Teaching models to express their uncertainty in words Teaching models to express their uncertainty in words

Abstract

We show that a GPT‑3 model can learn to express uncertainty about its own answers in natural language—without use of model logits. When given a question, the model generates both an answer and a level of confidence (e.g. "90% confidence" or "high confidence"). These levels map to probabilities that are well calibrated. The model also remains moderately calibrated under distribution shift, and is sensitive to uncertainty in its own answers, rather than imitating human examples. To our knowledge, this is the first time a model has been shown to express calibrated uncertainty about its own answers in natural language. For testing calibration, we introduce the CalibratedMath suite of tasks. We compare the calibration of uncertainty expressed in words ("verbalized probability") to uncertainty extracted from model logits. Both kinds of uncertainty are capable of generalizing calibration under distribution shift. We also provide evidence that GPT‑3's ability to generalize calibration depends on pre-trained latent representations that correlate with epistemic uncertainty over its answers.

We show that a GPT‑3 model can learn to express uncertainty about its own answers in natural language—without use of model logits. When given a question, the model generates both an answer and a level of confidence (e.g. "90% confidence" or "high confidence"). These levels map to probabilities that are well calibrated. The model also remains moderately calibrated under distribution shift, and is sensitive to uncertainty in its own answers, rather than imitating human examples. To our knowledge, this is the first time a model has been shown to express calibrated uncertainty about its own answers in natural language. For testing calibration, we introduce the CalibratedMath suite of tasks. We compare the calibration of uncertainty expressed in words ("verbalized probability") to uncertainty extracted from model logits. Both kinds of uncertainty are capable of generalizing calibration under distribution shift. We also provide evidence that GPT‑3's ability to generalize calibration depends on pre-trained latent representations that correlate with epistemic uncertainty over its answers.

* GPT

* Language

Authors

Stephanie Lin, Jacob Hilton, Owain Evans

Related articles

View all

Building agricultural database for farmers Jan 12, 2024

Creating websites in minutes with AI Website Builder May 29, 2025

Delivering LLM-powered health solutions Jan 4, 2024

Delivering LLM-powered health solutions Jan 4, 2024