A Rubric Reward Model evaluates LLM reasoning steps against problem rubrics to reduce Miracle Steps and improve verified math performance over outcome-only rewards.
side-length method,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Curing Miracle Steps in LLM Mathematical Reasoning with Rubric Rewards
A Rubric Reward Model evaluates LLM reasoning steps against problem rubrics to reduce Miracle Steps and improve verified math performance over outcome-only rewards.