Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Evaluating text-to-image/video/3D models with VQAScore
| Date | Stars |
|---|---|
| 2026-07-24 | 596 |
| 2026-07-25 | 597 |
| 2026-07-28 | 598 |
| 2026-07-30 | 598 |
| 2026-08-06 | 598 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# VQAScore for Evaluating Text-to-Visual Models [[Project Page]](https://linzhiqiu.github.io/papers/vqascore/) *VQAScore allows researchers to automatically evaluate text-to-image/video/3D models using one line of Python code!* [[VQAScore Page](https://linzhiqiu.github.io/papers/vqascore/)] [[VQAScore Demo](https://huggingface.co/spaces/zhiqiulin/VQAScore)] [[GenAI-Bench Page](https://linzhiqiu.github.io/papers/genai_bench/)] [[GenAI-Bench Demo](https://huggingface.co/spaces/BaiqiL/GenAI-Bench-DataViewer)] [[CLIP-FlanT5 Model Zoo](https://github.com/linzhiqiu/CLIP-FlanT5/blob/master/docs/MODEL_ZOO.md)] **VQAScore: Evaluating Text-to-Visual Generation with Image-to-Text Generation** (ECCV 2024) [[Paper](https://arxiv.org/pdf/2404.01291)] [[HF](https://huggingface.co/zhiqiulin/clip-flant5-xxl)] [Zhiqiu Lin](https://linzhiqiu.github.io/), [Deepak Pathak](https://www.cs.cmu.edu/~dpathak/), Baiqi Li, Jiayao Li, [Xide Xia](https://scholar.google.com/citations?user=FHLTntIAAAAJ&hl=en), [Graham Neubig](https://www.phontron.com/), [Pengchuan Zhang*](https://scholar.google.com/citations?user=3VZ_E64AAAAJ&hl=en), [Deva Ramanan*](https://www.cs.cmu.edu/~deva/) (*Co-First and co-senior authors) **GenAI-Bench: Evaluating and Improving Compositional Text-to-Visual Generation** (CVPR 2024, **Best Short Paper @ SynData Workshop**) [[Paper](https://arxiv.org/abs/2406.13743)] [[HF](https://huggingface.co/spaces/BaiqiL/GenAI-Bench-DataViewer)] Baiqi Li*, [Zhiqiu Lin*](https://linzhiqiu.github.io/), [Deepak Pathak](https://www.cs.cmu.edu/~dpathak/), Jiayao Li, Yixin Fei, Kewen Wu, Tiffany Ling, [Xide Xia*](https://scholar.google.com/citations?user=FHLTntIAAAAJ&hl=en), [Pengchuan Zhang*](https://scholar.google.com/citations?user=3VZ_E64AAAAJ&hl=en), [Graham Neubig*](https://www.phontron.com/), [Deva Ramanan*](https://www.cs.cmu.edu/~deva/) (*Co-First and co-senior authors) **CameraBench: Towards Understanding Camera Motions in Any Video** (arXiv 2025) [[Paper](https://arxiv.org/abs/2504.15376)] [[Site](https://linzhiqiu.github.io/papers/camerabench/)] [Zhiqiu Lin\*](https://linzhiqiu.github.io/), Siyuan Cen\*, Daniel Jiang, Jay Karhade, Hewei Wang, [Chancharik Mitra](https://chancharikmitra.github.io/), Tiffany Yu Tong Ling, Yuhan Huang, Sifan Liu, Mingyu Chen, Rushikesh Zawar, Xue Bai, Yilun Du, Chuang Gan, [Deva Ramanan](https://www.cs.cmu.edu/~deva/) (\*Co-First Authors) **CHAI: Building a Precise Video Language with Human–AI Oversight** (CVPR 2026, **Highlight · Top 3%**) [[Paper](https://arxiv.org/abs/2604.21718)] [[Code](https://github.com/chancharikmitra/CHAI)] [[HF](https://huggingface.co/datasets/chancharikm/CHAI_testset)] [[Site](https://linzhiqiu.github.io/papers/chai/)] [Zhiqiu Lin](https://linzhiqiu.github.io/)\*, [Chancharik Mitra](https://chancharikmitra.github.io/)\*, Siyuan Cen, Isaac Li, Yuhan Huang, Yu Tong Tiffany Ling, Hewei Wang, Irene Pi, Shihang Zhu, Ryan Rao, George Liu, Jiaxi Li, Ruojin Li, Yili Han, [Yilun Du](https://yilundu.github.io/), [Deva Ramanan](https://www.cs.cmu.edu/~deva/) (\*Co-First Authors) --- ## ⚠️ Reproducing Paper Results (Legacy Version) > **If you need to reproduce results from the original VQAScore or GenAI-Bench papers**, please use the legacy v3.0 release, which includes CLIP-FlanT5, InstructBLIP, LLaVA-1.5, and other models used in those works. > > See [`V_3.0_README.md`](V_3.0_README.md) for full documentation, or install the legacy version directly: > ```bash > pip install t2v-metrics==3.0 > ``` > v3.1 targets `torch>=2.7.0` and `transformers>=5.0.0` to support the latest > frontier models. Models using `trust_remote_code` (InternVL, Molmo2) are temporarily > unavailable due to transformers 5.x breaking changes and will return in a future release. --- ## News - [2026/06/04] 🚀 **VQAScore v3.1** — major model refresh with support for **Gemma 3**, **Qwen3.5**, and updated **Qwen3-VL** and **Qwen3-Omni** families. Adds `forward_with_trace` for full token-level scoring transparency
Excerpt of 19,844 characters
Read on GitHub
122
85
13
6
Jean de Dieu Nyandwi · Carnegie Mellon
2
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:19639549bd71bff4, topic:vision-language-model