ESSENTIAL READING · PAPER 89
Learning What to Keep: Transfer-Aware Personal Skill Learning from User Feedback
Wang, L., Zhang, C., Yang, Y., Chen, X., et al. (2025). Learning What to Keep: Transfer-Aware Personal Skill Learning from User Feedback. arXiv preprint arXiv:2610.19380.
FOUND
A SkillGrain framework splitting feedback into atomic rules, keeping only transferable preferences to avoid negative transfer.
MEANS
Teaching AI must separate habit from accident when remembering preferences â or the memory drifts.
SI DATA
No direct SI dataset counterpart yet — flagged for future audit expansion.