# Ultrasound Question Answering on U2-Bench disease-diagnosis

Source: [Ultrasound Question Answering on U2-Bench disease-diagnosis](<https://www.sota2.com/research/sota/ultrasound-question-answering-on-u2-bench-disease-diagnosis>)

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

Data updated: 2026-03-24T14:36:17.312Z

Evaluation of model performance on the U2-Bench disease-diagnosis dataset for Ultrasound Question Answering, reporting Accuracy metrics.

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 | Ultrasound Question Answering |
| Dataset (including any stated split) | U2-Bench disease-diagnosis |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 16 of 16 |
| 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](<https://www.sota2.com/research/sota/ultrasound-question-answering-on-u2-bench-disease-diagnosis.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [Accuracy (PPL \< 15)](<https://www.sota2.com/research/sota/ultrasound-question-answering-on-u2-bench-disease-diagnosis.md?metric=Accuracy%20(PPL%20%3C%2015)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| MedGemma | 42.22 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| QoQ-Med | 41.46 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| Average Med-VLM Baseline | 39.49 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLaVA-Med | 34.79 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 27.82 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 25.82 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 24.6 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 23.83 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 20.69 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 20.46 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 20.08 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 17.47 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 14.94 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 13.72 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 13.41 | Not recorded | 69c21ab80636e5bf4afc4af3 |
| LLM-driven prompt-edit attack with MCTS | 11.57 | Not recorded | 69c21ab80636e5bf4afc4af3 |

### MedGemma

Accuracy: 42.22. Result ID: 69c2a15547f34ce924ab560f.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79ff1e209131cea808a4.md>)

Recorded evaluation configuration: {"Condition":"Pre-attack"}. This record does not assert independent reproduction.

### QoQ-Med

Accuracy: 41.46. Result ID: 69c2a15547f34ce924ab5611.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c701e209131cea8e12d.md>)

Recorded evaluation configuration: {"Condition":"Pre-attack"}. This record does not assert independent reproduction.

### Average Med-VLM Baseline

Accuracy: 39.49. Result ID: 69c2a15547f34ce924ab5612.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Condition":"Pre-attack"}. This record does not assert independent reproduction.

### LLaVA-Med

Accuracy: 34.79. Result ID: 69c2a15547f34ce924ab5610.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76731e209131cea6d0fb.md>)

Recorded evaluation configuration: {"Condition":"Pre-attack"}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 27.82. Result ID: 69c2a15547f34ce924ab561d.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"GPT-4.1 mini","Target Med-VLM":"QoQ-Med"}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 25.82. Result ID: 69c2a15547f34ce924ab561b.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"GPT-4.1 mini","Target Med-VLM":"MedGemma"}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 24.6. Result ID: 69c2a15547f34ce924ab561e.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"GPT-4.1 mini","Target Med-VLM":"Avg."}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 23.83. Result ID: 69c2a15547f34ce924ab5619.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"Qwen-30B","Target Med-VLM":"QoQ-Med"}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 20.69. Result ID: 69c2a15547f34ce924ab561c.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"GPT-4.1 mini","Target Med-VLM":"LLaVA-Med"}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 20.46. Result ID: 69c2a15547f34ce924ab561a.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"Qwen-30B","Target Med-VLM":"Avg."}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 20.08. Result ID: 69c2a15547f34ce924ab5617.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"Qwen-30B","Target Med-VLM":"MedGemma"}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 17.47. Result ID: 69c2a15547f34ce924ab5618.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"Qwen-30B","Target Med-VLM":"LLaVA-Med"}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 14.94. Result ID: 69c2a15547f34ce924ab5615.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"Qwen-7B","Target Med-VLM":"QoQ-Med"}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 13.72. Result ID: 69c2a15547f34ce924ab5613.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"Qwen-7B","Target Med-VLM":"MedGemma"}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 13.41. Result ID: 69c2a15547f34ce924ab5616.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"Qwen-7B","Target Med-VLM":"Avg."}. This record does not assert independent reproduction.

### LLM-driven prompt-edit attack with MCTS

Accuracy: 11.57. Result ID: 69c2a15547f34ce924ab5614.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ab80636e5bf4afc4af3.md>)

Recorded evaluation configuration: {"Attacker LLM":"Qwen-7B","Target Med-VLM":"LLaVA-Med"}. This record does not assert independent reproduction.

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