Beyond Logit Lens: Contextual Embeddings for Robust Hallucination Detection & Grounding in VLMs

Open in new window