# Sign Language Generation on CSL-Daily (Quality Metrics)

Source: [Sign Language Generation on CSL-Daily (Quality Metrics)](<https://www.sota2.com/research/sota/sign-language-generation-on-csl-daily-quality-metrics>)

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

Data updated: 2026-06-29T06:15:31.661Z

Evaluation of Sign Language Generation performance on the CSL-Daily dataset, focusing on sequence quality metrics (BLEU, ROUGE) and keypoint error metrics (DTW-JPE).

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 | Sign Language Generation |
| Dataset (including any stated split) | CSL-Daily |
| Metric | BLEU-4 |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 16 of 16 |
| 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

- [BLEU-4](<https://www.sota2.com/research/sota/sign-language-generation-on-csl-daily-quality-metrics.md?metric=BLEU-4>): Higher is better; unit: PERCENT.

- [ROUGE](<https://www.sota2.com/research/sota/sign-language-generation-on-csl-daily-quality-metrics.md?metric=ROUGE>): Higher is better; unit: PERCENT.

- [DTW-PA-JPE (Body)](<https://www.sota2.com/research/sota/sign-language-generation-on-csl-daily-quality-metrics.md?metric=DTW-PA-JPE%20(Body)>): Lower is better; unit: SCALAR.

- [DTW-PA-JPE (Hand)](<https://www.sota2.com/research/sota/sign-language-generation-on-csl-daily-quality-metrics.md?metric=DTW-PA-JPE%20(Hand)>): Lower is better; unit: SCALAR.

- [DTW-JPE (Body)](<https://www.sota2.com/research/sota/sign-language-generation-on-csl-daily-quality-metrics.md?metric=DTW-JPE%20(Body)>): Lower is better; unit: SCALAR.

- [DTW-JPE (Hand)](<https://www.sota2.com/research/sota/sign-language-generation-on-csl-daily-quality-metrics.md?metric=DTW-JPE%20(Hand)>): Lower is better; unit: SCALAR.

- [DTW-JPE (All)](<https://www.sota2.com/research/sota/sign-language-generation-on-csl-daily-quality-metrics.md?metric=DTW-JPE%20(All)>): Lower is better; unit: SCALAR.

- [WER](<https://www.sota2.com/research/sota/sign-language-generation-on-csl-daily-quality-metrics.md?metric=WER>): Lower is better; unit: PERCENT.

## Recorded results

| Method | BLEU-4 | Result date | Report paper |
| --- | --- | --- | --- |
| SIGNER | 15.6 | Not recorded | 68fa79451e209131cea7cfa0 |
| GARD | 12.73 | Not recorded | 6a3a12e6206542251d7c6e16 |
| SOKE\* | 11.21 | Not recorded | 6a3a12e6206542251d7c6e16 |
| T2S-GPT | 8.94 | Not recorded | 6a3a12e6206542251d7c6e16 |
| M-GPT | 8.84 | Not recorded | 6a3a12e6206542251d7c6e16 |
| Sing-IDD\* | 8.13 | Not recorded | 6a3a12e6206542251d7c6e16 |
| NAT-EA | 7.87 | Not recorded | 68fa79451e209131cea7cfa0 |
| G2P-DDM\* | 6.52 | Not recorded | 6a3a12e6206542251d7c6e16 |
| G2P-DDM | 6.44 | Not recorded | 68fa79451e209131cea7cfa0 |
| T2M | 5.11 | Not recorded | 6a3a12e6206542251d7c6e16 |
| SOKE | 4.09 | Not recorded | 68fa79451e209131cea7cfa0 |
| MoMask | 3.79 | Not recorded | 68fa79451e209131cea7cfa0 |
| PT\* | 3.07 | Not recorded | 6a3a12e6206542251d7c6e16 |
| MDM | 2.78 | Not recorded | 68fa79451e209131cea7cfa0 |
| NSA | 2.09 | Not recorded | 68fa79451e209131cea7cfa0 |
| MotionGPT | 1.79 | Not recorded | 68fa79451e209131cea7cfa0 |

### SIGNER

BLEU-4: 15.6. Result ID: 6a420d7cfbc373592c86251c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79451e209131cea7cfa0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79451e209131cea7cfa0.md>)

Recorded evaluation configuration: {"Setting":"Gloss-based"}. This record does not assert independent reproduction.

### GARD

BLEU-4: 12.73. Result ID: 6a3a422dfbc373592c34badb.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a12e6206542251d7c6e16.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3a12e6206542251d7c6e16.md>)

Recorded evaluation configuration: {"Category":"mesh-based","Gloss-dependent":"true"}. This record does not assert independent reproduction.

### SOKE\*

BLEU-4: 11.21. Result ID: 6a3a422dfbc373592c34bada.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a12e6206542251d7c6e16.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a4c1e209131cea82499.md>)

Recorded evaluation configuration: {"Category":"mesh-based","Gloss-dependent":"false"}. This record does not assert independent reproduction.

### T2S-GPT

BLEU-4: 8.94. Result ID: 6a3a422dfbc373592c34bad8.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a12e6206542251d7c6e16.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73981e209131cea5d762.md>)

[Implementation](<https://t2sgpt-demo.yinaoxiong.cn/>)

Recorded evaluation configuration: {"Category":"mesh-based","Gloss-dependent":"false"}. This record does not assert independent reproduction.

### M-GPT

BLEU-4: 8.84. Result ID: 6a3a422dfbc373592c34bad9.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a12e6206542251d7c6e16.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70201e209131cea493fc.md>)

[Implementation](<https://github.com/OpenMotionLab/MotionGPT>)

Recorded evaluation configuration: {"Category":"mesh-based","Gloss-dependent":"false"}. This record does not assert independent reproduction.

### Sing-IDD\*

BLEU-4: 8.13. Result ID: 6a3a422dfbc373592c34bad6.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a12e6206542251d7c6e16.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76471e209131cea6c1be.md>)

Recorded evaluation configuration: {"Category":"keypoint-based","Gloss-dependent":"true"}. This record does not assert independent reproduction.

### NAT-EA

BLEU-4: 7.87. Result ID: 6a420d7cfbc373592c86251b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79451e209131cea7cfa0.md>)

Recorded evaluation configuration: {"Setting":"Gloss-based"}. This record does not assert independent reproduction.

### G2P-DDM\*

BLEU-4: 6.52. Result ID: 6a3a422dfbc373592c34bad5.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a12e6206542251d7c6e16.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71781e209131cea516d9.md>)

Recorded evaluation configuration: {"Category":"keypoint-based","Gloss-dependent":"true"}. This record does not assert independent reproduction.

### G2P-DDM

BLEU-4: 6.44. Result ID: 6a420d7cfbc373592c86251a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79451e209131cea7cfa0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71781e209131cea516d9.md>)

Recorded evaluation configuration: {"Setting":"Gloss-based"}. This record does not assert independent reproduction.

### T2M

BLEU-4: 5.11. Result ID: 6a3a422dfbc373592c34bad7.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a12e6206542251d7c6e16.md>)

Recorded evaluation configuration: {"Category":"mesh-based","Gloss-dependent":"false"}. This record does not assert independent reproduction.

### SOKE

BLEU-4: 4.09. Result ID: 6a420d7cfbc373592c862519.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79451e209131cea7cfa0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a4c1e209131cea82499.md>)

Recorded evaluation configuration: {"Setting":"Gloss-free"}. This record does not assert independent reproduction.

### MoMask

BLEU-4: 3.79. Result ID: 6a420d7cfbc373592c862518.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79451e209131cea7cfa0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa715d1e209131cea50e45.md>)

[Implementation](<https://ericguo5513.github.io/momask/>)

Recorded evaluation configuration: {"Setting":"Gloss-free"}. This record does not assert independent reproduction.

### PT\*

BLEU-4: 3.07. Result ID: 6a3a422dfbc373592c34bad4.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a12e6206542251d7c6e16.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67f11e209131cea1943d.md>)

Recorded evaluation configuration: {"Category":"keypoint-based","Gloss-dependent":"true"}. This record does not assert independent reproduction.

### MDM

BLEU-4: 2.78. Result ID: 6a420d7cfbc373592c862516.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79451e209131cea7cfa0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d8f1e209131cea3a994.md>)

[Implementation](<https://github.com/GuyTevet/motion-diffusion-model>)

Recorded evaluation configuration: {"Setting":"Gloss-free"}. This record does not assert independent reproduction.

### NSA

BLEU-4: 2.09. Result ID: 6a420d7cfbc373592c862517.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79451e209131cea7cfa0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72d31e209131cea58c0b.md>)

[Implementation](<https://baltatzisv.github.io/neural-sign-actors/>)

Recorded evaluation configuration: {"Setting":"Gloss-free"}. This record does not assert independent reproduction.

### MotionGPT

BLEU-4: 1.79. Result ID: 6a420d7cfbc373592c862515.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79451e209131cea7cfa0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70201e209131cea493fc.md>)

[Implementation](<https://github.com/OpenMotionLab/MotionGPT>)

Recorded evaluation configuration: {"Setting":"Gloss-free"}. This record does not assert independent reproduction.

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