Skip to main content
ScienceReportedMedium
4.9

AI Models Solve 35% of Theoretical Problems, Up from 2%

MIT researchers report a dramatic improvement in AI's ability to solve theoretical problems, rising from ~2% to ~35% across disciplines. This signals accelerating AI capability in scientific reasoning, with implications for research productivity and potential disruption of knowledge work.

La Nación (Argentina)1 day agoUSCredibility 31%View source

Score Breakdown

Mosaic Score4.9
Model confidence0.5
Significance0.5
Source credibility0.3

Intelligence Tags

Locations

MIT

Entities

country
Source

Related signals

8 found