ROMA: Recursive Open Meta-Agent Framework for Long-Horizon Multi-Agent Systems
Salaheddin Alzu’bi, Baran Nama, Arda Kaz, Anushri Eswaran, Weiyuan Chen, Sarvesh Khetan, Rishab Bala, Tu Vu, Sewoong Oh · 2026
Read ROMA: Recursive Open Meta-Agent Framework for Long-Horizon Multi-Agent Systems by Salaheddin Alzu’bi, Baran Nama, Arda Kaz, Anushri Eswaran, Weiyuan Chen, Sarvesh Khetan, Rishab Bala, Tu Vu, Sewoong Oh (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-2602-01848.txt.
Description
Salaheddin Alzu'bi and colleagues fix agentic frameworks for long tasks, where orchestration turns brittle and traces hide failures. Their ROMA framework decomposes goals into subtask trees run in parallel, while aggregation compresses and validates intermediates to control context. Four roles separate orchestration from model choice: Atomizer, Planner, Executor and Aggregator, giving transparent traces and letting systems mix models and tools by cost and capability. On SEAL-0 with GLM-4 it beats Kimi-Researcher by nearly ten percent on web evidence, and on EQ-Bench DeepSeek-V3 matches leading closed source models.
Catalog record
| CREATOR | Salaheddin Alzu’bi, Baran Nama, Arda Kaz, Anushri Eswaran, Weiyuan Chen, Sarvesh Khetan, Rishab Bala, Tu Vu, Sewoong Oh |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 77913 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.01848