A critical look at self-distillation for reasoning
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
A critical reproduction study of self-distillation for LLM reasoning, which trains a model to match its own behavior when given privileged information such as reference solutions.