Large Language Diffusion Models(ml-gsai.github.io)

29 points by SerCe 1 year ago | 7 comments

Can large language models reason?(arnaldur.be)

4 points by Hugsun 2 years ago | 0 comments

Toddlerbot: Open-Source Humanoid Robot(toddlerbot.github.io)

99 points by base698 1 year ago | 23 comments

High-resolution postmortem human brain MRI at 7 tesla(pulkit-khandelwal.github.io)

109 points by pulks 2 years ago | 16 comments

Generating Music Tracks with Unified Representation and Diffusion Framework(ai-muzic.github.io)

82 points by laurieg 3 years ago | 33 comments

Lumos: Learning Agents with Unified Data, Modular Design, and Open-Source LLMs(allenai.github.io)

82 points by milliondreams 2 years ago | 7 comments

Stealing Reasoning Traces from Proprietary LLM APIs(stolen-thoughts.com)

665 points by quantumgarbage 2 months ago | 295 comments

Project page for a research paper on extracting hidden reasoning traces from proprietary LLM APIs, by authors from MATS, ELLIS, Max Planck and Tübingen.

LLMs use a surprisingly simple mechanism to retrieve some stored knowledge(news.mit.edu)

336 points by CharlesW 2 years ago | 121 comments

MIT News article on research showing LLMs often retrieve stored facts via simple linear functions, and a technique to probe what a model knows about subjects.

CatalaLang/catala: Programming language for law specification(github.com)

420 points by gorenb 3 years ago | 251 comments

GitHub repository for Catala, a programming language for literate programming of law specifications, such as tax and benefits legislation.

Vision language models are blind(vlmsareblind.github.io)

425 points by taesiri 2 years ago | 172 comments

Project page for the research paper "Vision language models are blind" from Auburn University and University of Alberta, with links to the arXiv paper, code, and dataset.

Publish fake stats, get cited in LLMs(peec.ai)

315 points by robtherobber 6 days ago | 204 comments

Peec AI blog post by a GEO researcher describing how publishing statistics led to the author's data being widely cited across the web and by LLMs.

LLMs are making formal software verification feasible at a much larger scale(anthropic.com)

429 points by johnbender 3 days ago | 265 comments

Anthropic research post describing how Claude produced formal proofs of crash-freedom for RISC-V binaries of 123 OSS-Fuzz open-source libraries, finding bugs along the way.