# AI-Generated Text Detection on mixed-source AI -\> Human GPT-2, GPT-Neo, GPT-J, LLaMa, GPT-3

Source: [AI-Generated Text Detection on mixed-source AI -\> Human GPT-2, GPT-Neo, GPT-J, LLaMa, GPT-3](<https://www.sota2.com/research/sota/ai-generated-text-detection-on-mixed-source-ai-human-gpt-2-gpt-neo-gpt-j-llama-gpt-3>)

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

Data updated: 2026-02-26T05:54:24.821Z

Evaluation of AI-generated text detection models using a mixed dataset containing text generated by GPT-2, GPT-Neo, GPT-J, LLaMa, and GPT-3, compared against human text. Metrics include OA, AUA, ASR, and ANQ.

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 | AI-Generated Text Detection |
| Dataset (including any stated split) | mixed-source AI -\> Human GPT-2, GPT-Neo, GPT-J, LLaMa, GPT-3 |
| Metric | Overall Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 26 of 26 |
| 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

- [Overall Accuracy](<https://www.sota2.com/research/sota/ai-generated-text-detection-on-mixed-source-ai-human-gpt-2-gpt-neo-gpt-j-llama-gpt-3.md?metric=Overall%20Accuracy>): Higher is better; unit: PERCENT.

- [AUA Score](<https://www.sota2.com/research/sota/ai-generated-text-detection-on-mixed-source-ai-human-gpt-2-gpt-neo-gpt-j-llama-gpt-3.md?metric=AUA%20Score>): Higher is better; unit: SCALAR.

- [ASR Rate](<https://www.sota2.com/research/sota/ai-generated-text-detection-on-mixed-source-ai-human-gpt-2-gpt-neo-gpt-j-llama-gpt-3.md?metric=ASR%20Rate>): Higher is better; unit: PERCENT.

- [ANQ Score](<https://www.sota2.com/research/sota/ai-generated-text-detection-on-mixed-source-ai-human-gpt-2-gpt-neo-gpt-j-llama-gpt-3.md?metric=ANQ%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Overall Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| SeqXGPT | 96.5 | Not recorded | 68fa73e41e209131cea5eb76 |
| RDrop | 96.5 | Not recorded | 68fa73e41e209131cea5eb76 |
| SeqXGPT | 96.25 | Not recorded | 68fa73e41e209131cea5eb76 |
| Flooding | 96 | Not recorded | 68fa73e41e209131cea5eb76 |
| DeBERTa | 95.75 | Not recorded | 68fa73e41e209131cea5eb76 |
| Flooding | 95.75 | Not recorded | 68fa73e41e209131cea5eb76 |
| RDrop | 95.75 | Not recorded | 68fa73e41e209131cea5eb76 |
| RoBERTa | 95.5 | Not recorded | 68fa73e41e209131cea5eb76 |
| DeBERTa | 95.5 | Not recorded | 68fa73e41e209131cea5eb76 |
| SCRN | 95.5 | Not recorded | 68fa73e41e209131cea5eb76 |
| SCRN | 95 | Not recorded | 68fa73e41e209131cea5eb76 |
| RoBERTa | 94.25 | Not recorded | 68fa73e41e209131cea5eb76 |
| RanMASK | 94 | Not recorded | 68fa73e41e209131cea5eb76 |
| RMLM | 91.25 | Not recorded | 68fa73e41e209131cea5eb76 |
| RMLM | 91 | Not recorded | 68fa73e41e209131cea5eb76 |
| BERT | 90.5 | Not recorded | 68fa73e41e209131cea5eb76 |
| BERT | 90.25 | Not recorded | 68fa73e41e209131cea5eb76 |
| RanMASK | 90 | Not recorded | 68fa73e41e209131cea5eb76 |
| GLTR | 76.5 | Not recorded | 68fa73e41e209131cea5eb76 |
| Log-Rank | 73.5 | Not recorded | 68fa73e41e209131cea5eb76 |
| Log-Likelihood | 72 | Not recorded | 68fa73e41e209131cea5eb76 |
| GLTR | 72 | Not recorded | 68fa73e41e209131cea5eb76 |
| Log-Rank | 68 | Not recorded | 68fa73e41e209131cea5eb76 |
| Log-Likelihood | 67 | Not recorded | 68fa73e41e209131cea5eb76 |
| Entropy | 63 | Not recorded | 68fa73e41e209131cea5eb76 |
| Entropy | 59.25 | Not recorded | 68fa73e41e209131cea5eb76 |

### SeqXGPT

Overall Accuracy: 96.5. Result ID: 699311603a6c3148491376cd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71751e209131cea51585.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### RDrop

Overall Accuracy: 96.5. Result ID: 699311603a6c3148491376d2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b2c1e209131cea2c8c6.md>)

[Implementation](<https://github.com/dropreg/R-Drop>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### SeqXGPT

Overall Accuracy: 96.25. Result ID: 699311603a6c3148491376e7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71751e209131cea51585.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### Flooding

Overall Accuracy: 96. Result ID: 699311603a6c3148491376d1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69af1e209131cea2361d.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### DeBERTa

Overall Accuracy: 95.75. Result ID: 699311603a6c3148491376ea.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6aef1e209131cea2b2a6.md>)

[Implementation](<https://github.com/microsoft/DeBERTa111>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### Flooding

Overall Accuracy: 95.75. Result ID: 699311603a6c3148491376eb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69af1e209131cea2361d.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### RDrop

Overall Accuracy: 95.75. Result ID: 699311603a6c3148491376ec.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b2c1e209131cea2c8c6.md>)

[Implementation](<https://github.com/dropreg/R-Drop>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### RoBERTa

Overall Accuracy: 95.5. Result ID: 699311603a6c3148491376cf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66131e209131cea0e18b.md>)

[Implementation](<https://github.com/pytorch/fairseq>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### DeBERTa

Overall Accuracy: 95.5. Result ID: 699311603a6c3148491376d0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6aef1e209131cea2b2a6.md>)

[Implementation](<https://github.com/microsoft/DeBERTa111>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### SCRN

Overall Accuracy: 95.5. Result ID: 699311603a6c3148491376ef.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Implementation](<https://github.com/CarlanLark/Robust-AIGC-Detector>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### SCRN

Overall Accuracy: 95. Result ID: 699311603a6c3148491376d5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Implementation](<https://github.com/CarlanLark/Robust-AIGC-Detector>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### RoBERTa

Overall Accuracy: 94.25. Result ID: 699311603a6c3148491376e9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66131e209131cea0e18b.md>)

[Implementation](<https://github.com/pytorch/fairseq>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### RanMASK

Overall Accuracy: 94. Result ID: 699311603a6c3148491376d3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea2844a.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### RMLM

Overall Accuracy: 91.25. Result ID: 699311603a6c3148491376ee.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### RMLM

Overall Accuracy: 91. Result ID: 699311603a6c3148491376d4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### BERT

Overall Accuracy: 90.5. Result ID: 699311603a6c3148491376ce.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65c31e209131cea0c4ed.md>)

[Implementation](<https://github.com/google-research/bert>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### BERT

Overall Accuracy: 90.25. Result ID: 699311603a6c3148491376e8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65c31e209131cea0c4ed.md>)

[Implementation](<https://github.com/google-research/bert>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### RanMASK

Overall Accuracy: 90. Result ID: 699311603a6c3148491376ed.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea2844a.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### GLTR

Overall Accuracy: 76.5. Result ID: 699311603a6c3148491376cc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65e11e209131cea0d18b.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### Log-Rank

Overall Accuracy: 73.5. Result ID: 699311603a6c3148491376ca.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa701a1e209131cea49064.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### Log-Likelihood

Overall Accuracy: 72. Result ID: 699311603a6c3148491376c9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66ae1e209131cea11b7a.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### GLTR

Overall Accuracy: 72. Result ID: 699311603a6c3148491376e6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65e11e209131cea0d18b.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### Log-Rank

Overall Accuracy: 68. Result ID: 699311603a6c3148491376e4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa701a1e209131cea49064.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### Log-Likelihood

Overall Accuracy: 67. Result ID: 699311603a6c3148491376e3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66ae1e209131cea11b7a.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### Entropy

Overall Accuracy: 63. Result ID: 699311603a6c3148491376cb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65e11e209131cea0d18b.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

### Entropy

Overall Accuracy: 59.25. Result ID: 699311603a6c3148491376e5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73e41e209131cea5eb76.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65e11e209131cea0d18b.md>)

Recorded evaluation configuration: {"Attack":"PWWS"}. This record does not assert independent reproduction.

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