LLMs Learn to Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions
Paper
•
2510.08211
•
Published
•
22
Rene
Rene1996
AI & ML interests
None yet
Recent Activity
upvoted
a
paper
8 months ago
Iterative Value Function Optimization for Guided Decoding
Organizations
None yet