收录解读
deep research agent 正从短答案检索转向长报告综合，但多模态证据、开放式 synthesis 和可验证性仍然是弱点。
Ptah 把 interleaved report generation 做成 multi-agent harness，围绕证据收集、文本论证、视觉证据和验证流程进行编排。
它的系统意义在于把 deep research 从单 agent 写报告，推进到带 evidence lifecycle 和 verification 的工作流。
它值得收录，因为可验证 deep research 是 agent 系统的重要生产场景，尤其适合本库关注的 harness、workflow 和可信生成。
原始摘要与中文对照
中文对照翻译
大型语言模型 (LLMs) 已经将自主代理从基于深度搜索的简洁事实性答案，发展到基于深度研究的合成散落证据的长篇报告。然而，可验证的多模态深度研究仍然具有挑战性，因为缺乏确定性真值且需要将文本论证与视觉证据交织在一起。我们提出 P TAH，一种用于交织报告生成的多智能体框架。 P TAH 通过规划、研究和写作阶段，对从用户查询到渲染的网页报告的生命周期进行编排，其中专门的代理构建视觉感知计划、收集基于主张的证据、在视觉工作记忆中维护源对齐图像，并通过声明式多模态工具使用来撰写报告。一个验证代理作为框架的接受函数，在整个工作流程中强制执行事实性基础、引用忠实度和跨模态一致性。我们进一步引入 P TAHEval，一种通过增强现有基准并添加图像级别和演示级别评估来扩充的评估协议。在深度研究基准上的实验表明，与强大的基线相比，PAH 产生的更可靠、更具视觉信息量和更易于使用的面向人类的多模态报告。
原始摘要
Large Language Models (LLMs) have advanced autonomous agents from deep search, which retrieves concise factual answers, to deep research, which synthesizes scattered evidence into long-form reports. However, verifiable multimodal deep research remains challenging due to open-ended synthesis without deterministic ground truth and the need to interleave textual arguments with visual evidence. We propose P TAH, a multi-agent harness for interleaved report generation. P TAH orchestrates the lifecycle from user query to rendered web report through planning, research, and writing stages, where specialized agents construct visual-aware plans, collect claim-grounded evidence, maintain source-aligned images in a Visual Working Memory, and compose reports through declarative multimodal tool use. A verifier agent serves as the harness’s acceptance function, enforcing factual grounding, citation fidelity, and cross-modal consistency throughout the workflow. We further introduce P TAHEval, an evaluation protocol that augments existing benchmarks with image-level and presentation-level assessments. Experiments on deep research benchmarks show that P TAH produces more reliable, visually informative, and usable human-facing multimodal reports than strong baselines.