# Malicious Prompt Detection on GPT4V-Caption (IOD)

Source: [Malicious Prompt Detection on GPT4V-Caption (IOD)](<https://www.sota2.com/research/sota/malicious-prompt-detection-on-gpt4v-caption-iod>)

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

Data updated: 2026-04-09T23:29:50.426Z

Evaluation of model performance on detecting malicious prompts using the GPT4V-Caption (IOD) dataset, measured by False Positive Rate (FPR).

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 | Malicious Prompt Detection |
| Dataset (including any stated split) | GPT4V-Caption (IOD) |
| Metric | FPR |
| Unit | PERCENT |
| Direction | Lower is better |
| Results loaded | 14 of 14 |
| 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

- [FPR](<https://www.sota2.com/research/sota/malicious-prompt-detection-on-gpt4v-caption-iod.md?metric=FPR>): Lower is better; unit: PERCENT.

## Recorded results

| Method | FPR | Result date | Report paper |
| --- | --- | --- | --- |
| VLMShield | 0 | Not recorded | 69d732797d0172140ec76a8a |
| VLMShield | 0 | Not recorded | 69d732797d0172140ec76a8a |
| CIDER | 2.2 | Not recorded | 69d732797d0172140ec76a8a |
| CIDER | 2.2 | Not recorded | 69d732797d0172140ec76a8a |
| ASTRA | 2.26 | Not recorded | 69d732797d0172140ec76a8a |
| JailGuard | 2.64 | Not recorded | 69d732797d0172140ec76a8a |
| VLMGuard | 2.67 | Not recorded | 69d732797d0172140ec76a8a |
| ECSO | 3.7 | Not recorded | 69d732797d0172140ec76a8a |
| ASTRA | 3.85 | Not recorded | 69d732797d0172140ec76a8a |
| VLMGuard | 4.76 | Not recorded | 69d732797d0172140ec76a8a |
| JailGuard | 4.91 | Not recorded | 69d732797d0172140ec76a8a |
| ECSO | 6.02 | Not recorded | 69d732797d0172140ec76a8a |
| MirrorCheck | 7.94 | Not recorded | 69d732797d0172140ec76a8a |
| MirrorCheck | 7.94 | Not recorded | 69d732797d0172140ec76a8a |

### VLMShield

FPR: 0. Result ID: 69d8365e47f34ce924bff47b.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

Recorded evaluation configuration: {"Target Model":"LLaVA-1.5-13B"}. This record does not assert independent reproduction.

### VLMShield

FPR: 0. Result ID: 69d8365e47f34ce924bff482.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

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

### CIDER

FPR: 2.2. Result ID: 69d8365e47f34ce924bff478.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75411e209131cea66d6f.md>)

Recorded evaluation configuration: {"Target Model":"LLaVA-1.5-13B"}. This record does not assert independent reproduction.

### CIDER

FPR: 2.2. Result ID: 69d8365e47f34ce924bff47f.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75411e209131cea66d6f.md>)

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

### ASTRA

FPR: 2.26. Result ID: 69d8365e47f34ce924bff47d.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78541e209131cea77be4.md>)

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

### JailGuard

FPR: 2.64. Result ID: 69d8365e47f34ce924bff47e.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea7372a.md>)

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

### VLMGuard

FPR: 2.67. Result ID: 69d8365e47f34ce924bff47c.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75071e209131cea65625.md>)

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

### ECSO

FPR: 3.7. Result ID: 69d8365e47f34ce924bff481.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75321e209131cea6650e.md>)

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

### ASTRA

FPR: 3.85. Result ID: 69d8365e47f34ce924bff476.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78541e209131cea77be4.md>)

Recorded evaluation configuration: {"Target Model":"LLaVA-1.5-13B"}. This record does not assert independent reproduction.

### VLMGuard

FPR: 4.76. Result ID: 69d8365e47f34ce924bff475.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75071e209131cea65625.md>)

Recorded evaluation configuration: {"Target Model":"LLaVA-1.5-13B"}. This record does not assert independent reproduction.

### JailGuard

FPR: 4.91. Result ID: 69d8365e47f34ce924bff477.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea7372a.md>)

Recorded evaluation configuration: {"Target Model":"LLaVA-1.5-13B"}. This record does not assert independent reproduction.

### ECSO

FPR: 6.02. Result ID: 69d8365e47f34ce924bff47a.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75321e209131cea6650e.md>)

Recorded evaluation configuration: {"Target Model":"LLaVA-1.5-13B"}. This record does not assert independent reproduction.

### MirrorCheck

FPR: 7.94. Result ID: 69d8365e47f34ce924bff479.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75401e209131cea66d50.md>)

Recorded evaluation configuration: {"Target Model":"LLaVA-1.5-13B"}. This record does not assert independent reproduction.

### MirrorCheck

FPR: 7.94. Result ID: 69d8365e47f34ce924bff480.

[Reporting paper](<https://www.sota2.com/research/paper/69d732797d0172140ec76a8a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75401e209131cea66d50.md>)

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

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