# Cross-lingual retrieval on MKQA

Source: [Cross-lingual retrieval on MKQA](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa>)

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

Data updated: 2026-03-20T09:45:08.765Z

Evaluation of model performance on the MKQA dataset for cross-lingual retrieval tasks, reporting Recall@K scores across various source and target languages.

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 | Cross-lingual retrieval |
| Dataset (including any stated split) | MKQA |
| Metric | Avg. Recall@100 |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 27 of 27 |
| 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

- [Avg. Recall@100](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Avg.%20Recall%40100>): Higher is better; unit: PERCENT.

- [Avg. Recall@20](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Avg.%20Recall%4020>): Higher is better; unit: PERCENT.

- [Recall@K (en)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40K%20(en)>): Higher is better; unit: PERCENT.

- [Recall@K (ar)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40K%20(ar)>): Higher is better; unit: PERCENT.

- [Recall@K (ja)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40K%20(ja)>): Higher is better; unit: PERCENT.

- [Recall@K (ko)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40K%20(ko)>): Higher is better; unit: PERCENT.

- [Recall@K (es)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40K%20(es)>): Higher is better; unit: PERCENT.

- [Recall@K (he)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40K%20(he)>): Higher is better; unit: PERCENT.

- [Recall@K (de)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40K%20(de)>): Higher is better; unit: PERCENT.

- [Recall@100 (ar)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(ar)>): Higher is better; unit: PERCENT.

- [Recall@100 (da)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(da)>): Higher is better; unit: PERCENT.

- [Recall@100 (de)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(de)>): Higher is better; unit: PERCENT.

- [Recall@100 (es)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(es)>): Higher is better; unit: PERCENT.

- [Recall@100 (fi)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(fi)>): Higher is better; unit: PERCENT.

- [Recall@100 (fr)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(fr)>): Higher is better; unit: PERCENT.

- [Recall@100 (he)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(he)>): Higher is better; unit: PERCENT.

- [Recall@100 (hu)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(hu)>): Higher is better; unit: PERCENT.

- [Recall@100 (it)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(it)>): Higher is better; unit: PERCENT.

- [Recall@100 (ja)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(ja)>): Higher is better; unit: PERCENT.

- [Recall@100 (km)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(km)>): Higher is better; unit: PERCENT.

- [Recall@100 (ko)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(ko)>): Higher is better; unit: PERCENT.

- [Recall@100 (ms)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(ms)>): Higher is better; unit: PERCENT.

- [Recall@100 (nl)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(nl)>): Higher is better; unit: PERCENT.

- [Recall@100 (no)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(no)>): Higher is better; unit: PERCENT.

- [Recall@100 (pl)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(pl)>): Higher is better; unit: PERCENT.

- [Recall@100 (pt)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(pt)>): Higher is better; unit: PERCENT.

- [Recall@100 (ru)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(ru)>): Higher is better; unit: PERCENT.

- [Recall@100 (sv)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(sv)>): Higher is better; unit: PERCENT.

- [Recall@100 (th)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(th)>): Higher is better; unit: PERCENT.

- [Recall@100 (tr)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(tr)>): Higher is better; unit: PERCENT.

- [Recall@100 (vi)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(vi)>): Higher is better; unit: PERCENT.

- [Recall@100 (zh\_cn)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(zh_cn)>): Higher is better; unit: PERCENT.

- [Recall@100 (zh\_hk)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(zh_hk)>): Higher is better; unit: PERCENT.

- [Recall@100 (zh\_tw)](<https://www.sota2.com/research/sota/cross-lingual-retrieval-on-mkqa.md?metric=Recall%40100%20(zh_tw)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Avg. Recall@100 | Result date | Report paper |
| --- | --- | --- | --- |
| MILCO | 76.6 | Not recorded | 68fa7bc41e209131cea8a4c0 |
| M3-Embedding | 75.5 | Not recorded | 68fa73d91e209131cea5e7ab |
| M3-Embedding | 75.3 | Not recorded | 68fa73d91e209131cea5e7ab |
| M3-Embedding | 75.3 | Not recorded | 68fa73d91e209131cea5e7ab |
| BGE-M3 Multi-Vec | 75.3 | Not recorded | 68fa7bc41e209131cea8a4c0 |
| BGE-M3 Dense+Sparse | 75.3 | Not recorded | 68fa7bc41e209131cea8a4c0 |
| M3-Embedding | 75.1 | Not recorded | 68fa73d91e209131cea5e7ab |
| BGE-M3 Dense | 75.1 | Not recorded | 68fa7bc41e209131cea8a4c0 |
| PLAID-X (multivector) | 73.4 | Not recorded | 68fa7bc41e209131cea8a4c0 |
| mE5large | 70.9 | Not recorded | 68fa73d91e209131cea5e7ab |
| E5-large | 70.9 | Not recorded | 68fa7bc41e209131cea8a4c0 |
| E5mistral-7b | 70.1 | Not recorded | 68fa73d91e209131cea5e7ab |
| E5-mistral-7b | 70.1 | Not recorded | 68fa7bc41e209131cea8a4c0 |
| OpenAI-3 | 69.5 | Not recorded | 68fa73d91e209131cea5e7ab |
| mContriever | 67.9 | Not recorded | 68fa73d91e209131cea5e7ab |
| Qwen3-Embed-8B | 67.9 | Not recorded | 68fa7bc41e209131cea8a4c0 |
| mContriever | 65.6 | Not recorded | 68fa6d4b1e209131cea38d50 |
| mDPR | 60.6 | Not recorded | 68fa73d91e209131cea5e7ab |
| CORA | 59.8 | Not recorded | 68fa6d4b1e209131cea38d50 |
| XLM-R | 59.6 | Not recorded | 68fa6d4b1e209131cea38d50 |
| mBERT | 57.9 | Not recorded | 68fa6d4b1e209131cea38d50 |
| Qwen3-Embed-0.6B | 54.4 | Not recorded | 68fa7bc41e209131cea8a4c0 |
| mContriever | 49.2 | Not recorded | 68fa6d4b1e209131cea38d50 |
| M3-Embedding | 45.3 | Not recorded | 68fa73d91e209131cea5e7ab |
| BGE-M3 Sparse | 45.3 | Not recorded | 68fa7bc41e209131cea8a4c0 |
| BM25 | 39.9 | Not recorded | 68fa73d91e209131cea5e7ab |
| BM25 | 39.9 | Not recorded | 68fa7bc41e209131cea8a4c0 |

### MILCO

Avg. Recall@100: 76.6. Result ID: 69bd171147f34ce924a7951b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

Recorded evaluation configuration: {"Zero-shot":"true","Components":"SAP, SCTKD, LexEcho"}. This record does not assert independent reproduction.

### M3-Embedding

Avg. Recall@100: 75.5. Result ID: 69930ea73a6c314849136473.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: {"Retrieval Strategy":"All"}. This record does not assert independent reproduction.

### M3-Embedding

Avg. Recall@100: 75.3. Result ID: 69930ea73a6c314849136471.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: {"Retrieval Strategy":"Multi-vec"}. This record does not assert independent reproduction.

### M3-Embedding

Avg. Recall@100: 75.3. Result ID: 69930ea73a6c314849136472.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: {"Retrieval Strategy":"Dense+Sparse"}. This record does not assert independent reproduction.

### BGE-M3 Multi-Vec

Avg. Recall@100: 75.3. Result ID: 69bd171147f34ce924a79514.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### BGE-M3 Dense+Sparse

Avg. Recall@100: 75.3. Result ID: 69bd171147f34ce924a79515.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### M3-Embedding

Avg. Recall@100: 75.1. Result ID: 69930ea73a6c31484913646f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: {"Retrieval Strategy":"Dense"}. This record does not assert independent reproduction.

### BGE-M3 Dense

Avg. Recall@100: 75.1. Result ID: 69bd171147f34ce924a79513.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### PLAID-X (multivector)

Avg. Recall@100: 73.4. Result ID: 69bd171147f34ce924a79516.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71d31e209131cea53306.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### mE5large

Avg. Recall@100: 70.9. Result ID: 69930ea73a6c31484913646c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72491e209131cea5595a.md>)

[Implementation](<https://github.com/microsoft/unilm>)

Recorded evaluation configuration: {"Method Group":"Baselines (Prior Work)"}. This record does not assert independent reproduction.

### E5-large

Avg. Recall@100: 70.9. Result ID: 69bd171147f34ce924a79511.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72251e209131cea54c85.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### E5mistral-7b

Avg. Recall@100: 70.1. Result ID: 69930ea73a6c31484913646d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73611e209131cea5c103.md>)

Recorded evaluation configuration: {"Method Group":"Baselines (Prior Work)"}. This record does not assert independent reproduction.

### E5-mistral-7b

Avg. Recall@100: 70.1. Result ID: 69bd171147f34ce924a79512.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73611e209131cea5c103.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### OpenAI-3

Avg. Recall@100: 69.5. Result ID: 69930ea73a6c31484913646e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

Recorded evaluation configuration: {"Method Group":"Baselines (Prior Work)"}. This record does not assert independent reproduction.

### mContriever

Avg. Recall@100: 67.9. Result ID: 69930ea73a6c31484913646b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d4b1e209131cea38d50.md>)

[Implementation](<https://github.com/facebookresearch/contriever>)

Recorded evaluation configuration: {"Method Group":"Baselines (Prior Work)"}. This record does not assert independent reproduction.

### Qwen3-Embed-8B

Avg. Recall@100: 67.9. Result ID: 69bd171147f34ce924a79518.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79601e209131cea7d516.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### mContriever

Avg. Recall@100: 65.6. Result ID: 698f9ace3a6c314849fa765a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d4b1e209131cea38d50.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d4b1e209131cea38d50.md>)

[Implementation](<https://github.com/facebookresearch/contriever>)

Recorded evaluation configuration: {"Fine-tuned on MS MARCO":true}. This record does not assert independent reproduction.

### mDPR

Avg. Recall@100: 60.6. Result ID: 69930ea73a6c31484913646a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c2c1e209131cea32868.md>)

Recorded evaluation configuration: {"Method Group":"Baselines (Prior Work)"}. This record does not assert independent reproduction.

### CORA

Avg. Recall@100: 59.8. Result ID: 698f9ace3a6c314849fa7656.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d4b1e209131cea38d50.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b181e209131cea2c1ff.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### XLM-R

Avg. Recall@100: 59.6. Result ID: 698f9ace3a6c314849fa7658.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d4b1e209131cea38d50.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67521e209131cea15706.md>)

Recorded evaluation configuration: {"Fine-tuned on MS MARCO":true}. This record does not assert independent reproduction.

### mBERT

Avg. Recall@100: 57.9. Result ID: 698f9ace3a6c314849fa7657.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d4b1e209131cea38d50.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65c31e209131cea0c4ed.md>)

[Implementation](<https://github.com/google-research/bert>)

Recorded evaluation configuration: {"Fine-tuned on MS MARCO":true}. This record does not assert independent reproduction.

### Qwen3-Embed-0.6B

Avg. Recall@100: 54.4. Result ID: 69bd171147f34ce924a79517.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79601e209131cea7d516.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### mContriever

Avg. Recall@100: 49.2. Result ID: 698f9ace3a6c314849fa7659.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d4b1e209131cea38d50.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d4b1e209131cea38d50.md>)

[Implementation](<https://github.com/facebookresearch/contriever>)

Recorded evaluation configuration: {"Fine-tuned on MS MARCO":false}. This record does not assert independent reproduction.

### M3-Embedding

Avg. Recall@100: 45.3. Result ID: 69930ea73a6c314849136470.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: {"Retrieval Strategy":"Sparse"}. This record does not assert independent reproduction.

### BGE-M3 Sparse

Avg. Recall@100: 45.3. Result ID: 69bd171147f34ce924a7951a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

[Implementation](<https://github.com/FlagOpen/FlagEmbedding>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### BM25

Avg. Recall@100: 39.9. Result ID: 69930ea73a6c314849136469.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73d91e209131cea5e7ab.md>)

Recorded evaluation configuration: {"Method Group":"Baselines (Prior Work)"}. This record does not assert independent reproduction.

### BM25

Avg. Recall@100: 39.9. Result ID: 69bd171147f34ce924a79519.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bc41e209131cea8a4c0.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

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