# Grounding FT: Bounding-Box Token Format + RefCOCO-Tarzı Görev

> Source: https://sukruyusufkaya.com/learn/fine-tuning-cookbook/ftc-grounding-fine-tuning-bbox
> Updated: 2026-08-13T06:44:03.753Z
> Category: Fine-Tuning Cookbook (Model-by-Model)
> Module: Part VI — Vision-Language Multimodal FT
**TLDR:** VLM'in 'göstermek' özelliği: 'köpeği işaret et' → [0.32, 0.45, 0.58, 0.71]. Bounding box (bbox) token format: <bbox>x1,y1,x2,y2</bbox> veya normalize 0-1000 koordinatlar. RefCOCO dataset, grounding evaluation (IoU), Qwen 2.5-VL'in native grounding desteği.

