
Scientific agent swarms that recursively self-improve by evolving shared hypotheses, programs and evidence. SOTA on OpenProblems, NatureBench and MLE-Bench Lite.
Oct 14, 2026

Note: this paper was used solely for internal project initiation and is unrelated to the open-source EvoScientist framework (evoscientist.ai). Preprint on multi-agent evolving AI scientists.
Mar 9, 2026

Harnessing vibe research with self-evolving AI scientists that autonomously explore, generate insights, and iteratively improve.
Mar 6, 2026