# Long-form Video Question Answering on LongVideoBench (test) Source: [Long-form Video Question Answering on LongVideoBench (test)]() Collection index: [llms.txt]() Data updated: 2026-05-12T16:44:24.827Z Evaluation of model performance on the test split of the LongVideoBench dataset, specifically focusing on long-form video question answering capabilities and computational efficiency. 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 | Long-form Video Question Answering | | Dataset (including any stated split) | LongVideoBench (test) | | Metric | Long Video Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 9 of 9 | | 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 - [Long Video Accuracy](): Higher is better; unit: PERCENT. - [Overall Accuracy](): Higher is better; unit: PERCENT. - [Computational Cost (TFLOPs)](): Lower is better; unit: SCALAR. ## Recorded results | Method | Long Video Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Qwen3-VL-4B | 57.3 | Not recorded | 6a02b4addbc836b87d824f22 | | GP-4B | 55.9 | Not recorded | 6a02b4addbc836b87d824f22 | | Qwen3-VL-8B | 55.1 | Not recorded | 6a02b4addbc836b87d824f22 | | GP-2B-\>4B | 54.3 | Not recorded | 6a02b4addbc836b87d824f22 | | GP-8B | 53.4 | Not recorded | 6a02b4addbc836b87d824f22 | | GP-2B-\>8B | 52 | Not recorded | 6a02b4addbc836b87d824f22 | | GP-2B | 51.4 | Not recorded | 6a02b4addbc836b87d824f22 | | Qwen3-VL-2B | 49.8 | Not recorded | 6a02b4addbc836b87d824f22 | | Uniform-Meff -\> 8B | 49.8 | Not recorded | 6a02b4addbc836b87d824f22 | ### Qwen3-VL-4B Long Video Accuracy: 57.3. Result ID: 6a0358cf47f34ce924e8e6f5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Pipeline":"Monolithic Baselines","K (Frame Pool Size)":"144","Model Size":"4B"}. This record does not assert independent reproduction. ### GP-4B Long Video Accuracy: 55.9. Result ID: 6a0358cf47f34ce924e8e6f8. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Pipeline":"Single-Model GridProbe","Selector Model":"Qwen3-VL-4B","QA Model":"Qwen3-VL-4B","K (Frame Pool Size)":"144"}. This record does not assert independent reproduction. ### Qwen3-VL-8B Long Video Accuracy: 55.1. Result ID: 6a0358cf47f34ce924e8e6f6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Pipeline":"Monolithic Baselines","K (Frame Pool Size)":"144","Model Size":"8B"}. This record does not assert independent reproduction. ### GP-2B-\>4B Long Video Accuracy: 54.3. Result ID: 6a0358cf47f34ce924e8e6fb. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Pipeline":"Cross-Model GridProbe","Selector Model":"Qwen3-VL-2B","QA Model":"Qwen3-VL-4B","K (Frame Pool Size)":"144"}. This record does not assert independent reproduction. ### GP-8B Long Video Accuracy: 53.4. Result ID: 6a0358cf47f34ce924e8e6f9. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Pipeline":"Single-Model GridProbe","Selector Model":"Qwen3-VL-8B","QA Model":"Qwen3-VL-8B","K (Frame Pool Size)":"144"}. This record does not assert independent reproduction. ### GP-2B-\>8B Long Video Accuracy: 52. Result ID: 6a0358cf47f34ce924e8e6fc. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Pipeline":"Cross-Model GridProbe","Selector Model":"Qwen3-VL-2B","QA Model":"Qwen3-VL-8B","K (Frame Pool Size)":"144"}. This record does not assert independent reproduction. ### GP-2B Long Video Accuracy: 51.4. Result ID: 6a0358cf47f34ce924e8e6f7. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Pipeline":"Single-Model GridProbe","Selector Model":"Qwen3-VL-2B","QA Model":"Qwen3-VL-2B","K (Frame Pool Size)":"144"}. This record does not assert independent reproduction. ### Qwen3-VL-2B Long Video Accuracy: 49.8. Result ID: 6a0358cf47f34ce924e8e6f4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Pipeline":"Monolithic Baselines","K (Frame Pool Size)":"144","Model Size":"2B"}. This record does not assert independent reproduction. ### Uniform-Meff -\> 8B Long Video Accuracy: 49.8. Result ID: 6a0358cf47f34ce924e8e6fa. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Pipeline":"Cross-Model GridProbe (Control)","Selection":"Uniform","QA Model":"Qwen3-VL-8B","K (Frame Pool Size)":"144"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.