TY  - RPRT
TI  - Self-Supervised Alignment with Mutual Information: Learning to Follow Principles without Preference Labels
AU  - Jan-Philipp Fränken
AU  - Eric Zelikman
AU  - Rafael Rafailov
AU  - Kanishk Gandhi
AU  - Tobias Gerstenberg
AU  - Noah D. Goodman
PY  - 2024
UR  - https://arxiv.org/abs/2404.14313
ID  - 2404.14313
ER  -