Benchmarking vision-language models on OCR in dynamic video environments February 14, 2025 by kamal Comments