ParallelMuse: Agentic Parallel Thinking for Deep Information Seeking
This work addresses performance bottlenecks for deep information-seeking agents, offering incremental improvements in efficiency and reasoning integration.
The paper tackled the inefficiency and integration challenges of parallel thinking in deep information-seeking agents by introducing ParallelMuse, a two-stage paradigm that improved performance by up to 62% while reducing exploratory token consumption by 10–30%.
Parallel thinking expands exploration breadth, complementing the deep exploration of information-seeking (IS) agents to further enhance problem-solving capability. However, conventional parallel thinking faces two key challenges in this setting: inefficiency from repeatedly rolling out from scratch, and difficulty in integrating long-horizon reasoning trajectories during answer generation, as limited context capacity prevents full consideration of the reasoning process. To address these issues, we propose ParallelMuse, a two-stage paradigm designed for deep IS agents. The first stage, Functionality-Specified Partial Rollout, partitions generated sequences into functional regions and performs uncertainty-guided path reuse and branching to enhance exploration efficiency. The second stage, Compressed Reasoning Aggregation, exploits reasoning redundancy to losslessly compress information relevant to answer derivation and synthesize a coherent final answer. Experiments across multiple open-source agents and benchmarks demonstrate up to 62% performance improvement with a 10--30% reduction in exploratory token consumption.