Ever waited 10+ minutes for an agent to finish deep researching, only to get results different from what you intended? … we built this for you.

EDR is transparent, steerable and modular — you can intervene mid-run, correct course, and stay aligned in real time.

Code: https://github.com/SalesforceAIResearch/enterprise-deep-rese...

Paper: https://arxiv.org/pdf/2510.17797

Beside the framework and evaluation code, we've also released EDR-200 dataset, which contains 201 complete agentic research trajectories generated by Enterprise Deep Research. These trajectories expose the full reasoning process across search, reflection, and synthesis steps, enabling fine-grained analysis of agentic planning and decision-making dynamics. Check it out!

HuggingFace: https://lnkd.in/gYNH2ZF8