PANINI-DR: Controlled Deep Research through Generative Semantic Workspaces

Key points

• Deep research requires connecting multiple steps of search, reasoning, and computation, with intermediate findings guiding what to investigate next.1
• As investigations grow, agents must maintain research state across model calls and check findings before errors propagate into subsequent work.1
• A shared workspace, built on the Generative Semantic Workspace representation, records subquestions, their dependencies, findings, and supporting evidence.1
• We evaluate PANINI-DR on BrowseComp-Plus, FRAMES, and MoNaCo against seven state-of-the-art deep-research harnesses, all using a shared language-model backbone and frozen corpora.1

Sources (2)

• [1]PANINI-DR: Controlled Deep Research through Generative Semantic Workspaces
  Hugging Face Daily Papers · Oct 10, 12:00 AM

  We introduce PANINI-DR, a multi-agent harness that separates coordination of the overall investigation from the research and verification of individual subquestions.

  Deep research requires connecting multiple steps of search, reasoning, and computation, with intermediate findings guiding what to investigate next.

• [2]PANINI-DR: Controlled Deep Research through Generative Semantic Workspaces
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 10, 10:42 PM · same content

Extractive summary: sentences quoted from the sources.