CLAILGMay 28, 2022

Teaching Models to Express Their Uncertainty in Words

Berkeley
arXiv:2205.14334v2719 citationsh-index: 24
Originality Highly original
AI Analysis

This addresses the need for AI models to communicate uncertainty transparently to users, representing a novel method for a known bottleneck in interpretability.

The paper tackles the problem of enabling GPT-3 to express its own uncertainty in natural language without using model logits, achieving well-calibrated probabilities (e.g., 90% confidence) and moderate calibration under distribution shift.

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.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes