FIELD NOTES01
How vision-language models learn where things are
Spatial grounding depends on preserving coordinates, local visual tokens, and training the decoder to speak in regions.
Essays, notebooks, and field notes connected by vlm.
Spatial grounding depends on preserving coordinates, local visual tokens, and training the decoder to speak in regions.
Phi-3.5 Vision and Florence-2 look similar at the input, but they are built for rather different jobs.