Tencent Open-Sources WeVisOne-VL, Unifying Detection, OCR, Segmentation in One Model

Tencent's WeChat Vision team open-sourced WeVisOne-VL (2B/4B, Apache 2.0, built on Qwen3-VL), a single model unifying object-level and pixel-level visual perception including detection, counting, OCR, and segmentation. The paper argues existing VLMs are unreliable at fine-grained perception and addresses it via structured next-token prediction.

2026-10-26 ~ 2026-10-26 · 3 related posts

• Tencent open-sources WeVisOne-VL: one VLM for detection, OCR, segmentation and matting — lmoroney · 2026-10-26
• WeVisOne-VL paper: unified object-to-pixel visual perception on arXiv — lmoroney · 2026-10-26

1 near-duplicate retellings: lmoroney