Persistent 'Rock Tokens' in on-policy distillation resist teacher corrections, consume large gradient norms, yet add negligible value to reasoning, allowing targeted bypassing to streamline alignment.
These open new reasoning steps and signal turns in the chain of thought
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Cornerstones or Stumbling Blocks? Deciphering the Rock Tokens in On-Policy Distillation
Persistent 'Rock Tokens' in on-policy distillation resist teacher corrections, consume large gradient norms, yet add negligible value to reasoning, allowing targeted bypassing to streamline alignment.