Thermometer: Towards Universal Calibration for Large Language Models

TL;DR

This paper introduces THERMOMETER, a novel calibration method for large language models that efficiently improves their response calibration across diverse tasks without sacrificing accuracy.

Contribution

THERMOMETER is a task-agnostic calibration approach that learns an auxiliary model to enhance LLM calibration efficiently and effectively.

Findings

THERMOMETER improves calibration across multiple benchmarks.

The method maintains the original accuracy of LLMs.

It is computationally efficient for large-scale models.

Abstract

We consider the issue of calibration in large language models (LLM). Recent studies have found that common interventions such as instruction tuning often result in poorly calibrated LLMs. Although calibration is well-explored in traditional applications, calibrating LLMs is uniquely challenging. These challenges stem as much from the severe computational requirements of LLMs as from their versatility, which allows them to be applied to diverse tasks. Addressing these challenges, we propose THERMOMETER, a calibration approach tailored to LLMs. THERMOMETER learns an auxiliary model, given data from multiple tasks, for calibrating a LLM. It is computationally efficient, preserves the accuracy of the LLM, and produces better-calibrated responses for new tasks. Extensive empirical evaluations across various benchmarks demonstrate the effectiveness of the proposed method.