taste

KO | EN

gitlite — search

#kubernetes #ai #ios #database #css #agent #git #vue #data-science #self-hosted

(★ 9)

TASTE: A Designer-Annotated Multi-Dimensional Preference Dataset for AI-Generated Graphic Design

파일 탐색기

ZIP 다운로드

- - aggregate_report.py
  - distribution_plots.py
  - evaluator_agreeableness.py
  - generate_reference_stats.py
  - null_kendall.py
  - pair_matrices.py
  - README.md
  - run_dimension.py
  - taste_stats.py
  - verify.py
- - aggregate_slate.py
  - analyze_results.py
  - input_audit.py
  - prepare_pairs.py
  - README.md
  - run_vlm_judge.py
  - single_image_diagnostic.py
- expected_results.md
- hallucination_agreement.py
- README.md
- requirements.txt
- process.py
- README.md
- requirements.txt
- - input_example.csv
- - score.py
- .gitignore
- LICENSE
- README.md
- requirements.txt
.gitignore
CITATION.cff
LICENSE
README.md
requirements.txt

# CDN으로 사용하기

jsDelivr

jsDelivr는 공개 GitHub 리포지토리를 별도 설정 없이 CDN으로 즉시 서빙합니다. 버전과 파일을 고르면 웹페이지에 바로 붙일 수 있는 링크와 예시 코드가 만들어집니다.

버전

파일

링크

예시

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

This is the code and data repository for the paper

> **TASTE: A Designer-Annotated Multi-Dimensional Preference Dataset for AI-Generated Graphic Design**
> [arXiv:2605.20731](https://arxiv.org/abs/2605.20731) · [PDF](https://arxiv.org/pdf/2605.20731)

and the accompanying **TASTE dataset**, hosted on the Hugging Face Hub:

> [`purvanshi/TASTE`](https://huggingface.co/datasets/purvanshi/TASTE)

TASTE is a corpus of designer-panel rankings of AI-generated graphic designs
across multiple aesthetic and description-faithfulness dimensions, plus a small
preference model trained on it.

This is a joint work of **[Lica World](https://lica.world)** and **[Contra](https://contra.com)**.

## Repository layout

The repo is organized into three top-level parts, each self-contained with its
own README:

| Directory | What it is | README |
|---|---|---|
| [`analysis/`](analysis/) | The dataset-paper analysis — signal-validation distribution tests, the open-weight VLM-as-judge benchmark, and hallucination-flag agreement. | [`analysis/README.md`](analysis/README.md) |
| [`data/`](data/) | Processing scripts for the TASTE dataset on the Hugging Face Hub — download a local snapshot and derive the per-dimension ranking CSVs and pairwise battles used by `analysis/` and `taste-scorer/`. | [`data/README.md`](data/README.md) |
| [`taste-scorer/`](taste-scorer/) | The pip-installable **TASTE preference model** — inference for per-dimension "which design does the panel prefer?" probabilities from an image pair and a prompt. | [`taste-scorer/README.md`](taste-scorer/README.md) |

## Quick start

The three parts are independent; install only what you need.

```bash
# Score image pairs with the preference model
cd taste-scorer && pip install -e .

# Reproduce the paper analysis
cd analysis && pip install -r requirements.txt

# Fetch and process the dataset
cd data && pip install -r requirements.txt
```

The dataset itself is **not** committed here; it is fetched from the Hugging
Face Hub (see [`data/README.md`](data/README.md)). Designer identities are
masked before release — do not attempt to re-identify raters.

## Links

- Code: [github.com/purvanshi-lica/taste](https://github.com/purvanshi-lica/taste)
- Paper: [arXiv:2605.20731](https://arxiv.org/abs/2605.20731) ([PDF](https://arxiv.org/pdf/2605.20731))
- Dataset: [`purvanshi/TASTE`](https://huggingface.co/datasets/purvanshi/TASTE)
- Model checkpoint: [TASTE_Checkpoint.zip](https://storage.googleapis.com/lica-assets/TASTE/TASTE_Checkpoint.zip)
- Backbone used by the scorer: [Qwen3-VL-Embedding-2B](https://huggingface.co/Qwen/Qwen3-VL-Embedding-2B)
- Lica World: [lica.world](https://lica.world) · Contra: [contra.com](https://contra.com)

## Citing

If you use the TASTE dataset or the TASTE-scorer, please cite the paper
([arXiv:2605.20731](https://arxiv.org/abs/2605.20731)).

```
@article{zhu2026taste,
  title={TASTE: A Designer-Annotated Multi-Dimensional Preference Dataset for AI-Generated Graphic Design},
  author={Zhu, Haonan and Hirsch, Elad and Minetti, Alexandria and Nulty, Allison and Mehta, Purvanshi},
  journal={arXiv preprint arXiv:2605.20731},
  year={2026}
}
```

## License

MIT — see [LICENSE](LICENSE).

taste

오너의 다른 레포

파일 탐색기

# CDN으로 사용하기

Was this content helpful?

Recent Feedback

taste

오너의 다른 레포

파일 탐색기

# CDN으로 사용하기

Was this content helpful?

Recent Feedback

Download README