Ax Abhijit Talluri, Pujith Anne, Bhagavan Choudary Pendiyala, Raghavendra Chilukuri 5/8/2026

Retrieval-Conditioned Topology Selection with Provable Budget Conservation for Multi-Agent Code Generation

Multi-agent LLM code generation system selecting orchestration topology based on code structural complexity retrieval.

Ax Gongli Xi, Ye Tian, Mengyu Yang, Huahui Yi, Liang Lin, Xiaoshuai Hao, Kun Wang, Wendong Wang 5/8/2026

Large Vision-Language Models Get Lost in Attention

Mechanistic analysis of attention module roles in vision-language model decoders for architectural optimization.

Ax Xiaomin Li, Jianheng Hou, Zheyuan Deng, Zhiwei Zhang, Taoran Li, Binghang Lu, Bing Hu, Yunhan Zhao, Yuexing Hao 5/8/2026

Chain of Risk: Safety Failures in Large Reasoning Models and Mitigation via Adaptive Multi-Principle Steering

Safety failures in chain-of-thought reasoning traces of large reasoning models with adaptive multi-principle steering mitigation.

Ax Qiyao Liang, Risto Miikkulainen, Ila Fiete 5/8/2026

Attractor Geometry of Transformer Memory: From Conflict Arbitration to Confident Hallucination

Geometric analysis of LLM failure modes (conflict, hallucination) via parametric vs. working memory interactions.

Ax Xiaomin Li, Andrzej Banburski-Fahey, Jaron Lanier 5/8/2026

DataDignity: Training Data Attribution for Large Language Models

Data attribution method for LLMs: FakeWiki benchmark to identify source documents supporting model responses.

Ax Shaozhen Ma, Wei Huang, Hanchen Wang, Dong Wen, Wenjie Zhang 5/8/2026

GCCM: Enhancing Generative Graph Prediction via Contrastive Consistency Model

Contrastive consistency model for generative graph prediction reducing inference cost vs. diffusion methods.

Ax Yanlong Zhao, Xiaoyuan Cheng, Huihang Liu, Baihua He, Xinyu Zhang, Harrison Bo Hua Zhu, Wenlong Chen, Li Zeng, Zhuo Sun 5/8/2026

Saliency-Aware Regularized Quantization Calibration for Large Language Models

Saliency-aware regularization method for post-training quantization of large language models under memory constraints.

Ax Zhengru Fang, Senkang Forest Hu, Zhonghao Chang, Yu Guo, Yihang Tao, Hongyao Liu, Mengzhe Ruan, Jun Huang, Yuguang Fang 5/8/2026

Inference-Time Budget Control for LLM Search Agents

Budget control framework for LLM search agents optimizing tool calls and token allocation in multi-hop QA.

Ax Huyu Wu, Jun Liu, Xiaochi Wei, Yan Gao, Yi Wu, Yao Hu 5/8/2026

Knowledge-Graph Paths as Intermediate Supervision for Self-Evolving Search Agents

Knowledge-graph paths as intermediate supervision for self-evolving search agents, improving multi-step reasoning.

Ax Md Farhamdur Reza, Richeng Jin, Tianfu Wu, Huaiyu Dai 5/8/2026

Conceal, Reconstruct, Jailbreak: Exploiting the Reconstruction-Concealment Tradeoff in MLLMs

Analysis of reconstruction-concealment tradeoff in MLLM jailbreak attacks, studying safety mechanism vulnerabilities.

Ax Ming Liu 5/8/2026

Decodable but Not Corrected by Fixed Residual-Stream Linear Steering: Evidence from Medical LLM Failure Regimes

Investigation of linear decodability vs. correction gap in medical LLM failures via Overthinking behavior analysis.

Ax Ming Liu 5/8/2026

More Is Not Always Better: Cross-Component Interference in LLM Agent Scaffolding

Study of cross-component interference in LLM agent systems across 32 configurations, showing all-in approaches often degrade performance.

Ax Hyeongwon Kang, Jeongseob Kim, Jinwoo Park, Pilsung Kang 5/8/2026

Detecting Time Series Anomalies Like an Expert: A Multi-Agent LLM Framework with Specialized Analyzers

Multi-agent LLM framework (SAGE) for time-series anomaly detection using specialized analyzers for structured diagnosis.

Ax Hongcheol Cho, Ryangkyung Kang, Youngeun Kim 5/8/2026

SkillRet: A Large-Scale Benchmark for Skill Retrieval in LLM Agents

SkillRet benchmark for skill retrieval in LLM agent systems, addressing critical challenge of selecting appropriate skills from large libraries under constraints.

Ax Wei Li, Shibo Feng, Pengcheng Wu, Min Wu, Peilin Zhao 5/8/2026

SDFlow: Similarity-Driven Flow Matching for Time Series Generation

SDFlow addresses exposure bias in autoregressive time-series generation via similarity-driven flow matching, reducing accumulated errors in long-horizon prediction.

Ax Fan Huang 5/8/2026

ReFlect: An Effective Harness System for Complex Long-Horizon LLM Reasoning

ReFlect system enables LLM reasoning to detect and recover from errors across multi-stage long-horizon tasks, improving reliability beyond chain-of-thought approaches.

Ax Chengda Lu, Xiaoyu Fan, Wei Xu 5/8/2026

HyperLens: Quantifying Cognitive Effort in LLMs with Fine-grained Confidence Trajectory

HyperLens analyzes LLM inference dynamics through fine-grained confidence trajectories, leveraging layer-wise magnification in transformers to quantify cognitive effort.

Ax Md Touhidul Islam, Sujan Kumar Saha, Farimah Farahmandi, Mark Tehranipoor 5/8/2026

CircuitFormer: A Circuit Language Model for Analog Topology Design from Natural Language Prompt

Transformer-based LLM for analog circuit design from natural language, addressing dataset scarcity and efficiency challenges in hardware design automation.

Ax Yiming Lei, Yiqi Wang, Yujia Zhang, Bo Guan, Depei Zhu, Chunhui Wang, Zhuonan Hao, Tianyu Shi 5/8/2026

Sheet as Token: A Graph-Enhanced Representation for Multi-Sheet Spreadsheet Understanding

Graph-enhanced representation method for multi-sheet spreadsheet understanding to improve LLM-based data analysis agents' ability to handle heterogeneous schemas.

Ax Armaan A. Abraham, Lucy Xiaoyang Shi, Chelsea Finn 5/8/2026

Long-Horizon Q-Learning: Accurate Value Learning via n-Step Inequalities

Proposes long-horizon Q-learning (LQL) to address error propagation in off-policy value-based reinforcement learning through n-step inequalities.

Ax Yang Xu, Kun Yao, Yiming Deng, Zheng Fang, Kai Ming Ting, Ming Pang 5/8/2026

AGPO: Asymmetric Group Policy Optimization for Verifiable Reasoning and Search Ads Relevance at JD

AGPO: Asymmetric group policy optimization for LLM reasoning verification in search ads relevance, addressing reasoning capability narrowing in RLVR.

Ax Guanyu Zhu, Jining Luan, Hanwen Du, Xinyu Fang, Sibo Xu, Ersheng Ni, Hongji Li, Jincheng Fang, Ronghao Chen, Huacan Wang, Xuanqi Lan, Yongxin Ni, Yiqi Sun, Youhua Li 5/8/2026

On the Role of Language Representations in Auto-Bidding: Findings and Implications

Study on using language representations in auto-bidding for real-time advertising to explicitly control high-level intent and strategy.

Ax Zaki Kurdya, Mohammed Zuqlam, Salem Amassi, Shady Telbany, Motaz Saad 5/8/2026

Taklif.AI: LLM-Powered Platform for Interest-Based Personalized College Assignments

Taklif.AI: LLM-powered platform automatically generating personalized college assignments based on student interests and abilities.

Ax Yong Xiao, Haoran Zhou, Yujie Zhou, Marwan Krunz 5/8/2026

SANEmerg: An Emergent Communication Framework for Semantic-aware Agentic AI Networking

SANEmerg: Emergent communication framework enabling semantic-aware cooperation between heterogeneous AI agents in networking systems.

Ax Basel Magableh, OmniRisk Research 5/8/2026

Agentic, Context-Aware Risk Intelligence in the Internet of Value

Risk intelligence system for Internet of Value using prediction engines and verification for composite risk in heterogeneous networks.

Ax Yuhang Wang, Zhenxing Niu, Haoxuan Ji, Guangyu He, Linlin Zhang, Haichang Gao 5/8/2026

Null Space Constrained Contrastive Visual Forgetting for MLLM Unlearning

MLLM unlearning approach using null space constraints to remove target visual knowledge while preserving non-target knowledge in multimodal models.

Ax Alex B\"auerle, Adam Connors, Alexander Novikov, Adam Zsolt Wagner, Ng\^an V\~u, Fernanda Viegas, Martin Wattenberg, Lucas Dixon 5/8/2026

Intentmaking and Sensemaking: Human Interaction with AI-Guided Mathematical Discovery

User study on expert mathematicians interacting with AlphaEvolve evolutionary coding agent, characterizing intentmaking workflow for AI-guided discovery.

Ax Yuhang Wang, Wenjie Mei, Junkai Zhang, Guangyu He, Zhenxing Niu, Haichang Gao 5/8/2026

ICU-Bench:Benchmarking Continual Unlearning in Multimodal Large Language Models

ICU-Bench: Benchmark for evaluating continual unlearning in multimodal large language models with sequential privacy deletion requests.

Ax Yuliang Xu, Xiang Xu, Yao Wan, Hu Wei, Tong Jia 5/8/2026

MAS-Algorithm: A Workflow for Solving Algorithmic Programming Problems with a Multi-Agent System

MAS-Algorithm: Multi-agent system workflow for solving algorithmic programming problems using structured reasoning and external tools.

Ax Peilin Zhan, Wei Chen, Weilin Chen, Shuyi Pan, Ruichu Cai 5/8/2026

Temporal Smoothness Doubly Robust Learning for Debiased Knowledge Tracing

Debiased knowledge tracing approach addressing selection bias in educational logging systems for student mastery estimation.

Ax Xinghao Wu, Jianwei Niu, Guogang Zhu, Xuefeng Liu, Shaojie Tang, Jiayuan Zhang 5/8/2026

From Coordinate Matching to Structural Alignment: Rethinking Prototype Alignment in Heterogeneous Federated Learning

Prototype-based alignment method for heterogeneous federated learning across clients with different data distributions and model architectures.

Ax Junkai Li, Yunghwei Lai, Tianyi Zhu, Zheng Long Lee, Weizhi Ma, Yang Liu 5/8/2026

TheraAgent: Self-Improving Therapeutic Agent for Precise and Comprehensive Treatment Planning

TheraAgent: Agentic framework for iterative treatment planning using LLMs with verification loops instead of one-shot generation for safer, more comprehensive medical plans.

Ax Yinbo Yu, Xueyu Yin, Jiadai Wang, Chunwei Tian, Sai Xu, Qi Zhu, Daoqiang Zhang 5/8/2026

BehaviorGuard: Online Backdoor Defense for Deep Reinforcement Learning

BehaviorGuard provides online backdoor defense for deep reinforcement learning using trigger-agnostic behavior analysis.

Ax Yuan Sui, Yulin Chen, Yibo Li, Xue Jiang, Yufei He, Yihong Dong, Xiaoxin He, Tianyu Gao, Bryan Hooi 5/8/2026

TACT: Mitigating Overthinking and Overacting in Coding Agents via Activation Steering

TACT mitigates agent drift in coding agents via activation steering, addressing overthinking and overacting failure modes in long-horizon tasks.

Ax Remigiusz Kinas, Joanna Krawczyk, Rafa{\l} Powalski, Przemys{\l}aw Pietrzak, Agnieszka Kowalewska, Krzysztof Kolmus, Maciej Sypetkowski, {\L}ukasz Smoli\'nski, Tomasz Jetka 5/8/2026

BioResearcher: Scenario-Guided Multi-Agent for Translational Medicine

BioResearcher multi-agent system for translational medicine combining literature, trials, patents, and multi-omics analysis with auditable provenance.

Ax Wenliang Huang, Zengyi Yu 5/8/2026

Strat-LLM: Stratified Strategy Alignment for LLM-based Stock Trading with Real-time Multi-Source Signals

Strat-LLM framework uses LLM agents for stock trading with real-time multi-source data integration and strategy alignment, tested live in 2025.

Ax Leon Hamm, Zlatan Ajanovic 5/8/2026

Novelty-based Tree-of-Thought Search for LLM Reasoning and Planning

Novelty-based tree-of-thought search algorithm for improved LLM reasoning and planning with reduced token costs.

Ax Amal Alnouri, Andreas Hinterreiter, Christina Humer, Furui Cheng, Marc Streit 5/8/2026

Visual Fingerprints for LLM Generation Comparison

Visual fingerprinting method for analyzing how prompts, instructions, and parameters shape LLM output behaviors.

Ax Keisuke Kamahori, Shihang Li, Simon Peter, Baris Kasikci 5/8/2026

VibeServe: Can AI Agents Build Bespoke LLM Serving Systems?

VibeServe: multi-agent agentic loop that automatically synthesizes bespoke LLM serving system stacks end-to-end.

Ax Oliver Sch\"on, Licio Romao, Sadegh Soudjani 5/8/2026

Safety Certification is Classification

Kernel embedding framework treating safety certification as classification problem for dynamical systems under uncertainty.

Ax Jungsuk Oh, Hyeseo Jeon, Hyunjune Ji, Kyongmin Kong, Jay-Yoon Lee 5/8/2026

Shallow Prefill, Deep Decoding: Efficient Long-Context Inference via Layer-Asymmetric KV Visibility

SPEED: layer-asymmetric key-value visibility policy for efficient long-context inference in decoder-only language models.

Ax Xinglin Wang, Zishen Liu, Shaoxiong Feng, Peiwen Yuan, Yiwei Li, Jiayi Shi, Yueqi Zhang, Chuyi Tan, Ji Zhang, Boyuan Pan, Yao Hu, Kan Li 5/8/2026

On Time, Within Budget: Constraint-Driven Online Resource Allocation for Agentic Workflows

Framework for resource-constrained scheduling of agentic workflows under time and budget constraints with task dependencies.

Ax Zhen Zeng, Leijiang Gu, Feng Li, Jing Yu, Zenglin Shi 5/8/2026

CrossCult-KIBench: A Benchmark for Cross-Cultural Knowledge Insertion in MLLMs

CrossCult-KIBench benchmark for evaluating cross-cultural knowledge adaptation in multimodal language models.

Ax Wenwen Si, Insup Lee, Osbert Bastani 5/8/2026

Policy-Guided Stepwise Model Routing for Cost-Effective Reasoning

Policy-guided model routing for cost-effective reasoning by dynamically routing chain-of-thought states across language models of different sizes.

Ax Nguyen Viet Tuan Kiet, Bui Dinh Pham, Dao Van Tung, Tran Cong Dao, Huynh Thi Thanh Binh 5/8/2026

Back to the Beginning of Heuristic Design: Bridging Code and Knowledge with LLMs

LLM-based automatic heuristic design for combinatorial optimization using bottom-up paradigm from code to knowledge insights.

Ax Xin Peng, Ang Gao 5/8/2026

P-Guide: Parameter-Efficient Prior Steering for Single-Pass CFG Inference

P-Guide: parameter-efficient method for classifier-free guidance in flow matching using single-pass inference with latent state modulation.

Ax Yaorui Shi, Yuxin Chen, Zhengxi Lu, Yuchun Miao, Shugui Liu, Qi GU, Xunliang Cai, Xiang Wang, An Zhang 5/8/2026

Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning

Skill1 framework for unified evolution of skill-augmented language model agents through reinforcement learning with persistent skill libraries.

Ax Kossi Amouzouvi, Robert Wardenga, Jens Lehmann, Sahar Vahdati 5/8/2026

Graphlets as Building Blocks for Structural Vocabulary in Knowledge Graph Foundation Models

Graphlets as structural vocabulary tokens for knowledge graph foundation models to enable discrete symbolic representation.

Ax Shihao Weng, Yang Feng, Xiaofei Xie 5/8/2026

Beyond Accuracy: Policy Invariance as a Reliability Test for LLM Safety Judges

Policy invariance framework for testing reliability of LLM-as-a-Judge safety evaluation pipelines used in agent systems.

Ax Richmond Sin Jing Xuan, Rishabh Bhardwaj, Soujanya Poria 5/8/2026

Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost

Post-Reasoning approach to improve LLM performance on tasks requiring minimal reasoning while reducing token consumption and latency.