Papers by Ashutosh Chaubey
Can VLMs Recall Factual Associations From Visual References? (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a systematic deficiency in the multimodal grounding of Vision Language Models is identified . VLMs can recall factual associations when provided a textual reference to an entity . |
| Approach: | They identify a systematic deficiency in the multimodal grounding of Vision Language Models . they show that VLMs struggle to link their internal knowledge of an entity with its image representation . |
| Outcome: | The study shows that VLMs struggle to link internal knowledge with image representations . the findings provide recommendations for future research . |