AI 中文总结
提出图引导深度搜索框架G-ReAct，通过结构-状态协同演化解决现有LLM深度搜索的上下文遗忘等问题，仅用少量轨迹微调即提升模型在BrowseComp-ZH、XBench上的准确率，且推理时可增强现有LLM性能。
AI 中文摘要
深度搜索已成为大型语言模型（LLMs）解决开放域复杂任务的基础能力。然而，现有方法通常依赖线性序列推理进行轨迹生成和推理，难以在长 horizon 多跳搜索中一致保留中间状态和约束，因此常出现上下文遗忘、搜索漂移和探索效率低下的问题。为解决这些局限，我们提出 G-ReAct，一种深度搜索的推理框架，将推理组织为「固定拓扑查询图上的状态演化」。演化的图状态明确跟踪搜索进度并指导后续决策，将文本历史驱动的探索性搜索转变为显式约束下的图引导推理。G-ReAct 支持训练和推理：它生成高质量深度搜索轨迹用于监督微调，并在推理时为搜索提供结构化指导，无需额外微调。实验表明，仅用 1900 条生成轨迹微调，Qwen3-30B-A3B-Thinking-2507 在 BrowseComp-ZH 上准确率达 52.6%，在 XBench 上达 79.0%，优于在大得多的数据集上训练的可比开源方法，包括强化学习增强的方法。此外，在推理时应用 G-ReAct，可持续提升现有强大 LLMs 在深度搜索任务上的性能。我们将公开发布所有代码和模型权重。
英文摘要
Deep search has become a fundamental capability of large language models (LLMs) for solving open-domain complex tasks. However, existing approaches typically rely on linear sequential reasoning for both trajectory generation and inference, making it difficult to consistently preserve intermediate states and constraints throughout long-horizon multi-hop search. Consequently, they often suffer from context forgetting, search drift, and inefficient exploration. To address these limitations, we propose $\textbf{G-ReAct}$, a reasoning framework for deep search that organizes reasoning as $\textbf{state evolution over a fixed-topology query graph}$. The evolving graph state explicitly tracks search progress and guides subsequent decisions, transforming exploratory search driven by textual history into graph-guided reasoning under explicit constraints. G-ReAct supports both training and inference: it generates high-quality deep-search trajectories for supervised fine-tuning and provides structured guidance for inference-time search without additional fine-tuning. Experiments demonstrate that with only 1.9K generated trajectories for fine-tuning, Qwen3-30B-A3B-Thinking-2507 achieves $52.6\%$ accuracy on BrowseComp-ZH and $79.0\%$ on XBench, outperforming comparable open-source methods trained on substantially larger datasets, including RL-enhanced methods. Furthermore, when applied at inference time, G-ReAct consistently improves the performance of existing strong LLMs on deep-search tasks. We will publicly release all code and model weights.
