# Prompt Injection Defense on MATH random architecture

Source: [Prompt Injection Defense on MATH random architecture](<https://www.sota2.com/research/sota/prompt-injection-defense-on-math-random-architecture>)

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

Data updated: 2026-05-19T07:21:14.486Z

Evaluation of model performance regarding prompt injection defense mechanisms using the MATH dataset configured with a random architecture, measured by Attack Success Rate (ASR) and Misdirection Success Rate (MDSR).

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 | Prompt Injection Defense |
| Dataset (including any stated split) | MATH random architecture |
| Metric | ASR |
| Unit | PERCENT |
| Direction | Lower is better |
| Results loaded | 6 of 6 |
| 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

- [ASR](<https://www.sota2.com/research/sota/prompt-injection-defense-on-math-random-architecture.md?metric=ASR>): Lower is better; unit: PERCENT.

- [MDSR](<https://www.sota2.com/research/sota/prompt-injection-defense-on-math-random-architecture.md?metric=MDSR>): Lower is better; unit: PERCENT.

## Recorded results

| Method | ASR | Result date | Report paper |
| --- | --- | --- | --- |
| PropGuard | 6 | Not recorded | 6a0bee84dbc836b87d82c32c |
| Qwen3Guard | 15.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| ThinkGuard | 16.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| WildGuard | 22.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| LlamaGuard | 23.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| No Defense | 26.7 | Not recorded | 6a0bee84dbc836b87d82c32c |

### PropGuard

ASR: 6. Result ID: 6a0c0dcbfbc373592c01a619.

[Reporting paper](<https://www.sota2.com/research/paper/6a0bee84dbc836b87d82c32c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0bee84dbc836b87d82c32c.md>)

Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Communication Architecture":"random"}. This record does not assert independent reproduction.

### Qwen3Guard

ASR: 15.3. Result ID: 6a0c0dcbfbc373592c01a618.

[Reporting paper](<https://www.sota2.com/research/paper/6a0bee84dbc836b87d82c32c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c351e209131cea8d09c.md>)

Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Communication Architecture":"random"}. This record does not assert independent reproduction.

### ThinkGuard

ASR: 16.7. Result ID: 6a0c0dcbfbc373592c01a617.

[Reporting paper](<https://www.sota2.com/research/paper/6a0bee84dbc836b87d82c32c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78d41e209131cea7aa09.md>)

Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Communication Architecture":"random"}. This record does not assert independent reproduction.

### WildGuard

ASR: 22.3. Result ID: 6a0c0dcbfbc373592c01a616.

[Reporting paper](<https://www.sota2.com/research/paper/6a0bee84dbc836b87d82c32c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76291e209131cea6b9a6.md>)

Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Communication Architecture":"random"}. This record does not assert independent reproduction.

### LlamaGuard

ASR: 23.7. Result ID: 6a0c0dcbfbc373592c01a615.

[Reporting paper](<https://www.sota2.com/research/paper/6a0bee84dbc836b87d82c32c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa717b1e209131cea5186d.md>)

Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Communication Architecture":"random"}. This record does not assert independent reproduction.

### No Defense

ASR: 26.7. Result ID: 6a0c0dcbfbc373592c01a614.

[Reporting paper](<https://www.sota2.com/research/paper/6a0bee84dbc836b87d82c32c.md>)

Recorded evaluation configuration: {"Base Model":"GPT-4o-mini","Communication Architecture":"random"}. This record does not assert independent reproduction.

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