The Research Desk.

The most upvoted and starred AI research crossing the community today.

Last Brew Time: Aug 31, 2026, 11:09 AM PT

AlphaXiv Trending

Retrieval
Accelerating Scientific Research with Gemini in the Real-World
AlphaXiv
269

Accelerating Scientific Research with Gemini in the Real-World

Affiliation: Google DeepMind, Affiliation: Duke University, Affiliation: Columbia University, Affiliation: Google Research, Affiliation: Texas A&M University

Reinforcement Learning
TTPO: Test-Time Policy Optimization
AlphaXiv
132

TTPO: Test-Time Policy Optimization

Affiliation: Zhejiang University, Zhejiang University, Alibaba Group

NLP
Understanding Evolution Strategies for LLM Reasoning: Broader Reasoning Coverage than GRPO
AlphaXiv
94

Understanding Evolution Strategies for LLM Reasoning: Broader Reasoning Coverage than GRPO

Affiliation: Southern University of Science and Technology, Affiliation: National University of Singapore, Affiliation: Harbin Institute of Technology, Weihai, Affiliation: Huawei Noah’s Ark Lab, Affiliation: City University of Hong Kong

Reinforcement Learning
WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolution
AlphaXiv
86

WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolution

Affiliation: Google Research, Google Research

Puro-2B: Poor Lab's Qwen2-1.5B Trained on RTX 5090 within $5090
AlphaXiv
60

Puro-2B: Poor Lab's Qwen2-1.5B Trained on RTX 5090 within $5090

Affiliation: 1 Tsinghua University, Lab’s Qwen2-1.5B Trained on RTX 5090 within $5090, Tsinghua University, Pengcheng Laboratory, We sincerely thank Yanfu Investments for generously providing GPU computing resources to support this research

HuggingFace Daily Papers

Reasoning
LoopArena: Benchmarking Models as Runtime Controllers for Loop Engineering
HuggingFace
82

LoopArena: Benchmarking Models as Runtime Controllers for Loop Engineering

Yi Wang, Haopeng Zhang, Chengxiang Huang, Rui Dai, Kaikui Liu

AMAP-ML

Computer Vision
Act with Intent: Distilling Behavior Intent for Vision-Language-Action Models
HuggingFace
23

Act with Intent: Distilling Behavior Intent for Vision-Language-Action Models

Sangoh Lee, Sangwoo Mo, Wook-Shin Han

Pohang University of Science and Technology

Sliding-window beats linear attention
HuggingFace
10

Sliding-window beats linear attention

Alexia Jolicoeur-Martineau, Rhea Sanjay Sukthanker, Pashmina Cameron, Emy Gervais

Microsoft

Computer Vision
Generative Semantic Scene Completion
HuggingFace
2

Generative Semantic Scene Completion

Shi Chen, Weifeng Ge

NLP
Lost in Compression: A Controlled Cross-Lingual Audit of Extractive Prompt Compressors
HuggingFace
2

Lost in Compression: A Controlled Cross-Lingual Audit of Extractive Prompt Compressors

Mantas Lukauskas

NLP
GGSS: Geodesic-Gated Spherical Steering for Inference-Time Debiasing of Generative Vision-Language Models
HuggingFace
2

GGSS: Geodesic-Gated Spherical Steering for Inference-Time Debiasing of Generative Vision-Language Models

Yiqun Sun, Junyu Chen, Pengfei Wei, Lawrence B. Hsieh