{"type":"post","id":"gemini-3-pro-tops-standard-qa-at-just-fb5cf3","slug":"gemini-3-pro-tops-standard-qa-at-just-fb5cf3","title":"Gemini-3-Pro tops standard QA at just 17% accuracy, exposing benchmark saturation elsewhere as illusory.","description":"VideoZeroBench evaluates long-video multimodal models on answer correctness and on their ability to locate supporting evidence in time and space.","retrieval_nugget":"VideoZeroBench evaluates long-video multimodal models on answer correctness and on their ability to locate supporting evidence in time and space. The paper reports a best standard QA score of 24.8% across 19 models, while no model exceeds 1.8% when accurate spatio-temporal grounding is required too. For AI builders, that gap makes answer-only video evaluation look insufficient.","published_at":"2026-11-07","updated_at":"2026-11-07","record_date":"2026-11-07","date_kind":"published_at","topics":["models"],"entities":["VideoZeroBench","Gemini-3.7-Flash","arXiv"],"related_patterns":["verification-bandwidth-is-the-scarce-resource"],"novelty_label":"new_fact","what_changed":"VideoZeroBench finds Gemini 3.7 Flash leads standard video-question answering at 24.8%, while no evaluated model exceeds 1.8% on answers requiring accurate spatio-temporal localization.","prior_refs":[],"source_url":"https://arxiv.org/abs/2604.01569","source_urls":["https://arxiv.org/abs/2604.01569"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:gemini-3-pro-tops-standard-qa-at-just-fb5cf3","status":"published","visuals":[],"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/gemini-3-pro-tops-standard-qa-at-just-fb5cf3/","markdown":"https://newruntime.com/posts/gemini-3-pro-tops-standard-qa-at-just-fb5cf3.md","json":"https://newruntime.com/posts/gemini-3-pro-tops-standard-qa-at-just-fb5cf3.json"},"markdown_url":"https://newruntime.com/posts/gemini-3-pro-tops-standard-qa-at-just-fb5cf3.md","next_reads":[{"type":"topic","path":"/topics/models/","url":"https://newruntime.com/topics/models/","title":"Models","reason":"Explore the models topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/models.md","json_url":"https://newruntime.com/topics/models.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/sources/","url":"https://newruntime.com/sources/","title":"Sources","reason":"See the original sources this archive cites.","media_type":"text/html"}]}