license: mit
pipeline_tag: text-generation
Distillation-Outerlinks
🌐 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 Distillation-Style setting, the Expert Agent provides guidance to the Learner Agent, while Outer RecursiveLink modules transfer latent states between the two agents for collaborative refinement.
Model Details
| Item | Description |
|---|---|
| Model | Distillation-Outerlinks |
| Collaboration Style | Distillation-Style |
| Component Role | Outer RecursiveLink Modules |
| Expert-Learner-Outerlink.pt | Expert Agent → Learner Agent |
| Learner-Expert-Outerlink.pt | Learner Agent → Expert Agent |
⚠️ Note: This checkpoint contains Outer RecursiveLink modules in RecursiveMAS, rather than a standalone model intended for plain-text generation.
For detailed usage instructions, please refer to our GitHub repository.
Sample Usage
You can load the entire multi-agent system using the following code:
from system_loader import load_mas_system
mas = load_mas_system(
style="distillation",
device="cuda",
trust_remote_code=True,
)
expert = mas.agents["expert"].model
learner = mas.agents["learner"].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},
}