AgentDisCo: Towards Disentanglement and Collaboration in Open-ended Deep Research Agents

Jiarui Jin, Zexuan Yan, Shijian Wang, Wenxiang Jiao, Yuan Lu · 2026

Read AgentDisCo: Towards Disentanglement and Collaboration in Open-ended Deep Research Agents by Jiarui Jin, Zexuan Yan, Shijian Wang, Wenxiang Jiao, Yuan Lu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-11732.txt.

Introduction

Jiarui Jin and four colleagues split deep research into two cooperating roles instead of one tangled module. In AgentDisCo, a critic agent judges outlines and sharpens search queries. A generator agent fetches fresh results and revises the outlines, and a writer then turns the refined outline into a report. The system matched or beat leading closed systems on three deep research benchmarks.

Catalog record

CREATORJiarui Jin, Zexuan Yan, Shijian Wang, Wenxiang Jiao, Yuan Lu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE147947
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2605.11732

Browse the full library · About the Tower of Babel