Deep Researcher with Test-Time Diffusion
Topics: AI Mode, Query Fan Out, Search Query Processing
The research paper by Google presents the Test-Time Diffusion Deep Researcher (TTD-DR), a novel framework designed to enhance the performance of deep research agents powered by Large Language Models (LLMs). By modeling the generation of research reports as a diffusion process akin to human writing cycles of planning, drafting, and revising, TTD-DR utilizes a preliminary draft that is continuously refined through a retrieval mechanism and self-evolutionary algorithms. This approach improves coherence and reduces information loss during the research process, leading to superior outcomes on complex research tasks.
