Flat-Consensus Diffusion for Robust Data Reshaping under Noisy Evaluator
The results show that robust data reshaping requires searching for flat, consensus-supported regions rather than sharp single-trajectory optima, and proposes FCDiff, a flat-consensus diffusion framework that addresses both failures through a micro-macro decomposition.
Hong-Yu Cao, Kun-Peng Liu, Fei Xie et al.
· 1 citation