The Year in AI Papers
The year/Independent research

Paper 2509.13313

ReSum: Unlocking Long-Horizon Search Intelligence via Context Summarization

Xixi Wu, Kuan Li, Yida Zhao, Liwen Zhang, Litu Ou, Huifeng Yin, Zhongwang Zhang, Xinmiao Yu, Dingchu Zhang, Yong Jiang, Pengjun Xie, Fei Huang, et al.

Published
Sep 2025
Research lab
Independent
Citations
95
GitHub
Not linked

01 In brief

Summary

ReSum is a plug-and-play paradigm that enables LLM-based web agents to perform unbounded exploration by periodically invoking an external summarization tool to condense interaction histories into compact summaries, addressing the conflict between extensive exploration and limited context windows.

The authors developed ReSumTool-30B, a specialized summary model fine-tuned from Qwen3-30B-A3B-Thinking on high-quality conversation-summary pairs, which outperforms larger models like DeepSeek-R1-671B in summarization quality.

They also propose ReSum-GRPO, an adaptation of Group Relative Policy Optimization that segments long trajectories at summarization points and broadcasts trajectory-level advantages across segments to enable credit assignment over long horizons.

Experiments on GAIA, BrowseComp, and BrowseComp-zh show ReSum achieves a 4.5% average improvement over ReAct in training-free settings, with ReSum-GRPO yielding a further 8.2% gain.

Notably, with only 1K training samples, a ReSum-enhanced 30B agent achieves competitive performance with leading open-source models trained on 10K+ samples, demonstrating ReSum's effectiveness and data efficiency.

02 From the paper

Abstract

Large Language Model (LLM)-based web agents excel at knowledge-intensive tasks but face a fundamental conflict between the need for extensive exploration and the constraints of limited context windows. Current solutions typically rely on architectural modifications, e.g., internal memory tokens, which break compatibility with pre-existing agents and necessitate costly end-to-end retraining. To overcome these limitations, we introduce ReSum, a lightweight, plug-and-play paradigm that enables unbounded exploration by periodically invoking an external tool to condense interaction histories into compact summaries. Although this paradigm functions without training, standard agents are not inherently aligned to reason over such compressed contexts. To bridge this gap, we propose ReSum-GRPO, which adapts Group Relative Policy Optimization (GRPO) via advantage broadcasting to propagate final rewards across segmented trajectories, enabling credit assignments over long-horizons. Extensive experiments show that ReSum achieves a 4.5% improvement over ReAct in training-free settings, with ReSum-GRPO yielding a further 8.2% gain. Notably, with only 1K training samples, a ReSum-enhanced 30B agent achieves competitive performance with leading open-source models, showing ReSum's effectiveness.