Few-shot Calibration
Also called contextual calibration, calibrate before use.
Few-shot calibration corrects a model's systematic biases toward particular answers, such as the most frequent or most recent label in the examples, by adjusting output probabilities measured on a content-free input.
Description
Zhao et al. identified majority-label bias, recency bias, and common-token bias in few-shot prompting, and proposed contextual calibration, which estimates the bias from an input such as "N/A" and rescales predictions to neutralize it.
Sources
- Zhao et al. (2021). Calibrate Before Use: Improving Few-Shot Performance of Language Models.
Cite this entry
Protologue. (2026). Few-shot Calibration. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.0.0, PTL-0021). https://protologue.com/t/few-shot-calibration/
BibTeX
@misc{protologue_few_shot_calibration,
title = {Few-shot Calibration},
author = {{Protologue}},
year = {2026},
howpublished = {Protologue: A Taxonomy of Prompting and LLM Techniques, v1.0.0},
note = {Entry PTL-0021},
url = {https://protologue.com/t/few-shot-calibration/}
}