Search More, Think Less: Rethinking Long-Horizon Agentic Search for Efficiency and Generalization

TL;DR

This paper introduces SMTL, a framework for long-horizon agentic search that improves efficiency and generalization by parallel evidence acquisition and a unified data synthesis pipeline, achieving state-of-the-art results across multiple benchmarks.

Contribution

SMTL replaces sequential reasoning with parallel evidence gathering and introduces a unified data synthesis pipeline for diverse tasks, enhancing efficiency and generalization.

Findings

Reduces reasoning steps by 70.7% on BrowseComp

Achieves state-of-the-art performance on multiple benchmarks

Improves accuracy while decreasing inference cost

Abstract

Recent deep research agents primarily improve performance by scaling reasoning depth, but this leads to high inference cost and latency in search-intensive scenarios. Moreover, generalization across heterogeneous research settings remains challenging. In this work, we propose \emph{Search More, Think Less} (SMTL), a framework for long-horizon agentic search that targets both efficiency and generalization. SMTL replaces sequential reasoning with parallel evidence acquisition, enabling efficient context management under constrained context budgets. To support generalization across task types, we further introduce a unified data synthesis pipeline that constructs search tasks spanning both deterministic question answering and open-ended research scenarios with task appropriate evaluation metrics. We train an end-to-end agent using supervised fine-tuning and reinforcement learning, achieving strong and often state of the art performance across benchmarks including BrowseComp (48.6\%), GAIA (75.7\%), Xbench (82.0\%), and DeepResearch Bench (45.9\%). Compared to Mirothinker-v1.0, SMTL with maximum 100 interaction steps reduces the average number of reasoning steps on BrowseComp by 70.7\%, while improving accuracy.