license: mit
pipeline_tag: text-generation
Sequential-Scaled-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 Sequential-Scaled setting, the Planner Agent, Critic Agent, and Solver Agent collaborate through Outer RecursiveLink modules for recursive planning, critique, and solution generation.
Model Details
| Item | Description |
|---|---|
| Model | Sequential-Scaled-Outerlinks |
| Collaboration Style | Sequential-Scaled |
| Component Role | Outer RecursiveLink Modules |
| Planner-Critic-Outerlink.pt | Planner Agent → Critic Agent |
| Critic-Solver-Outerlink.pt | Critic Agent → Solver Agent |
| Solver-Planner-Outerlink.pt | Solver Agent → Planner Agent |
⚠️ Note: This checkpoint contains Outer RecursiveLink modules in RecursiveMAS, rather than a standalone model intended for plain-text generation.
Sample Usage
You can load the full multi-agent system (MAS) including these modules using the following code snippet from the official repository:
from system_loader import load_mas_system
mas = load_mas_system(
style="sequential_scaled",
device="cuda",
trust_remote_code=True,
)
planner = mas.agents["planner"].model
critic = mas.agents["critic"].model
solver = mas.agents["solver"].model
For detailed usage instructions, please refer to the GitHub repository.
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},
}