DECOR: Learning to Decompose and Collaborate in Deep Search via Multi-Agent Reinforcement Learning

Ruiqing Chen, Zekun Zhang, Gong-Duo Zhang, Lihong Gu, Lin Zhou
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:18398-18416, 2026.

Abstract

Monolithic agents in deep search often suffer from "cognitive overload," while existing multi-agent approaches mostly rely on frozen models that cannot learn from collaboration failures. To bridge this gap, we propose $\textbf{DECOR}$ ($\textbf{DE}$compose and $\textbf{CO}$llaborate via $\textbf{R}$ole-specialized agents), a framework formulating deep search as a Multi-Agent Reinforcement Learning (MARL) problem. DECOR functionally decomposes the task into three specialized roles: a $\textit{Planner}$ to navigate, a $\textit{Filter}$ to curate a noise-reduced memory, and an $\textit{Answerer}$ for synthesis. Unlike training-free orchestration, we jointly optimize these agents using a hybrid reward strategy that harmonizes role-specific intrinsic feedback with team-level outcome signals. Experiments on seven benchmarks show that DECOR significantly outperforms strong monolithic baselines, demonstrating the necessity of learning-based functional decomposition in handling cognitive overload.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-chen26gp, title = {{DECOR}: Learning to Decompose and Collaborate in Deep Search via Multi-Agent Reinforcement Learning}, author = {Chen, Ruiqing and Zhang, Zekun and Zhang, Gong-Duo and Gu, Lihong and Zhou, Lin}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {18398--18416}, 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/chen26gp/chen26gp.pdf}, url = {https://proceedings.mlr.press/v306/chen26gp.html}, abstract = {Monolithic agents in deep search often suffer from "cognitive overload," while existing multi-agent approaches mostly rely on frozen models that cannot learn from collaboration failures. To bridge this gap, we propose $\textbf{DECOR}$ ($\textbf{DE}$compose and $\textbf{CO}$llaborate via $\textbf{R}$ole-specialized agents), a framework formulating deep search as a Multi-Agent Reinforcement Learning (MARL) problem. DECOR functionally decomposes the task into three specialized roles: a $\textit{Planner}$ to navigate, a $\textit{Filter}$ to curate a noise-reduced memory, and an $\textit{Answerer}$ for synthesis. Unlike training-free orchestration, we jointly optimize these agents using a hybrid reward strategy that harmonizes role-specific intrinsic feedback with team-level outcome signals. Experiments on seven benchmarks show that DECOR significantly outperforms strong monolithic baselines, demonstrating the necessity of learning-based functional decomposition in handling cognitive overload.} }
Endnote
%0 Conference Paper %T DECOR: Learning to Decompose and Collaborate in Deep Search via Multi-Agent Reinforcement Learning %A Ruiqing Chen %A Zekun Zhang %A Gong-Duo Zhang %A Lihong Gu %A Lin Zhou %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-chen26gp %I PMLR %P 18398--18416 %U https://proceedings.mlr.press/v306/chen26gp.html %V 306 %X Monolithic agents in deep search often suffer from "cognitive overload," while existing multi-agent approaches mostly rely on frozen models that cannot learn from collaboration failures. To bridge this gap, we propose $\textbf{DECOR}$ ($\textbf{DE}$compose and $\textbf{CO}$llaborate via $\textbf{R}$ole-specialized agents), a framework formulating deep search as a Multi-Agent Reinforcement Learning (MARL) problem. DECOR functionally decomposes the task into three specialized roles: a $\textit{Planner}$ to navigate, a $\textit{Filter}$ to curate a noise-reduced memory, and an $\textit{Answerer}$ for synthesis. Unlike training-free orchestration, we jointly optimize these agents using a hybrid reward strategy that harmonizes role-specific intrinsic feedback with team-level outcome signals. Experiments on seven benchmarks show that DECOR significantly outperforms strong monolithic baselines, demonstrating the necessity of learning-based functional decomposition in handling cognitive overload.
APA
Chen, R., Zhang, Z., Zhang, G., Gu, L. & Zhou, L.. (2026). DECOR: Learning to Decompose and Collaborate in Deep Search via Multi-Agent Reinforcement Learning. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:18398-18416 Available from https://proceedings.mlr.press/v306/chen26gp.html.

Related Material