# High-Precision Image Captioning

PixelReasoner is a vision-language model based on Qwen2.5-VL-7B-Instruct, trained with curiosity-driven reinforcement learning, focusing on image-text-to-text tasks.

The quantized version of Qwen2.5-VL-3B-Instruct, with weights quantized to INT4 and activations quantized to FP16, designed for efficient vision-text task inference.

Asagi-4B is a large-scale Japanese Vision-Language Model (VLM) trained on extensive Japanese datasets, incorporating diverse data sources.