CANON turns a model's majority-vote consensus into dense per-token supervision by distilling a frozen teacher conditioned on a consensus solution, improving label-free LLM reasoning by up to about 12 points and transferring to held-out benchmarks.
ISBN 979-8-89176-390-6
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Consensus as Privileged Context for Label-Free Self-Distillation
CANON turns a model's majority-vote consensus into dense per-token supervision by distilling a frozen teacher conditioned on a consensus solution, improving label-free LLM reasoning by up to about 12 points and transferring to held-out benchmarks.