MetaRSI-v1 Unifies Three Recurrent Self-Improvement Paradigms

CosmosMind, with Stanford, Berkeley, MIT, Tsinghua and other universities, released MetaRSI-v1, the first framework unifying Model-RSI, Data-RSI and Harness-RSI. It delivers average gains of 10.9 points for a 3B model and 7.3 points for flagship models across four benchmarks.

2026-09-13 ~ 2026-09-14 · 2 related posts

• MetaRSI Report Makes Self-Improvement Recursive on Itself, Boosting Flagship Models 7.3 Points — 青稞AI · 2026-09-13
• MetaRSI-v1 unifies model/data/harness recursive self-improvement, boosting a 3B-active model by 10.9 points — 量子位 · 2026-09-14