Self-Adapting Language Models
Comments
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Summary
MIT researchers Adam Zweiger, Jyothish Pari, and colleagues introduce Self-Adapting Language Models (SEAL), a framework where an LLM generates its own finetuning data and update instructions, then learns via reinforcement learning, using downstream performance as reward, to produce better self-edits.