IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis
Xingyu Wu and colleagues at Zhejiang University and Tencent propose IterSynth, a deep-search agent that splits work between a Planner and a Synthesizer that maintains an evolving summary instead of a growing history.
Ask this paper
Role decoupling. The Planner identifies information needs; the Synthesizer folds evidence into a summary state that serves as the persistent state of search.
RDPO. Role-Decoupled Policy Optimization combines terminal outcome rewards with turn-level rubric scores and computes separate advantages per role.
Results. IterSynth-8B averages 50.7 across five long-horizon benchmarks including BrowseComp and Xbench-DS, 4.2% above the strongest prior agent at 8B or below.
Prompting transfer. Used as a prompting pattern on frontier proprietary models, it gives zero-shot gains over ReAct.
Abstract
Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle planning, evidence use, and synthesis; and context accumulation, where growing search histories introduce noise and obscure useful information. To address these issues, we propose IterSynth, a role-decoupled and summary-based paradigm that alternates between a Planner for identifying information needs and a Synthesizer for integrating evidence into an evolving summary state. This design separates planning from synthesis while using the summary as the persistent state of search, reducing both capability coupling and context noise. To train IterSynth effectively, we further introduce Role-Decoupled Policy Optimization (RDPO) for reinforcement learning, which combines terminal outcome rewards with turn-level rubric evaluations and computes role-specific advantages for more precise credit assignment. Experiments on five long-horizon deep-search benchmarks such as BrowseComp and Xbench-DS show that IterSynth-8B achieves an average score of 50.7, surpassing the strongest prior $\leq$8B agent by +4.2\%. Moreover, IterSynth serves as a model-agnostic prompting paradigm, delivering substantial zero-shot gains over ReAct and similar prompting paradigms on frontier proprietary models.