Internal feedback rewards (entropy and self-certainty) improve base LLM math reasoning only in early training and degrade later, with little benefit for instruct models.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
No Free Lunch: Rethinking Internal Feedback for LLM Reasoning
Internal feedback rewards (entropy and self-certainty) improve base LLM math reasoning only in early training and degrade later, with little benefit for instruct models.