Mind DeepResearch Technical Report

TL;DR

Mind DeepResearch (MindDR) is an efficient multi-agent deep research framework that achieves high performance with ~30B parameters through a specialized training pipeline and collaborative architecture, outperforming larger models.

Contribution

The paper introduces a novel multi-agent architecture and a four-stage training pipeline for efficient research models, demonstrating competitive results with smaller-scale models.

Findings

Achieves 45.7% on BrowseComp-ZH and 52.5 on DeepResearch Bench.

Outperforms comparable open-source systems and rivals larger models.

Introduces MindDR Bench, a new Chinese query benchmark with a state-of-the-art score of 51.8.

Abstract

We present Mind DeepResearch (MindDR), an efficient multi-agent deep research framework that achieves leading performance with only ~30B-parameter models through a meticulously designed data synthesis and multi-stage training pipeline. The core innovation of MindDR lies in a collaborative three-agent architecture (Planning Agent, DeepSearch Agent, and Report Agent) and a four-stage agent-specialized training pipeline comprising SFT cold-start, Search-RL, Report-RL and preference alignment. With this regime, MindDR demonstrates competitive performance even with ~30B-scale models. Specifically, MindDR achieves 45.7% on BrowseComp-ZH, 42.8% on BrowseComp, 46.5% on WideSearch, 75.0% on xbench-DS, and 52.5 on DeepResearch Bench, outperforming comparable-scale open-source agent systems and rivaling larger-scale models. MindDR has been deployed as an online product in Li Auto. Furthermore, we introduce MindDR Bench, a curated benchmark of 500 real-world Chinese queries from our internal product user interactions, evaluated through a comprehensive multi-dimensional rubric system rather than relying on a single RACE metric. On MindDR Bench, MindDR achieves a state-of-the-art score of 51.8.