EvoCF: Multi-Agent Collaboration via Agentic Memory-Driven Evolutionary Counterfactual Planning

Haotian Chi, Zeyu Feng, Xingrui Yu, Linbo Luo, Yew-Soon Ong, Ivor Tsang, Hechang Chen, Yi Chang, Haiyan Yin
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:19424-19437, 2026.

Abstract

Planning collaboration strategies for multi-agent embodied systems remains a core challenge for LLM-based planners, which often fail to capture the physical and coordination constraints of realworld environments. To address this, we present EvoCF, an agentic memory-driven evolutionary counterfactual planning framework for discovering improved multi-agent collaboration strategies through counterfactual plan generation and evaluation. First, we propose a symbolic constraint inductor that induces reusable symbolic constraints from failures, forming an evolving rule library. Then, we propose an evolutionary counterfactual plan generator that systematically explores semantically consistent plan variants through rule-conditioned mutations, enabling robust collaboration strategies beyond short-sighted one-shot LLM plans. Finally, we design an agentic memory-grounded evaluator that ranks candidate plans using retrieval-augmented evidence, producing interpretable, constraint-aware selections. Across multi-agent embodied simulation benchmarks, EvoCF consistently discovers more robust and executable plans compared to baseline approaches. Our results demonstrate that grounding multi-agent planning in agentic memory and counterfactual reasoning significantly enhances both effectiveness and robustness.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-chi26b, title = {{E}vo{CF}: Multi-Agent Collaboration via Agentic Memory-Driven Evolutionary Counterfactual Planning}, author = {Chi, Haotian and Feng, Zeyu and Yu, Xingrui and Luo, Linbo and Ong, Yew-Soon and Tsang, Ivor and Chen, Hechang and Chang, Yi and Yin, Haiyan}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {19424--19437}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/chi26b/chi26b.pdf}, url = {https://proceedings.mlr.press/v306/chi26b.html}, abstract = {Planning collaboration strategies for multi-agent embodied systems remains a core challenge for LLM-based planners, which often fail to capture the physical and coordination constraints of realworld environments. To address this, we present EvoCF, an agentic memory-driven evolutionary counterfactual planning framework for discovering improved multi-agent collaboration strategies through counterfactual plan generation and evaluation. First, we propose a symbolic constraint inductor that induces reusable symbolic constraints from failures, forming an evolving rule library. Then, we propose an evolutionary counterfactual plan generator that systematically explores semantically consistent plan variants through rule-conditioned mutations, enabling robust collaboration strategies beyond short-sighted one-shot LLM plans. Finally, we design an agentic memory-grounded evaluator that ranks candidate plans using retrieval-augmented evidence, producing interpretable, constraint-aware selections. Across multi-agent embodied simulation benchmarks, EvoCF consistently discovers more robust and executable plans compared to baseline approaches. Our results demonstrate that grounding multi-agent planning in agentic memory and counterfactual reasoning significantly enhances both effectiveness and robustness.} }
Endnote
%0 Conference Paper %T EvoCF: Multi-Agent Collaboration via Agentic Memory-Driven Evolutionary Counterfactual Planning %A Haotian Chi %A Zeyu Feng %A Xingrui Yu %A Linbo Luo %A Yew-Soon Ong %A Ivor Tsang %A Hechang Chen %A Yi Chang %A Haiyan Yin %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-chi26b %I PMLR %P 19424--19437 %U https://proceedings.mlr.press/v306/chi26b.html %V 306 %X Planning collaboration strategies for multi-agent embodied systems remains a core challenge for LLM-based planners, which often fail to capture the physical and coordination constraints of realworld environments. To address this, we present EvoCF, an agentic memory-driven evolutionary counterfactual planning framework for discovering improved multi-agent collaboration strategies through counterfactual plan generation and evaluation. First, we propose a symbolic constraint inductor that induces reusable symbolic constraints from failures, forming an evolving rule library. Then, we propose an evolutionary counterfactual plan generator that systematically explores semantically consistent plan variants through rule-conditioned mutations, enabling robust collaboration strategies beyond short-sighted one-shot LLM plans. Finally, we design an agentic memory-grounded evaluator that ranks candidate plans using retrieval-augmented evidence, producing interpretable, constraint-aware selections. Across multi-agent embodied simulation benchmarks, EvoCF consistently discovers more robust and executable plans compared to baseline approaches. Our results demonstrate that grounding multi-agent planning in agentic memory and counterfactual reasoning significantly enhances both effectiveness and robustness.
APA
Chi, H., Feng, Z., Yu, X., Luo, L., Ong, Y., Tsang, I., Chen, H., Chang, Y. & Yin, H.. (2026). EvoCF: Multi-Agent Collaboration via Agentic Memory-Driven Evolutionary Counterfactual Planning. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:19424-19437 Available from https://proceedings.mlr.press/v306/chi26b.html.

Related Material