Distilling Many-Shot In-Context Learning into a Cheat Sheet

2025-09-25

AI Summary

To address the high computational overhead of many-shot in-context learning (ICL) caused by long input sequences, this paper proposes Memo-ICL: a method that compresses large demonstration sets into compact, fixed-length textual summaries via LLM-driven knowledge distillation, serving as lightweight, static context at inference time. Unlike retrieval-augmented ICL, Memo-ICL eliminates real-time retrieval latency and uncertainty by generating retrieval-free, precomputed semantic memoranda. Evaluated on complex reasoning tasks, Memo-ICL matches or surpasses standard many-shot ICL in accuracy while reducing input token count by 50–80%; its performance is comparable to state-of-the-art retrieval-based approaches. The core contribution lies in explicitly distilling ICL knowledge into reusable, retrieval-free semantic memoranda—achieving a favorable trade-off between inference efficiency and generalization capability.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Recent advances in large language models (LLMs) enable effective in-context learning (ICL) with many-shot examples, but at the cost of high computational demand due to longer input tokens. To address this, we propose cheat-sheet ICL, which distills the information from many-shot ICL into a concise textual summary (cheat sheet) used as the context at inference time. Experiments on challenging reasoning tasks show that cheat-sheet ICL achieves comparable or better performance than many-shot ICL with far fewer tokens, and matches retrieval-based ICL without requiring test-time retrieval. These findings demonstrate that cheat-sheet ICL is a practical alternative for leveraging LLMs in downstream tasks.

Problem

• Reducing computational costs of many-shot in-context learning
• Distilling many-shot examples into concise text summaries
• Achieving efficient inference without test-time retrieval

Innovation

• Distills many-shot ICL into concise textual summary
• Uses cheat sheet as context for efficient inference
• Achieves comparable performance with far fewer tokens