# Science Question Answering on ScienceQA (Image Acc.)

Source: [Science Question Answering on ScienceQA (Image Acc.)](<https://www.sota2.com/research/sota/science-question-answering-on-scienceqa-image-acc>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-25T07:07:05.208Z

Evaluation of model performance on the ScienceQA dataset for Science Question Answering, specifically reporting Image 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) | ScienceQA |
| Metric | Image Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 26 of 26 |
| 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

- [Image Accuracy](<https://www.sota2.com/research/sota/science-question-answering-on-scienceqa-image-acc.md?metric=Image%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Image Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Qwen2.5-VL-7B-Instruct | 88.6 | Not recorded | 68fa7a321e209131cea81adb |
| SENTINEL | 88.5 | Not recorded | 68fa7a321e209131cea81adb |
| SENTINEL | 86.9 | Not recorded | 68fa7a321e209131cea81adb |
| Qwen2-VL-7B-Instruct | 85.7 | Not recorded | 68fa7a321e209131cea81adb |
| SENTINEL | 78 | Not recorded | 68fa7a321e209131cea81adb |
| LLaVA-v1.6-vicuna-13B | 77.7 | Not recorded | 68fa7a321e209131cea81adb |
| SENTINEL | 77.4 | Not recorded | 68fa7a321e209131cea81adb |
| Qwen2-VL-2B-Instruct | 76.9 | Not recorded | 68fa7a321e209131cea81adb |
| LLaVA-v1.6-vicuna-7B | 74.3 | Not recorded | 68fa7a321e209131cea81adb |
| SENTINEL | 74.2 | Not recorded | 68fa7a321e209131cea81adb |
| SENTINEL | 72.8 | Not recorded | 68fa7a321e209131cea81adb |
| VCD | 72 | Not recorded | 68fa7a321e209131cea81adb |
| vanilla-DPO | 71.8 | Not recorded | 68fa7a321e209131cea81adb |
| baseline | 71.6 | Not recorded | 68fa7a321e209131cea81adb |
| HSA-DPO | 71.3 | Not recorded | 68fa7a321e209131cea81adb |
| HA-DPO | 69.7 | Not recorded | 68fa7a321e209131cea81adb |
| SENTINEL | 69.2 | Not recorded | 68fa7a321e209131cea81adb |
| POVID | 68.8 | Not recorded | 68fa7a321e209131cea81adb |
| VCD | 68.7 | Not recorded | 68fa7a321e209131cea81adb |
| OPERA | 68.2 | Not recorded | 68fa7a321e209131cea81adb |
| RLAIF-V | 68.2 | Not recorded | 68fa7a321e209131cea81adb |
| CLIP-DPO | 67.6 | Not recorded | 68fa7a321e209131cea81adb |
| DoLa | 67.5 | Not recorded | 68fa7a321e209131cea81adb |
| TPO | 67.1 | Not recorded | 68fa7a321e209131cea81adb |
| baseline | 66.8 | Not recorded | 68fa7a321e209131cea81adb |
| EFUF | 66.4 | Not recorded | 68fa7a321e209131cea81adb |

### Qwen2.5-VL-7B-Instruct

Image Accuracy: 88.6. Result ID: 6a13f49cfbc373592c0aa061.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Mitigation":"None (baseline)"}. This record does not assert independent reproduction.

### SENTINEL

Image Accuracy: 88.5. Result ID: 6a13f49cfbc373592c0aa062.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

Recorded evaluation configuration: {"Base Model":"Qwen2.5-VL-7B-Instruct"}. This record does not assert independent reproduction.

### SENTINEL

Image Accuracy: 86.9. Result ID: 6a13f49cfbc373592c0aa060.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

Recorded evaluation configuration: {"Base Model":"Qwen2-VL-7B-Instruct"}. This record does not assert independent reproduction.

### Qwen2-VL-7B-Instruct

Image Accuracy: 85.7. Result ID: 6a13f49cfbc373592c0aa05f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Mitigation":"None (baseline)"}. This record does not assert independent reproduction.

### SENTINEL

Image Accuracy: 78. Result ID: 6a13f49cfbc373592c0aa05c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

Recorded evaluation configuration: {"Base Model":"LLaVA-v1.6-vicuna-13B"}. This record does not assert independent reproduction.

### LLaVA-v1.6-vicuna-13B

Image Accuracy: 77.7. Result ID: 6a13f49cfbc373592c0aa05b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

Recorded evaluation configuration: {"Mitigation":"None (baseline)"}. This record does not assert independent reproduction.

### SENTINEL

Image Accuracy: 77.4. Result ID: 6a13f49cfbc373592c0aa05e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

Recorded evaluation configuration: {"Base Model":"Qwen2-VL-2B-Instruct"}. This record does not assert independent reproduction.

### Qwen2-VL-2B-Instruct

Image Accuracy: 76.9. Result ID: 6a13f49cfbc373592c0aa05d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Mitigation":"None (baseline)"}. This record does not assert independent reproduction.

### LLaVA-v1.6-vicuna-7B

Image Accuracy: 74.3. Result ID: 6a13f49cfbc373592c0aa059.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

Recorded evaluation configuration: {"Mitigation":"None (baseline)"}. This record does not assert independent reproduction.

### SENTINEL

Image Accuracy: 74.2. Result ID: 6a13f49cfbc373592c0aa05a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

Recorded evaluation configuration: {"Base Model":"LLaVA-v1.6-vicuna-7B"}. This record does not assert independent reproduction.

### SENTINEL

Image Accuracy: 72.8. Result ID: 6a13f49cfbc373592c0aa0c5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-13B"}. This record does not assert independent reproduction.

### VCD

Image Accuracy: 72. Result ID: 6a13f49cfbc373592c0aa0c2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71461e209131cea5053d.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-13B"}. This record does not assert independent reproduction.

### vanilla-DPO

Image Accuracy: 71.8. Result ID: 6a13f49cfbc373592c0aa0c3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa767b1e209131cea6d52a.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-13B"}. This record does not assert independent reproduction.

### baseline

Image Accuracy: 71.6. Result ID: 6a13f49cfbc373592c0aa0c1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa732e1e209131cea5af59.md>)

[Implementation](<https://llava-vl.github.io/>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-13B"}. This record does not assert independent reproduction.

### HSA-DPO

Image Accuracy: 71.3. Result ID: 6a13f49cfbc373592c0aa0c4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa767b1e209131cea6d52a.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-13B"}. This record does not assert independent reproduction.

### HA-DPO

Image Accuracy: 69.7. Result ID: 6a13f49cfbc373592c0aa0bb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72111e209131cea544b5.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

### SENTINEL

Image Accuracy: 69.2. Result ID: 6a13f49cfbc373592c0aa0c0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

### POVID

Image Accuracy: 68.8. Result ID: 6a13f49cfbc373592c0aa0bc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa723a1e209131cea55174.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

### VCD

Image Accuracy: 68.7. Result ID: 6a13f49cfbc373592c0aa0b7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71461e209131cea5053d.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

### OPERA

Image Accuracy: 68.2. Result ID: 6a13f49cfbc373592c0aa0b8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72761e209131cea56869.md>)

[Implementation](<https://www.sota2.com/This%20link>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

### RLAIF-V

Image Accuracy: 68.2. Result ID: 6a13f49cfbc373592c0aa0be.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76731e209131cea6d104.md>)

[Implementation](<https://github.com/RLHF-V/RLAIF-V>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

### CLIP-DPO

Image Accuracy: 67.6. Result ID: 6a13f49cfbc373592c0aa0bd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74831e209131cea62173.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

### DoLa

Image Accuracy: 67.5. Result ID: 6a13f49cfbc373592c0aa0b9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72711e209131cea56577.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

### TPO

Image Accuracy: 67.1. Result ID: 6a13f49cfbc373592c0aa0bf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79941e209131cea7e625.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

### baseline

Image Accuracy: 66.8. Result ID: 6a13f49cfbc373592c0aa0b6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa732e1e209131cea5af59.md>)

[Implementation](<https://llava-vl.github.io/>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

### EFUF

Image Accuracy: 66.4. Result ID: 6a13f49cfbc373592c0aa0ba.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a321e209131cea81adb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74e71e209131cea64983.md>)

Recorded evaluation configuration: {"Model":"LLaVA-v1.5-7B"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
