RE-TRAC: REcursive TRAjectory Compression for Deep Search Agents
This addresses inefficiencies in deep search agents for AI research, offering a novel framework that improves exploration and planning, though it is incremental relative to existing agent designs.
The paper tackled the problem of LLM-based deep research agents getting stuck in local optima and inefficient search due to linear designs like ReAct, by proposing Re-TRAC, a framework that enables cross-trajectory exploration and iterative reflection, resulting in 15-20% performance improvements over ReAct on BrowseComp and reduced tool calls and token usage.
LLM-based deep research agents are largely built on the ReAct framework. This linear design makes it difficult to revisit earlier states, branch into alternative search directions, or maintain global awareness under long contexts, often leading to local optima, redundant exploration, and inefficient search. We propose Re-TRAC, an agentic framework that performs cross-trajectory exploration by generating a structured state representation after each trajectory to summarize evidence, uncertainties, failures, and future plans, and conditioning subsequent trajectories on this state representation. This enables iterative reflection and globally informed planning, reframing research as a progressive process. Empirical results show that Re-TRAC consistently outperforms ReAct by 15-20% on BrowseComp with frontier LLMs. For smaller models, we introduce Re-TRAC-aware supervised fine-tuning, achieving state-of-the-art performance at comparable scales. Notably, Re-TRAC shows a monotonic reduction in tool calls and token usage across rounds, indicating progressively targeted exploration driven by cross-trajectory reflection rather than redundant search.