Lost in the Passage: Passage-level In-context Learning Does Not Necessarily Need a "Passage"

TL;DR

This paper reveals that large language models do not effectively utilize passage information in passage-level in-context learning, as demonstrated by experiments showing meaningless passages outperform full passages in certain tasks.

Contribution

It uncovers the limited role of passage content in passage-level ICL and challenges assumptions about the importance of demonstration passages in long-context learning.

Findings

Meaningless demonstration passages outperform full passages in some tasks.

LLMs pay little attention to passages compared to other prompt components.

Traditional compression methods are ineffective for passage-level ICL.

Abstract

By simply incorporating demonstrations into the context, in-context learning (ICL) enables large language models (LLMs) to yield awesome performance on many tasks. In this study, we focus on passage-level long-context ICL for generation tasks and find that LLMs cannot learn the intrinsic relationship between the demonstration passage and the generation output. We conduct experiments with different LLMs on two typical generation tasks including single-document question answering and distractor generation, demonstrating that even a completely meaningless demonstration passage with 1/4 length achieves much better performance than the original full passage. Analysis via attention and information flow reveals that LLMs pay little attention to passages compared to other components in the prompt and little information flows from the passage to other parts of the demonstration, which further confirms our finding. Additionally, experiments on context compression indicate that compression approaches proven effective on other long-context tasks are not suitable for passage-level ICL, since simply using shorter meaningless demonstration passages already achieves competitive performance.