rsi.watch

Sakana AI on RSI

3 claims, newest first, each quoted from Sakana AI's own pages or papers.

2026-10-12

Sakana AI

MASS: Recursive Self-Improvement through Multi-Agent Self-Supervision

“After the second cycle, score per output token reached 1.2-1.6x the base model's level across four research benchmarks.”

quote checked 2026-10-14paper or codeno replication recordedsource

  • contradicted Gains across three research benchmarks. Four: MLR-Bench, DSBench, ScienceAgentBench, AstaBench E2E-Hard.
  • left out Coding results. Coding scores were lower on Terminal-Bench 2.0 and SWE-bench Verified.
2026-09

Sakana AI · Jürgen Schmidhuber

RSI Lab

“That is why the RSI Lab is being established in Tokyo.”

quote checked 2026-10-14statement onlyno replication recordedsource

  • no primary source The lab will focus on physical AI. The page does not mention physical AI; that framing is from press.

← Full ledger