Xiaomi open-sources Xiaomi-Lens-0, a 4B model that marks image regions it cites
Overview
Xiaomi has open-sourced Xiaomi-Lens-0, a 4-billion-parameter multimodal model that marks the exact image region it refers to while answering.
The model is released under Apache 2.0, with weights on Hugging Face under SeerRay-Lab, code on GitHub and a technical report on arXiv. It is built on Qwen3.5-4B with a lightweight pixel-mask decoder.
According to the technical report, the model scores 5.6 points higher on average than the Qwen3.5 base on RefCOCO referring splits. Those scores come from the authors' own evaluations.
Written by AI from the articles below · updated Oct 21, 3:40 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- PandailyNewsXiaomi open-sources Xiaomi-Lens-0, a 4B multimodal model that points to image regions
AIXiaomi has open-sourced Xiaomi-Lens-0, a 4-billion-parameter multimodal model under Apache 2.0 that marks the image region it refers to while answering. The model is built on Qwen3.5-4B with a lightweight pixel-mask decoder. According to the technical report, it gains 5.6 points on average over the Qwen3.5 base on RefCOCO referring splits, and these scores are the authors' own evaluations.
Heat Trend
Not enough continuous observations to show a trend yet.