SOURCE-LINKED INTELLIGENCE

What Do Hallucinations Reveal About Multimodal Reasoning? Diagnosing Visual Grounding Failures via Contrastive Decoding Probes

arXiv · AI, language, vision and robotics · article · Sep 15, 2026 · UTC

When strong multimodal models are widely available, progress requires new scientific methodologies beyond benchmark scores---using models as instruments for understanding behavior. We address this by asking: can we use large vision-language models (LVLMs) as experimental instruments for studying their own failure dynamics? Focusing on visual hallucination, we introduce SAFE, a training-free decoding framework that contrasts visually-grounded and vision-ablated generation paths to produce a token-level contrastive grounding score that identifies when the model favors linguistic priors over visu

Read original source ↗ Open in workspace

recordType

paper

region

Global