Video Mistake Detection Model
The method generalizes better to unseen procedures, making it valuable for real-world applications where task variations are common.
The method generalizes better to unseen procedures, making it valuable for real-world applications where task variations are common.
用可执行真值逐帧审计视频模型计数,发现Gemini 3.6 Flash处理瞬时事件几乎失效,加帧只虚涨分数,值得看。