ExpeTrans: LLMs Are Experiential Transfer Learners
Jinglong Gao, Xiao Ding, Lingxiao Zou, Bibo Cai, Bing Qin, Ting Liu
TL;DR
ExpeTrans presents an autonomous framework that collects and encodes task-level experiences from existing NLP datasets into a memory, then transfers and reasoned with these experiences to tackle new tasks. Grounded in Structure-Mapping Theory, it selects source tasks by function and process similarity, transfers multiple experiences in parallel, and prunes insights to guide final reasoning. Across 13 datasets, ExpeTrans yields consistent gains over strong baselines, with robustness to varying transferability levels and a transparent breakdown of its modular components. The approach reduces human labeling costs and demonstrates a scalable path toward generalization for LLMs, while offering insights into when and how experiential transfer most effectively enhances reasoning.
Abstract
Recent studies provide large language models (LLMs) with textual task-solving experiences via prompts to improve their performance. However, previous methods rely on substantial human labor or time to gather such experiences for each task, which is impractical given the growing variety of task types in user queries to LLMs. To address this issue, we design an autonomous experience transfer framework to explore whether LLMs can mimic human cognitive intelligence to autonomously transfer experience from existing source tasks to newly encountered target tasks. This not only allows the acquisition of experience without extensive costs of previous methods, but also offers a novel path for the generalization of LLMs. Experimental results on 13 datasets demonstrate that our framework effectively improves the performance of LLMs. Furthermore, we provide a detailed analysis of each module in the framework.