base_model:
- Qwen/Qwen3.5-4B
license: mit
pipeline_tag: text-generation
Deliberation-Reflector-Qwen3.5-4B
🌐 Project Page | 💻 Code | 📄 Paper
We introduce RecursiveMAS, a multi-agent framework that scales agent collaboration through latent-space recursion. RecursiveMAS treats a multi-agent system as a unified recursive computation, where heterogeneous agents iteratively exchange, refine, and evolve their latent states across recursion rounds. In the Deliberation-Style setting, the Reflector Agent reviews and refines the reasoning process, while collaborating with the Tool-Caller Agent through RecursiveLink modules for iterative improvement.
Model Details
| Item | Description |
|---|---|
| Model | Deliberation-Reflector-Qwen3.5-4B |
| Collaboration Style | Deliberation-Style |
| Agent Role | Reflector Agent |
| Base Model | Qwen3.5-4B |
⚠️ Note: This checkpoint is a role-specific agent in RecursiveMAS, rather than a standalone model intended for plain-text generation.
Sample Usage
To use this model, you should load the entire multi-agent system using the official system loader as described in the GitHub repository:
from system_loader import load_mas_system
# Load the full deliberation multi-agent system
mas = load_mas_system(
style="deliberation",
device="cuda",
trust_remote_code=True,
)
# Access the individual reflector agent
reflector = mas.agents["reflector"].model
Model Collections for RecursiveMAS
| Style | Model Collection |
|---|---|
| Sequential-Style | 🤗 HuggingFace |
| Mixture-Style | 🤗 HuggingFace |
| Distillation-Style | 🤗 HuggingFace |
| Deliberation-Style | 🤗 HuggingFace |
Experiment Results
Citation
@misc{recursivemas,
title={Recursive Multi-Agent Systems},
author={Xiyuan Yang and Jiaru Zou and Rui Pan and Ruizhong Qiu and Pan Lu and Shizhe Diao and Jindong Jiang and Hanghang Tong and Tong Zhang and Markus J. Buehler and Jingrui He and James Zou},
year={2026},
eprint={2604.25917},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2604.25917},
}