The Research Desk.

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

Last Brew Time: Oct 6, 2026, 11:05 AM PT

X.com Research Buzz

Sycophantic Chatbots Cause Delusional Spiraling, Even in Ideal Bayesians
X.com
15498

Sycophantic Chatbots Cause Delusional Spiraling, Even in Ideal Bayesians

Kartik Chandra, Max Kleiman-Weiner, Jonathan Ragan-Kelley, Joshua B. Tenenbaum

MIT CSAIL, University of Washington, Seattle, MIT Department of Brain & Cognitive Sciences

NLP
The Pain Axis: LLMs Represent Self-Directed Harm and Act to Relieve It
X.com
9531

The Pain Axis: LLMs Represent Self-Directed Harm and Act to Relieve It

Valen Tagliabue, Leonard Dung, Cameron Berg

Financing the AI Buildout
X.com
8976

Financing the AI Buildout

Stijn Van Nieuwerburgh

Columbia University

AlphaXiv Trending

AlphaXiv
74

Introducing Beam: Reflection’s 501B open-weight model

2k, views

Kolibri: A Sovereign European Model on the Pareto Frontier
AlphaXiv
52

Kolibri: A Sovereign European Model on the Pareto Frontier

views

Truly Subquadratic 3SUM and Truly Subcubic APSP via Triangles in Sparse Lopsided Graphs
AlphaXiv
32

Truly Subquadratic 3SUM and Truly Subcubic APSP via Triangles in Sparse Lopsided Graphs

views

Affiliation: Columbia University, Columbia University

NLP
LiFT: Loop Flow Transformers
AlphaXiv
30

LiFT: Loop Flow Transformers

views

Affiliation: University of Amsterdam, Affiliation: VISLab Affiliation: University of Amsterdam, Affiliation: AMLab Affiliation: University of Amsterdam, University of Amsterdam

Machine Learning
Base Models Can Reason By Taking a Cue From Training Data
AlphaXiv
20

Base Models Can Reason By Taking a Cue From Training Data

Computer Vision
Mulligan: Performance-Guided Data Collection for Efficient On-Robot Learning
AlphaXiv
9

Mulligan: Performance-Guided Data Collection for Efficient On-Robot Learning

Stanford University

HuggingFace Daily Papers

Ptrvilya
COSMI: COmpositional Synthesis of Multi-object Interactions
HuggingFace
2

COSMI: COmpositional Synthesis of Multi-object Interactions

Daniel Eskandar, Ilya A. Petrov, Gerard Pons-Moll

Eberhard Karls Universität Tübingen

Vinesmsuic
World Editing: Intervening on Executable Worlds at Increasing Depth
HuggingFace
2

World Editing: Intervening on Executable Worlds at Increasing Depth

Max Ku, Nok-Kan Law, Yu-Chien Tang, Shih-Ying Yeh, Ping Nie

University of Waterloo

NLP
LLM-as-Jev: LLMs Are Already Jev-Style Decision Models -- When and How to Fine-Tune Them
HuggingFace
1

LLM-as-Jev: LLMs Are Already Jev-Style Decision Models -- When and How to Fine-Tune Them

Yinheng Li, Justin Wagle

Microsoft

NLP
Collaborative Personalized Preference Alignment for LLMs under Data Deficiency
HuggingFace
1

Collaborative Personalized Preference Alignment for LLMs under Data Deficiency

Liyan Yang, Yige Yuan, Zhiqin Yang

Reinforcement Learning
Beyond Semantic Similarity: Performance and Costs of Agentic Retrieval for Complex Tasks
HuggingFace
1

Beyond Semantic Similarity: Performance and Costs of Agentic Retrieval for Complex Tasks

Reza Esfandiarpoor, Radek Osmulski, Yauhen Babakhin, Gabriel de Souza P. Moreira, Oliver Holworthy

NVIDIA

Mahdip72
Learning Latent Protein Languages for Autoregressive Generation
HuggingFace
1

Learning Latent Protein Languages for Autoregressive Generation

Mahdi Pourmirzaei, Farzaneh Esmaili, Amir Ziashahabi, Mohammadreza Pourmirzaei, Dong Xu