René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Request for Official MVBench Evaluation Protocol for InternVideo2_5_Chat_8B","articleBody":"Hi OpenGVLab Team,\n\nThank you for releasing InternVideo2.5.\n\nI am trying to reproduce the **official MVBench score of 75.7** reported for *OpenGVLab/InternVideo2_5_Chat_8B*.\nI am using the HuggingFace reference implementation as a starting point:\n\n[https://huggingface.co/OpenGVLab/InternVideo2_5_Chat_8B#%F0%9F%9A%80-how-to-use-the-model](https://huggingface.co/OpenGVLab/InternVideo2_5_Chat_8B#%F0%9F%9A%80-how-to-use-the-model)\n\nTo ensure my evaluation matches yours, I would appreciate clarification on the exact protocol, including:\n\n* the number of frames and the frame-sampling strategy,\n* preprocessing details (input resolution, dynamic tiling, thumbnail usage),\n* the prompt template used for multiple-choice questions,\n* decoding settings and the rule for extracting the predicted option.\n\nMost importantly, could you share the **official evaluation script** or configuration used to obtain the reported **75.7** score?\nIf available, reference evaluation settings for other datasets (LongVideoBench, VideoMME, EgoSchema, Perception Test, LVBench) would also be very helpful.\n\nThank you in advance!\n\n~Sergey\n","author":{"url":"https://github.com/sergeyprokudin","@type":"Person","name":"sergeyprokudin"},"datePublished":"2025-11-25T16:52:59.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":1},"url":"https://github.com/310/InternVideo/issues/310"}