🤖 AI Summary
Existing chain-of-thought (CoT) prompting approaches rely on either human-annotated examples or large language model (LLM)-generated reasoning steps for example selection, resulting in high annotation costs and poor scalability.
Method: We propose the first unsupervised, implicit reasoning skill modeling framework. It learns a mapping from questions to intermediate reasoning steps—termed “reasoning skills”—via latent variable modeling, jointly optimizing skill representation learning and skill-aligned retrieval. Crucially, it requires no human annotations or auxiliary LLM calls.
Contribution/Results: Theoretically grounded and computationally efficient, our method enables end-to-end, skill-driven in-context learning (ICL) example selection. Experiments across multiple reasoning benchmarks demonstrate superior performance over state-of-the-art skill selection baselines, 4× faster example library processing, 50% reduction in LLM inference cost during selection, and enhanced robustness to suboptimal example collections.
📝 Abstract
Chain-of-thought (CoT) prompting is a popular in-context learning (ICL) approach for large language models (LLMs), especially when tackling complex reasoning tasks. Traditional ICL approaches construct prompts using examples that contain questions similar to the input question. However, CoT prompting, which includes crucial intermediate reasoning steps (rationales) within its examples, necessitates selecting examples based on these rationales rather than the questions themselves. Existing methods require human experts or pre-trained LLMs to describe the skill, a high-level abstraction of rationales, to guide the selection. These methods, however, are often costly and difficult to scale. Instead, this paper introduces a new approach named Latent Reasoning Skills (LaRS) that employs unsupervised learning to create a latent space representation of rationales, with a latent variable called a reasoning skill. Concurrently, LaRS learns a reasoning policy to determine the required reasoning skill for a given question. Then the ICL examples are selected by aligning the reasoning skills between past examples and the question. This approach is theoretically grounded and compute-efficient, eliminating the need for auxiliary LLM inference or manual prompt design. Empirical results demonstrate that LaRS consistently outperforms SOTA skill-based selection methods, processing example banks four times faster, reducing LLM inferences during the selection stage by half, and showing greater robustness to sub-optimal example banks.
