Benchmarking Zero-Shot Recognition with Vision-Language Models: Challenges on Granularity and Specificity
This work addresses challenges in open-world settings for vision-language models, providing insights into limitations for researchers and developers, though it is incremental as it builds on existing evaluation methods.
The paper tackled the problem of evaluating vision-language models in zero-shot recognition by introducing benchmarks focusing on granularity and specificity, finding that models favor moderately fine-grained concepts and struggle with specificity, often misjudging texts differing from training data.
This paper presents novel benchmarks for evaluating vision-language models (VLMs) in zero-shot recognition, focusing on granularity and specificity. Although VLMs excel in tasks like image captioning, they face challenges in open-world settings. Our benchmarks test VLMs' consistency in understanding concepts across semantic granularity levels and their response to varying text specificity. Findings show that VLMs favor moderately fine-grained concepts and struggle with specificity, often misjudging texts that differ from their training data. Extensive evaluations reveal limitations in current VLMs, particularly in distinguishing between correct and subtly incorrect descriptions. While fine-tuning offers some improvements, it doesn't fully address these issues, highlighting the need for VLMs with enhanced generalization capabilities for real-world applications. This study provides insights into VLM limitations and suggests directions for developing more robust models.