[TOC]

  1. Title: Strength in Numbers: Estimating Confidence of Large Language Models by Prompt Agreement
  2. Author: Gwenyth Portillo Wightman et. al.
  3. Publish Year: TrustNLP 2023
  4. Review Date: Tue, Feb 27, 2024
  5. url: https://aclanthology.org/2023.trustnlp-1.28.pdf

Summary of paper

image-20240227154510805

Motivation

Contribution

Some key terms

calibrated confidence score

summarized method

consider two approaches

image-20240228154938411

  1. measure the log probability of the response across multiple prompts that agree on the same answer.
  2. measure the diversity in answer across different prompts in the model output, concluding that answers which appear in more responses have relatively higher confidence.
    • the diversity measures the confidence, for example, suppose that for a given question queried across ten prompts, the model always replies eggplant. For a second question queried with the same prompts, the model answers potato (5 times) and eggplant, cucumber, squash, carrot and kale. We would say the model is more confident in its answer to the first question.

directly ask GPT about the confidence

ref: Stephanie Lin, Jacob Hilton, and Owain Evans. 2022. Teaching models to express their uncertainty in words.

it suggests that model have some notion of confidence in MCQ tasks

Results

  1. the confidence estimate based on multiple prompts more accurately reflects the chance that a model is correct as compared to log probabilities from a single prompt.

Summary

Our experiments with T0++, FLAN-T5-XXL, and GPT-3 suggest that prompt agreement provides a more calibrated confidence estimate than the typical approach of log probability from a single prompt