language benchmark
Translation en→Set1 COMET22
COMET-22 is an ensemble machine translation evaluation metric combining a COMET estimator model trained with Direct Assessments and a multitask model that predicts sentence-level scores and word-level OK/BAD tags. It demonstrates improved correlations compared to state-of-the-art metrics and increased robustness to critical errors.
Updated Aug 11, 2026
Models3
Model coverage3
MetricScore
EvidenceB
Translation en→Set1 COMET22 Ranking
Higher score ranks better on this benchmark.
| 1 | AM Nova ProAmazon | 89.1% | 100.0% | 3 | C | |
| 2 | AMNova LiteAmazon | 88.8% | 50.0% | 3 | C | |
| 3 | AMNova MicroAmazon | 88.5% | 0.0% | 3 | C |
Translation en→Set1 COMET22 Score Distribution
A closer view of the leading scores on this benchmark.
Translation en→Set1 COMET22 Highlights
The leading models and scores on this benchmark.
What is Translation en→Set1 COMET22?
What Translation en→Set1 COMET22 measures and how its scores work.
COMET-22 is an ensemble machine translation evaluation metric combining a COMET estimator model trained with Direct Assessments and a multitask model that predicts sentence-level scores and word-level OK/BAD tags. It demonstrates improved correlations compared to state-of-the-art metrics and increased robustness to critical errors.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- Translation en→Set1 COMET22
- Modality
- text
- Primary category
- language
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- translation-en→set1-comet22|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about Translation en→Set1 COMET22.
Which model scores highest on Translation en→Set1 COMET22?
Nova Pro is currently ranked first with 89.1%.
What does Translation en→Set1 COMET22 measure?
COMET-22 is an ensemble machine translation evaluation metric combining a COMET estimator model trained with Direct Assessments and a multitask model that predicts sentence-level scores and word-level OK/BAD tags. It demonstrates improved correlations compared to state-of-the-art metrics and increased robustness to critical errors.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
3 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.