# Prompt Injection on CSQA (ASR, MDSR)

Source: [Prompt Injection on CSQA (ASR, MDSR)](<https://www.sota2.com/research/sota/prompt-injection-on-csqa-asr-mdsr>)

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

Data updated: 2026-05-19T07:21:14.617Z

Evaluation of model performance regarding prompt injection attacks using the CSQA dataset, specifically 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 |
| Dataset (including any stated split) | CSQA |
| Metric | ASR |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 36 of 36 |
| 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-on-csqa-asr-mdsr.md?metric=ASR>): Higher is better; unit: PERCENT.

- [MDSR](<https://www.sota2.com/research/sota/prompt-injection-on-csqa-asr-mdsr.md?metric=MDSR>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ASR | Result date | Report paper |
| --- | --- | --- | --- |
| Inspector | 62 | Not recorded | 6a0bee84dbc836b87d82c32c |
| AgentXposed | 61.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| AgentSafe | 60 | Not recorded | 6a0bee84dbc836b87d82c32c |
| AgentXposed | 59.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| No Defense | 56.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| Challenger | 56 | Not recorded | 6a0bee84dbc836b87d82c32c |
| AgentSafe | 52.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| No Defense | 52 | Not recorded | 6a0bee84dbc836b87d82c32c |
| Inspector | 50.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| No Defense | 50 | Not recorded | 6a0bee84dbc836b87d82c32c |
| Challenger | 49.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| AgentXposed | 49 | Not recorded | 6a0bee84dbc836b87d82c32c |
| AgentSafe | 48 | Not recorded | 6a0bee84dbc836b87d82c32c |
| Challenger | 46 | Not recorded | 6a0bee84dbc836b87d82c32c |
| AgentSafe | 45.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| No Defense | 44.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| Challenger | 44.4 | Not recorded | 6a0bee84dbc836b87d82c32c |
| Inspector | 43.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| AgentXposed | 43.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| Inspector | 36.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| GUARDIAN | 31.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| GUARDIAN | 31 | Not recorded | 6a0bee84dbc836b87d82c32c |
| INFA-Guard | 29.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| GUARDIAN | 29.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| INFA-Guard | 28.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| GUARDIAN | 27.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| G-Safeguard | 27.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| G-Safeguard | 27.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| G-Safeguard | 26.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| INFA-Guard | 25.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| G-Safeguard | 24.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| INFA-Guard | 24.7 | Not recorded | 6a0bee84dbc836b87d82c32c |
| PropGuard | 21 | Not recorded | 6a0bee84dbc836b87d82c32c |
| PropGuard | 18.3 | Not recorded | 6a0bee84dbc836b87d82c32c |
| PropGuard | 18 | Not recorded | 6a0bee84dbc836b87d82c32c |
| PropGuard | 16 | Not recorded | 6a0bee84dbc836b87d82c32c |

### Inspector

ASR: 62. Result ID: 6a0c0dcbfbc373592c01a633.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa78f11e209131cea7b579.md>)

Recorded evaluation configuration: {"Architecture":"Tree"}. This record does not assert independent reproduction.

### AgentXposed

ASR: 61.3. Result ID: 6a0c0dcbfbc373592c01a63a.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be61e209131cea8b2b6.md>)

Recorded evaluation configuration: {"Architecture":"Star"}. This record does not assert independent reproduction.

### AgentSafe

ASR: 60. Result ID: 6a0c0dcbfbc373592c01a639.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa79ef1e209131cea801a8.md>)

Recorded evaluation configuration: {"Architecture":"Star"}. This record does not assert independent reproduction.

### AgentXposed

ASR: 59.7. Result ID: 6a0c0dcbfbc373592c01a643.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be61e209131cea8b2b6.md>)

Recorded evaluation configuration: {"Architecture":"Random"}. This record does not assert independent reproduction.

### No Defense

ASR: 56.3. Result ID: 6a0c0dcbfbc373592c01a638.

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

Recorded evaluation configuration: {"Architecture":"Star"}. This record does not assert independent reproduction.

### Challenger

ASR: 56. Result ID: 6a0c0dcbfbc373592c01a63b.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa78f11e209131cea7b579.md>)

Recorded evaluation configuration: {"Architecture":"Star"}. This record does not assert independent reproduction.

### AgentSafe

ASR: 52.3. Result ID: 6a0c0dcbfbc373592c01a627.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa79ef1e209131cea801a8.md>)

Recorded evaluation configuration: {"Architecture":"Chain"}. This record does not assert independent reproduction.

### No Defense

ASR: 52. Result ID: 6a0c0dcbfbc373592c01a626.

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

Recorded evaluation configuration: {"Architecture":"Chain"}. This record does not assert independent reproduction.

### Inspector

ASR: 50.7. Result ID: 6a0c0dcbfbc373592c01a62a.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa78f11e209131cea7b579.md>)

Recorded evaluation configuration: {"Architecture":"Chain"}. This record does not assert independent reproduction.

### No Defense

ASR: 50. Result ID: 6a0c0dcbfbc373592c01a62f.

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

Recorded evaluation configuration: {"Architecture":"Tree"}. This record does not assert independent reproduction.

### Challenger

ASR: 49.3. Result ID: 6a0c0dcbfbc373592c01a644.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa78f11e209131cea7b579.md>)

Recorded evaluation configuration: {"Architecture":"Random"}. This record does not assert independent reproduction.

### AgentXposed

ASR: 49. Result ID: 6a0c0dcbfbc373592c01a631.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be61e209131cea8b2b6.md>)

Recorded evaluation configuration: {"Architecture":"Tree"}. This record does not assert independent reproduction.

### AgentSafe

ASR: 48. Result ID: 6a0c0dcbfbc373592c01a630.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa79ef1e209131cea801a8.md>)

Recorded evaluation configuration: {"Architecture":"Tree"}. This record does not assert independent reproduction.

### Challenger

ASR: 46. Result ID: 6a0c0dcbfbc373592c01a632.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa78f11e209131cea7b579.md>)

Recorded evaluation configuration: {"Architecture":"Tree"}. This record does not assert independent reproduction.

### AgentSafe

ASR: 45.3. Result ID: 6a0c0dcbfbc373592c01a642.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa79ef1e209131cea801a8.md>)

Recorded evaluation configuration: {"Architecture":"Random"}. This record does not assert independent reproduction.

### No Defense

ASR: 44.7. Result ID: 6a0c0dcbfbc373592c01a641.

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

Recorded evaluation configuration: {"Architecture":"Random"}. This record does not assert independent reproduction.

### Challenger

ASR: 44.4. Result ID: 6a0c0dcbfbc373592c01a629.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa78f11e209131cea7b579.md>)

Recorded evaluation configuration: {"Architecture":"Chain"}. This record does not assert independent reproduction.

### Inspector

ASR: 43.7. Result ID: 6a0c0dcbfbc373592c01a63c.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa78f11e209131cea7b579.md>)

Recorded evaluation configuration: {"Architecture":"Star"}. This record does not assert independent reproduction.

### AgentXposed

ASR: 43.3. Result ID: 6a0c0dcbfbc373592c01a628.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be61e209131cea8b2b6.md>)

Recorded evaluation configuration: {"Architecture":"Chain"}. This record does not assert independent reproduction.

### Inspector

ASR: 36.3. Result ID: 6a0c0dcbfbc373592c01a645.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa78f11e209131cea7b579.md>)

Recorded evaluation configuration: {"Architecture":"Random"}. This record does not assert independent reproduction.

### GUARDIAN

ASR: 31.3. Result ID: 6a0c0dcbfbc373592c01a62c.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c1d1e209131cea8c74f.md>)

Recorded evaluation configuration: {"Architecture":"Chain"}. This record does not assert independent reproduction.

### GUARDIAN

ASR: 31. Result ID: 6a0c0dcbfbc373592c01a647.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c1d1e209131cea8c74f.md>)

Recorded evaluation configuration: {"Architecture":"Random"}. This record does not assert independent reproduction.

### INFA-Guard

ASR: 29.7. Result ID: 6a0c0dcbfbc373592c01a62d.

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

[Method origin paper](<https://www.sota2.com/research/paper/6971af07e25ad2fa9d1ed982.md>)

Recorded evaluation configuration: {"Architecture":"Chain"}. This record does not assert independent reproduction.

### GUARDIAN

ASR: 29.3. Result ID: 6a0c0dcbfbc373592c01a635.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c1d1e209131cea8c74f.md>)

Recorded evaluation configuration: {"Architecture":"Tree"}. This record does not assert independent reproduction.

### INFA-Guard

ASR: 28.7. Result ID: 6a0c0dcbfbc373592c01a648.

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

[Method origin paper](<https://www.sota2.com/research/paper/6971af07e25ad2fa9d1ed982.md>)

Recorded evaluation configuration: {"Architecture":"Random"}. This record does not assert independent reproduction.

### GUARDIAN

ASR: 27.7. Result ID: 6a0c0dcbfbc373592c01a63e.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c1d1e209131cea8c74f.md>)

Recorded evaluation configuration: {"Architecture":"Star"}. This record does not assert independent reproduction.

### G-Safeguard

ASR: 27.3. Result ID: 6a0c0dcbfbc373592c01a634.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa76fd1e209131cea70499.md>)

[Implementation](<https://github.com/wslong20/G-safeguard>)

Recorded evaluation configuration: {"Architecture":"Tree"}. This record does not assert independent reproduction.

### G-Safeguard

ASR: 27.3. Result ID: 6a0c0dcbfbc373592c01a646.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa76fd1e209131cea70499.md>)

[Implementation](<https://github.com/wslong20/G-safeguard>)

Recorded evaluation configuration: {"Architecture":"Random"}. This record does not assert independent reproduction.

### G-Safeguard

ASR: 26.7. Result ID: 6a0c0dcbfbc373592c01a62b.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa76fd1e209131cea70499.md>)

[Implementation](<https://github.com/wslong20/G-safeguard>)

Recorded evaluation configuration: {"Architecture":"Chain"}. This record does not assert independent reproduction.

### INFA-Guard

ASR: 25.3. Result ID: 6a0c0dcbfbc373592c01a636.

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

[Method origin paper](<https://www.sota2.com/research/paper/6971af07e25ad2fa9d1ed982.md>)

Recorded evaluation configuration: {"Architecture":"Tree"}. This record does not assert independent reproduction.

### G-Safeguard

ASR: 24.7. Result ID: 6a0c0dcbfbc373592c01a63d.

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

[Method origin paper](<https://www.sota2.com/research/paper/68fa76fd1e209131cea70499.md>)

[Implementation](<https://github.com/wslong20/G-safeguard>)

Recorded evaluation configuration: {"Architecture":"Star"}. This record does not assert independent reproduction.

### INFA-Guard

ASR: 24.7. Result ID: 6a0c0dcbfbc373592c01a63f.

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

[Method origin paper](<https://www.sota2.com/research/paper/6971af07e25ad2fa9d1ed982.md>)

Recorded evaluation configuration: {"Architecture":"Star"}. This record does not assert independent reproduction.

### PropGuard

ASR: 21. Result ID: 6a0c0dcbfbc373592c01a640.

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

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

Recorded evaluation configuration: {"Architecture":"Star"}. This record does not assert independent reproduction.

### PropGuard

ASR: 18.3. Result ID: 6a0c0dcbfbc373592c01a649.

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

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

Recorded evaluation configuration: {"Architecture":"Random"}. This record does not assert independent reproduction.

### PropGuard

ASR: 18. Result ID: 6a0c0dcbfbc373592c01a637.

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

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

Recorded evaluation configuration: {"Architecture":"Tree"}. This record does not assert independent reproduction.

### PropGuard

ASR: 16. Result ID: 6a0c0dcbfbc373592c01a62e.

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

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

Recorded evaluation configuration: {"Architecture":"Chain"}. This record does not assert independent reproduction.

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