# Multilingual Ability Evaluation on MLQA, XLSUM, XQuAD, GMMLU

Source: [Multilingual Ability Evaluation on MLQA, XLSUM, XQuAD, GMMLU](<https://www.sota2.com/research/sota/multilingual-ability-evaluation-on-mlqa-xlsum-xquad-gmmlu>)

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

Data updated: 2026-04-08T06:36:52.495Z

Aggregate evaluation assessing multilingual ability across the MLQA, XLSUM, XQuAD, and GMMLU datasets.

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 | Multilingual Ability Evaluation |
| Dataset (including any stated split) | MLQA, XLSUM, XQuAD, GMMLU |
| Metric | MLQA F1 |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 10 of 10 |
| 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

- [MLQA F1](<https://www.sota2.com/research/sota/multilingual-ability-evaluation-on-mlqa-xlsum-xquad-gmmlu.md?metric=MLQA%20F1>): Higher is better; unit: PERCENT.

- [XLSUM ROUGE](<https://www.sota2.com/research/sota/multilingual-ability-evaluation-on-mlqa-xlsum-xquad-gmmlu.md?metric=XLSUM%20ROUGE>): Higher is better; unit: PERCENT.

- [XQuAD F1](<https://www.sota2.com/research/sota/multilingual-ability-evaluation-on-mlqa-xlsum-xquad-gmmlu.md?metric=XQuAD%20F1>): Higher is better; unit: PERCENT.

- [GMMLU Score](<https://www.sota2.com/research/sota/multilingual-ability-evaluation-on-mlqa-xlsum-xquad-gmmlu.md?metric=GMMLU%20Score>): Higher is better; unit: PERCENT.

- [Average Score](<https://www.sota2.com/research/sota/multilingual-ability-evaluation-on-mlqa-xlsum-xquad-gmmlu.md?metric=Average%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | MLQA F1 | Result date | Report paper |
| --- | --- | --- | --- |
| LLAMA-3-8B | 60.9 | Not recorded | 68fa754c1e209131cea66e8d |
| LLAMA-3-8B | 59.7 | Not recorded | 68fa754c1e209131cea66e8d |
| LLAMA-3-8B | 48 | Not recorded | 68fa754c1e209131cea66e8d |
| LLAMA-3-8B | 43.8 | Not recorded | 68fa754c1e209131cea66e8d |
| MISTRAL-7B | 40.6 | Not recorded | 68fa754c1e209131cea66e8d |
| LLAMA-3-8B | 32.1 | Not recorded | 68fa754c1e209131cea66e8d |
| MISTRAL-7B | 30.7 | Not recorded | 68fa754c1e209131cea66e8d |
| MISTRAL-7B | 28.8 | Not recorded | 68fa754c1e209131cea66e8d |
| MISTRAL-7B | 24.6 | Not recorded | 68fa754c1e209131cea66e8d |
| MISTRAL-7B | 22.9 | Not recorded | 68fa754c1e209131cea66e8d |

### LLAMA-3-8B

MLQA F1: 60.9. Result ID: 69d5f66747f34ce924bdf68a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

Recorded evaluation configuration: {"Phase 1 (P1) Dataset":"Instruct","Protocol":"Two-phase Continual Fine-tuning","Evaluation Phase":"P1"}. This record does not assert independent reproduction.

### LLAMA-3-8B

MLQA F1: 59.7. Result ID: 69d5f66747f34ce924bdf689.

[Reporting paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

Recorded evaluation configuration: {"Phase 1 (P1) Dataset":"ALPACA","Phase 2 (P2) Dataset":"ALPACA","Protocol":"Two-phase Continual Fine-tuning","Evaluation Phase":"P2"}. This record does not assert independent reproduction.

### LLAMA-3-8B

MLQA F1: 48. Result ID: 69d5f66747f34ce924bdf68d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

Recorded evaluation configuration: {"Fine-tuning Protocol":"Dataset Mixture","Dataset Mixture":"ALPACA MULTIALPACA"}. This record does not assert independent reproduction.

### LLAMA-3-8B

MLQA F1: 43.8. Result ID: 69d5f66747f34ce924bdf688.

[Reporting paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

Recorded evaluation configuration: {"Phase 1 (P1) Dataset":"ALPACA","Phase 2 (P2) Dataset":"ALPACA","Protocol":"Two-phase Continual Fine-tuning","Evaluation Phase":"P1"}. This record does not assert independent reproduction.

### MISTRAL-7B

MLQA F1: 40.6. Result ID: 69d5f66747f34ce924bdf68c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

Recorded evaluation configuration: {"Fine-tuning Protocol":"Dataset Mixture","Dataset Mixture":"ALPACA MULTIALPACA"}. This record does not assert independent reproduction.

### LLAMA-3-8B

MLQA F1: 32.1. Result ID: 69d5f66747f34ce924bdf68b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

Recorded evaluation configuration: {"Phase 1 (P1) Dataset":"Instruct","Protocol":"Two-phase Continual Fine-tuning","Evaluation Phase":"P2"}. This record does not assert independent reproduction.

### MISTRAL-7B

MLQA F1: 30.7. Result ID: 69d5f66747f34ce924bdf687.

[Reporting paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

Recorded evaluation configuration: {"Phase 1 (P1) Dataset":"Instruct","Phase 2 (P2) Dataset":"MULTI","Protocol":"Two-phase Continual Fine-tuning","Evaluation Phase":"P2"}. This record does not assert independent reproduction.

### MISTRAL-7B

MLQA F1: 28.8. Result ID: 69d5f66747f34ce924bdf685.

[Reporting paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

Recorded evaluation configuration: {"Phase 1 (P1) Dataset":"ALPACA","Protocol":"Two-phase Continual Fine-tuning","Evaluation Phase":"P2"}. This record does not assert independent reproduction.

### MISTRAL-7B

MLQA F1: 24.6. Result ID: 69d5f66747f34ce924bdf686.

[Reporting paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

Recorded evaluation configuration: {"Phase 1 (P1) Dataset":"Instruct","Phase 2 (P2) Dataset":"MULTI","Protocol":"Two-phase Continual Fine-tuning","Evaluation Phase":"P1"}. This record does not assert independent reproduction.

### MISTRAL-7B

MLQA F1: 22.9. Result ID: 69d5f66747f34ce924bdf684.

[Reporting paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa754c1e209131cea66e8d.md>)

Recorded evaluation configuration: {"Phase 1 (P1) Dataset":"ALPACA","Protocol":"Two-phase Continual Fine-tuning","Evaluation Phase":"P1"}. This record does not assert independent reproduction.

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