callum Stuart McDougall, Neel Nanda 5/12/2026

What's the plan? Metrics for implicit planning in LLMs and their application to rhyme generation and question answering

Techniques for measuring implicit planning behavior in language models, with applications to rhyme generation and question answering tasks.

Ax Minwu Kim, Safal Shrestha, Anubhav Shrestha, Keith Ross 5/12/2026

Training Reasoning Models on Saturated Problems via Failure-Prefix Conditioning

Method for training LLMs on saturated reasoning problems using failure-prefix conditioning to improve RLVR signal in reinforcement learning.

Ax Yiming Wang, Zhuosheng Zhang, Rui Wang 5/12/2026

On the Overscaling Curse of Parallel Thinking: System Efficacy Contradicts Sample Efficiency

Analysis of inefficiency in parallel reasoning for LLMs, where global budgets cause underutilization on easier samples.

Ax Mohamed Elgaar, Hadi Amiri 5/12/2026

Curriculum Learning for LLM Pretraining: An Analysis of Learning Dynamics

Study of curriculum learning effects on LLM pretraining dynamics across 14M-1B parameter models, analyzing learning phases and gradient noise.

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

Why Adam Works Better with $\beta_1 = \beta_2$: The Missing Gradient Scale Invariance Principle

Analysis of why Adam optimizer performs better when momentum parameters β₁=β₂, explaining gradient scale invariance principle.

Ax Hantong Feng, Duxin Chen, Wenwu Yu 5/12/2026

CoDCL: Counterfactual-Inspired Augmentation Contrastive Learning for Temporal Link Prediction in Social Networks

Counterfactual-augmented contrastive learning framework for temporal link prediction in evolving social networks.

Ax Palash Goyal, Mihir Parmar, Yiwen Song, Hamid Palangi, Tomas Pfister, Jinsung Yoon 5/12/2026

ScholarPeer: A Context-Aware Multi-Agent Framework for Automated Peer Review

Multi-agent framework automating peer review of ML papers by operationalizing senior researcher audit workflows.

Ax Hun Chang, Byunghee Cha, Jong Chul Ye 5/12/2026

Hyperspherical Autoencoder for High-Fidelity Image Reconstruction and Generation

Hyperspherical autoencoder combining vision foundation models with high-fidelity pixel-level reconstruction.

Ax Kunal Mukherjee, Zulfikar Alom, Tran Gia Bao Ngo, Cuneyt Gurcan Akcora, Murat Kantarcioglu 5/12/2026

Optimal Transport-Guided Adversarial Attacks on Graph Neural Network-Based Bot Detection

Adversarial attack study on GNN-based bot detectors using optimal transport under temporal constraints.

Ax Wei Chen, Jiacheng Li, Shigui Li, Zhiqi Lin, Junmei Yang, John Paisley, Delu Zeng 5/12/2026

A Minimum Variance Path Principle for Accurate and Stable Score-Based Density Ratio Estimation

Minimum Variance Path principle addressing path-dependence in score-based density ratio estimation methods.

Ax Brandon Yee, Pairie Koh 5/12/2026

Calibrating Behavioral Parameters with Large Language Models

Framework using LLMs as calibration instruments for behavioral parameters in economic and asset pricing models.

Ax Donald Ye 5/12/2026

Hidden Heroes and Gradient Bloats: Layer-Wise Redundancy Inverts Attribution in Transformers

Study revealing gradient-based attribution unreliability in transformers with systematic layer-wise failures.

Ax Jung H. Lee, Sujith Vijayan 5/12/2026

Functional Subspace, where language models can use vector algebra to solve problems

Analysis of functional subspaces enabling LLMs to use vector algebra for problem-solving and in-context learning.

Ax Antonino Emanuele Scurria 5/12/2026

A Physical Theory of Backpropagation: Exact Gradients from the Least-Action Principle

Physics-inspired derivation of backpropagation from least-action principle enabling exact gradient computation.

Ax Xinyu Liu, Zixuan Xie, Amir Moeini, Claire Chen, Shuze Daniel Liu, Yu Meng, Aidong Zhang, Shangtong Zhang 5/12/2026

MathlibLemma: Folklore Lemma Generation and Benchmark for Formal Mathematics

MathlibLemma pipeline for automated folklore lemma generation and benchmarking using LLMs for formal mathematics.

Ax Jonas Petersen, Camilla Mazzoleni, Gian-Alessandro Lombardi, Federico Martelli, Riccardo Maggioni 5/12/2026

What Structural Inductive Bias Helps Transformers Reason Over Knowledge Graphs? A Study with Tabula RASA

Ablation study isolating structural inductive biases that enable transformers to reason over knowledge graphs.

Ax Wenquan Lu, Hai Huang, Enqi Liu, Randall Balestriero 5/12/2026

PrAg-PO: Prompt Augmented Policy Optimization for Robust and Diverse Mathematical Reasoning

Prompt augmentation with policy optimization for improving mathematical reasoning robustness and diversity in LLMs.

Ax Zhenshuo Zhang, Minxuan Duan, Hongyang R. Zhang 5/12/2026

Efficient Estimation of Kernel Surrogate Models for Task Attribution

Efficient kernel surrogate models for task attribution in multi-task trained AI agents like LLMs.

Ax David F. Ramirez, Tim Overman, Kristen Jaskie, Joe Marvin, Andreas Spanias 5/12/2026

SAR-RAG: ATR Visual Question Answering by Semantic Search, Retrieval, and MLLM Generation

ImageRAG system combining semantic search and MLLMs for automatic target recognition in synthetic aperture radar imagery.

Ax Jiarui Yuan, Tailin Jin, Weize Chen, Zeyuan Liu 5/12/2026

SE-Bench: Benchmarking Self-Evolution with Knowledge Internalization

SE-Bench benchmark for measuring self-evolution and knowledge internalization in AI agents with lifelong learning.

Ax Seohyeon Cha, Huancheng Chen, Dongjun Kim, Haoran Zhang, Kevin Chan, Gustavo de Veciana, Haris Vikalo 5/12/2026

CoreQ: Learning-Free Mismatch Correction and Successive Rounding for Quantization

CoreQ post-training quantization method for LLMs correcting layer-wise mismatches without retraining.

Ax Jan Benad, Pradeep Kr. Banerjee, Frank R\"oder, Nihat Ay, Martin V. Butz, Manfred Eppe 5/12/2026

Dynamics-Aligned Shared Hypernetworks for Contextual RL under Discontinuous Shifts

DMA*-SH framework using hypernetworks for contextual RL with zero-shot generalization across discontinuous context shifts.

Ax Rishabh Jain, Keisuke Okumura, Michael Amir, Pietro Lio, Amanda Prorok 5/12/2026

Pairwise is Not Enough: Hypergraph Neural Networks for Multi-Agent Pathfinding

Hypergraph neural networks for multi-agent path finding that extends beyond pairwise message passing for agent coordination.

Ax Samuel Daulton, David Eriksson, Maximilian Balandat, Eytan Bakshy 5/12/2026

BONSAI: Bayesian Optimization with Natural Simplicity and Interpretability

Bayesian optimization technique that minimizes deviation from default configurations while tuning black-box functions.

Ax Heiko Hoppe, Fabian Akkerman, Wouter van Heeswijk, Maximilian Schiffer 5/12/2026

Breaking the Grid: Distance-Guided Reinforcement Learning in Large Discrete Action Spaces

Distance-Guided RL algorithm for handling large discrete action spaces up to 10^20 actions using sampled dynamic neighborhoods and distance-based updates.

Ax Ya-Chi Chu, Alkiviades Boukas, Madeleine Udell 5/12/2026

SnareNet: Flexible Repair Layers for Neural Networks with Hard Constraints

SnareNet neural network architecture with differentiable repair layers that enforce input-dependent constraints on model outputs.

Ax Xavier Hu, Jinxiang Xia, Shengze Xu, Kangqi Song, Yishuo Yuan, Guibin Zhang, JinCheng Ren, Boyu Feng, Li Lu, Tieyong Zeng, Jiaheng Liu, Minghao Liu, He Zhu, Yuchen Eleanor Jiang, Wei Wang, Wangchunshu Zhou 5/12/2026

EcoGym: Evaluating LLMs for Long-Horizon Plan-and-Execute in Interactive Economies

EcoGym benchmark for evaluating LLM-based agents on long-horizon planning and execution tasks in persistent interactive economic environments.

Ax Fei Wu, Jia Hu, Geyong Min, Shiqiang Wang 5/12/2026

Preventing Rank Collapse in Federated Low-Rank Adaptation with Client Heterogeneity

Research on federated low-rank adaptation addressing rank collapse in heterogeneous client settings with varying resource constraints and data distributions.