SlimSearcher reduces web agent tool calls by 17-58% via adaptive reward gating
SlimSearcher: Training Efficiency-Aware Web Agents via Adaptive Reward Gating
Hugging FaceTrending 8
Summary
SlimSearcher uses Pareto-efficient trajectory filtration and adaptive reward gating to train web agents that reduce tool-call rounds by 17-58% while maintaining or improving accuracy on GAIA, BrowseComp, and XBenchDeepSearch.