jerlendds
7 hours ago
VLM != LLM. Vision language models basically treat text tokens and image tokens the same. Post-training an LLM on images+text can improve its capabilities. Id recommend searching around the keyword VLM to find more resources on how multi-modal AI works.