# Translation on Translation (COMET evaluation)

Source: [Translation on Translation (COMET evaluation)](<https://www.sota2.com/research/sota/translation-on-translation-comet-evaluation>)

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

Data updated: 2026-06-23T11:59:08.656Z

Evaluation of general Translation performance on the Translation dataset, measured using the COMET metric.

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 | Translation |
| Dataset (including any stated split) | Translation |
| Metric | COMET Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 34 of 34 |
| 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

- [COMET Score](<https://www.sota2.com/research/sota/translation-on-translation-comet-evaluation.md?metric=COMET%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | COMET Score | Result date | Report paper |
| --- | --- | --- | --- |
| Teacher (70B SFT) | 86.31 | Not recorded | 6a3a13ad206542251d7c77db |
| 2-Stage | 83.63 | Not recorded | 6a3a13ad206542251d7c77db |
| Mix | 83.15 | Not recorded | 6a3a13ad206542251d7c77db |
| GKD | 83.1 | Not recorded | 6a3a13ad206542251d7c77db |
| IGDS | 83.07 | Not recorded | 69f190ae55a3c32af89fbb8d |
| SFT | 82.87 | Not recorded | 6a3a13ad206542251d7c77db |
| Full | 82.71 | Not recorded | 69f190ae55a3c32af89fbb8d |
| IFD | 82.69 | Not recorded | 69f190ae55a3c32af89fbb8d |
| 2-Stage | 82.67 | Not recorded | 6a3a13ad206542251d7c77db |
| Loss | 82.66 | Not recorded | 69f190ae55a3c32af89fbb8d |
| Random | 82.61 | Not recorded | 69f190ae55a3c32af89fbb8d |
| ZIP | 82.57 | Not recorded | 69f190ae55a3c32af89fbb8d |
| Mix | 82.41 | Not recorded | 6a3a13ad206542251d7c77db |
| Original | 81.3 | Not recorded | 69f190ae55a3c32af89fbb8d |
| GKD | 79.39 | Not recorded | 6a3a13ad206542251d7c77db |
| SFT | 78.91 | Not recorded | 6a3a13ad206542251d7c77db |
| 2-Stage | 78.65 | Not recorded | 6a3a13ad206542251d7c77db |
| IGDS | 78.45 | Not recorded | 69f190ae55a3c32af89fbb8d |
| Mix | 78.13 | Not recorded | 6a3a13ad206542251d7c77db |
| Full | 77.68 | Not recorded | 69f190ae55a3c32af89fbb8d |
| IFD | 76.82 | Not recorded | 69f190ae55a3c32af89fbb8d |
| ZIP | 74.49 | Not recorded | 69f190ae55a3c32af89fbb8d |
| Loss | 72.29 | Not recorded | 69f190ae55a3c32af89fbb8d |
| Random | 70.78 | Not recorded | 69f190ae55a3c32af89fbb8d |
| GKD | 69.87 | Not recorded | 6a3a13ad206542251d7c77db |
| IGDS | 68.14 | Not recorded | 69f190ae55a3c32af89fbb8d |
| SFT | 68 | Not recorded | 6a3a13ad206542251d7c77db |
| Full | 65.93 | Not recorded | 69f190ae55a3c32af89fbb8d |
| IFD | 64.19 | Not recorded | 69f190ae55a3c32af89fbb8d |
| ZIP | 63.37 | Not recorded | 69f190ae55a3c32af89fbb8d |
| Random | 62.92 | Not recorded | 69f190ae55a3c32af89fbb8d |
| Loss | 55.31 | Not recorded | 69f190ae55a3c32af89fbb8d |
| Original | 31.11 | Not recorded | 69f190ae55a3c32af89fbb8d |
| Original | 26.92 | Not recorded | 69f190ae55a3c32af89fbb8d |

### Teacher (70B SFT)

COMET Score: 86.31. Result ID: 6a3a74f1fbc373592c385d5e.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d11e209131cea6a028.md>)

[Implementation](<https://github.com/meta-llama/llama-agentic-system>)

Recorded evaluation configuration: {"Model Size":"70B"}. This record does not assert independent reproduction.

### 2-Stage

COMET Score: 83.63. Result ID: 6a3a74f1fbc373592c385d6a.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"8B"}. This record does not assert independent reproduction.

### Mix

COMET Score: 83.15. Result ID: 6a3a74f1fbc373592c385d69.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"8B"}. This record does not assert independent reproduction.

### GKD

COMET Score: 83.1. Result ID: 6a3a74f1fbc373592c385d68.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"8B"}. This record does not assert independent reproduction.

### IGDS

COMET Score: 83.07. Result ID: 69f1b8a047f34ce924d7fd0a.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B-Base","Data Selection Strategy":"IGDS"}. This record does not assert independent reproduction.

### SFT

COMET Score: 82.87. Result ID: 6a3a74f1fbc373592c385d67.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"8B"}. This record does not assert independent reproduction.

### Full

COMET Score: 82.71. Result ID: 69f1b8a047f34ce924d7fd05.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B-Base","Data Selection Strategy":"Full"}. This record does not assert independent reproduction.

### IFD

COMET Score: 82.69. Result ID: 69f1b8a047f34ce924d7fd08.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72d61e209131cea58d61.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B-Base","Data Selection Strategy":"IFD"}. This record does not assert independent reproduction.

### 2-Stage

COMET Score: 82.67. Result ID: 6a3a74f1fbc373592c385d66.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"3B"}. This record does not assert independent reproduction.

### Loss

COMET Score: 82.66. Result ID: 69f1b8a047f34ce924d7fd07.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B-Base","Data Selection Strategy":"Loss"}. This record does not assert independent reproduction.

### Random

COMET Score: 82.61. Result ID: 69f1b8a047f34ce924d7fd06.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B-Base","Data Selection Strategy":"Random"}. This record does not assert independent reproduction.

### ZIP

COMET Score: 82.57. Result ID: 69f1b8a047f34ce924d7fd09.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74131e209131cea5fad7.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B-Base","Data Selection Strategy":"ZIP"}. This record does not assert independent reproduction.

### Mix

COMET Score: 82.41. Result ID: 6a3a74f1fbc373592c385d65.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"3B"}. This record does not assert independent reproduction.

### Original

COMET Score: 81.3. Result ID: 69f1b8a047f34ce924d7fd04.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Qwen3-8B-Base","Data Selection Strategy":"Original"}. This record does not assert independent reproduction.

### GKD

COMET Score: 79.39. Result ID: 6a3a74f1fbc373592c385d64.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"3B"}. This record does not assert independent reproduction.

### SFT

COMET Score: 78.91. Result ID: 6a3a74f1fbc373592c385d63.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"3B"}. This record does not assert independent reproduction.

### 2-Stage

COMET Score: 78.65. Result ID: 6a3a74f1fbc373592c385d62.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"1B"}. This record does not assert independent reproduction.

### IGDS

COMET Score: 78.45. Result ID: 69f1b8a047f34ce924d7fd03.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Data Selection Strategy":"IGDS"}. This record does not assert independent reproduction.

### Mix

COMET Score: 78.13. Result ID: 6a3a74f1fbc373592c385d61.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"1B"}. This record does not assert independent reproduction.

### Full

COMET Score: 77.68. Result ID: 69f1b8a047f34ce924d7fcfe.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Data Selection Strategy":"Full"}. This record does not assert independent reproduction.

### IFD

COMET Score: 76.82. Result ID: 69f1b8a047f34ce924d7fd01.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72d61e209131cea58d61.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Data Selection Strategy":"IFD"}. This record does not assert independent reproduction.

### ZIP

COMET Score: 74.49. Result ID: 69f1b8a047f34ce924d7fd02.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74131e209131cea5fad7.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Data Selection Strategy":"ZIP"}. This record does not assert independent reproduction.

### Loss

COMET Score: 72.29. Result ID: 69f1b8a047f34ce924d7fd00.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Data Selection Strategy":"Loss"}. This record does not assert independent reproduction.

### Random

COMET Score: 70.78. Result ID: 69f1b8a047f34ce924d7fcff.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Data Selection Strategy":"Random"}. This record does not assert independent reproduction.

### GKD

COMET Score: 69.87. Result ID: 6a3a74f1fbc373592c385d60.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"1B"}. This record does not assert independent reproduction.

### IGDS

COMET Score: 68.14. Result ID: 69f1b8a047f34ce924d7fcfc.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Gemma-2-2B","Data Selection Strategy":"IGDS"}. This record does not assert independent reproduction.

### SFT

COMET Score: 68. Result ID: 6a3a74f1fbc373592c385d5f.

[Reporting paper](<https://www.sota2.com/research/paper/6a3a13ad206542251d7c77db.md>)

Recorded evaluation configuration: {"Model Size":"1B"}. This record does not assert independent reproduction.

### Full

COMET Score: 65.93. Result ID: 69f1b8a047f34ce924d7fcf7.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Gemma-2-2B","Data Selection Strategy":"Full"}. This record does not assert independent reproduction.

### IFD

COMET Score: 64.19. Result ID: 69f1b8a047f34ce924d7fcfa.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72d61e209131cea58d61.md>)

Recorded evaluation configuration: {"Model":"Gemma-2-2B","Data Selection Strategy":"IFD"}. This record does not assert independent reproduction.

### ZIP

COMET Score: 63.37. Result ID: 69f1b8a047f34ce924d7fcfb.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74131e209131cea5fad7.md>)

Recorded evaluation configuration: {"Model":"Gemma-2-2B","Data Selection Strategy":"ZIP"}. This record does not assert independent reproduction.

### Random

COMET Score: 62.92. Result ID: 69f1b8a047f34ce924d7fcf8.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Gemma-2-2B","Data Selection Strategy":"Random"}. This record does not assert independent reproduction.

### Loss

COMET Score: 55.31. Result ID: 69f1b8a047f34ce924d7fcf9.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Gemma-2-2B","Data Selection Strategy":"Loss"}. This record does not assert independent reproduction.

### Original

COMET Score: 31.11. Result ID: 69f1b8a047f34ce924d7fcfd.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Llama-3.1-8B","Data Selection Strategy":"Original"}. This record does not assert independent reproduction.

### Original

COMET Score: 26.92. Result ID: 69f1b8a047f34ce924d7fcf6.

[Reporting paper](<https://www.sota2.com/research/paper/69f190ae55a3c32af89fbb8d.md>)

Recorded evaluation configuration: {"Model":"Gemma-2-2B","Data Selection Strategy":"Original"}. This record does not assert independent reproduction.

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