让AI搜索能自动验证证据真伪,提升回答可信度
EviGraph: Towards Verifiable Evidence Construction for Information-Seeking Agents

- 分离搜索与证据记录,用共享策略统一管理
- 在BrowseComp-Plus上准确率达35.9%,显著优于基线
- 适合需要可验证推理的智能搜索系统开发者
当前智能体在网页搜索中虽能获取信息,却难以验证内容是否支持答案。现有方法将搜索与证据记录混在同一线性流程中,主要优化最终答案正确性,对中间环节的可信性监督不足。我们提出EviGraph,一个深度搜索框架,将搜索执行与证据记录分离,使用共享策略训练角色。执行器生成简洁查询,冻结的证据验证器检查网页并返回带明确极性的原文片段,策略将这些片段映射为添加/支持图请求,并由确定性结构验证器检验。生成的图既作为持久工作记忆,又提供密集过程奖励,使强化学习可直接监督证据构建而非仅最终答案。在BrowseComp-Plus上,基于Qwen3-8B的EviGraph agent在匹配交互预算下达到35.9%准确率,优于无强化学习的双角色架构(26.9%)和单体智能体(2.7%),且每轮迭代生成更少令牌。在BrowseComp、GAIA和XBench上均表现一致提升,表明显式结构化并奖励证据记录可显著改善智能体搜索性能。
原文摘要 · Abstract (English)
Agentic Web search can retrieve relevant information without establishing that the retrieved content actually supports the claims used in an answer. Existing agents typically keep search and evidence recording in a linear interaction trace and optimize primarily for final-answer correctness, providing limited supervision for intermediate grounding. We present EviGraph, a deep-search framework that separates search execution from evidence recording while using a shared policy for the trainable roles. An executor plans concise queries, a frozen evidence verifier inspects source pages and returns verbatim evidence items with an explicit polarity, and the policy maps those items to add/support graph requests that are checked by a deterministic structural validator. The resulting graph serves both as persistent working memory and as a source of dense process rewards, enabling reinforcement learning to directly supervise evidence construction rather than only the final answer. On BrowseComp-Plus, a Qwen3-8B EviGraph agent achieves 35.9% accuracy under a matched interaction budget, compared with 26.9% for the same dual-role architecture without reinforcement learning and 2.7% for a monolithic agent, while generating fewer tokens per rollout. Consistent gains on BrowseComp, GAIA, and XBench indicate that explicitly structuring and rewarding evidence recording improves agentic search
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。