Process Supervision of Confidence Margin for Calibrated LLM Reasoning
For LLM practitioners, this addresses the overconfidence problem in outcome-based RL, enabling more reliable confidence-based control and efficient compute allocation.
RLCM jointly optimizes LLM reasoning correctness and confidence calibration via a margin-enhanced process reward, improving calibration while maintaining or improving accuracy across math, code, logic, and science benchmarks.
Scaling test-time computation with reinforcement learning (RL) has emerged as a reliable path to improve large language models (LLM) reasoning ability. Yet, outcome-based reward often incentivizes models to be overconfident, leading to hallucinations, unreliable confidence-based control, and unnecessary compute allocation. We introduce Reinforcement Learning with Confidence Margin (\textbf{RLCM}), a calibration-aware RL framework that jointly optimizes correctness and confidence reliability via a margin-enhanced process reward over intermediate-budget completions. Rather than aligning confidence to correctness likelihoods, RLCM encourages to widen the confidence margin between correct and incorrect steps within a single reasoning trajectory. Across mathematical, code, logic and science benchmarks, our method substantially improves calibration while maintaining or improving accuracy. We further show that, with calibrated confidence signals, the resulting models enable more efficient conformal risk control and effective confidence-weighted aggregation.