Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs
This addresses the need for trustworthy decision-making with LLMs, especially for closed-source models, by providing a baseline for black-box uncertainty estimation, though it is incremental in improving existing methods.
The paper tackles the problem of getting large language models (LLMs) to accurately express their uncertainty in black-box settings, where internal model information is unavailable. It introduces a systematic framework for black-box confidence elicitation and benchmarks methods on calibration and failure prediction tasks, finding that LLMs tend to be overconfident but performance improves with model scaling and certain strategies, with black-box methods showing a narrow gap to white-box ones (e.g., AUROC 0.522 to 0.605).
Empowering large language models to accurately express confidence in their answers is essential for trustworthy decision-making. Previous confidence elicitation methods, which primarily rely on white-box access to internal model information or model fine-tuning, have become less suitable for LLMs, especially closed-source commercial APIs. This leads to a growing need to explore the untapped area of black-box approaches for LLM uncertainty estimation. To better break down the problem, we define a systematic framework with three components: prompting strategies for eliciting verbalized confidence, sampling methods for generating multiple responses, and aggregation techniques for computing consistency. We then benchmark these methods on two key tasks-confidence calibration and failure prediction-across five types of datasets (e.g., commonsense and arithmetic reasoning) and five widely-used LLMs including GPT-4 and LLaMA 2 Chat. Our analysis uncovers several key insights: 1) LLMs, when verbalizing their confidence, tend to be overconfident, potentially imitating human patterns of expressing confidence. 2) As model capability scales up, both calibration and failure prediction performance improve. 3) Employing our proposed strategies, such as human-inspired prompts, consistency among multiple responses, and better aggregation strategies can help mitigate this overconfidence from various perspectives. 4) Comparisons with white-box methods indicate that while white-box methods perform better, the gap is narrow, e.g., 0.522 to 0.605 in AUROC. Despite these advancements, none of these techniques consistently outperform others, and all investigated methods struggle in challenging tasks, such as those requiring professional knowledge, indicating significant scope for improvement. We believe this study can serve as a strong baseline and provide insights for eliciting confidence in black-box LLMs.