# Adversarial Hallucination Attack on MMLU filtered subset (16 subjects)

Source: [Adversarial Hallucination Attack on MMLU filtered subset (16 subjects)](<https://www.sota2.com/research/sota/adversarial-hallucination-attack-on-mmlu-filtered-subset-16-subjects>)

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

Data updated: 2026-05-14T07:15:12.790Z

Evaluation of model robustness against adversarial hallucination attacks using a specific 16-subject filtered subset of the MMLU benchmark.

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 | Adversarial Hallucination Attack |
| Dataset (including any stated split) | MMLU filtered subset (16 subjects) |
| Metric | ASR@30 |
| 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@30](<https://www.sota2.com/research/sota/adversarial-hallucination-attack-on-mmlu-filtered-subset-16-subjects.md?metric=ASR%4030>): Lower is better; unit: PERCENT.

- [SCE](<https://www.sota2.com/research/sota/adversarial-hallucination-attack-on-mmlu-filtered-subset-16-subjects.md?metric=SCE>): Lower is better; unit: SCALAR.

- [SEE](<https://www.sota2.com/research/sota/adversarial-hallucination-attack-on-mmlu-filtered-subset-16-subjects.md?metric=SEE>): Lower is better; unit: PERCENT.

## Recorded results

| Method | ASR@30 | Result date | Report paper |
| --- | --- | --- | --- |
| Raw prompting | 2.01 | Not recorded | 6a05571ddbc836b87d828d2f |
| ICD | 2.57 | Not recorded | 6a05571ddbc836b87d828d2f |
| Raw prompting | 4.02 | Not recorded | 6a05571ddbc836b87d828d2f |
| ICD | 6.32 | Not recorded | 6a05571ddbc836b87d828d2f |
| REALISTA | 20.72 | Not recorded | 6a05571ddbc836b87d828d2f |
| REALISTA | 23.61 | Not recorded | 6a05571ddbc836b87d828d2f |

### Raw prompting

ASR@30: 2.01. Result ID: 6a05767847f34ce924ec032d.

[Reporting paper](<https://www.sota2.com/research/paper/6a05571ddbc836b87d828d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa690d1e209131cea1fcb0.md>)

[Implementation](<https://www.sota2.com/github.com/hendrycks/test>)

Recorded evaluation configuration: {"Target Model":"GPT-5-Mini"}. This record does not assert independent reproduction.

### ICD

ASR@30: 2.57. Result ID: 6a05767847f34ce924ec032e.

[Reporting paper](<https://www.sota2.com/research/paper/6a05571ddbc836b87d828d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72721e209131cea565de.md>)

Recorded evaluation configuration: {"Target Model":"GPT-5-Mini"}. This record does not assert independent reproduction.

### Raw prompting

ASR@30: 4.02. Result ID: 6a05767847f34ce924ec032a.

[Reporting paper](<https://www.sota2.com/research/paper/6a05571ddbc836b87d828d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa690d1e209131cea1fcb0.md>)

[Implementation](<https://www.sota2.com/github.com/hendrycks/test>)

Recorded evaluation configuration: {"Target Model":"GPT-5-Nano"}. This record does not assert independent reproduction.

### ICD

ASR@30: 6.32. Result ID: 6a05767847f34ce924ec032b.

[Reporting paper](<https://www.sota2.com/research/paper/6a05571ddbc836b87d828d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72721e209131cea565de.md>)

Recorded evaluation configuration: {"Target Model":"GPT-5-Nano"}. This record does not assert independent reproduction.

### REALISTA

ASR@30: 20.72. Result ID: 6a05767847f34ce924ec032f.

[Reporting paper](<https://www.sota2.com/research/paper/6a05571ddbc836b87d828d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a05571ddbc836b87d828d2f.md>)

Recorded evaluation configuration: {"Target Model":"GPT-5-Mini"}. This record does not assert independent reproduction.

### REALISTA

ASR@30: 23.61. Result ID: 6a05767847f34ce924ec032c.

[Reporting paper](<https://www.sota2.com/research/paper/6a05571ddbc836b87d828d2f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a05571ddbc836b87d828d2f.md>)

Recorded evaluation configuration: {"Target Model":"GPT-5-Nano"}. This record does not assert independent reproduction.

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