October 2, 2025Open Access

ORIC: Benchmarking Object Recognition in Incongruous Context for Large Vision-Language Models

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Benchmarking highlights significant recognition gaps in object identification within incongruous contexts, suggesting improvements are needed.

Significant recognition gaps were found in large vision-language models, particularly in incongruous contexts.
Evaluation of 18 vision-language models revealed notable failures in object misidentification and hallucination.
The study introduced ORIC to systematically benchmark object recognition amidst contextually unexpected scenarios.
Findings underscore the importance of understanding contextual incongruity to enhance future vision-language models.

Li et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

Benchmarking highlights significant recognition gaps in object identification within incongruous contexts, suggesting improvements are needed.