# Evaluation Pitfalls and Sparsity Limitations in LLM-based Confidence Estimates for Classification

arXiv: https://arxiv.org/abs/2608.04899 · HTML: https://mlsanity.com/paper/2608.04899/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 22 of its 35 references that are in the map.
- First posted on arXiv 5 Aug 2026.

## Builds on (most cited first, 22 of 22)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Large Language Models are Zero-Shot Reasoners](https://mlsanity.com/paper/2205.11916/) (2022)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs](https://mlsanity.com/paper/2306.13063/) (2023)
- [Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedback](https://mlsanity.com/paper/2305.14975/) (2023)
- [Prometheus 2: An Open Source Language Model Specialized in Evaluating Other Language Models](https://mlsanity.com/paper/2405.01535/) (2024)
- [A Survey of Confidence Estimation and Calibration in Large Language Models](https://mlsanity.com/paper/2311.08298/) (2023)
- [Kernel Language Entropy: Fine-grained Uncertainty Quantification for LLMs from Semantic Similarities](https://mlsanity.com/paper/2405.20003/) (2024)
- [Benchmarking Uncertainty Quantification Methods for Large Language Models with LM-Polygraph](https://mlsanity.com/paper/2406.15627/) (2024)
- [CheckEval: A reliable LLM-as-a-Judge framework for evaluating text generation using checklists](https://mlsanity.com/paper/2403.18771/) (2024)
- [Shopping Queries Dataset: A Large-Scale ESCI Benchmark for Improving Product Search](https://mlsanity.com/paper/2206.06588/) (2022)
- [Adaptation with Self-Evaluation to Improve Selective Prediction in LLMs](https://mlsanity.com/paper/2310.11689/) (2023)
- [Calibrating the Confidence of Large Language Models by Eliciting Fidelity](https://mlsanity.com/paper/2404.02655/) (2024)
- [Beyond Semantic Entropy: Boosting LLM Uncertainty Quantification with Pairwise Semantic Similarity](https://mlsanity.com/paper/2506.00245/) (2025)
- [BitsAI-CR: Automated Code Review via LLM in Practice](https://mlsanity.com/paper/2501.15134/) (2025)
- [BingoGuard: LLM Content Moderation Tools with Risk Levels](https://mlsanity.com/paper/2503.06550/) (2025)
- [Seeing is Believing, but How Much? A Comprehensive Analysis of Verbalized Calibration in Vision-Language Models](https://mlsanity.com/paper/2505.20236/) (2025)
- [MetaFaith: Faithful Natural Language Uncertainty Expression in LLMs](https://mlsanity.com/paper/2505.24858/) (2025)
- [Geometric Uncertainty for Detecting and Correcting Hallucinations in LLMs](https://mlsanity.com/paper/2509.13813/) (2025)
- [Unconditional Truthfulness: Learning Unconditional Uncertainty of Large Language Models](https://mlsanity.com/paper/2408.10692/) (2024)
- [Benford's Curse: Tracing Digit Bias to Numerical Hallucination in LLMs](https://mlsanity.com/paper/2506.01734/) (2025)
- [Scaling Policy Compliance Assessment in Language Models with Policy Reasoning Traces](https://mlsanity.com/paper/2509.23291/) (2025)

## Abstract (arXiv)

Confidence estimation is essential when LLMs are used for classification, indicating when predictions can be trusted. However, common approaches such as verbalization produce extremely sparse outputs. For instance, Qwen3-32B verbalizes only eight unique confidence values on SST-2, with over half being exactly 95%, a pattern we observe consistently across four datasets and two LLMs. Besides limiting practical utility, we show that this sparsity critically affects evaluation: the choice of interpolation in area under the accuracy-rejection curve (AUARC) dramatically alters rankings, with consistency sampling dropping from best to worst under stepwise versus linear interpolation. We advocate for standardizing stepwise interpolation for a fairer comparison. Under such a fair evaluation, we find that weighting verbalized digits by token probabilities, a method we term verbalization logprobs, addresses sparsity and achieves the best AUARC (+2.3 points over vanilla verbalization) without incurring additional inference cost.