Prompt-MII: Meta-Learning Instruction Induction for LLMs

2025-10-19

AI Summary

To address the high inference overhead of large language models (LLMs) in long-context tasks, this paper proposes PROMPT-MII, a meta-learning and reinforcement learning–based instruction induction framework that automatically compresses in-context learning (ICL) examples into compact, generalizable instructions—replacing redundant demonstrations. It introduces the first meta-training protocol across thousands of HuggingFace classification datasets, enabling dynamic, task-adaptive instruction generation in zero-shot settings. Evaluated on 90 unseen tasks, PROMPT-MII achieves 4–9 percentage-point (10%–20% relative) F1 improvements over standard ICL while matching or exceeding its accuracy, reducing token consumption by 3–13×. Key contributions include: (1) the first instruction compression framework supporting large-scale, cross-task meta-learning; (2) efficient zero-shot instruction generation without model fine-tuning; and (3) substantial inference cost reduction without performance degradation.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

A popular method to adapt large language models (LLMs) to new tasks is in-context learning (ICL), which is effective but incurs high inference costs as context length grows. In this paper we propose a method to perform instruction induction, where we take training examples and reduce them to a compact but descriptive prompt that can achieve performance comparable to ICL over the full training set. Specifically, we propose PROMPT-MII, a reinforcement learning (RL) based framework to meta-learn an instruction induction model that can generate compact instructions on the fly for an arbitrary new dataset. We train on over 3,000 diverse classification datasets from the HuggingFace hub, and evaluate on 90 unseen tasks. PROMPT-MII improves downstream model quality by 4-9 F1 points (10-20% relative), matching ICL performance while requiring 3-13x fewer tokens.

Problem

• Reducing high inference costs of in-context learning
• Generating compact prompts matching full training performance
• Meta-learning instruction induction for diverse classification tasks

Innovation

• Meta-learns instruction induction via reinforcement learning
• Generates compact prompts from training examples
• Reduces token usage while matching ICL performance