Autoresearch with coding agents can now run for days on a single problem, but researchers've seen an increasing number of cases where they stop improving early.

Researchers address this issue with a simple idea: forgetting.

Meet new work: "Scaling Discovery Requires Forgetting"