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.
[ECCV 2026] Official repo of "Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens"
| Date | Stars |
|---|---|
| 2026-07-24 | 376 |
| 2026-07-25 | 376 |
| 2026-07-28 | 380 |
| 2026-07-30 | 386 |
| 2026-08-10 | 390 |
| 2026-08-18 | 393 |
| 2026-08-19 | 393 |
| 2026-08-23 | 394 |
| 2026-08-24 | 395 |
| 2026-08-26 | 394 |
| 2026-08-29 | 395 |
| 2026-08-31 | 396 |
| 2026-09-04 | 397 |
| 2026-09-07 | 398 |
| 2026-09-08 | 399 |
| 2026-09-10 | 400 |
| 2026-09-11 | 401 |
| 2026-09-15 | 403 |
| 2026-09-16 | 404 |
| 2026-09-18 | 405 |
| 2026-09-20 | 405 |
Today
— stars today
This week
+4 stars this week
This month
+12 stars this month
Momentum
0.0
growth rate 1.00%/day
<!-- # CoVT: Chain-of-Visual-Thought -->
<div align="center">
<h1 style="margin: 0; font-size: 1.8em;">
Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens
</h1>
<h4 style="margin: 15px 0; color:rgb(50, 151, 234);">
⭐️ CoVT enriches VLMs’ vision-centric reasoning capabilities. ⭐️
</h4>
[](https://arxiv.org/abs/2511.19418)
[](https://huggingface.co/collections/Wakals/covt-chain-of-visual-thought)
[](https://wakalsprojectpage.github.io/covt-website/)
</div>
<div align="center">
<b>
<a href="https://wakals.github.io/" target="_blank">Yiming Qin</a><sup>1,4</sup>,
<a href="https://github.com/David-BominWei" target="_blank">Bomin Wei</a><sup>2</sup>,
<a href="https://gejiaxin.org/" target="_blank">Jiaxin Ge</a><sup>1</sup>,
<a href="https://tech-ai.panasonic.com/en/researcher_introduction/048/" target="_blank">Konstantinos Kallidromitis</a><sup>3</sup>,<br>
<a href="https://stephanie-fu.github.io/" target="_blank">Stephanie Fu</a><sup>1</sup>,
<a href="https://people.eecs.berkeley.edu/~trevor/" target="_blank">Trevor Darrell</a><sup>1</sup>,
<a href="https://xudongfrankwang.github.io/" target="_blank">XuDong Wang</a><sup>1,4*</sup>
</b><br>
<span style="font-size: 1em; color: #555;">
University of California, Berkeley<sup>1</sup><br>
University of California, Los Angeles<sup>2</sup><br>
Panasonic AI Research<sup>3</sup><br>
Duke University<sup>4</sup>
</span>
<p style="color: #555; font-size: 0.9em; margin-top: 8px; margin-bottom: 0;">
*Corresponding author
</p>
</div>
<div align="center">
<img src="./assets/DEMO.jpg" alt="" style="width: 100%; margin: 10px 0;">
<img src="./assets/edit_demo.jpg" alt="" style="width: 100%; margin: 10px 0;">
</div>
## 🔥 News
[2026-07-28] 🔥 CoVT was selected for an oral presentation at ECCV 2026!
[2026-02-18] 🏆 The evaluation code is supported in [VLMEvalKit](https://github.com/open-compass/VLMEvalKit).
[2025-11-24] 🥃 Released the [training data](https://huggingface.co/datasets/Wakals/CoVT-Dataset) and code. Give it a shot!
[2025-11-24] ⭐️ The evaluation and Gradio demo are available NOW!
[2025-11-24] 🤗 Our finetuned weights are available. [Check it here!](https://huggingface.co/collections/Wakals/covt-chain-of-visual-thought)
## 📑 Table of Contents
- [👀 Overview](#-overview)
- [🚀 Quick Start](#-quick-start)
- [🤗 Model Zoo](#-model-zoo)
- [🏖️ TODO](#-todo)
- [🪪 License](#-license)
- [📮 Contact](#-contact)
## 👀 Overview
> Rather than restricting VLM reasoning to a discrete language space with limited representational capacity, **CoVT** forms a visual thought chain that enables VLMs to reason in continuous visual space. By introducing *continuous visual tokens* that encode perceptual cues (e.g., segmentation, depth, instance, and edge structure), CoVT composes *chains of textual and visual thoughts* that link semantic reasoning with perceptual grounding. These visual “thought chains” bridge language and vision, enabling fine-grained understanding, spatial precision, and geometric awareness beyond the reach of text-based reasoning.
<details>
<summary><strong>💡 Abstract</strong></summary>
<br>
Vision–Language Models (VLMs) excel at reasoning in linguistic space but struggle with perceptual understanding that requires dense visual perception, *e.g.*, spatial reasoning and geometric awaExcerpt of 11,854 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:9f7a4b92bee96886, topic:vision-language-model, topic:vlm
matched fp:9f7a4b92bee96886, topic:qwen
matched fp:9f7a4b92bee96886, topic:computer-vision