# Science Question Answering on SQA-I (Accuracy) Source: [Science Question Answering on SQA-I (Accuracy)]() Collection index: [llms.txt]() Data updated: 2026-05-27T09:10:48.034Z Evaluation of model performance on the SQA-I subset for the Science Question Answering task, measured specifically by Accuracy. 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 | Science Question Answering | | Dataset (including any stated split) | SQA-I | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 13 of 13 | | 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 - [Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | OFA | 67.9 | Not recorded | 6a167a85dbd7150506f7ff6b | | PreSel | 66.2 | Not recorded | 6a167a85dbd7150506f7ff6b | | CoIDO | 66.2 | Not recorded | 6a167a85dbd7150506f7ff6b | | COINCIDE | 63.9 | Not recorded | 6a167a85dbd7150506f7ff6b | | EL2N | 62.7 | Not recorded | 6a167a85dbd7150506f7ff6b | | CLIP-Score | 62.3 | Not recorded | 6a167a85dbd7150506f7ff6b | | Random | 62.1 | Not recorded | 6a167a85dbd7150506f7ff6b | | ICONS | 60.2 | Not recorded | 6a167a85dbd7150506f7ff6b | | Self-Filter | 59.3 | Not recorded | 6a167a85dbd7150506f7ff6b | | TypiClust | 59.2 | Not recorded | 6a167a85dbd7150506f7ff6b | | XMAS | 58.1 | Not recorded | 6a167a85dbd7150506f7ff6b | | IFD | 57.8 | Not recorded | 6a167a85dbd7150506f7ff6b | | Full-Data | 55.6 | Not recorded | 6a167a85dbd7150506f7ff6b | ### OFA Accuracy: 67.9. Result ID: 6a16b3fdfbc373592c100c2b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### PreSel Accuracy: 66.2. Result ID: 6a16b3fdfbc373592c100c25. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### CoIDO Accuracy: 66.2. Result ID: 6a16b3fdfbc373592c100c29. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### COINCIDE Accuracy: 63.9. Result ID: 6a16b3fdfbc373592c100c27. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### EL2N Accuracy: 62.7. Result ID: 6a16b3fdfbc373592c100c22. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### CLIP-Score Accuracy: 62.3. Result ID: 6a16b3fdfbc373592c100c28. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### Random Accuracy: 62.1. Result ID: 6a16b3fdfbc373592c100c20. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### ICONS Accuracy: 60.2. Result ID: 6a16b3fdfbc373592c100c2a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### Self-Filter Accuracy: 59.3. Result ID: 6a16b3fdfbc373592c100c21. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### TypiClust Accuracy: 59.2. Result ID: 6a16b3fdfbc373592c100c23. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### XMAS Accuracy: 58.1. Result ID: 6a16b3fdfbc373592c100c26. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### IFD Accuracy: 57.8. Result ID: 6a16b3fdfbc373592c100c24. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"15%"}. This record does not assert independent reproduction. ### Full-Data Accuracy: 55.6. Result ID: 6a16b3fdfbc373592c100c1f. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.5-7B","Fine-tuning Dataset":"Vision-Flan-186K","Sampling Ratio":"100%"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.