# Multilingual Retrieval on WSDM Cup 2026 (test)

Source: [Multilingual Retrieval on WSDM Cup 2026 (test)](<https://www.sota2.com/research/sota/multilingual-retrieval-on-wsdm-cup-2026-test>)

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

Data updated: 2026-02-26T06:11:10.172Z

Evaluation of Multilingual Retrieval performance on the test split of the WSDM Cup 2026 challenge.

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 Retrieval |
| Dataset (including any stated split) | WSDM Cup 2026 (test) |
| Metric | nDCG@20 |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 9 of 9 |
| 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

- [nDCG@20](<https://www.sota2.com/research/sota/multilingual-retrieval-on-wsdm-cup-2026-test.md?metric=nDCG%4020>): Higher is better; unit: SCALAR.

- [Judged@20](<https://www.sota2.com/research/sota/multilingual-retrieval-on-wsdm-cup-2026-test.md?metric=Judged%4020>): Higher is better; unit: SCALAR.

## Recorded results

| Method | nDCG@20 | Result date | Report paper |
| --- | --- | --- | --- |
| Low-Cost Multi-Stage Retrieval Pipeline | 0.403 | Not recorded | 6997eab509c33900fcd76755 |
| MILCO | 0.395 | Not recorded | 6997eab509c33900fcd76755 |
| JinaV3 | 0.355 | Not recorded | 6997eab509c33900fcd76755 |
| Arctic-Embed Large v2 | 0.352 | Not recorded | 6997eab509c33900fcd76755 |
| BM25 | 0.349 | Not recorded | 6997eab509c33900fcd76755 |
| Qwen3 0.6B Embed | 0.309 | Not recorded | 6997eab509c33900fcd76755 |
| RepLlama | 0.255 | Not recorded | 6997eab509c33900fcd76755 |
| e5 Large | 0.205 | Not recorded | 6997eab509c33900fcd76755 |
| BGE-M3 Sparse | 0.054 | Not recorded | 6997eab509c33900fcd76755 |

### Low-Cost Multi-Stage Retrieval Pipeline

nDCG@20: 0.403. Result ID: 699b02ba5698665ec494d224.

[Reporting paper](<https://www.sota2.com/research/paper/6997eab509c33900fcd76755.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6997eab509c33900fcd76755.md>)

Recorded evaluation configuration: {"Pipeline":"GRF expansion + BM25 + jina-embeddings-v4 + Qwen3-Reranker-4B"}. This record does not assert independent reproduction.

### MILCO

nDCG@20: 0.395. Result ID: 699b02ba5698665ec494d223.

[Reporting paper](<https://www.sota2.com/research/paper/6997eab509c33900fcd76755.md>)

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

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

### JinaV3

nDCG@20: 0.355. Result ID: 699b02ba5698665ec494d222.

[Reporting paper](<https://www.sota2.com/research/paper/6997eab509c33900fcd76755.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74d91e209131cea6424f.md>)

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

### Arctic-Embed Large v2

nDCG@20: 0.352. Result ID: 699b02ba5698665ec494d221.

[Reporting paper](<https://www.sota2.com/research/paper/6997eab509c33900fcd76755.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76381e209131cea6befd.md>)

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

### BM25

nDCG@20: 0.349. Result ID: 699b02ba5698665ec494d220.

[Reporting paper](<https://www.sota2.com/research/paper/6997eab509c33900fcd76755.md>)

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

### Qwen3 0.6B Embed

nDCG@20: 0.309. Result ID: 699b02ba5698665ec494d21f.

[Reporting paper](<https://www.sota2.com/research/paper/6997eab509c33900fcd76755.md>)

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

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

### RepLlama

nDCG@20: 0.255. Result ID: 699b02ba5698665ec494d21e.

[Reporting paper](<https://www.sota2.com/research/paper/6997eab509c33900fcd76755.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70db1e209131cea4dba5.md>)

Recorded evaluation configuration: {"Retrieval Type":"LLM-based"}. This record does not assert independent reproduction.

### e5 Large

nDCG@20: 0.205. Result ID: 699b02ba5698665ec494d21d.

[Reporting paper](<https://www.sota2.com/research/paper/6997eab509c33900fcd76755.md>)

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

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

### BGE-M3 Sparse

nDCG@20: 0.054. Result ID: 699b02ba5698665ec494d21c.

[Reporting paper](<https://www.sota2.com/research/paper/6997eab509c33900fcd76755.md>)

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

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

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

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