# Video Question Answering on INST-IT Bench Video Source: [Video Question Answering on INST-IT Bench Video]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:49:44.107Z Evaluation of model performance on the INST-IT Bench Video dataset for Video Question Answering capabilities, reporting metrics for Open-Ended (OE Q&A) and Multiple-Choice (MC Q&A) questions. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Video Question Answering | | Dataset (including any stated split) | INST-IT Bench Video | | Metric | MC Q&A Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 12 of 12 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [MC Q&A Accuracy](): Higher is better; unit: PERCENT. - [OE Q&A Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | MC Q&A Accuracy | Result date | Report paper | | --- | --- | --- | --- | | GPT-4o | 81 | Not recorded | 68fa76001e209131cea6ad69 | | Gemini-1.5-flash | 75.8 | Not recorded | 68fa76001e209131cea6ad69 | | Qwen2-VL-Instruct | 74.6 | Not recorded | 68fa76001e209131cea6ad69 | | LLaVA-NeXT-INST-IT | 53.3 | Not recorded | 68fa76001e209131cea6ad69 | | LLaVA-Video | 53.2 | Not recorded | 68fa76001e209131cea6ad69 | | LLaVA-OV | 45.6 | Not recorded | 68fa76001e209131cea6ad69 | | InternVL2 | 45.5 | Not recorded | 68fa76001e209131cea6ad69 | | LLaVA-NeXT-INST-IT | 42.1 | Not recorded | 68fa76001e209131cea6ad69 | | LLaVA-OV (SI) | 36.4 | Not recorded | 68fa76001e209131cea6ad69 | | Random Guess | 25 | Not recorded | 68fa76001e209131cea6ad69 | | LLaVA-NeXT-Video | 24.8 | Not recorded | 68fa76001e209131cea6ad69 | | ShareGPT4Video | 16.1 | Not recorded | 68fa76001e209131cea6ad69 | ### GPT-4o MC Q&A Accuracy: 81. Result ID: 699409883a6c3148491a93dd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#IT":"N/A"}. This record does not assert independent reproduction. ### Gemini-1.5-flash MC Q&A Accuracy: 75.8. Result ID: 699409883a6c3148491a93de. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"#IT":"N/A"}. This record does not assert independent reproduction. ### Qwen2-VL-Instruct MC Q&A Accuracy: 74.6. Result ID: 699409883a6c3148491a93e5. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Qwen2-72B","#IT":"N/A"}. This record does not assert independent reproduction. ### LLaVA-NeXT-INST-IT MC Q&A Accuracy: 53.3. Result ID: 699409883a6c3148491a93e7. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Qwen2-7B","#IT":"920K"}. This record does not assert independent reproduction. ### LLaVA-Video MC Q&A Accuracy: 53.2. Result ID: 699409883a6c3148491a93e3. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Qwen2-7B","#IT":"~7.4M"}. This record does not assert independent reproduction. ### LLaVA-OV MC Q&A Accuracy: 45.6. Result ID: 699409883a6c3148491a93e2. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Qwen2-7B","#IT":"~8.8M"}. This record does not assert independent reproduction. ### InternVL2 MC Q&A Accuracy: 45.5. Result ID: 699409883a6c3148491a93e4. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"InternLM2.5-7B","#IT":"N/A"}. This record does not assert independent reproduction. ### LLaVA-NeXT-INST-IT MC Q&A Accuracy: 42.1. Result ID: 699409883a6c3148491a93e6. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Vicuna-7B","#IT":"920K"}. This record does not assert independent reproduction. ### LLaVA-OV (SI) MC Q&A Accuracy: 36.4. Result ID: 699409883a6c3148491a93e1. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Qwen2-7B","#IT":"~7.2M"}. This record does not assert independent reproduction. ### Random Guess MC Q&A Accuracy: 25. Result ID: 699409883a6c3148491a93dc. [Reporting paper]() Recorded evaluation configuration: {"#IT":"N/A"}. This record does not assert independent reproduction. ### LLaVA-NeXT-Video MC Q&A Accuracy: 24.8. Result ID: 699409883a6c3148491a93df. [Reporting paper]() Recorded evaluation configuration: {"LLM":"Vicuna-7B","#IT":"860K"}. This record does not assert independent reproduction. ### ShareGPT4Video MC Q&A Accuracy: 16.1. Result ID: 699409883a6c3148491a93e0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"LLM":"Llama3-8B","#IT":"~1.0M"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.