Search papers, labs, and topics across Lattice.
This paper introduces SlimSearcher, a novel framework designed to enhance the efficiency of web agents by balancing accuracy and computational cost through two key stages: Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL). By implementing Pareto-efficient filtration in the SFT phase and Adaptive Reward Gating during RL, SlimSearcher effectively reduces unnecessary tool calls and token consumption while maintaining or improving task accuracy. Experimental results on benchmarks such as GAIA, BrowseComp, and XBenchDeepSearch reveal a significant reduction in average tool-call rounds by 17%-58%, underscoring the framework's effectiveness in promoting efficiency-aware search behaviors.
SlimSearcher cuts tool-call rounds by up to 58% without sacrificing accuracy, redefining efficiency in web agent training.
Deep research agents have demonstrated remarkable capabilities in complex information-seeking tasks, yet this power comes at a steep computational cost. Driven by accuracy-focused training paradigms, current models adopt brute-force strategies characterized by blind tool dependency and performative reasoning-generating long, redundant trajectories that are far from necessary for resolving these tasks, leading to wasteful tool calls and excessive token consumption. To overcome this efficiency trap, we propose SlimSearcher, a principled framework that pushes the Pareto frontier between accuracy and computational cost across both Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL). In the SFT stage, SlimSearcher employs Pareto-efficient filtration to distill trajectories that are both successful and economical, guiding the model toward inherently efficiency-aware search behaviors. During RL, we introduce Adaptive Reward Gating, a dynamic reward-shaping mechanism that evaluates relative tool and token efficiency within a sampled cohort. By cascading these adaptive efficiency metrics with a strict correctness gate, our approach effectively avoids the brevity bias associated with absolute penalties and mitigates reward hacking. Extensive experiments on long-horizon benchmarks, including GAIA, BrowseComp, and XBenchDeepSearch, demonstrate that SlimSearcher reduces average tool-call rounds by 17%-58% while maintaining or improving accuracy.
