mlsanityData to 7 Oct 2026

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

arXiv 2608.04899 ↗First posted 5 Aug 2026cs.CLcited by 0

Not on the map. Why?

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.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 22 of its 35 references that are in the map. First posted on arXiv 5 Aug 2026.

Builds on

22 of 35 references.

References · 22

  1. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  2. 2022Large Language Models are Zero-Shot Reasonerscited by 4,323
  3. 2015Character-level Convolutional Networks for Text Classificationcited by 2,293
  4. 2023Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMscited by 946
  5. 2023Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedbackcited by 788
  6. 2024Prometheus 2: An Open Source Language Model Specialized in Evaluating Other Language Modelscited by 431
  7. 2023A Survey of Confidence Estimation and Calibration in Large Language Modelscited by 253
  8. 2024Kernel Language Entropy: Fine-grained Uncertainty Quantification for LLMs from Semantic Similaritiescited by 156
  9. 2024Benchmarking Uncertainty Quantification Methods for Large Language Models with LM-Polygraphcited by 105
  10. 2024CheckEval: A reliable LLM-as-a-Judge framework for evaluating text generation using checklistscited by 87
  11. 2022Shopping Queries Dataset: A Large-Scale ESCI Benchmark for Improving Product Searchcited by 58
  12. 2023Adaptation with Self-Evaluation to Improve Selective Prediction in LLMscited by 49
10 more
  1. 2024Calibrating the Confidence of Large Language Models by Eliciting Fidelitycited by 47
  2. 2025Beyond Semantic Entropy: Boosting LLM Uncertainty Quantification with Pairwise Semantic Similaritycited by 26
  3. 2025BitsAI-CR: Automated Code Review via LLM in Practicecited by 22
  4. 2025BingoGuard: LLM Content Moderation Tools with Risk Levelscited by 20
  5. 2025Seeing is Believing, but How Much? A Comprehensive Analysis of Verbalized Calibration in Vision-Language Modelscited by 16
  6. 2025MetaFaith: Faithful Natural Language Uncertainty Expression in LLMscited by 15
  7. 2025Geometric Uncertainty for Detecting and Correcting Hallucinations in LLMscited by 12
  8. 2024Unconditional Truthfulness: Learning Unconditional Uncertainty of Large Language Modelscited by 9
  9. 2025Benford's Curse: Tracing Digit Bias to Numerical Hallucination in LLMscited by 6
  10. 2025Scaling Policy Compliance Assessment in Language Models with Policy Reasoning Tracescited by 4

Built on by

Not cited yet.