Ax Seunghyun Lee, David Brumley 5/15/2026

ExploitBench: A Capability Ladder Benchmark for LLM Cybersecurity Agents

Benchmark for evaluating LLM-based cybersecurity agents on exploitation tasks with granular capability levels, moving beyond binary success/failure metrics.

Ax Harshita Chopra, Krishna Kant Chintalapudi, Suman Nath, Ryen W. White, Chirag Shah 5/15/2026

Thinking Ahead: Prospection-Guided Retrieval of Memory with Language Models

Method for improving RAG systems in dialogue assistants by using prospection-guided retrieval to recover semantically distant but contextually relevant facts from interaction histories.

Ax Kai Guo, Xinnan Dai, Zhibo Zhang, Nuohan Lin, Shenglai Zeng, Jie Ren, Haoyu Han, Jiliang Tang 5/15/2026

Why Retrieval-Augmented Generation Fails: A Graph Perspective

Research study examining failure modes of RAG systems through a graph-based lens, analyzing how retrieved evidence influences LLM answer generation.

Ax Hrushitha Goud Tigulla, Marco Vieira 5/15/2026

LLM-Based Robustness Testing of Microservice Applications: An Empirical Study

Empirical study of LLM-based robustness testing for microservice APIs, comparing model and prompt strategies for failure detection.

Ax Yingying Fang, Haijie Xu, Shuang Wu, Mariathasan Anish, Guang Yang 5/15/2026

Towards Fine-Grained and Verifiable Concept Bottleneck Models

Fine-grained concept bottleneck models with visual grounding for interpretable predictions with verifiable concept evidence.

Ax Andrew Lanpouthakoun, Aryaman Arora, Zhengxuan Wu, Dhruv Pai, Ben Keigwin, Dan Jurafsky, Christopher Potts 5/15/2026

PreFT: Prefill-only finetuning for efficient inference

Prefill-only finetuning method enabling efficient personalized LLM serving without throughput degradation from user-specific adapters.

Ax Tianle Zhong, Neiwen Ling, Yifan Pi, Zijun Wei, Tianshu Yu, Geoffrey Fox, Peng Wu, Xiao Yu 5/15/2026

Diagnosing Training Inference Mismatch in LLM Reinforcement Learning

Identifies and diagnoses training-inference mismatch in LLM RL systems where rollout and optimization stages produce inconsistent token probabilities.

Ax Jerem\'ias Figueiredo Paschmann, Juan Kaplan, Francisco Nattero Santiago Mauricio Barron Bucolo, Juan Wisznia, Luciano del Corro 5/15/2026

Active Learners as Efficient PRP Rerankers

Active learning approach to improve pairwise ranking prompting from LLMs by reframing as efficient reranking problem.

Ax Kai Sun, Peibo Duan, Yongsheng Huang, Guowei Zhang, Benjamin Smith, Nanxu Gong, Levin Kuhlmann 5/15/2026

Not All Timesteps Matter Equally: Selective Alignment Knowledge Distillation for Spiking Neural Networks

Selective alignment knowledge distillation for spiking neural networks that weights timesteps unequally to improve performance over ANNs.

Ax Jesseba Fernando, Grigori Guitchounts 5/15/2026

Dynamics of the Transformer Residual Stream: Coupling Spectral Geometry to Network Topology

Analyzes spectral geometry of transformer residual streams across layers, revealing full eigenvalue distributions and coupling to network topology.

Ax Juho Kim, Fei Fang, Tuomas Sandholm 5/15/2026

Watermarking Game-Playing Agents in Perfect-Information Extensive-Form Games

Watermarking techniques for game-playing agents in perfect-information games to detect unauthorized use and cheating in gaming platforms.

Ax Weisen Jiang, Shuhao Chen, Sinno Jialin Pan 5/15/2026

MetaMoE: Diversity-Aware Proxy Selection for Privacy-Preserving Mixture-of-Experts Unification

MetaMoE: Privacy-preserving framework for unifying independently trained domain-specialized experts into centralized MoE using public proxy data.

Ax Julien Piet, Annabella Chow, Yiwei Hou, Muxi Lyu, Sylvie Venuto, Jinhao Zhu, Raluca Ada Popa, David Wagner 5/15/2026

Web Agents Should Adopt the Plan-Then-Execute Paradigm

Web agents should use plan-then-execute paradigm instead of ReAct, committing to task-specific programs before observing web content to avoid injection attacks.

Ax Chengshuai Zhao, Zhen Tan, Dawei Li, Zhiyuan Yu, Huan Liu 5/15/2026

To See is Not to Learn: Protecting Multimodal Data from Unauthorized Fine-Tuning of Large Vision-Language Model

MMGuard: Proactive protection mechanism for multimodal data preventing unauthorized fine-tuning of vision-language models before training.

Ax Matias Alvo, Daniel Russo, Yash Kanoria 5/15/2026

Policy Optimization in Hybrid Discrete-Continuous Action Spaces via Mixed Gradients

Policy optimization for hybrid discrete-continuous action spaces using mixed gradients to improve credit assignment in robotics and control.

Ax Zuyuan Zhang, Carlee Joe-Wong, Tian Lan 5/15/2026

Matrix-Space Reinforcement Learning for Reusing Local Transition Geometry

MSRL: Geometric reinforcement learning that reuses local transition geometry via matrix descriptors for compositional generalization across tasks.

Ax Shen Lin, Jing Lin, Junhao Dong, Piotr Koniusz, Li Xu 5/15/2026

ICED: Concept-level Machine Unlearning via Interpretable Concept Decomposition

ICED: Concept-level machine unlearning for vision-language models using interpretable concept decomposition to remove specific knowledge.

Ax Yuchen Sun, Pei Fu, Shaojie Zhang, Anan Du, Xiuwen Xi, Ruoceng Zhang, Zhenbo Luo, Jian Luan, Chongyang Zhang 5/15/2026

Beyond Binary: Reframing GUI Critique as Continuous Semantic Alignment

Continuous semantic alignment for GUI agents: Improves test-time scaling by replacing binary critic classification with fine-grained ranking ability.

Ax Fangyuan Yu, Xin Su, Amir Abdullah 5/15/2026

Dynamic Latent Routing

Dynamic Latent Routing: Language model post-training method using General Dijkstra Search for temporal composition of sub-policies in MDPs.

Ax Zhengjia Zhong, Shuyan Ke, Zaizhou Lin, Jiaqi Song, Hongyi Lan, Hui Li 5/15/2026

RQ-MoE: Residual Quantization via Mixture of Experts for Efficient Input-Dependent Vector Compression

RQ-MoE: Dynamic vector quantization method using mixture of experts for efficient input-dependent compression of high-dimensional embeddings.

Ax Shweta Mishra 5/15/2026

Correctness-Aware Repository Filtering Under Maximum Effective Context Window Constraints

Context window filtering for LLM-based developer tools using correctness-aware repository filtering to maximize effective context within practical constraints.

Ax Changryeol Choi, Hyewon Park, Yujin Kwon, Gowun Jeong 5/15/2026

LoMETab: Beyond Rank-1 Ensembles for Tabular Deep Learning

LoMETab: Rank-r generalization of multiplicative ensembles for tabular deep learning that improves on gradient boosting and attention-based architectures.

Ax Injin Kong, Hyoungjoon Lee, Yohan Jo 5/15/2026

Where Should Diffusion Enter a Language Model? Geometry-Guided Hidden-State Replacement

DiHAL: A diffusion-transformer hybrid that identifies optimal layers for integrating continuous diffusion into pretrained language models for improved denoising.

Ax Chen Liang, Xiatao Sun, Qian Wang, Daniel Rakita 5/15/2026

Turning Stale Gradients into Stable Gradients: Coherent Coordinate Descent with Implicit Landscape Smoothing for Lightweight Zeroth-Order Optimization

Coherent Coordinate Descent optimization method for zeroth-order scenarios where backpropagation unavailable, improving sample efficiency and variance.

Ax Young-Chae Hong, Yangho Chen 5/15/2026

Optimal Pattern Detection Tree for Symbolic Rule-Based Classification

Optimal Pattern Detection Tree for symbolic rule-based classification providing interpretable rules for pattern discovery tasks.

Ax JB Lanier, Nathan Monette, Pierre Baldi, Roy Fox 5/15/2026

Data-Augmented Game Starts for Accelerating Self-Play Exploration in Imperfect Information Games

Multi-agent starting-state sampling strategy to accelerate exploration in imperfect-information competitive games like StarCraft and Dota.

Ax Taebong Kim, Youngsik Hong, Minsik Kim, Sunyoung Choi, Jaewon Jang, Junghoon Shin, Minseo Kim 5/15/2026

Darwin Family: MRI-Trust-Weighted Evolutionary Merging for Training-Free Scaling of Language-Model Reasoning

Darwin Family framework for training-free evolutionary merging of LLMs via gradient-free weight-space recombination to improve reasoning performance.

Ax Xiang Shen, Yuhang Zhou, Yifan Wu, Zhuokai Zhao, Siyu Lin, Lei Huang, Qianqian Zhong, Lizhu Zhang, Benyu Zhang, Xiangjun Fan, Hong Yan 5/15/2026

Agentic Recommender System with Hierarchical Belief-State Memory

MARS framework for memory-augmented LLM agents in recommendation systems using structured belief-state memory with lifecycle management.

Ax Man Ho Lam, Chaozheng Wang, Hange Liu, Jingyu Xiao, Haau-sing Li, Jen-tse Huang, Terry Yue Zhuo, Michael R. Lyu 5/15/2026

SWE-Chain: Benchmarking Coding Agents on Chained Release-Level Package Upgrades

SWE-Chain benchmark for evaluating coding agents on realistic package upgrade chains, testing continuous maintenance tasks beyond isolated issue resolution.

Ax Jean-Philippe Monteuuis, Cong Chen, Jonathan Petit 5/15/2026

The Great Pretender: A Stochasticity Problem in LLM Jailbreak

Analysis of stochasticity problems in LLM jailbreak evaluation, showing reported adversarial attack methods fail to replicate promised performance against independent models.

Ax Ciyan Ouyang, Rui Hou 5/15/2026

MemLineage: Lineage-Guided Enforcement for LLM Agent Memory

MemLineage defense mechanism for LLM agent memory using cryptographic provenance and derivation lineage to prevent malicious state injection into agent sessions.

Ax Leo Muxing Wang, Pengkun Yang, Lili Su 5/15/2026

Collaborative Yet Personalized Policy Training: Single-Timescale Federated Actor-Critic

Federated actor-critic framework where agents share linear subspace representation while maintaining personalized local policies for collaborative training.

Ax Jianbo Zhu, Xing Fang, Jing Wang, Mingmin Jin, Bokang Wang, Guangxin Song, Zhenyu Xie, Junjie Bai 5/15/2026

Efficient Generative Retrieval for E-commerce Search with Semantic Cluster IDs and Expert-Guided RL

Generative retrieval framework for e-commerce search with semantic cluster IDs and expert-guided RL, optimized for industrial deployment.

Ax Siyang Yao, Erhu Feng, Yubin Xia 5/15/2026

When Answers Stray from Questions: Hallucination Detection via Question-Answer Orthogonal Decomposition

Single-pass framework (QAOD) for hallucination detection in LLMs using question-answer orthogonal decomposition without repeated inference.

Ax Yihang Chen, Pin Qian, Su Wang, Sipeng Zhang, Huan Xu, Shuhuai Lin, Xinpeng Wei 5/15/2026

Does RAG Know When Retrieval Is Wrong? Diagnosing Context Compliance under Knowledge Conflict

Framework for detecting when RAG systems inappropriately rely on retrieved context conflicting with model knowledge, with inference-time intervention.

Ax Haojun Weng, Qianqian Yang, Hao Fu, Haobin Pan, Xinwei Lv 5/15/2026

When Retrieval Hurts Code Completion: A Diagnostic Study of Stale Repository Context

Diagnostic study on retrieval-augmented code generation showing how temporally stale repository context degrades completion quality.

Ax Truong Thanh Hung Nguyen, Vo Thanh Khang Nguyen, Hoang-Loc Cao, Phuc Ho, Van Pham, Hung Cao 5/15/2026

Contestable Multi-Agent Debate with Arena-based Argumentative Computation for Multimedia Verification

Multi-agent framework with arena-based argumentation for transparent multimedia verification using multimodal LLMs and verification tools.

Ax Letian Yang (Shanghai Jiao Tong University, Shanghai, China), Xu Liu (Shanghai Jiao Tong University, Shanghai, China), Yiqiang Lu (Ant Group, Shanghai, China), Jian Liu (Ant Group, Shanghai, China), Weiqiang Wang (Ant Group, Shanghai, China), Shuai Li (Shanghai Jiao Tong University, Shanghai, China) 5/15/2026

ROAD: Adaptive Data Mixing for Offline-to-Online Reinforcement Learning via Bi-Level Optimization

Offline-to-online reinforcement learning approach using bi-level optimization for adaptive data mixing across training stages.

Ax Edoardo Scarpel, Alberto Pettena, Matteo Cederle, Federico Chiariotti, Marco Fabris, Gian Antonio Susto 5/15/2026

Fully Dynamic Rebalancing in Dockless Bike-Sharing Systems via Deep Reinforcement Learning

Deep reinforcement learning method for dynamic rebalancing of dockless bike-sharing systems with real-time truck routing.

Ax GAng Peng 5/15/2026

Dimension-Level Intent Fidelity Evaluation for Large Language Models: Evidence from Structured Prompt Ablation

Framework for dimension-level intent fidelity evaluation of LLMs through structured prompt ablation across multiple languages and domains.

Ax Shuhao Chen, Weisen Jiang, Changmiao Wang, Xiaoqing Wu, Xuanren Shi, Yu Zhang, James T. Kwok 5/15/2026

RxEval: A Prescription-Level Benchmark for Evaluating LLM Medication Recommendation

Prescription-level benchmark (RxEval) for evaluating LLM medication recommendation systems with temporally-aware medical decision-making.

Ax Donghao Li, Chengshuai Shi, Weijuan Ou, Cong Shen, Jing Yang 5/15/2026

Efficient Multi-objective Prompt Optimization via Pure-exploration Bandits

Framework for multi-objective prompt optimization using pure-exploration bandits to select effective LLM prompts across multiple performance metrics.

Ax Langzhou He, Junyou Zhu, Yue Zhou, Zhengyao Gu, Junhua Liu, Wei-Chieh Huang, Henry Peng Zou, David Wipf, Philip S. Yu, Qitian Wu 5/15/2026

Resolving Action Bottleneck: Agentic Reinforcement Learning Informed by Token-Level Energy

Method for improving agentic reinforcement learning by optimizing token-level credit assignment in LLM trajectories using energy-based approaches.

Ax Ho Hung Lim, Yi Yang 5/15/2026

A Picture is Worth a Thousand Words? An Empirical Study of Aggregation Strategies for Visual Financial Document Retrieval

Empirical study on aggregation strategies for visual document retrieval in RAG systems, evaluating information loss in financial document processing.

Ax Oubo Ma, Ruixiao Lin, Yang Dai, Jiahao Chen, Chunyi Zhou, Linkang Du, Shouling Ji 5/15/2026

Angel or Demon: Investigating the Plasticity Interventions' Impact on Backdoor Threats in Deep Reinforcement Learning

Research on backdoor attack vulnerabilities in deep reinforcement learning agents with plasticity interventions, examining security threats in DRL systems.

Ax Andreas Schlaginhaufen, Maryam Kamgarpour 5/15/2026

Fast Rates for Inverse Reinforcement Learning

Theoretical analysis establishing fast convergence rates for entropy-regularized inverse reinforcement learning with linear reward classes.

Ax Itay Zloczower, Eyal Lenga, Gilad Gressel, Yisroel Mirsky 5/15/2026

One Step to the Side: Why Defenses Against Malicious Finetuning Fail Under Adaptive Adversaries

Analysis showing defenses against malicious fine-tuning of foundation models fail under adaptive adversaries that account for defense mechanisms.

Ax Yaroslav Sokolov, Yury Khudyakov, Lenar Sharipov, Andrei Gasparian, Parth Tiwary, Artem Trofimov 5/15/2026

In-IDE Toolkit for Developers of AI-Based Features

JetBrains IDE plugin toolkit for developers building AI features with LLMs and agentic workflows, enabling tracing, debugging, and evaluation in development loop.

Ax Tian Qin, Junzhe Chen, Yuqing Shi, Tianshu Zhang, Qiang Ju, Lijie Wen 5/15/2026

Do We Really Need External Tools to Mitigate Hallucinations? SIRA: Shared-Prefix Internal Reconstruction of Attribution

SIRA: Training-free method to reduce hallucinations in vision-language models through internal contrastive decoding without external tools.

Ax Sohaib Afifi 5/15/2026

An Amortized Efficiency Threshold for Comparing Neural and Heuristic Solvers in Combinatorial Optimization

Energy efficiency analysis comparing neural combinatorial optimization solvers to CPU metaheuristics, accounting for amortized training costs.