Strategy-Induct: Task-Level Strategy Induction for Instruction Generation

2026-05-20

AI Summary

This work proposes a novel framework for task-level instruction induction that eliminates the need for expensive annotated answer pairs by leveraging only a small set of unlabeled questions. The approach first prompts a large language model to generate explicit reasoning strategies for each question, then induces generalizable task instructions from the resulting (strategy, question) pairs to guide subsequent reasoning. This is the first method to achieve task-level instruction induction without any answer supervision, while also exploring the synergistic interplay between large language models and large reasoning models in instruction generation and inference. Experimental results demonstrate that the proposed method significantly outperforms state-of-the-art baselines across diverse tasks and model scales, exhibiting superior generalization and reasoning capabilities under the challenging setting where only questions—without answers—are provided.

Abstract

Designing effective task-level prompts is crucial for improving the performance of Large Language Models (LLMs). While prior work on instruction induction demonstrates that LLMs can infer better instructions with limited examples, existing approaches often rely on input-output pairs, where obtaining labeled answers can be difficult or costly. To address this limitation, we propose Strategy-Induct, a framework that derives task-level instructions solely from a small set of example questions without requiring labeled answers. Our approach first prompts the model to generate explicit reasoning strategies for each question, forming (strategy, question) pairs. These pairs are then used to induce a task instruction that guides reasoning. Experiments across multiple tasks and model scales demonstrate that Strategy-Induct outperforms state-of-the-art methods in question-only settings. Furthermore, we observe that jointly utilizing LLMs and Large Reasoning Models across task instruction generation and inference may lead to further performance improvements.

Problem

• instruction induction
• task-level prompting
• label-free learning
• large language models
• reasoning strategies

Innovation

• strategy induction
• instruction generation
• question-only prompting
• large language models
• reasoning strategies