---
license: apache-2.0
language: en
---
# 🦒 The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic Distillation

This repository hosts the models from **PlanPhys**, a family of  foundation models designed for researching multi-turn long-horizon planning.

The PlanPhys model family explores different training strategies, including Random model, SFT model, OPD model.

---

## 🚀 Model Zoo

We provide multiple PlanPhys models (**The data is in each Files and version**):

| Model |
|---|
| 🔹 [PlanPhys-Model-Pre-Training-Random](https://huggingface.co/MultimodalAgent/PlanPhys-Model-Pre-Training-Random) |
| 🔹 [PlanPhys-Model-Pre-Training-Scaling](https://huggingface.co/MultimodalAgent/PlanPhys-Model-Pre-Training-Scaling) |
| 🔹 [PlanPhys-Model-Pre-Training-Distribution](https://huggingface.co/MultimodalAgent/PlanPhys-Model-Pre-Training-Distribution) |
| 🔹 [PlanPhys-Model-Pre-Training-Ensemble](https://huggingface.co/MultimodalAgent/PlanPhys-Model-Pre-Training-Ensemble) |

---
## 📝 Citation

If you find PlanPhys useful in your research, please cite:

```bibtex
@article{PlanPhys,
  title={The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic Distillation},
  author={},
  journal={},
  year={}
}