README.md

---
license: mit
library_name: boltz
tags:
  - biology
  - protein-structure-prediction
  - cofolding
  - protein-ligand
  - flow-matching
  - diffusion-distillation
---

# DeCAF-Boltz: Denoiser Cofolding All-atom Flowmap Model

**Distilling Boltz: Flow Maps for Fast All-Atom Cofolding**

[![arXiv](https://img.shields.io/badge/arXiv-2606.08375-b31b1b.svg)](https://arxiv.org/abs/2606.08375)
[![Blog](https://img.shields.io/badge/Blog-genesis.ml-1f6feb.svg)](https://www.genesis.ml/news/genesis-model-distillation)
[![Code](https://img.shields.io/badge/Code-GitHub-181717.svg)](https://github.com/genesistherapeutics/decaf)
[![Hugging Face](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-genesisml%2Fdecaf-ffd21e.svg)](https://huggingface.co/genesisml/decaf)

This repository hosts the **DeCAF-Boltz checkpoint** (`decaf_ckpt.ckpt`). The
inference/evaluation code lives in the
[genesistherapeutics/decaf](https://github.com/genesistherapeutics/decaf) GitHub repository.

## Overview

DeCAF is the first flow map model for all-atom cofolding. Instead of taking many steps along the denoising trajectory, a flow map learns to jump directly from one point on the trajectory to another, potentially traversing the entire generation process in just a handful of steps.

DeCAF-Boltz distills the [Boltz-1](https://github.com/jwohlwend/boltz) cofolding model into a fast few-step generator, achieving a **5x inference speedup** with near-parity in structure prediction quality. Using over 5x fewer compute steps, DeCAF-Boltz exceeds AlphaFold 3, Chai-1, Boltz-1x, and Boltz-2 on the Runs N' Poses benchmark success rate by 3 to 15 percentage points.

![Runs N' Poses benchmark](https://cdn.prod.website-files.com/69d6baade8258c456e8a100f/6a297ef4945575016df637c1_d6ba7352.png)

*Runs N' Poses (post-2023) success rate, best@5: DeCAF-Boltz nearly matches its full-budget teacher while outperforming AF3, Boltz-1x, Chai-1, and Boltz-2.*

## Usage

Download the checkpoint and run the bundled end-to-end example from the
[code repository](https://github.com/genesistherapeutics/decaf):

```bash
# 1. get the checkpoint (requires `pip install huggingface_hub`)
hf download genesisml/decaf decaf_ckpt.ckpt --local-dir .

# 2. run few-step DeCAF inference on the bundled example
bash scripts/run_decaf_example.sh ./decaf_ckpt.ckpt
```

Or call the predictor directly:

```bash
python -m boltz.main predict <input.yaml> \
    --checkpoint ./decaf_ckpt.ckpt \
    --model boltz1 \
    --sampling_steps 10 \
    --diffusion_samples 5 \
    --recycling_steps 3 \
    --accelerator gpu \
    --out_dir ./out \
    --no_kernels
```

The code auto-detects this checkpoint and switches to the few-step DecafSampler
(`Detected Decaf checkpoint — using DecafSampler for inference.`). See the
repository's `docs/decaf_prediction.md` for full prediction and evaluation instructions.

## Confidence outputs

Two checkpoints are provided:

- **`decaf_ckpt.ckpt`** — the DeCAF-Boltz few-step sampler (no confidence head). Structure only.
- **`decaf_conf_ckpt.ckpt`** — the same DeCAF head fused with the Boltz-1 confidence module. Because DeCAF shares the Boltz-1 trunk, the confidence head scores DeCAF poses in the *same* trunk pass, so pLDDT / ipTM / PAE come at essentially no extra cost.

Use the fused checkpoint exactly like the plain one — the CLI auto-detects the confidence module and writes real pLDDT into the B-factor column plus a `confidence_*.json`:

```bash
hf download genesisml/decaf decaf_conf_ckpt.ckpt --local-dir .
bash scripts/run_decaf_example.sh ./decaf_conf_ckpt.ckpt
```

You can also build the fused checkpoint yourself from a plain DeCAF checkpoint and `boltz1_conf.ckpt` with `scripts/merge_decaf_confidence.py` in the code repo.

## Citation

```bibtex
@misc{scarpellini2026fewstepcofoldingallatomflow,
      title={Few-step Cofolding with All-Atom Flow Maps},
      author={Gianluca Scarpellini and Ron Shprints and Peter Holderrieth and Juno Nam and Pranav Murugan and Rafael Gómez-Bombarelli and Tommi Jaakola and Maruan Al-Shedivat and Nicholas Matthew Boffi and Avishek Joey Bose},
      year={2026},
      eprint={2606.08375},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2606.08375},
}
```

## License

Released under the MIT License (Copyright (c) 2026 Genesis Molecular AI). The code
extends the open-source [Boltz](https://github.com/jwohlwend/boltz) project (Apache 2.0).

## Acknowledgments

The research team at Genesis is grateful to our collaborators from Massachusetts Institute of Technology: Ron Shprints, Peter Holderrieth, Juno Nam, Rafael Gomez-Bombarelli and Tommi Jaakola; Carnegie Mellon University: Nicholas Matthew Boffi, and Joey Bose from Imperial College London and Mila.