papers.joyehuang.me

Paper Recommendations

Daily AI-powered curation from arXiv and Hugging Face Daily Papers. Last updated May 12, 2026 · 23:00 UTC.

Archive joyehuang.me Source
87
Candidates
5
Top Picks
2
Blockbusters
16
Max Score
Today's Picks

Editor's Rationale

## Today's Recommendations ### 1. **Towards On-Policy Data Evolution for Visual-Native Multimodal Deep Search Agents** **Core Contribution**: Proposes a visual-native agent architecture with an Image Bank Reference Protocol that registers tool-returned images as addressable references, enabling intermediate visual evidence to be reused by later tools. Also introduces On-policy Data Evolution (ODE) for closed-loop data generation from policy rollouts. **Why Recommend**: Covers Agent architecture design, Tool Use (solves cross-tool visual context passing), and Context management (Image Bank enables multimodal context accumulation across tool calls). ODE's on-policy bootstrapping is valuable for Agent fine-tuning and RL training. **Link**: https://arxiv.org/abs/2605.10832v1 ### 2. **TMAS: Scaling Test-Time Compute via Multi-Agent Synergy** **Core Contribution**: Proposes a multi-agent framework with hierarchical memory systems - Experience Bank for reusing low-level reasoning conclusions and Guideline Bank for recording high-level strategies. Uses hybrid reward RL training to outperform existing test-time scaling methods. **Why Recommend**: Directly related to Multi-Agent orchestration and Context management. The Experience Bank and Guideline Bank design is practical - not simply adding more agents, but using structured memory modules for effective knowledge transfer between agents. **Link**: https://arxiv.org/abs/2605.10344 ### 3. **Rethinking Agentic Search with Pi-Serini: Is Lexical Retrieval Sufficient?** **Core Contribution**: Challenges the assumption that agentic search requires dense retrievers. BM25 + frontier LLMs achieves 83.1% accuracy and 94.7% evidence recall on BrowseComp-Plus, surpassing dense retriever-based search agents. Open-sources Pi-Serini search agent framework. **Why Recommend**: Provides counterintuitive insight for Tool Use and Agent retrieval architecture - stronger agent reasoning capabilities mean simpler search backends suffice. Directly impacts agent search system architecture decisions. **Link**: https://arxiv.org/abs/2605.10848v1 ### 4. **Engineering Robustness into Personal Agents with the AI Workflow Store** **Core Contribution**: Critiques the dominant on-the-fly loop paradigm for agents, proposing to adopt software engineering best practices through an AI Workflow Store supporting iterative design, rigorous testing, adversarial evaluation, and staged deployment. **Why Recommend**: Directly addresses Agent architecture design for production environments. The Workflow Store approach is practical for building reliable production-grade agent systems rather than demos. **Link**: https://arxiv.org/abs/2605.10907v1

Top Picks

5 papers
#01 16

Towards On-Policy Data Evolution for Visual-Native Multimodal Deep Search Agents

Shijue Huang, Hangyu Guo, Chenxin Li et al.
cs.CL
Multimodal deep search requires an agent to solve open-world problems by chaining search, tool use, and visual reasoning over evolving textual and visual context. Two bottlenecks limit current systems. First, existing tool-use harnesses treat images returned by search, browsing, …
agenttool usetool-usereasoningworkflowRAG
#02 14

AssayBench: An Assay-Level Virtual Cell Benchmark for LLMs and Agents

Edward De Brouwer, Carl Edwards, Alexander Wu et al.
cs.LG, cs.AI, q-bio.QM
Recent advances in machine learning and large-scale biological data collections have revived the prospect of building a virtual cell, a computational model of cellular behavior that could accelerate biological discovery. One of the most compelling promises of this vision is the a…
agentagenticworkflowfine-tuningevaluationbenchmark
#03 13

Rethinking Agentic Search with Pi-Serini: Is Lexical Retrieval Sufficient?

Tz-Huan Hsu, Jheng-Hong Yang, Jimmy Lin
cs.IR, cs.AI, cs.CL
Does a lexical retriever suffice as large language models (LLMs) become more capable in an agentic loop? This question naturally arises when building deep research systems. We revisit it by pairing BM25 with frontier LLMs that have better reasoning and tool-use abilities. To supp…
agentagentictool-usereasoningPPO
#04 13

TMAS: Scaling Test-Time Compute via Multi-Agent Synergy

George Wu, Nan Jing, Qing Yi et al.
Test-time scaling has become an effective paradigm for improving the reasoning ability of large language models by allocating additional computation during inference. Recent structured approaches have further advanced this paradigm by organizing inference across multiple trajecto…
agentmulti-agentreasoningRAGbenchmarkPPO
#05 12

WildClawBench: A Benchmark for Real-World, Long-Horizon Agent Evaluation

Shuangrui Ding, Xuanlang Dai, Long Xing et al.
cs.CL
Large language and vision-language models increasingly power agents that act on a user's behalf through command-line interface (CLI) harnesses. However, most agent benchmarks still rely on synthetic sandboxes, short-horizon tasks, mock-service APIs, and final-answer checks, leavi…
agentRAGevaluationbenchmarkPPO

All Candidates

87 papers
ScorePaperAuthorsCategory
12 AgentForesight: Online Auditing for Early Failure Prediction in Multi-Agent Systems Boxuan Zhang, Jianing Zhu et al.
11 Engineering Robustness into Personal Agents with the AI Workflow Store Roxana Geambasu, Mariana Raykova et al. cs.CR, cs.AI
10 Dynamic Skill Lifecycle Management for Agentic Reinforcement Learning Junhao Shen, Teng Zhang et al. cs.LG, cs.CL
9 PlantMarkerBench: A Multi-Species Benchmark for Evidence-Grounded Plant Marker Reasoning Sajib Acharjee Dip, Song Li et al.
8 Clin-JEPA: A Multi-Phase Co-Training Framework for Joint-Embedding Predictive Pretraining on EHR Patient Trajectories Yixuan Yang, Mehak Arora et al. cs.LG, cs.AI, q-bio.QM
8 From Controlled to the Wild: Evaluation of Pentesting Agents for the Real-World Pedro Conde, Henrique Branquinho et al. cs.AI, cs.CR
8 RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards Gaotang Li, Bhavana Dalvi Mishra et al. cs.CL, cs.LG
8 Safe, or Simply Incapable? Rethinking Safety Evaluation for Phone-Use Agents Zhengyang Tang, Yi Zhang et al.
7 Qwen-Image-2.0 Technical Report 🚀 Bing Zhao, Chenfei Wu et al.
7 DataMaster: Towards Autonomous Data Engineering for Machine Learning Yaxin Du, Xiyuan Yang et al. cs.LG, cs.AI
7 The Last Word Often Wins: A Format Confound in Chain-of-Thought Corruption Studies Gabriel Garcia cs.LG, cs.AI, cs.CL
7 Conformal Agent Error Attribution Naihe Feng, Yi Sui et al.
7 FORTIS: Benchmarking Over-Privilege in Agent Skills Shawn Li, Chenxiao Yu et al.
7 SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning Haoqiang Kang, Xiaokang Ye et al.
6 SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training 🚀 Shengkun Tang, Zekun Wang et al.
6 Shepherd: A Runtime Substrate Empowering Meta-Agents with a Formalized Execution Trace Simon Yu, Derek Chong et al. cs.AI, cs.PL, cs.SE
6 CADBench: A Multimodal Benchmark for AI-Assisted CAD Program Generation Anna C. Doris, Jacob Thomas Sony et al. cs.CV, cs.AI
6 Remember the Decision, Not the Description: A Rate-Distortion Framework for Agent Memory Mingxi Zou, Zhihan Guo et al. cs.AI
6 BenchCAD: A Comprehensive, Industry-Standard Benchmark for Programmatic CAD Haozhe Zhang, Kaichen Liu et al. cs.AI, cs.CV, cs.SE
6 The Generalized Turing Test: A Foundation for Comparing Intelligence Daniel Mitropolsky, Susan S. Hong et al. cs.AI, cs.CL, cs.LG
6 Grounded Satirical Generation with RAG Oona Itkonen, Yuxin Su et al. cs.CL
6 DeepRefine: Agent-Compiled Knowledge Refinement via Reinforcement Learning Haoyu Huang, Jiaxin Bai et al.
6 Crosslingual On-Policy Self-Distillation for Multilingual Reasoning Yihong Liu, Raoyuan Zhao et al.
6 RoboMemArena: A Comprehensive and Challenging Robotic Memory Benchmark Huashuo Lei, Wenxuan Song et al.
6 LLiMba: Sardinian on a Single GPU -- Adapting a 3B Language Model to a Vanishing Romance Language Luca Ballore
6 NanoResearch: Co-Evolving Skills, Memory, and Policy for Personalized Research Automation Jinhang Xu, Qiyuan Zhu et al.
5 Shields to Guarantee Probabilistic Safety in MDPs Linus Heck, Filip Macák et al. cs.LO, cs.AI
5 V4FinBench: Benchmarking Tabular Foundation Models, LLMs, and Standard Methods on Corporate Bankruptcy Prediction Marcin Kostrzewa, Sebastian Tomczak et al. cs.LG
5 DGPO: Beyond Pairwise Preferences with Directional Consistent Groupwise Optimization Mengyi Deng, Zhiwei Li et al. cs.CL
5 Reasoning Is Not Free: Robust Adaptive Cost-Efficient Routing for LLM-as-a-Judge Wenbo Zhang, Lijinghua Zhang et al. cs.AI, cs.CL, stat.ML
5 Metal-Sci: A Scientific Compute Benchmark for Evolutionary LLM Kernel Search on Apple Silicon Víctor Gallego
5 CollabVR: Collaborative Video Reasoning with Vision-Language and Video Generation Models Joowon Kim, Seungho Shin et al.
5 MuSS: A Large-Scale Dataset and Cinematic Narrative Benchmark for Multi-Shot Subject-to-Video Generation Haojie Zhang, Di Wu et al.
4 Training-Free Cultural Alignment of Large Language Models via Persona Disagreement Huynh Trung Kiet, Dao Sy Duy Minh et al. cs.CL, cs.AI, cs.CY
4 Optimal and Scalable MAPF via Multi-Marginal Optimal Transport and Schrödinger Bridges Usman A. Khan, Joseph W. Durham cs.LG, cs.MA, cs.RO
4 Revisiting Policy Gradients for Restricted Policy Classes: Escaping Myopic Local Optima with $k$-step Policy Gradients Alex DeWeese, Guannan Qu cs.LG, stat.ML
4 Neural at ArchEHR-QA 2026: One Method Fits All: Unified Prompt Optimization for Clinical QA over EHRs Abrar Majeedi, Viswanatha Reddy Gajjala et al. cs.CL, cs.IR
4 RUBEN: Rule-Based Explanations for Retrieval-Augmented LLM Systems Joel Rorseth, Parke Godfrey et al. cs.CL
4 SLIM: Sparse Latent Steering for Interpretable and Property-Directed LLM-Based Molecular Editing Mingxu Zhang, Yuhan Li et al. cs.LG, cs.AI, cs.CE, cs.CL
3 Unmasking On-Policy Distillation: Where It Helps, Where It Hurts, and Why Mohammadreza Armandpour, Fatih Ilhan et al. cs.LG, cs.AI
3 Equivariant Reinforcement Learning for Clifford Quantum Circuit Synthesis Richie Yeung, Aleks Kissinger et al. quant-ph, cs.LG
3 Masked Generative Transformer Is What You Need for Image Editing Wei Chow, Linfeng Li et al. cs.CV, cs.LG
3 Grounded or Guessing? LVLM Confidence Estimation via Blind-Image Contrastive Ranking Reza Khanmohammadi, Erfan Miahi et al. cs.CL
3 BabelDOC: Better Layout-Preserving PDF Translation via Intermediate Representation Qi Yang, Xiangyao Ma et al. cs.CV, cs.CL
3 GridProbe: Posterior-Probing for Adaptive Test-Time Compute in Long-Video VLMs Mohamed Eltahir, Lama Ayash et al.
3 jina-embeddings-v5-omni: Text-Geometry-Preserving Multimodal Embeddings via Frozen-Tower Composition Florian Hönicke, Michael Günther et al.
3 Training-Free Dense Hand Contact Estimation with Multi-Modal Large Language Models Daniel Sungho Jung, Kyoung Mu Lee
3 Memory-Efficient Looped Transformer: Decoupling Compute from Memory in Looped Language Models Victor Conchello Vendrell, Arnau Padres Masdemont et al.
3 PaperFit: Vision-in-the-Loop Typesetting Optimization for Scientific Documents Bihui Yu, Xinglong Xu et al.
3 WorldReasonBench: Human-Aligned Stress Testing of Video Generators as Future World-State Predictors Keming Wu, Yijing Cui et al.
2 LoKA: Low-precision Kernel Applications for Recommendation Models At Scale Liang Luo, Yinbin Ma et al. cs.LG, cs.AI
2 BEACON: A Multimodal Dataset for Learning Behavioral Fingerprints from Gameplay Data Ishpuneet Singh, Gursmeep Kaur et al. cs.CR, cs.AI, cs.CV, cs.LG, cs.NI
2 Learning More from Less: Exploiting Counterfactuals for Data-Efficient Chart Understanding Jianzhu Bao, Haozhen Zhang et al. cs.CL
2 Addressing Performance Saturation for LLM RL via Precise Entropy Curve Control Bolian Li, Yifan Wang et al.
2 A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models Hamid Kazemi, Atoosa Chegini et al.
2 Prompt-Activation Duality: Improving Activation Steering via Attention-Level Interventions Diancheng Kang, Zheyuan Liu et al.
2 Key-Value Means Daniel Goldstein, Eugene Cheah
2 Queryable LoRA: Instruction-Regularized Routing Over Shared Low-Rank Update Atoms Omatharv Bharat Vaidya, Connor T. Jerzak et al.
2 SlimSpec: Low-Rank Draft LM-Head for Accelerated Speculative Decoding Anton Plaksin, Sergei Krutikov et al.
2 LLaVA-UHD v4: What Makes Efficient Visual Encoding in MLLMs? Kechen Fang, Yihua Qin et al.
2 Scratchpad Patching: Decoupling Compute from Patch Size in Byte-Level Language Models Lin Zheng, Vasilisa Bashlovkina et al.
2 Injecting Distributional Awareness into MLLMs via Reinforcement Learning for Deep Imbalanced Regression Yao Du, Shanshan Song et al.
2 Dystruct: Dynamically Structured Diffusion Language Model Decoding via Bayesian Inference Bian Sun, Kevin Zhai et al.
1 ELF: Embedded Language Flows Keya Hu, Linlu Qiu et al. cs.CL, cs.AI, cs.LG
1 Confidence-Guided Diffusion Augmentation for Enhanced Bangla Compound Character Recognition Md. Sultan Al Rayhan, Maheen Islam cs.CV, cs.AI
1 Attractor-Vascular Coupling Theory: Formal Grounding and Empirical Validation for AAMI-Standard Cuffless Blood Pressure Estimation from Smartphone Photoplethysmography Timothy Oladunni, Farouk Ganiyu Adewumi physics.med-ph, cs.AI, cs.LG
1 DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices Chenyang Song, Weilin Zhao et al. cs.LG, cs.CL
1 Quantifying Concentration Phenomena of Mean-Field Transformers in the Low-Temperature Regime Albert Alcalde, Leon Bungert et al. math.AP, cs.LG, math.DS
1 Beyond Red-Teaming: Formal Guarantees of LLM Guardrail Classifiers Nikita Kezins, Urbas Ekka et al. cs.LG
1 Compute Where it Counts: Self Optimizing Language Models Yash Akhauri, Mohamed S. Abdelfattah cs.LG, cs.CL
1 Can Muon Fine-tune Adam-Pretrained Models? Xingyu Qu, Peigeng Huang et al.
1 SplatWeaver: Learning to Allocate Gaussian Primitives for Generalizable Novel View Synthesis Yecong Wan, Fan Li et al.
1 Shaping Schema via Language Representation as the Next Frontier for LLM Intelligence Expanding Zhiqin Yang, Yuhan Liu et al.
1 The Alpha Blending Hypothesis: Compositing Shortcut in Deepfake Detection Andrii Yermakov, Jan Cech et al.
1 Mela: Test-Time Memory Consolidation based on Transformation Hypothesis Lungchuan Chen
0 Variational Inference for Lévy Process-Driven SDEs via Neural Tilting Yaman Kindap, Manfred Opper et al. cs.LG, cs.AI, cs.CV, cs.RO, stat.ML
0 Neural Weight Norm = Kolmogorov Complexity Tiberiu Musat cs.LG, cs.IT
0 InfoLaw: Information Scaling Laws for Large Language Models with Quality-Weighted Mixture Data and Repetition Fengze Liu, Weidong Zhou et al.
0 A Closed-Form Upper Bound for Admissible Learning-Rate Steps in Belief-Space Dynamics Zixi Li, Youzhen Li
0 CapVector: Learning Transferable Capability Vectors in Parametric Space for Vision-Language-Action Models Wenxuan Song, Han Zhao et al.
0 SEIF: Self-Evolving Reinforcement Learning for Instruction Following Qingyu Ren, Qianyu He et al.
0 Pixal3D: Pixel-Aligned 3D Generation from Images Dong-Yang Li, Wang Zhao et al.