The Research Desk.

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

Last Brew Time: Sep 7, 2026, 11:15 AM PT

AlphaXiv Trending

Machine Learning
Rethinking On-Policy Distillation of Large Language Models II: One Training Example
AlphaXiv
134

Rethinking On-Policy Distillation of Large Language Models II: One Training Example

Affiliation: University of Chinese Academy of Sciences, Affiliation: Tsinghua University, Affiliation: Northeastern University, Affiliation: University of Illinois Urbana-Champaign, University of Chinese Academy of Sciences

NLP
Unlocking Lossless Speedups in LLMs via Discrete Diffusion
AlphaXiv
64

Unlocking Lossless Speedups in LLMs via Discrete Diffusion

Reinforcement Learning
Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments
AlphaXiv
61

Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments

Affiliation: Qwen Team, Alibaba Group Tsinghua University, Affiliation: Qwen Team, Alibaba Group Tsinghua University Affiliation: Corresponding authors, Qwen Team, Alibaba Group Tsinghua University

Efficiency
Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning
AlphaXiv
57

Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning

Affiliation: Salesforce AI Research, Affiliation: University of Illinois Urbana-Champaign, Affiliation: Salesforce AI Research Affiliation: University of Illinois Urbana-Champaign, Salesforce AI Research, University of Illinois Urbana-Champaign

Continual Learning Mechanisms Compose for Long-Horizon Memorization
AlphaXiv
24

Continual Learning Mechanisms Compose for Long-Horizon Memorization

HuggingFace Daily Papers

Friedrich M
UniMate: One Unified Model to Animate Diverse Skeletons
HuggingFace
9

UniMate: One Unified Model to Animate Diverse Skeletons

Linzhan Mou, Jiahui Lei, Zhiyang Dou, Chenyue Cai, Chaoyue Song

Princeton University

Retrieval
Dr. Claw: An AI Scientist Workspace for Vibe Research
HuggingFace
5

Dr. Claw: An AI Scientist Workspace for Vibe Research

Dingjie Song, Hanrong Zhang, Dawei Liu, Yixin Liu, Zongxia Li

NLP
Knowing What Not to Answer: Selective Non-Compliance in Vision-Language Models
HuggingFace
1

Knowing What Not to Answer: Selective Non-Compliance in Vision-Language Models

Minji Kim, Jihyoung Jang, Hyounghun Kim

Pohang University of Science and Technology

NLP
Refuse without Refusal: A Structural Analysis of Safety-Tuning Responses for Reducing False Refusals in Language Models
HuggingFace
1

Refuse without Refusal: A Structural Analysis of Safety-Tuning Responses for Reducing False Refusals in Language Models

Minji Kim, Hyounghun Kim

Pohang University of Science and Technology

Efficiency
The 2026 PNPL Competition: Word Classification and Efficient Cross-Subject Generalisation in LibriBrain100
HuggingFace
1

The 2026 PNPL Competition: Word Classification and Efficient Cross-Subject Generalisation in LibriBrain100

Francesco Mantegna, Gereon Elvers, Dulhan Jayalath, Gilad Landau, Tasha Kim

Parker Jones Neural Processing Lab (PNPL)

NLP
HarvestBench: Measuring Whether LLM Agents Will Pay to Avoid Killing Animals
HuggingFace
0

HarvestBench: Measuring Whether LLM Agents Will Pay to Avoid Killing Animals

Jasmine Brazilek, Miles Tidmarsh, Matthias Endres, Anshuman Singh, Jeremiah Miller

Compassion aligned machine learning