Second Thoughts are Best: Learning to Re-Align With Human Values from Text Edits

Date:

This paper present SECOND THOUGHTS, a new learning paradigm that enables language models (LMs) to re-align with human values.

By modeling the chain-of-edits between value-unaligned and value-aligned text, with LM fine-tuning and additional refinement through reinforcement learning.

The generated editing steps also offer better interpretability and ease for interactive error correction.

Powerpoint for this talk

Powerpoint for this talk

Leave a Comment