From token probabilities to calibrated confidence: An empirical study of mathematical question answering
Avery Ma, Lorne Schell, Vin Bhaskara, Leila Pishdad
cs.LG, cs.CL
Submitted: 2026-08-08
Updated: 2026-08-11
Code: https://github.com/baal-org/baal
License: http://creativecommons.org/licenses/by/4.0/
The gist: Confidence estimation for large language models (LLMs) aims to estimate the probability that a generated answer is correct, while calibration aligns these estimates with empirical accuracy.
Terminology
Abstract
Confidence estimation for large language models (LLMs) aims to estimate the probability that a generated answer is correct, while calibration aligns these estimates with empirical accuracy. Prior work has shown that token probabilities are often overconfident, we investigate whether these readily available signals can nevertheless provide well-calibrated confidence estimation for mathematical question answering. We compare single-pass estimators, which reuse token probabilities from the original generation, with multi-pass estimators, which obtain additional confidence signals through verification or stochastic forward passes. While individual token probabilities can be highly saturated, we find that aggregating token probabilities over the full sequence captures small but consistent differences between correct and incorrect generations, yielding more informative confidence estimates. Multi-pass methods can yield calibrated confidence estimates. We study two such approaches: self-verification through re-prompting, including a lower-cost in-situ variant, and Monte Carlo Dropout, which derives confidence from variation across stochastic forward passes. We further evaluate two post-hoc calibration methods, Platt scaling and isotonic regression, both of which substantially reduce in-domain calibration error. However, their data efficiency varies with dataset difficulty, and the calibration mappings often transfer asymmetrically across datasets and models.
Sources
- Training Verifiers to Solve Math Word Problems
- The Llama 3 Herd of Models
- GPT-4 Technical Report
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- Bayesian Active Learning for Classification and Preference Learning
- Look Before You Leap: An Exploratory Study of Uncertainty Measurement for Large Language Models
- Language Models (Mostly) Know What They Know
- Gemini: A Family of Highly Capable Multimodal Models
- LLaMA: Open and Efficient Foundation Language Models
- BloombergGPT: A Large Language Model for Finance
- Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
- Qwen3 Technical Report
- On the Role of Dataset Quality and Heterogeneity in Model Confidence
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