# Mathematical Information Retrieval on NTCIR Wiki-Formula 12

Source: [Mathematical Information Retrieval on NTCIR Wiki-Formula 12](<https://www.sota2.com/research/sota/mathematical-information-retrieval-on-ntcir-wiki-formula-12>)

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

Data updated: 2026-02-26T05:37:25.541Z

Evaluation of Mathematical Information Retrieval performance on the NTCIR Wiki-Formula 12 dataset, measured using BPref and Judged % metrics.

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 | Mathematical Information Retrieval |
| Dataset (including any stated split) | NTCIR Wiki-Formula 12 |
| Metric | BPref (Full) |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 17 of 17 |
| 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

- [BPref (Full)](<https://www.sota2.com/research/sota/mathematical-information-retrieval-on-ntcir-wiki-formula-12.md?metric=BPref%20(Full)>): Higher is better; unit: SCALAR.

- [BPref (Partial)](<https://www.sota2.com/research/sota/mathematical-information-retrieval-on-ntcir-wiki-formula-12.md?metric=BPref%20(Partial)>): Higher is better; unit: SCALAR.

- [Judged %](<https://www.sota2.com/research/sota/mathematical-information-retrieval-on-ntcir-wiki-formula-12.md?metric=Judged%20%25>): Higher is better; unit: PERCENT.

## Recorded results

| Method | BPref (Full) | Result date | Report paper |
| --- | --- | --- | --- |
| Approach0 + DPR | 0.7 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| Approach0 + ColBERT | 0.65 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| Tangent-S | 0.64 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| GNN | 0.63 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| Approach0 | 0.63 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| MathBERT | 0.614 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| MathBERT | 0.61 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| Tangent-CFT | 0.6 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| MCAT | 0.57 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| Approach0 + ColBERT | 0.56 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| ColBERT | 0.545 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| Approach0 + DPR | 0.53 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| further-pretrained BERT | 0.522 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| DPR | 0.516 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| SciBERT | 0.512 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| BERT | 0.505 | Not recorded | 68fa6dbe1e209131cea3bca8 |
| further-pretrained BERT | 0.486 | Not recorded | 68fa6dbe1e209131cea3bca8 |

### Approach0 + DPR

BPref (Full): 0.7. Result ID: 698fe35a3a6c314849fc9523.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Implementation](<https://github.com/approach0/math-dense-retrievers/tree/emnlp2022>)

Recorded evaluation configuration: {"strategy":"end-to-end fusion"}. This record does not assert independent reproduction.

### Approach0 + ColBERT

BPref (Full): 0.65. Result ID: 698fe35a3a6c314849fc9522.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Implementation](<https://github.com/approach0/math-dense-retrievers/tree/emnlp2022>)

Recorded evaluation configuration: {"strategy":"end-to-end fusion"}. This record does not assert independent reproduction.

### Tangent-S

BPref (Full): 0.64. Result ID: 698fe35a3a6c314849fc951b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

Recorded evaluation configuration: {"type":"Previous system"}. This record does not assert independent reproduction.

### GNN

BPref (Full): 0.63. Result ID: 698fe35a3a6c314849fc951d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

Recorded evaluation configuration: {"type":"Previous system"}. This record does not assert independent reproduction.

### Approach0

BPref (Full): 0.63. Result ID: 698fe35a3a6c314849fc951f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

Recorded evaluation configuration: {"type":"Structure search baseline"}. This record does not assert independent reproduction.

### MathBERT

BPref (Full): 0.614. Result ID: 698fe35a3a6c314849fc950c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69d21e209131cea246a8.md>)

Recorded evaluation configuration: {"Ensemble system":true}. This record does not assert independent reproduction.

### MathBERT

BPref (Full): 0.61. Result ID: 698fe35a3a6c314849fc951e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69d21e209131cea246a8.md>)

Recorded evaluation configuration: {"type":"Previous system"}. This record does not assert independent reproduction.

### Tangent-CFT

BPref (Full): 0.6. Result ID: 698fe35a3a6c314849fc951c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

Recorded evaluation configuration: {"type":"Previous system"}. This record does not assert independent reproduction.

### MCAT

BPref (Full): 0.57. Result ID: 698fe35a3a6c314849fc951a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

Recorded evaluation configuration: {"type":"Previous system"}. This record does not assert independent reproduction.

### Approach0 + ColBERT

BPref (Full): 0.56. Result ID: 698fe35a3a6c314849fc9521.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Implementation](<https://github.com/approach0/math-dense-retrievers/tree/emnlp2022>)

Recorded evaluation configuration: {"strategy":"reranking"}. This record does not assert independent reproduction.

### ColBERT

BPref (Full): 0.545. Result ID: 698fe35a3a6c314849fc950e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Implementation](<https://github.com/approach0/math-dense-retrievers/tree/emnlp2022>)

Recorded evaluation configuration: {"Training stage":"fully-trained"}. This record does not assert independent reproduction.

### Approach0 + DPR

BPref (Full): 0.53. Result ID: 698fe35a3a6c314849fc9520.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Implementation](<https://github.com/approach0/math-dense-retrievers/tree/emnlp2022>)

Recorded evaluation configuration: {"strategy":"reranking"}. This record does not assert independent reproduction.

### further-pretrained BERT

BPref (Full): 0.522. Result ID: 698fe35a3a6c314849fc950b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

Recorded evaluation configuration: {"Pre-trained backbone":"BERT","Fine-tuning epochs":"1","Further-pretraining epochs":"7"}. This record does not assert independent reproduction.

### DPR

BPref (Full): 0.516. Result ID: 698fe35a3a6c314849fc950d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Implementation](<https://github.com/approach0/math-dense-retrievers/tree/emnlp2022>)

Recorded evaluation configuration: {"Training stage":"fully-trained"}. This record does not assert independent reproduction.

### SciBERT

BPref (Full): 0.512. Result ID: 698fe35a3a6c314849fc9509.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66551e209131cea0f8c9.md>)

[Implementation](<https://github.com/allenai/scibert/>)

Recorded evaluation configuration: {"Pre-trained backbone":"SciBERT","Fine-tuning epochs":"1"}. This record does not assert independent reproduction.

### BERT

BPref (Full): 0.505. Result ID: 698fe35a3a6c314849fc9508.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

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

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

Recorded evaluation configuration: {"Pre-trained backbone":"BERT","Fine-tuning epochs":"1"}. This record does not assert independent reproduction.

### further-pretrained BERT

BPref (Full): 0.486. Result ID: 698fe35a3a6c314849fc950a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6dbe1e209131cea3bca8.md>)

Recorded evaluation configuration: {"Pre-trained backbone":"BERT","Fine-tuning epochs":"1","Further-pretraining epochs":"3"}. This record does not assert independent reproduction.

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