NEW BENCHMARK FOR VIDEO LANGUAGE MODELS

Summary from the TLDR AI newsletter

E.T. Bench is a new benchmark designed to evaluate video language models on fine-grained, event-level tasks. Unlike previous benchmarks that focus on video-level questions, E.T. Bench covers a range of time-sensitive tasks across multiple domains.