#vision-language

'Baidu launches ERNIE-4.5-VL-28B-A3B-Thinking, a compact open-source multimodal model that activates 3B parameters per token while offering strong document, chart and video reasoning capabilities.'

Hugging Face open-sourced FineVision — a 24M-sample multimodal dataset that boosts VLM performance across benchmarks while keeping data leakage minimal.