ExpEvolve: Structuring and Reusing Experience for Open-Ended Discovery
Abstract
Evolutionary coding agents such as AlphaEvolve are one of the notable successes in using Large Language Models (LLMs) to build AI Scientists. These agents tackle open-ended scientific problems by iteratively improving and evolving programs, leveraging the prior knowledge and reasoning capabilities of LLMs. Despite their success, mutations over programs are mostly low-level upon surface heuristics, without a persistent unit of control. In contrast, science compounds through ideas. In this work, we propose ExpEvolve, which injects idea-level extraction and control into evolution: it discovers and leverages the high-level factors from programs that explain scoring differences among different programs. We characterize theoretical conditions on when discovery with idea factors improves sample efficiency. We also build a practical framework to show that the extracted high-level idea factors are helpful for guiding exploitation from experience and exploration when discovery comes to a plateau. Across 11 open-ended mathematical tasks, ExpEvolve significantly improves the final performance and sample efficiency up to 5.0x. The improvements also transfer to 4 other LLMs and to 6 machine learning tasks from MLS-Bench. Analysis shows that ExpEvolve can discover idea-level factors to provide better exploration and exploitation for evolutionary discovery.
Related
- Open-ended Scientific Discovery with Possibilistic Reasoning
- MUON-RT Reusing the Teacher's Trajectory for Momentum-Free Muon for LLM Self-Distillation
- Reusing Statistical Guarantees Under Change: Selection Costs and Event Ownership
- Optimizing Open-Ended Interview Assessment Using Response-Based Semantic Indices
- Nonamenable subforests of multi‐ended quasi‐pmp graphs
BibTeX
@misc{chen2026expevolve,
title = {{ExpEvolve: Structuring and Reusing Experience for Open-Ended Discovery}},
author = {Yongqiang Chen and Chenxi Liu and Deyu Zou and Zhenhao Chen and Tongliang Liu and Bo Han and Kun Zhang},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.29303},
eprint = {2610.29303},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.29303},
note = {CC BY-SA 4.0},
}
RIS
TY - UNPB
TI - ExpEvolve: Structuring and Reusing Experience for Open-Ended Discovery
AU - Yongqiang Chen
AU - Chenxi Liu
AU - Deyu Zou
AU - Zhenhao Chen
AU - Tongliang Liu
AU - Bo Han
AU - Kun Zhang
PY - 2026
DA - 2026/10/21
PB - arXiv
DO - 10.48550/arxiv.2610.29303
UR - https://arxiv.org/abs/2610.29303
AB - Evolutionary coding agents such as AlphaEvolve are one of the notable successes in using Large Language Models (LLMs) to build AI Scientists. These agents tackle open-ended scientific problems by iteratively improving and evolving programs, leveraging the prior knowledge and reasoning capabilities of LLMs. Despite their success, mutations over programs are mostly low-level upon surface heuristics, without a persistent unit of control. In contrast, science compounds through ideas. In this work, we propose ExpEvolve, which injects idea-level extraction and control into evolution: it discovers and leverages the high-level factors from programs that explain scoring differences among different programs. We characterize theoretical conditions on when discovery with idea factors improves sample efficiency. We also build a practical framework to show that the extracted high-level idea factors are helpful for guiding exploitation from experience and exploration when discovery comes to a plateau. Across 11 open-ended mathematical tasks, ExpEvolve significantly improves the final performance and sample efficiency up to 5.0x. The improvements also transfer to 4 other LLMs and to 6 machine learning tasks from MLS-Bench. Analysis shows that ExpEvolve can discover idea-level factors to provide better exploration and exploitation for evolutionary discovery.
ER -