The Research Desk.

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

Last Brew Time: Jun 8, 2026, 7:38 AM PT

X.com Research Buzz

Memory Caching: RNNs with Growing Memory
X.com
6703

Memory Caching: RNNs with Growing Memory

Ali Behrouz, Zeman Li, Yuan Deng, Peilin Zhong, Meisam Razaviyayn, Vahab Mirrokni

Google Research, Cornell University

AlphaXiv Trending

Cosmos 3: Omnimodal World Models for Physical AI
AlphaXiv
163

Cosmos 3: Omnimodal World Models for Physical AI

Aditi, Niket Agarwal, Arslan Ali

Reinforcement Learning
Agents' Last Exam
AlphaXiv
104

Agents' Last Exam

MAI-Thinking-1: Building a Hill-Climbing Machine
AlphaXiv
83

MAI-Thinking-1: Building a Hill-Climbing Machine

Microsoft, The Microsoft AI Team

Reinforcement Learning
OPRD: On-Policy Representation Distillation
AlphaXiv
78

OPRD: On-Policy Representation Distillation

Shenzhi Yang, Guangcheng Zhu, Bowen Song

Zhejiang University

Reinforcement Learning
OrderGrad: Optimizing Beyond the Mean with Order-Statistic Policy Gradient Estimation
AlphaXiv
53

OrderGrad: Optimizing Beyond the Mean with Order-Statistic Policy Gradient Estimation

Paavo Parmas, Yongmin Kim, Kohsei Matsutani

the University of Tokyo

Reasoning
Latent Reasoning with Normalizing Flows
AlphaXiv
52

Latent Reasoning with Normalizing Flows

Guancheng Tu, Xiangjun Fu, Suhao Yu

†]University of Pennsylvania, §]Meta

HuggingFace Daily Papers

Darlednik
GENEB: Why Genomic Models Are Hard to Compare
HuggingFace
19

GENEB: Why Genomic Models Are Hard to Compare

Daria Ledneva, Mikhail Nuridinov, Denis Kuznetsov

Reinforcement Learning
Towards Retrieving Interaction Spaces for Agentic Search
HuggingFace
1

Towards Retrieving Interaction Spaces for Agentic Search

Shengyao Zhuang, Yuansheng Ni, Hengxin Fun, Jimmy Lin, Xueguang Ma

Machine Learning
LayerRoute: Input-Conditioned Adaptive Layer Skipping via LoRA Fine-Tuning for Agentic Language Models
HuggingFace
1

LayerRoute: Input-Conditioned Adaptive Layer Skipping via LoRA Fine-Tuning for Agentic Language Models

Prateek Kumar Sikdar

Accenture

Efficiency
Compress-Distill: Reasoning Trace Compression for Efficient Knowledge Distillation
HuggingFace
1

Compress-Distill: Reasoning Trace Compression for Efficient Knowledge Distillation

Maxime Griot, Paul Steven Scotti, Tanishq Mathew Abraham

Université Catholique de Louvain

Reinforcement Learning
Towards Human-Like Interactive Speech Recognition With Agentic Correction and Semantic Evaluation
HuggingFace
1

Towards Human-Like Interactive Speech Recognition With Agentic Correction and Semantic Evaluation

Zixuan Jiang, Yanqiao Zhu, Peng Wang, Qinyuan Chen, Xinjian Zhao

SJTU Cross Media Language Intelligence Lab

Wimh966
Augmenting Attention with Exponentially Decaying Memory Improves Query-Aware KV Sparsity
HuggingFace
0

Augmenting Attention with Exponentially Decaying Memory Improves Query-Aware KV Sparsity

Xiuying Wei, Caglar Gulcehre