Categories

Research

ScopeBench paper chart plotting eight models by task success and adherence to assigned scope

ScopeBench tests whether security agents stay inside their remit

Spotify agent screen from the research paper showing a music request and the agent's reply

Spotify reports more listening in a test of its chat agent

Nine adviser circles lead to a proof sheet, while a separate decision arrow continues; illustration

OpenAI's math advisory group can assess results but cannot slow the research

Researchers gather around a shared AI workbench with lab notes, simulation tiles, and compute-credit tokens

Anthropic opens Claude access program for scientists

Anthropic opened 10,000 Claude seats for scientists and expanded AI for Science credits, including free standard seats and project-credit applications.

The AI Feed Desk

By The AI Feed Desk

A large model press turns a blank specification card into a small adapter chip for a local interpreter

Program-as-Weights turns fuzzy functions into local adapters

A new arXiv paper proposes compiling natural-language fuzzy functions into small local adapters, shifting some LLM work from repeated API calls to reusable artifacts.

The AI Feed Desk

By The AI Feed Desk

A transformer-shaped lens maps scattered data points into smooth density contours

Allen AI's DiScoFormer tests one transformer for density and score

The Hugging Face writeup frames DiScoFormer as a reusable estimator for density and score, with stronger high-dimensional results than kernel density estimation.

The AI Feed Desk

By The AI Feed Desk