Benchmarking Uncertainty Quantification Methods for Large Language Models with LM-Polygraph

Abstract

The rapid proliferation of large language models (LLMs) has stimulated researchers to seek effective and efficient approaches to deal with LLM hallucinations and low-quality outputs. Uncertainty quantification (UQ) is a key element of machine learning applications in dealing with such challenges. However, research to date on UQ for LLMs has been fragmented in terms of techniques and evaluation methodologies. In this work, we address this issue by introducing a novel benchmark that implements a collection of state-of-the-art UQ baselines and offers an environment for controllable and consistent evaluation of novel UQ techniques over various text generation tasks. Our benchmark also supports the assessment of confidence normalization methods in terms of their ability to provide interpretable scores. Using our benchmark, we conduct a large-scale empirical investigation of UQ and normalization techniques across eleven tasks, identifying the most effective approaches. Code: https://github.com/IINemo/lm-polygraph Benchmark: https://huggingface.co/LM-Polygraph

Published at TACL 2025, presented at ACL 2025. Roman Vashurin, Ekaterina Fadeeva, Artem Vazhentsev contributed equally

References in corpus (12)

• Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model
• Language Models (Mostly) Know What They Know
• Confident Adaptive Language Modeling
• Generating with Confidence: Uncertainty Quantification for Black-box Large Language Models
• Accelerating Large Language Model Decoding with Speculative Sampling
• Jais and Jais-chat: Arabic-Centric Foundation and Instruction-Tuned Open Generative Large Language Models
• Stable LM 2 1.6B Technical Report
• Factcheck-Bench: Fine-Grained Evaluation Benchmark for Automatic Fact-checkers
• Rainproof: An Umbrella To Shield Text Generators From Out-Of-Distribution Data
• From Risk to Uncertainty: Generating Predictive Uncertainty Measures via Bayesian Estimation
• Quantifying Aleatoric and Epistemic Uncertainty with Proper Scoring Rules

Cited by in corpus (2)

• A Survey on Recent Advances in Conversational Data Generation
• Self-Reported Confidence of Large Language Models in Gastroenterology: Analysis of Commercial, Open-Source, and Quantized Models