MemGPT: Towards LLMs as Operating Systems(arxiv.org)

203 points by belter 3 years ago | 105 comments

Matryoshka Representation Learning(arxiv.org)

71 points by fzliu 2 years ago | 10 comments

Neural-Symbolic Recursive Machine for Systematic Generalization(arxiv.org)

63 points by beefman 3 years ago | 2 comments

Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence (2025)(arxiv.org)

140 points by robin_reala 2 months ago | 77 comments

Artificial Needles to Real Haystacks: Improving Retrieval Capabilities in LLMs(arxiv.org)

92 points by veryluckyxyz 2 years ago | 18 comments

LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics(arxiv.org)

61 points by nothrowaways 11 months ago | 14 comments

Study identifies weaknesses in how AI systems are evaluated(oii.ox.ac.uk)

397 points by pseudolus 11 months ago | 186 comments

Oxford Internet Institute news item on a study finding that most of 445 LLM benchmarks lack rigorous methods and clear definitions of what they measure.

Do not fear AI. Fear AI companies(df7sc6o35ljoz.cloudfront.net)

14 points by rDr4g0n 1 month ago | 2 comments

An essay arguing that people should fear the companies building AI rather than AI itself, appealing to readers' intuition that something is wrong.

AI models can provably hide arbitrary capabilities(arxiv.org)

252 points by sbulaev 13 days ago | 118 comments

An arXiv paper showing AI models can embed encrypted neural network backdoors that provably hide arbitrary capabilities from elicitation-based safety evaluations.

AI tools are spotting errors in research papers(nature.com)

555 points by kgwgk 1 year ago | 200 comments

Nature news article on a growing movement using AI and large language models to detect errors in research papers, sparked by a flawed black plastic utensils study.

AI companies in race to demonstrate their model most threatening to humanity(thecivilian.co.nz)

388 points by ljewalsh 28 days ago | 304 comments

Satirical news article from The Civilian claiming AI companies are competing to show their models are the most existentially threatening to humanity.

AI demands more engineering discipline. Not less(charitydotwtf.substack.com)

321 points by BerislavLopac 4 months ago | 157 comments

Charity Majors' Substack essay arguing that AI-assisted development requires more engineering discipline, comparing it to the shift from pet servers to immutable infrastructure.