{"paper":{"title":"Video-ChatGPT: Towards Detailed Video Understanding via Large Vision and Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Video-ChatGPT combines a video-adapted visual encoder with a large language model to support detailed conversations about video content.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fahad Shahbaz Khan, Hanoona Rasheed, Muhammad Maaz, Salman Khan","submitted_at":"2023-06-08T17:59:56Z","abstract_excerpt":"Conversation agents fueled by Large Language Models (LLMs) are providing a new way to interact with visual data. While there have been initial attempts for image-based conversation models, this work addresses the under-explored field of \\emph{video-based conversation} by introducing Video-ChatGPT. It is a multimodal model that merges a video-adapted visual encoder with an LLM. The resulting model is capable of understanding and generating detailed conversations about videos. We introduce a new dataset of 100,000 video-instruction pairs used to train Video-ChatGPT acquired via manual and semi-a"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The resulting model is capable of understanding and generating detailed conversations about videos.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The semi-automated pipeline for creating the 100,000 video-instruction pairs produces sufficiently clean training data without label noise that would degrade the model's ability to generate accurate conversations.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Video-ChatGPT is a multimodal model that combines a video visual encoder with an LLM to understand and generate conversations about videos, trained on a new dataset of 100,000 video-instruction pairs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Video-ChatGPT combines a video-adapted visual encoder with a large language model to support detailed conversations about video content.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"533f3463748ac8910ef18fcb487e600635e42d9003b546018bbba6b1a4d37d38"},"source":{"id":"2306.05424","kind":"arxiv","version":2},"verdict":{"id":"a1200b5c-d98e-43aa-ac40-21ef3e152a62","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T03:29:56.240351Z","strongest_claim":"The resulting model is capable of understanding and generating detailed conversations about videos.","one_line_summary":"Video-ChatGPT is a multimodal model that combines a video visual encoder with an LLM to understand and generate conversations about videos, trained on a new dataset of 100,000 video-instruction pairs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The semi-automated pipeline for creating the 100,000 video-instruction pairs produces sufficiently clean training data without label noise that would degrade the model's ability to generate accurate conversations.","pith_extraction_headline":"Video-ChatGPT combines a video-adapted visual encoder with a large language model to support detailed conversations about video content."},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":3,"snapshot_sha256":"a45098cacb1b8d38046a9f62e97c2526b9000811dbcbe5618d22fde14339f022"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}