Beneath the Diff: Diagnosing and Mitigating Algorithmic Mode Collapse in Code-Level Autonomous Research Loops
This work identifies a robust failure mode that is accompanied by a widening gap between in-loop metric gains and gains measured on independent held-out evaluations and proposes Diversity-Aware Proposal Sampling (DAPS), a lightweight mitigation that combines category-coverage reweighting, persistent edit memory, and a...
Bowei He, Weixu Zhang, Yi Jin et al.
· 0 citations