papers.joyehuang.me

Paper Recommendations

Daily AI-powered curation from arXiv and Hugging Face Daily Papers. Last updated May 19, 2026 · 23:00 UTC.

Archive joyehuang.me Source
88
Candidates
5
Top Picks
1
Blockbusters
19
Max Score
Today's Picks

Editor's Rationale

**1. EnvFactory: Scaling Tool-Use Agents via Executable Environments Synthesis and Robust RL** EnvFactory proposes a fully automated framework that autonomously explores and verifies executable multi-turn tool environments from authentic resources, generating natural multi-turn trajectories through topology-aware sampling and calibrated refinement. With only 85 verified environments across 7 domains, it produces 2,575 SFT and RL trajectories, improving Qwen3-series models by up to +15% on BFCLv3, +8.6% on MCP-Atlas. This work is highly practical for Tool-Use Agent training data production, especially its approach to automated environment synthesis + MCP integration, directly addressing the two key bottlenecks in Agentic RL: scarcity of training data and unreliable simulators. **2. TOBench: A Task-Oriented Omni-Modal Benchmark for Real-World Tool-Using Agents** TOBench is a benchmark for real-world tool-using agents with 100 executable tasks supported by 27 MCP servers and 324 tools. Its core innovation is closed-loop multimodal verification: agents must execute tools, inspect intermediate artifacts, and self-correct. Claude Opus 4.6 only achieves 32% task success (human baseline 94%), showing huge room for improvement. The MCP protocol practices and tool-use evaluation design are directly relevant. **3. Code as Agent Harness** A comprehensive survey on agent infrastructure architecture, proposing code as agent harness as a unified perspective: code is not just the output target of agents but the operational substrate for agent reasoning, action, environment modeling, and execution-based verification. Organized around three layers: harness interface (connecting reasoning/action/environment), harness mechanisms (planning/memory/tool use/feedback-driven optimization), and scaling (single-agent to multi-agent coordination). Covers coding assistants, GUI/OS automation, scientific discovery, DevOps and other real-world applications.

Top Picks

5 papers
#01 19

TOBench: A Task-Oriented Omni-Modal Benchmark for Real-World Tool-Using Agents

Zhiqiang Liu, Wenhui Dong, Yilang Tan et al.
Tool-using agents are increasingly expected to operate across realistic professional workflows, where they must interpret multimodal inputs, coordinate external tools, inspect intermediate artifacts, and revise their actions before producing a final result. Existing benchmarks, h…
agentagentictool useMCPreasoningworkflow
#02 18

Code as Agent Harness

Xuying Ning, Katherine Tieu, Dongqi Fu et al.
cs.CL, cs.AI
Recent large language models (LLMs) have demonstrated strong capabilities in understanding and generating code, from competitive programming to repository-level software engineering. In emerging agentic systems, code is no longer only a target output. It increasingly serves as an…
agentmulti-agentagentictool useplanningreasoning
#03 18

EnvFactory: Scaling Tool-Use Agents via Executable Environments Synthesis and Robust RL

Minrui Xu, Zilin Wang, Mengyi DENG et al.
cs.CL, cs.LG
Equipping LLMs with tool-use capabilities via Agentic Reinforcement Learning (Agentic RL) is bottlenecked by two challenges: the lack of scalable, robust execution environments and the scarcity of realistic training data that captures implicit human reasoning. Existing approaches…
agentagentictool-useMCPreasoningautonomous
#04 16

MA$^{2}$P: A Meta-Cognitive Autonomous Intelligent Agents Framework for Complex Persuasion

Dingyi Zhang, Ziqing Zhuang, Linhai Zhang et al.
cs.CL
Persuasive dialogue generation plays a vital role in decision-making, negotiation, counseling, and behavior change, yet it remains a challenging problem. In complex persuasion where the persuadee's internal states are not expressed clearly, the persuader must interpret responses,…
agentmulti-agentplanningreasoningautonomousRAG
#05 15

AstraFlow: Dataflow-Oriented Reinforcement Learning for Agentic LLMs

Haizhong Zheng, Yizhuo Di, Jiahui Wang et al.
Reinforcement learning (RL) is increasingly used to improve the reasoning, coding, and tool-use capabilities of large language models, but agentic RL remains prohibitively expensive. Scaling RL to agentic LLMs requires supporting complex workloads, including multi-policy collabor…
agentagentictool-usereasoningautonomousPPO

All Candidates

88 papers
ScorePaperAuthorsCategory
14 From Runnable to Shippable: Multi-Agent Test-Driven Development for Generating Full-Stack Web Applications from Requirements Yuxuan Wan, Tingshuo Liang et al.
14 Agent Bazaar: Enabling Economic Alignment in Multi-Agent Marketplaces Seth Karten, Cameron Crow et al.
12 VideoSeeker: Incentivizing Instance-level Video Understanding via Native Agentic Tool Invocation Yiming Zhao, Yu Zeng et al.
12 AgentKernelArena: Generalization-Aware Benchmarking of GPU Kernel Optimization Agents Sharareh Younesian, Wenwen Ouyang et al.
11 CHI-Bench: Can AI Agents Automate End-to-End, Long-Horizon, Policy-Rich Healthcare Workflows? Haolin Chen, Deon Metelski et al.
10 LongMINT: Evaluating Memory under Multi-Target Interference in Long-Horizon Agent Systems Hyunji Lee, Justin Chih-Yao Chen et al. cs.CL, cs.AI
10 Evaluating Cognitive Age Alignment in Interactive AI Agents Yifan Shen, Jiawen Zhang et al.
10 Code-as-Room: Generating 3D Rooms from Top-Down View Images via Agentic Code Synthesis Yixuan Yang, Zhen Luo et al.
10 Model-Adaptive Tool Necessity Reveals the Knowing-Doing Gap in LLM Tool Use Yize Cheng, Chenrui Fan et al.
9 A2RBench: An Automatic Paradigm for Formally Verifiable Abstract Reasoning Benchmark Generation Qingchuan Ma, Yuexiao Ma et al.
9 OProver: A Unified Framework for Agentic Formal Theorem Proving David Ma, Kaijing Ma et al.
8 SkillGenBench: Benchmarking Skill Generation Pipelines for LLM Agents Yifan Zhou, Zhentao Zhang et al. cs.AI
8 Reversa: A Reverse Documentation Engineering Framework for Converting Legacy Software into Operational Specifications for AI Agents Sanderson Oliveira de Macedo, Ronaldo Martins da Costa cs.SE, cs.AI
8 Overeager Coding Agents: Measuring Out-of-Scope Actions on Benign Tasks Yubin Qu, Ying Zhang et al. cs.SE, cs.AI, cs.CL, cs.CR
8 SCICONVBENCH: Benchmarking LLMs on Multi-Turn Clarification for Task Formulation in Computational Science Nithin Somasekharan, Youssef Hassan et al.
8 MementoGUI: Learning Agentic Multimodal Memory Control for Long-Horizon GUI Agents Ziyun Zeng, Hang Hua et al.
8 FINESSE-Bench: A Hierarchical Benchmark Suite for Financial Domain Knowledge and Technical Analysis in Large Language Models Dmitry Stanishevskii, Nini Kamkia et al.
8 Auditing Agent Harness Safety Chengzhi Liu, Yichen Guo et al.
7 Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation Qianhao Yuan, Jie Lou et al. cs.CV, cs.AI, cs.CL, cs.LG
7 Monitoring the Internal Monologue: Probe Trajectories Reveal Reasoning Dynamics Maciej Chrabąszcz, Aleksander Szymczyk et al. cs.CL, cs.CR
7 GRASP: Learning to Ground Social Reasoning in Multi-Person Non-Verbal Interactions Junho Kim, Xu Cao et al.
7 Stop When Reasoning Converges: Semantic-Preserving Early Exit for Reasoning Models Dehai Min, Giovanni Vaccarino et al.
6 DexHoldem: Playing Texas Hold'em with Dexterous Embodied System Feng Chen, Tianzhe Chu et al. cs.RO, cs.AI
6 Democratizing Large-Scale Re-Optimization with LLM-Guided Model Patches Tinghan Ye, Arnaud Deza et al. cs.AI, math.OC
6 Efficient Lookahead Encoding and Abstracted Width for Learning General Policies in Classical Planning Michael Aichmüller, Simon Ståhlberg et al. cs.AI
6 Position: A Three-Layer Probabilistic Assume-Guarantee Architecture Is Structurally Required for Safe LLM Agent Deployment S. Bensalem, Y. Dong et al. cs.AI
6 AI for Auto-Research: Roadmap & User Guide Lingdong Kong, Xian Sun et al.
5 Learning Quantifiable Visual Explanations Without Ground-Truth 🚀 Amritpal Singh, Andrey Barsky et al. cs.AI, cs.LG
5 ESI-Bench: Towards Embodied Spatial Intelligence that Closes the Perception-Action Loop Yining Hong, Jiageng Liu et al. cs.CV, cs.AI, cs.CL, cs.LG, cs.RO
5 Generative AI Advertising as a Problem of Trustworthy Commercial Intervention Jingyi Qiu, Qiaozhu Mei cs.CY, cs.CL
5 Auditing Multimodal LLM Raters: Central Tendency Bias in Clinical Ordinal Scoring Jiaqing Zhang, Sandeep Elluri et al.
5 SkillsVote: Lifecycle Governance of Agent Skills from Collection, Recommendation to Evolution Hongyi Liu, Haoyan Yang et al.
5 AtlasVA: Self-Evolving Visual Skill Memory for Teacher-Free VLM Agents Pan Wang, Yihao Hu et al.
4 PopPy: Opportunistically Exploiting Parallelism in Python Compound AI Applications Stephen Mell, David Mell et al. cs.DC, cs.AI, cs.PL
4 GIM: Evaluating models via tasks that integrate multiple cognitive domains Rohit Patel, Alexandre Rezende et al. cs.AI, cs.CL, cs.LG
4 AR-VLA: True Autoregressive Action Expert for Vision-Language-Action Models Yutong Hu, Jan-Nico Zaech et al.
4 MixSD: Mixed Contextual Self-Distillation for Knowledge Injection Jiarui Liu, Lechen Zhang et al.
4 KVPO: ODE-Native GRPO for Autoregressive Video Alignment via KV Semantic Exploration Ruicheng Zhang, Kaixi Cong et al.
4 EndPrompt: Efficient Long-Context Extension via Terminal Anchoring Han Tian, Luxuan Chen et al.
4 No One Knows the State of the Art in Geospatial Foundation Models Isaac Corley, Nils Lehmann et al.
3 What Does the AI Doctor Value? Auditing Pluralism in the Clinical Ethics of Language Models Payal Chandak, Victoria Alkin et al. cs.AI
3 Semantic Generative Tuning for Unified Multimodal Models Songsong Yu, Yuxin Chen et al. cs.CV, cs.AI
3 COOPO: Cyclic Offline-Online Policy Optimization Algorithm Qisai Liu, Zhanhong Jiang et al. cs.LG, cs.AI
3 Forecasting Downstream Performance of LLMs With Proxy Metrics Arkil Patel, Siva Reddy et al. cs.CL, cs.LG
3 AMARIS: A Memory-Augmented Rubric Improvement System for Rubric-Based Reinforcement Learning Peilin Wu, Xinlu Zhang et al. cs.LG, cs.AI, cs.CL
3 OSCAR: Offline Spectral Covariance-Aware Rotation for 2-bit KV Cache Quantization Zhongzhu Zhou, Donglin Zhuang et al.
3 WavFlow: Audio Generation in Waveform Space Feiyan Zhou, Luyuan Wang et al.
3 NGM: A Plug-and-Play Training-Free Memory Module for LLMs Yuwen Qu, Wenhui Dong et al.
3 Targeted Neuron Modulation via Contrastive Pair Search Sam Herring, Jake Naviasky et al.
3 LongLive-2.0: An NVFP4 Parallel Infrastructure for Long Video Generation Yukang Chen, Luozhou Wang et al.
3 Geometric Phase Transition Enables Extreme Hippocampal Memory Capacity Prashant C. Raju
2 DashAttention: Differentiable and Adaptive Sparse Hierarchical Attention Yuxiang Huang, Nuno M. T. Gonçalves et al. cs.CL, cs.AI, cs.LG
2 Lance: Unified Multimodal Modeling by Multi-Task Synergy Fengyi Fu, Mengqi Huang et al. cs.CV, cs.AI
2 A Readiness-Driven Runtime for Pipeline-Parallel Training under Runtime Variability Ruitao Liu, Xinyang Tian et al. cs.DC, cs.LG
2 SURGE: Approximation-free Training Free Particle Filter for Diffusion Surrogate Lifu Wei, Yinuo Ren et al. stat.ML, cs.LG, math.NA, math.PR, q-fin.MF, stat.C
2 General Preference Reinforcement Learning Muhammad Umer, Muhammad Ahmed Mohsin et al. cs.LG, cs.CL
2 Better Together: Evaluating the Complementarity of Earth Embedding Models Thijs L van der Plas, Jacob JW Bakermans et al. cs.CV, cs.LG
2 LiteFrame: Efficient Vision Encoders Unlock Frame Scaling in Video LLMs Jihwan Kim, Nikhil Parthasarathy et al.
1 Distilling Tabular Foundation Models for Structured Health Data Aditya Tanna, Nassim Bouarour et al. cs.LG, cs.AI
1 Ensembling Tabular Foundation Models - A Diversity Ceiling And A Calibration Trap Aditya Tanna, Yash Desai et al. cs.LG, cs.AI
1 Learned Memory Attenuation in Sage-Husa Kalman Filters for Robust UAV State Estimation Kenan Majewski, Marcin Żugaj eess.SP, cs.LG
1 Learning Normal Representations for Blood Biomarkers Aashna P. Shah, Michelle M. Li et al. cs.LG, q-bio.QM
1 An Assessment of Human vs. Model Uncertainty in Soft-Label Learning and Calibration Maja Pavlovic, Silviu Paun et al. cs.LG, cs.AI, cs.CL
1 Post-Trained MoE Can Skip Half Experts via Self-Distillation Xingtai Lv, Li Sheng et al. cs.LG, cs.AI, cs.CL
1 GUT-IS: A Data-Driven Approach to Integrating Constructs and Their Relations in Information Systems Maximilian Reinhardt, Jonas Scharfenberger et al. cs.CL, cs.LG
1 Symmetry-Compatible Principle for Optimizer Design: Embeddings, LM Heads, SwiGLU MLPs, and MoE Routers Tim Tsz-Kit Lau, Weijie Su
1 TopoPrimer: The Missing Topological Context in Forecasting Models Zara Zetlin, Kayhan Moharreri et al.
1 SafeDiffusion-R1: Online Reward Steering for Safe Diffusion Post-Training Komal Kumar, Ankan Deria et al.
1 StableVLA: Towards Robust Vision-Language-Action Models without Extra Data Yiyang Fu, Chubin Zhang et al.
1 Incantation: Natural Language as the Action Interface for Multi-Entity Video World Models Shangwen Zhu, Qianyu Peng et al.
1 CompactAttention: Accelerating Chunked Prefill with Block-Union KV Selection Jiwon Song, Dongwon Jo et al.
1 Measuring Maximum Activations in Open Large Language Models Luxuan Chen, Han Tian et al.
0 Actionable World Representation Kunqi Xu, Jitao Li et al. cs.AI
0 Predictable Confabulations: Factual Recall by LLMs Scales with Model Size and Topic Frequency Matthew L. Smith, Jonathan P. Shock et al. cs.CL, cs.AI, cs.LG
0 PIXLRelight: Controllable Relighting via Intrinsic Conditioning Miguel Farinha, Ronald Clark cs.CV, cs.GR, cs.LG
0 Can Adaptive Gradient Methods Converge under Heavy-Tailed Noise? A Case Study of AdaGrad Zijian Liu math.OC, cs.LG, stat.ML
0 Can machine learning for quantum-gas experiments be explainable? I. B. Spielman amd J. P. Zwolak cond-mat.quant-gas, cs.LG, physics.atom-ph, quant-
0 A No-Defense Defense Against Gradient-Based Adversarial Attacks on ML-NIDS: Is Less More? Mohamed elShehaby, Ashraf Matrawy cs.LG, cs.CR
0 Language-Switching Triggers Take a Latent Detour Through Language Models Francis Kulumba, Wissam Antoun et al. cs.CL
0 Readers make targeted regressions to plausible errors in reanalysis of "noisy-channel garden-path" sentences Thomas Hikaru Clark, Roger Levy et al. cs.CL
0 E-PMQ: Expert-Guided Post-Merge Quantization with Merged-Weight Anchoring Wenjun Wang, Yanggan Gu et al.
0 Where Should Diffusion Enter a Language Model? Geometry-Guided Hidden-State Replacement Injin Kong, Hyoungjoon Lee et al.
0 SNLP: Layer-Parallel Inference via Structured Newton Corrections Ligong Han, Kai Xu et al.