# Video Question Answering on HD-EPIC VQA Source: [Video Question Answering on HD-EPIC VQA]() Collection index: [llms.txt]() Data updated: 2026-05-26T08:40:08.658Z Evaluation of model performance on the HD-EPIC VQA dataset for Video Question Answering, covering specialized metrics such as Recipe, Ingredient, Nutrition, Fine-grained understanding, 3D Perception, Object Motion, and Gaze tracking. 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) | HD-EPIC VQA | | Metric | Recipe Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 7 of 7 | | 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 - [Recipe Accuracy](): Higher is better; unit: PERCENT. - [Ingredient Accuracy](): Higher is better; unit: PERCENT. - [Nutrition Accuracy](): Higher is better; unit: PERCENT. - [Fine-grained Accuracy](): Higher is better; unit: PERCENT. - [3D Perception Accuracy](): Higher is better; unit: PERCENT. - [Object Motion Accuracy](): Higher is better; unit: PERCENT. - [Gaze Tracking Accuracy](): Higher is better; unit: PERCENT. - [Overall Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Recipe Accuracy | Result date | Report paper | | --- | --- | --- | --- | | EgoAdapt | 88 | Not recorded | 6a152930dbd7150506f7dfd1 | | HelloWorld | 64.75 | Not recorded | 6a152930dbd7150506f7dfd1 | | DeepFrames | 64.62 | Not recorded | 6a152930dbd7150506f7dfd1 | | Gemini Pro | 60.5 | Not recorded | 6a152930dbd7150506f7dfd1 | | LLaVA-Video | 36.25 | Not recorded | 6a152930dbd7150506f7dfd1 | | VideoLLaMA 2 | 30.75 | Not recorded | 6a152930dbd7150506f7dfd1 | | LongVA | 29.62 | Not recorded | 6a152930dbd7150506f7dfd1 | ### EgoAdapt Recipe Accuracy: 88. Result ID: 6a155c62fbc373592c0cd039. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### HelloWorld Recipe Accuracy: 64.75. Result ID: 6a155c62fbc373592c0cd037. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### DeepFrames Recipe Accuracy: 64.62. Result ID: 6a155c62fbc373592c0cd038. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Gemini Pro Recipe Accuracy: 60.5. Result ID: 6a155c62fbc373592c0cd036. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Original Challenge Result":"true"}. This record does not assert independent reproduction. ### LLaVA-Video Recipe Accuracy: 36.25. Result ID: 6a155c62fbc373592c0cd035. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Original Challenge Result":"true"}. This record does not assert independent reproduction. ### VideoLLaMA 2 Recipe Accuracy: 30.75. Result ID: 6a155c62fbc373592c0cd033. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Original Challenge Result":"true"}. This record does not assert independent reproduction. ### LongVA Recipe Accuracy: 29.62. Result ID: 6a155c62fbc373592c0cd034. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Original Challenge Result":"true"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.