# Watermarking on eli5-category (test)

Source: [Watermarking on eli5-category (test)](<https://www.sota2.com/research/sota/watermarking-on-eli5-category-test>)

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

Data updated: 2026-02-26T05:55:11.195Z

Evaluation of Watermarking performance, likely related to text generation or classification, on the test split of the eli5-category dataset, using metrics like Perplexity (PPL) and Likelihood (LH).

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 | Watermarking |
| Dataset (including any stated split) | eli5-category (test) |
| Metric | PPL |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 28 of 28 |
| 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

- [PPL](<https://www.sota2.com/research/sota/watermarking-on-eli5-category-test.md?metric=PPL>): Lower is better; unit: SCALAR.

- [LH](<https://www.sota2.com/research/sota/watermarking-on-eli5-category-test.md?metric=LH>): Higher is better; unit: PERCENT.

- [AUC](<https://www.sota2.com/research/sota/watermarking-on-eli5-category-test.md?metric=AUC>): Higher is better; unit: PERCENT.

- [pAUC](<https://www.sota2.com/research/sota/watermarking-on-eli5-category-test.md?metric=pAUC>): Higher is better; unit: PERCENT.

## Recorded results

| Method | PPL | Result date | Report paper |
| --- | --- | --- | --- |
| Greedy Decoding | 1.313 | Not recorded | 68fa77c11e209131cea751b5 |
| Black-Box Watermarking | 1.61 | Not recorded | 68fa77c11e209131cea751b5 |
| Black-Box Watermarking | 1.61 | Not recorded | 68fa77c11e209131cea751b5 |
| Kuditipudi | 1.615 | Not recorded | 68fa77c11e209131cea751b5 |
| Black-Box Watermarking | 1.615 | Not recorded | 68fa77c11e209131cea751b5 |
| Recursive Black-Box Watermarking | 1.615 | Not recorded | 68fa77c11e209131cea751b5 |
| Aaronson | 1.619 | Not recorded | 68fa77c11e209131cea751b5 |
| Aaronson Cor. | 1.619 | Not recorded | 68fa77c11e209131cea751b5 |
| Black-Box Watermarking | 1.621 | Not recorded | 68fa77c11e209131cea751b5 |
| Recursive Black-Box Watermarking | 1.621 | Not recorded | 68fa77c11e209131cea751b5 |
| Black-Box Watermarking | 1.623 | Not recorded | 68fa77c11e209131cea751b5 |
| Recursive Black-Box Watermarking | 1.623 | Not recorded | 68fa77c11e209131cea751b5 |
| Random Sampling | 1.627 | Not recorded | 68fa77c11e209131cea751b5 |
| Recursive Black-Box Watermarking | 1.63 | Not recorded | 68fa77c11e209131cea751b5 |
| Black-Box Watermarking | 1.631 | Not recorded | 68fa77c11e209131cea751b5 |
| Black-Box Watermarking | 1.645 | Not recorded | 68fa77c11e209131cea751b5 |
| Kirchenbauer | 1.649 | Not recorded | 68fa77c11e209131cea751b5 |
| Black-Box Watermarking | 1.652 | Not recorded | 68fa77c11e209131cea751b5 |
| Black-Box Watermarking | 1.653 | Not recorded | 68fa77c11e209131cea751b5 |
| Recursive Black-Box Watermarking | 1.653 | Not recorded | 68fa77c11e209131cea751b5 |
| Black-Box Watermarking | 1.657 | Not recorded | 68fa77c11e209131cea751b5 |
| Recursive Black-Box Watermarking | 1.662 | Not recorded | 68fa77c11e209131cea751b5 |
| Recursive Black-Box Watermarking | 1.664 | Not recorded | 68fa77c11e209131cea751b5 |
| Recursive Black-Box Watermarking | 1.665 | Not recorded | 68fa77c11e209131cea751b5 |
| Kirchenbauer | 1.673 | Not recorded | 68fa77c11e209131cea751b5 |
| Kirchenbauer | 1.836 | Not recorded | 68fa77c11e209131cea751b5 |
| Kirchenbauer | 2.159 | Not recorded | 68fa77c11e209131cea751b5 |
| Kirchenbauer | 2.847 | Not recorded | 68fa77c11e209131cea751b5 |

### Greedy Decoding

PPL: 1.313. Result ID: 699e3d755698665ec4fa5c2f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct"}. This record does not assert independent reproduction.

### Black-Box Watermarking

PPL: 1.61. Result ID: 699e3d755698665ec4fa5c3d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"1","m":"512"}. This record does not assert independent reproduction.

### Black-Box Watermarking

PPL: 1.61. Result ID: 699e3d755698665ec4fa5c3e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"1","m":"1024"}. This record does not assert independent reproduction.

### Kuditipudi

PPL: 1.615. Result ID: 699e3d755698665ec4fa5c38.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa738a1e209131cea5d13e.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct"}. This record does not assert independent reproduction.

### Black-Box Watermarking

PPL: 1.615. Result ID: 699e3d755698665ec4fa5c3c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"1","m":"32"}. This record does not assert independent reproduction.

### Recursive Black-Box Watermarking

PPL: 1.615. Result ID: 699e3d755698665ec4fa5c46.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"1","m":"512"}. This record does not assert independent reproduction.

### Aaronson

PPL: 1.619. Result ID: 699e3d755698665ec4fa5c31.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct"}. This record does not assert independent reproduction.

### Aaronson Cor.

PPL: 1.619. Result ID: 699e3d755698665ec4fa5c32.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct"}. This record does not assert independent reproduction.

### Black-Box Watermarking

PPL: 1.621. Result ID: 699e3d755698665ec4fa5c3b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"1","m":"16"}. This record does not assert independent reproduction.

### Recursive Black-Box Watermarking

PPL: 1.621. Result ID: 699e3d755698665ec4fa5c44.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"1","m":"16"}. This record does not assert independent reproduction.

### Black-Box Watermarking

PPL: 1.623. Result ID: 699e3d755698665ec4fa5c3a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"1","m":"4"}. This record does not assert independent reproduction.

### Recursive Black-Box Watermarking

PPL: 1.623. Result ID: 699e3d755698665ec4fa5c43.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"1","m":"4"}. This record does not assert independent reproduction.

### Random Sampling

PPL: 1.627. Result ID: 699e3d755698665ec4fa5c30.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct"}. This record does not assert independent reproduction.

### Recursive Black-Box Watermarking

PPL: 1.63. Result ID: 699e3d755698665ec4fa5c45.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"1","m":"32"}. This record does not assert independent reproduction.

### Black-Box Watermarking

PPL: 1.631. Result ID: 699e3d755698665ec4fa5c39.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"1","m":"2"}. This record does not assert independent reproduction.

### Black-Box Watermarking

PPL: 1.645. Result ID: 699e3d755698665ec4fa5c42.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"50","m":"16"}. This record does not assert independent reproduction.

### Kirchenbauer

PPL: 1.649. Result ID: 699e3d755698665ec4fa5c33.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73131e209131cea5a4b7.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","delta (δ)":"0.5"}. This record does not assert independent reproduction.

### Black-Box Watermarking

PPL: 1.652. Result ID: 699e3d755698665ec4fa5c41.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"50","m":"4"}. This record does not assert independent reproduction.

### Black-Box Watermarking

PPL: 1.653. Result ID: 699e3d755698665ec4fa5c40.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"10","m":"16"}. This record does not assert independent reproduction.

### Recursive Black-Box Watermarking

PPL: 1.653. Result ID: 699e3d755698665ec4fa5c4a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"50","m":"16"}. This record does not assert independent reproduction.

### Black-Box Watermarking

PPL: 1.657. Result ID: 699e3d755698665ec4fa5c3f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"10","m":"4"}. This record does not assert independent reproduction.

### Recursive Black-Box Watermarking

PPL: 1.662. Result ID: 699e3d755698665ec4fa5c48.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"10","m":"16"}. This record does not assert independent reproduction.

### Recursive Black-Box Watermarking

PPL: 1.664. Result ID: 699e3d755698665ec4fa5c49.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"50","m":"4"}. This record does not assert independent reproduction.

### Recursive Black-Box Watermarking

PPL: 1.665. Result ID: 699e3d755698665ec4fa5c47.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","k":"10","m":"4"}. This record does not assert independent reproduction.

### Kirchenbauer

PPL: 1.673. Result ID: 699e3d755698665ec4fa5c34.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73131e209131cea5a4b7.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","delta (δ)":"1"}. This record does not assert independent reproduction.

### Kirchenbauer

PPL: 1.836. Result ID: 699e3d755698665ec4fa5c35.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73131e209131cea5a4b7.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","delta (δ)":"2"}. This record does not assert independent reproduction.

### Kirchenbauer

PPL: 2.159. Result ID: 699e3d755698665ec4fa5c36.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73131e209131cea5a4b7.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","delta (δ)":"3"}. This record does not assert independent reproduction.

### Kirchenbauer

PPL: 2.847. Result ID: 699e3d755698665ec4fa5c37.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77c11e209131cea751b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73131e209131cea5a4b7.md>)

Recorded evaluation configuration: {"Model":"Gemma-7B-instruct","delta (δ)":"4"}. This record does not assert independent reproduction.

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