Preference Instability in Reward Models: Detection and Mitigation via Sparse Autoencoders
Paper • 2605.16339 • Published • 1
Shunchang Liu
Shunchang
AI & ML interests
AI
Recent Activity
authored a paper 4 days ago
Narrow Multimodal Fine-Tuning Can Induce Emergent Misalignment upvoted a paper 4 days ago
Narrow Multimodal Fine-Tuning Can Induce Emergent MisalignmentOrganizations
None yet
upvoted a paper 2 days ago