On the effectiveness of reward functions in reinforcement learning for confidence calibration of large language models

arXiv:2607.04332 · cs.LG · Submitted 2026-07-05 · Read on arXiv

cs.LG

Submitted: 2026-07-05

Updated: 2026-10-06

Code: https://github.com/mangiucugna/json_

Terminology

• Calibration — 4× in this paper · explained in 7 episodes
• Accuracy — 2× in this paper · explained in 1 episode
• Confidence Calibration — 2× in this paper · explained in 2 episodes
• Reinforcement Learning — 2× in this paper · explained in 4 episodes
• Reward Hacking — 2× in this paper · explained in 8 episodes
• Large Language Models (LLMs) — 1× in this paper · explained in 6 episodes
• Reinforcement Learning (RL) — 1× in this paper · explained in 25 episodes
• Reward Function — 1× in this paper · explained in 3 episodes

Sources

• GPT-4 Technical Report
• Big-Math: A Large-Scale, High-Quality Math Dataset for Reinforcement Learning in Language Models
• Uncalibrated Reasoning: GRPO Induces Overconfidence for Stochastic Outcomes
• Trace Length is a Simple Uncertainty Signal in Reasoning Models
• Bias-Reduced Uncertainty Estimation for Deep Neural Classifiers
• Gemma 4 Technical Report
• The Llama 3 Herd of Models
• On Calibration of Modern Neural Networks
• DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoning
• LoRA: Low-Rank Adaptation of Large Language Models
• Consistency of Large Reasoning Models Under Multi-Turn Attacks
• Understanding R1-Zero-Like Training: A Critical Perspective
• gpt-oss-120b & gpt-oss-20b Model Card
• Defeating the Training-Inference Mismatch via FP16
• Qwen3 Technical Report
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• SCOPE-RL: Stable and Quantitative Control of Policy Entropy in RL Post-Training
• Mitigating LLM Hallucination via Behaviorally Calibrated Reinforcement Learning
• BAS: A Decision-Theoretic Approach to Evaluating Large Language Model Confidence
• It Takes Two: Your GRPO Is Secretly DPO

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks