Take One Step at a Time to Know Incremental Utility of Demonstration: An Analysis on Reranking for Few-Shot In-Context Learning
This work addresses the problem of optimizing demonstration selection for few-shot in-context learning in LLMs, offering incremental improvements for researchers and practitioners in NLP.
The paper analyzes different utility functions for selecting demonstrations in few-shot in-context learning, focusing on output probability and task-specific reward, and introduces incremental utility to estimate incremental knowledge from demonstrations. Results show that output probability works well for classification tasks with distributed values, while downstream metrics are more robust for segmentation and translation with nuanced rewards, and incremental utility further improves performance by contrasting LLM outputs with and without demonstrations.
In-Context Learning (ICL) is an emergent capability of Large Language Models (LLMs). Only a few demonstrations enable LLMs to be used as blackbox for new tasks. Previous studies have shown that using LLMs' outputs as labels is effective in training models to select demonstrations. Such a label is expected to estimate utility of a demonstration in ICL; however, it has not been well understood how different labeling strategies affect results on target tasks. This paper presents an analysis on different utility functions by focusing on LLMs' output probability given ground-truth output, and task-specific reward given LLMs' prediction. Unlike the previous work, we introduce a novel labeling method, incremental utility, which estimates how much incremental knowledge is brought into the LLMs by a demonstration. We conduct experiments with instruction-tuned LLMs on binary/multi-class classification, segmentation, and translation across Arabic, English, Finnish, Japanese, and Spanish. Our results show that (1) the probability is effective when the probability values are distributed across the whole value range (on the classification tasks), and (2) the downstream metric is more robust when nuanced reward values are provided with long outputs (on the segmentation and translation tasks). We then show that the proposed incremental utility further helps ICL by contrasting how the LLMs perform with and without the demonstrations.