Preprint
Aug 2026
Learning When to Trust via Selective Context Preference Optimization
This work recast the problem as selective trust and introduces MIST, a human-annotated benchmark that renders each reasoning item under four matched conditions, together with SC2W, a paired metric counting how often a misleading signal flips a clean-correct answer to wrong.
Xian-Shui Sun, Wei Chow, Ying-Shuo Wang et al.
· 2 citations