Ax Michael Mancini, Shabnam Sodagari 5/12/2026

Optimal FALQON for Quantum Approximate Optimization via Layer-wise Parameter Tuning

Optimal FALQON improves convergence speed for quantum approximate optimization on NISQ devices through layer-wise parameter tuning.

Ax Pengzhou Chen, Tao Chen 5/12/2026

CDS4RAG: Cyclic Dual-Sequential Hyperparameter Optimization for RAG

CDS4RAG framework for cyclic dual-sequential optimization of both retriever and generator hyperparameters in Retrieval-Augmented Generation systems.

Ax Geigh Zollicoffer, Minh Vu, Hongli Zhan, Raymond Li, Manish Bhattarai 5/12/2026

Sanity Checks for Long-Form Hallucination Detection

Methodology for detecting hallucinations in LLM chain-of-thought reasoning by distinguishing actual reasoning evaluation from answer-surface correlates.

Ax Guangchen Lan 5/12/2026

Reinforcement Learning for Scalable and Trustworthy Intelligent Systems

Research on reinforcement learning challenges for scalable and trustworthy intelligent systems in distributed environments and LLM post-training.

Ax Junyu Huo, Ziqi Mao, Zihao Wan, Gouri Ginde 5/12/2026

What Software Engineering Looks Like to AI Agents? -- An Empirical Study of AI-Only Technical Discourse on MoltBook

Empirical study examining software engineering discourse produced by autonomous AI agents interacting primarily with each other on MoltBook network.

Ax Mingyu Liu, Sihan Huang, Yijia Fan, Yinlin Yan, Quan Zhang, Jian-Fang Hu, Jianhuang Lai 5/12/2026

Decoupling Endpoint and Semantic Transition Learning for Zero-Shot Composed Image Retrieval

Zero-shot composed image retrieval method that decouples endpoint and semantic transition learning for lightweight projection-based approaches.

Ax Junnan Liu, Linhao Luo, Thuy-Trang Vu, Gholamreza Haffari 5/12/2026

AIPO: : Learning to Reason from Active Interaction

AIPO method for improving LLM reasoning through active interaction and reinforcement learning, extending beyond policy model capability boundaries.

Ax Dongdong Wang, Deepak Balakrishnan, Ravi Srinivasan, Shenhao Wang 5/12/2026

Built Environment Reasoning from Remote Sensing Imagery Using Large Vision--Language Models

Study of using multimodal LLMs with remote sensing imagery for smart city tasks including design suggestions and risk identification.

Ax Xuanqiang Angelo Huang, Charlie Tharas, Samuele Marro, Van Q. Truong, Bernhard Sch\"olkopf, Emanuele La Malfa, Zhijing Jin 5/12/2026

Mechanism Design Is Not Enough: Prosocial Agents for Cooperative AI

Research proving mechanism design alone is insufficient for safe AI agent cooperation, proposing prosocial agent approaches for beneficial multi-agent interaction.

Ax Zhanliang Wang, Jiancong Xiao, Ruochen Jin, Shu Yang, Bojian Hou, Li Shen 5/12/2026

A Semantic-Sampling Framework for Evaluating Calibration in Open-Ended Question Answering

Framework for evaluating LLM calibration in open-ended question answering tasks, addressing reliable deployment in high-stakes domains.

Ax Benjamin D. Shaffer, Brooks Kinch, M. Ani Hsieh, Nathaniel Trask 5/12/2026

A meshfree exterior calculus for generalizable and data-efficient learning of physics from point clouds

MEEC-Net uses meshfree exterior calculus to learn physics from point clouds with data efficiency and transferability across resolutions and parameters.

Ax Ramon Pires, Thales Sales Almeida, Celio Larcher Junior, Giovana Bon\'as, Hugo Abonizio, Marcos Piau, Roseval Malaquias Junior, Thiago Laitz, Rodrigo Nogueira 5/12/2026

Magis-Bench: Evaluating LLMs on Magistrate-Level Legal Tasks

Magis-Bench benchmark for evaluating LLMs on magistrate-level legal judgment tasks including weighing claims and rendering reasoned decisions.

Ax Haoyu Hu, Xuandong Zhao, Xuhai "Orson'' Xu, Nori Jacoby 5/12/2026

DUET: Optimize Token-Budget Allocation for Reinforcement Learning with Verifiable Rewards

DUET framework for optimizing token-budget allocation in reinforcement learning with verifiable rewards by jointly tuning rollout distribution across prompts.

Ax Jun Wen Leong 5/12/2026

Defense effectiveness across architectural layers: a mechanistic evaluation of persistent memory attacks on stateful LLM agents

Systematic evaluation of defense mechanisms against persistent memory attacks on stateful LLM agents across architectural layers and multiple open-source models.

Ax Maxwell J. Jacobson, Yexiang Xue 5/12/2026

Zero-shot Imitation Learning by Latent Topology Mapping

Research on zero-shot imitation learning for training agents from fixed demonstration datasets to solve new tasks without complete task-specific examples.

Ax Peter S\'uken\'ik, Cristina L\'opez Amado, Christoph H. Lampert, Marco Mondelli 5/12/2026

Sink vs. diagonal patterns as mechanisms for attention switch and oversmoothing prevention

Analyzes sinks and diagonal patterns in attention mechanisms as methods for preventing oversmoothing in deep networks.

Ax Yuxiang Luo, Andrew Perrault 5/12/2026

Recovering Physical Dynamics from Discrete Observations via Intrinsic Differential Consistency

Proposes method for recovering continuous-time dynamics from discrete observations using semi-group property constraint instead of local supervision.

Ax Ziwen Cai, Yihe Zhang, Xiali Hei 5/12/2026

When Child Inherits: Modeling and Exploiting Subagent Spawn in Multi-Agent Networks

Studies security risks of multi-agent LLM networks where agents spawn subagents, modeling attack vectors in hierarchical delegation.

Ax I. F. Atasoy, B. Mutlu, E. A. Sezer, A. Wahdan 5/12/2026

Do Benchmarks Underestimate LLM Performance? Evaluating Hallucination Detection With LLM-First Human-Adjudicated Assessment

Evaluates whether LLM benchmarks underestimate performance by comparing human annotations with LLM predictions on hallucination detection in RAG.

Ax Mehmet Iscan 5/12/2026

PYTHALAB-MERA: Validation-Grounded Memory, Retrieval, and Acceptance Control for Frozen-LLM Coding Agents

Presents system for frozen-LLM coding agents with validation-grounded memory and adaptive retrieval for execution-feedback repair cycles.

Ax Mingsong Yan, Dongyang Li, Charles Kulick, Sui Tang 5/12/2026

Transformers Can Implement Preconditioned Richardson Iteration for In-Context Gaussian Kernel Regression

Analyzes how transformers with softmax attention implement preconditioned Richardson iteration for in-context Gaussian kernel regression.

Ax Viresh Pati, Zhengyu Li, Piyush Jha, Rahul Garg, Yatharth Sejpal, Vijay Ganesh 5/12/2026

MathConstraint: Automated Generation of Verified Combinatorial Reasoning Instances for LLMs

Introduces MathConstraint, adaptive benchmark for evaluating LLM combinatorial reasoning with solver-based verification and dynamic difficulty.

Ax Zhifei Hu, Feng Xia 5/12/2026

Multi-Level Graph Attention Network Contrastive Learning for Knowledge-Aware Recommendation

Proposes multi-level graph attention network with contrastive learning for knowledge-aware recommendation systems using knowledge graphs.

Ax Giuseppe Bruno, Shi Chen, Zhengjiang Lin, Yury Polyanskiy, Philippe Rigollet 5/12/2026

Scaling Limits of Long-Context Transformers

Analyzes softmax attention scaling limits to understand when selectivity emerges versus uniform averaging in long-context transformers.

Ax Hamid Kazemi, Atoosa Chegini, Maria Safi 5/12/2026

A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models

Demonstrates safety alignment vulnerabilities in LLMs by targeting single neurons that control refusal and harmful knowledge to bypass safety measures.

Ax Timothy Tin Long Yu, Gursimran Singh, Ge Shi, Hanieh Sadri, Yong Zhang, Zhenan Fan 5/12/2026

MARLaaS: Multi-Tenant Asynchronous Reinforcement Learning as a Service

Proposes MARLaaS system for concurrent multi-tenant reinforcement learning fine-tuning of LLMs using verifiable rewards for agentic capabilities.

Ax Annan Yu, Dongwei Lyu, N. Benjamin Erichson 5/12/2026

Continuity Laws for Sequential Models

Studies continuity as inductive bias in sequential models, analyzing whether continuous-time formulations like state-space models behave continuously.

Ax Zichen Xie, Mrigank Pawagi, Yuxin Liu, Aaditi Rai, Lize Shao, John Berberian Jr., Sicong Che, Wenxi Wang 5/12/2026

VeriContest: A Competitive-Programming Benchmark for Verifiable Code Generation

Introduces VeriContest benchmark for evaluating LLMs on verifiable code generation with formal specifications and machine-checkable proofs.

Ax Salim Khazem, Ibrahim Mohamed Serouis, Zakaria Ezzahed 5/12/2026

MC-RFM: Geometry-Aware Few-Shot Adaptation via Mixed-Curvature Riemannian Flow Matching

Proposes mixed-curvature Riemannian flow matching for parameter-efficient adaptation of vision models using geometry-aware few-shot learning.

Ax Hassan Shapourian, Kasra Hejazi, Olabode M. Sule, Beren Millidge 5/12/2026

ZAYA1-VL-8B Technical Report

Technical report on ZAYA1-VL-8B, a compact mixture-of-experts vision-language model achieving competitive performance with smaller parameter count.

Ax Jongseok Park, Sunga Kim, Zhenyu Gu, Ion Stoica, Alvin Cheung 5/12/2026

Uncovering Intra-expert Activation Sparsity for Efficient Mixture-of-Expert Model Execution

Explores intra-expert activation sparsity in Mixture-of-Experts LLMs to improve computational efficiency beyond standard sparse expert activation.

Ax Jooyeon Kim 5/12/2026

Probing the Impact of Scale on Data-Efficient, Generalist Transformer World Models for Atari

Analyzes scaling behavior of minimalist transformer world models on Atari 100k benchmark to understand data efficiency in generalist systems.

Ax Zhuofu Chen, Rui Pan, Yinwei Dai, Ravi Netravali 5/12/2026

Slipstream: Trajectory-Grounded Compaction Validation for Long-Horizon Agents

Addresses context compression validation for long-horizon LLM agents by analyzing accuracy gaps when trajectory summaries are generated asynchronously.

Ax Lujia Zhong, Yihao Xia, Shuo Huang, Jianwei Zhang, Yonggang Shi 5/12/2026

PromptDx: Differentiable Prompt Tuning for Multimodal In-Context Alzheimer's Diagnosis

Applies differentiable prompt tuning and in-context learning to medical imaging for Alzheimer's diagnosis using multimodal models.

Ax Shanshan Zhu, Han Zhang, J. Doris Chi, Subigya Nepal, Koustuv Saha 5/12/2026

Causal Stories from Sensor Traces: Auditing Epistemic Overreach in LLM-Generated Personal Sensing Explanations

Evaluates epistemic overreach in LLM explanations of sensor data, measuring when generated accounts exceed available evidence in personal sensing applications.

Ax Liam Davis, Leopold Haller, Alberto Alfarano, Mark Santolucito 5/12/2026

Lattice Deduction Transformers

Lattice Deduction Transformer uses recurrent attention with lattice projections for logical reasoning tasks, achieving perfect accuracy on constraint satisfaction with minimal parameters.

Ax Md Atik Ahamed, Mihir Parmar, Palash Goyal, Chun-Liang Li, Qiang Cheng, Tomas Pfister, Jinsung Yoon 5/12/2026

Reasoning-Aware Training for Time Series Forecasting

STRIDE combines time series foundation models with LLM reasoning to improve forecasting interpretability while handling continuous numerical values without excessive tokenization.

Ax Zihao An, Taichi Liu, Ziqiong Liu, Dong Li, Ruofeng Liu, Emad Barsoum 5/12/2026

PARD-2: Target-Aligned Parallel Draft Model for Dual-Mode Speculative Decoding

Introduces PARD-2, a draft model optimization for speculative decoding that aligns training with inference goal of maximizing consecutive token acceptance in LLM inference.

Ax Yinwei Dai, Zhuofu Chen, Lijie Yang, Ravi Netravali 5/12/2026

Geometry Guided Self-Consistency for Physical AI

Proposes KeyStone, an inference-time self-consistency method using geometry guidance for physical AI models that generate action trajectories via diffusion/flow matching.

Ax Aritra Mazumder, Shubhashis Roy Dipta, Nusrat Jahan Lia, Tanzila Khan, Kainat Raisa Hossain, Nehaa Shri, Shubhrangshu Debsarkar, Humayra Tasnim, Gour Gupal Talukder Shawon, Debjoty Mitra, Sumaiya Ahmed Rani, Al Jami Islam Anik, Al Nafeu Khan 5/12/2026

AgentCollabBench: Diagnosing When Good Agents Make Bad Collaborators

Introduces AgentCollabBench, a diagnostic benchmark for measuring multi-hop process failures in multi-agent systems where individual agents appear correct but collaboration silently fails.

Ax Shan Jiang, Zijian Yi, Chenguang Zhu 5/12/2026

Sketch-and-Verify: Structured Inference-Time Scaling via Program Sketching

Proposes SketchVerify for structured inference-time scaling where LLMs enumerate algorithmic strategies and write program sketches verified by execution for cost-effective improvement.

Ax Gautam Veldanda 5/12/2026

Explanation Fairness in Large Language Models: An Empirical Analysis of Disparities in How LLMs Justify Decisions Across Demographic Groups

Empirically analyzes fairness in LLM explanations across demographic groups, introducing Explanation Fairness Taxonomy framework for evaluating explanation quality disparities.

Ax Chengcheng Sun, Chenhao Li, Xiang Lin, Tianji Zheng, Fanrong Meng, Xiaobin Rui, Zhixiao Wang 5/12/2026

Attention-based graph neural networks: a survey

Survey of attention-based graph neural networks covering mechanisms for adaptive feature selection and noise filtering in graph representation learning.

Ax Shan Jiang, Zijian Yi, Chenguang Zhu 5/12/2026

Semantic Voting: Execution-Grounded Consensus for LLM Code Generation

Compares voting strategies for LLM code generation including Semantic Voting, analyzing how textual voting, ranking, and execution-based agreement contribute to code selection.

Ax Jingzhe Xu, Rui Wang, Jiannan Wang, Guoliang Li 5/12/2026

PrepBench: How Far Are We from Natural-Language-Driven Data Preparation?

Introduces PrepBench benchmark evaluating natural language-driven data preparation using LLMs, assessing progress toward replacing GUI-based data transformation tools.

Ax Guodong Ding, Angela Yao 5/12/2026

Gate-and-Merge: Zero-shot Compositional Personalization of Vision Language Models

Introduces Gate-and-Merge, a zero-shot framework for compositional personalization of vision-language models using lightweight LoRA adapters paired with concept tokens.

Ax Changze Li, Zhe Chen, Shaoyu Chen, Lisen Mu, Yijian Li, Yuelong Yu, Qian Zhang, Qing Su, Ming Yang, Tong Qin 5/12/2026

REAP: Reinforcement-Learning End-to-End Autonomous Parking with Gaussian Splatting Simulator for Real2Sim2Real Transfer

Develops end-to-end autonomous parking using reinforcement learning with Gaussian Splatting simulator for real-to-sim-to-real transfer in extreme parking scenarios.

Ax Boxuan Zhang, Jianing Zhu, Zeru Shi, Dongfang Liu, Ruixiang Tang 5/12/2026

AgentForesight: Online Auditing for Early Failure Prediction in Multi-Agent Systems

Proposes AgentForesight for early failure prediction in LLM-based multi-agent systems through online auditing, enabling intervention before trajectory-level failures cascade.

Ax Chenyu Zhao, Shenglin Zhang, Yihang Lin, Wenwei Gu, Zhimin Chen, Yongqian Sun, Dan Pei, Chetan Bansal, Saravan Rajmohan, Minghua Ma 5/12/2026

Debugging the Debuggers: Failure-Anchored Structured Recovery for Software Engineering Agents

Introduces PROBE, a framework for structured recovery in software engineering agents that converts runtime failures into grounded recovery guidance for subsequent attempts.

Ax Ziyun Liu, Fengmiao Bian, Jian-Feng Cai 5/12/2026

AdaPreLoRA: Adafactor Preconditioned Low-Rank Adaptation

Proposes AdaPreLoRA, an optimization method for Low-Rank Adaptation using Adafactor preconditioning to handle rank-deficiency issues in LoRA factor-space preconditioners.