---
task_categories:
  - graph-ml
tags:
  - chemistry
  - molecular-design
  - in-context-learning
  - drug-discovery
language:
  - en
---

---
Paper: [Graph Diffusion Transformers are In-Context Molecular Designers](https://huggingface.co/papers/2510.08744)
Code: https://github.com/liugangcode/DemoDiff
---
# DemoDiff Downstream Context Data

This dataset contains context data for the DemoDiff project, a diffusion-based molecular foundation model for in-context inverse molecular design. It provides contextual examples to guide molecular generation, enabling few-shot molecular design across diverse chemical tasks without task-specific fine-tuning.

## Structure

Each task is organized as a separate folder in the repository root:

```
tasks/
├── Albuterol_Similarity/
│   ├── positive.csv
│   ├── medium.csv
│   └── negative.csv
├── Amlodipine_MPO/
│   ├── positive.csv
│   ├── medium.csv
│   └── negative.csv
└── ... (other tasks)
```