---
title: Stemma
emoji: 🌳
colorFrom: green
colorTo: indigo
sdk: static
app_file: index.html
pinned: false
license: apache-2.0
tags: [model-provenance, lineage, safetensors, ai-bom, model-merging, supply-chain]
short_description: Model lineage from weights alone, over HTTP Range
---

# Stemma — model provenance from weights alone

This is a **static** results page: the measured benchmark, the figures, and the honest
limitations. The interactive analysis runs locally, not here — a Gradio Space needs a Python
backend, which requires a PRO subscription to host.

```bash
pip install "git+https://github.com/NagaYu/stemma"
stemma direction Qwen/Qwen2.5-0.5B Qwen/Qwen2.5-0.5B-Instruct
```

- Code: https://github.com/NagaYu/stemma
- Model: https://huggingface.co/NagaYu/stemma-direction
- Dataset: https://huggingface.co/datasets/NagaYu/stemma-bench