# Multi-page Document Question Answering on MP-DocVQA (test)

Source: [Multi-page Document Question Answering on MP-DocVQA (test)](<https://www.sota2.com/research/sota/multi-page-document-question-answering-on-mp-docvqa-test>)

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

Data updated: 2026-02-26T05:22:34.252Z

Evaluation of model performance on the test split of the MP-DocVQA dataset for Multi-page Document Question Answering, measured using the ANLS metric.

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 | Multi-page Document Question Answering |
| Dataset (including any stated split) | MP-DocVQA (test) |
| Metric | ANLS |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 30 of 30 |
| 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

- [ANLS](<https://www.sota2.com/research/sota/multi-page-document-question-answering-on-mp-docvqa-test.md?metric=ANLS>): Higher is better; unit: SCALAR.

- [Page Pred. Acc](<https://www.sota2.com/research/sota/multi-page-document-question-answering-on-mp-docvqa-test.md?metric=Page%20Pred.%20Acc>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ANLS | Result date | Report paper |
| --- | --- | --- | --- |
| AVIR | 0.8458 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| M3DOCRAG | 0.8444 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| Qwen2.5-3B-AWQ | 0.8405 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| Arctic-TILT | 0.8122 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| GRAM | 0.8032 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| ScreenAI | 0.7711 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| ScreenAI | 0.771 | Not recorded | 68fa74061e209131cea5f4cc |
| ScreenAI | 0.729 | Not recorded | 68fa74061e209131cea5f4cc |
| RMT5 | 0.6401 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| Hi-VT5 | 0.6201 | Not recorded | 68fa707c1e209131cea4b65b |
| Hi-VT5 | 0.6201 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| SelfAttnScoring | 0.6199 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| Hi-VT5 | 0.618 | Not recorded | 68fa74061e209131cea5f4cc |
| Hi-VT5 | 0.618 | Not recorded | 68fa74061e209131cea5f4cc |
| Claude + LATIN-Prompt | 0.6129 | Not recorded | 68fa707c1e209131cea4b65b |
| Big Bird | 0.5854 | Not recorded | 68fa707c1e209131cea4b65b |
| Big Bird | 0.5854 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| LayoutLMv3 | 0.5513 | Not recorded | 68fa707c1e209131cea4b65b |
| LayoutLMv3 | 0.5513 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| Longformer | 0.5506 | Not recorded | 68fa707c1e209131cea4b65b |
| Longformer | 0.5506 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| BERT | 0.5347 | Not recorded | 68fa707c1e209131cea4b65b |
| BERT | 0.5347 | Not recorded | 69705d59e25ad2fa9d1ecdde |
| Longformer | 0.5287 | Not recorded | 68fa707c1e209131cea4b65b |
| T5 | 0.505 | Not recorded | 68fa707c1e209131cea4b65b |
| Big Bird | 0.4929 | Not recorded | 68fa707c1e209131cea4b65b |
| LayoutLMv3 | 0.4538 | Not recorded | 68fa707c1e209131cea4b65b |
| BERT | 0.4183 | Not recorded | 68fa707c1e209131cea4b65b |
| T5 | 0.4028 | Not recorded | 68fa707c1e209131cea4b65b |
| T5 | 0.4028 | Not recorded | 69705d59e25ad2fa9d1ecdde |

### AVIR

ANLS: 0.8458. Result ID: 6997b24e5698665ec4087beb.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

Recorded evaluation configuration: {"Year":null,"OCR":false,"Params.":"3B"}. This record does not assert independent reproduction.

### M3DOCRAG

ANLS: 0.8444. Result ID: 6997b24e5698665ec4087be9.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa759f1e209131cea69082.md>)

Recorded evaluation configuration: {"Year":2024,"OCR":false,"Params.":"8B"}. This record does not assert independent reproduction.

### Qwen2.5-3B-AWQ

ANLS: 0.8405. Result ID: 6997b24e5698665ec4087bea.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Year":2025,"OCR":false,"Params.":"3B"}. This record does not assert independent reproduction.

### Arctic-TILT

ANLS: 0.8122. Result ID: 6997b24e5698665ec4087be6.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74701e209131cea61c51.md>)

Recorded evaluation configuration: {"Year":2024,"OCR":true,"Params.":"800M"}. This record does not assert independent reproduction.

### GRAM

ANLS: 0.8032. Result ID: 6997b24e5698665ec4087be5.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72951e209131cea5749e.md>)

Recorded evaluation configuration: {"Year":2024,"OCR":true,"Params.":"859M"}. This record does not assert independent reproduction.

### ScreenAI

ANLS: 0.7711. Result ID: 6997b24e5698665ec4087be4.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74061e209131cea5f4cc.md>)

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

Recorded evaluation configuration: {"Year":2024,"OCR":true,"Params.":"5B"}. This record does not assert independent reproduction.

### ScreenAI

ANLS: 0.771. Result ID: 699315313a6c3148491394d7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74061e209131cea5f4cc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74061e209131cea5f4cc.md>)

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

Recorded evaluation configuration: {"OCR Usage":"With OCR"}. This record does not assert independent reproduction.

### ScreenAI

ANLS: 0.729. Result ID: 699315313a6c3148491394d5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74061e209131cea5f4cc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74061e209131cea5f4cc.md>)

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

Recorded evaluation configuration: {"OCR Usage":"Without OCR"}. This record does not assert independent reproduction.

### RMT5

ANLS: 0.6401. Result ID: 6997b24e5698665ec4087be7.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

Recorded evaluation configuration: {"Year":2024,"OCR":true,"Params.":"312M"}. This record does not assert independent reproduction.

### Hi-VT5

ANLS: 0.6201. Result ID: 6991445a3a6c31484906e87b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eec1e209131cea42e4e.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Multipage"}. This record does not assert independent reproduction.

### Hi-VT5

ANLS: 0.6201. Result ID: 6997b24e5698665ec4087be3.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eec1e209131cea42e4e.md>)

Recorded evaluation configuration: {"Year":2023,"OCR":true,"Params.":"316M"}. This record does not assert independent reproduction.

### SelfAttnScoring

ANLS: 0.6199. Result ID: 6997b24e5698665ec4087be8.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73091e209131cea59fb1.md>)

Recorded evaluation configuration: {"Year":2024,"OCR":false,"Params.":"273M"}. This record does not assert independent reproduction.

### Hi-VT5

ANLS: 0.618. Result ID: 699315313a6c3148491394d4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74061e209131cea5f4cc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eec1e209131cea42e4e.md>)

Recorded evaluation configuration: {"OCR Usage":"SOTA Overall"}. This record does not assert independent reproduction.

### Hi-VT5

ANLS: 0.618. Result ID: 699315313a6c3148491394d6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74061e209131cea5f4cc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eec1e209131cea42e4e.md>)

Recorded evaluation configuration: {"OCR Usage":"With OCR"}. This record does not assert independent reproduction.

### Claude + LATIN-Prompt

ANLS: 0.6129. Result ID: 6991445a3a6c31484906e87c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Implementation](<https://github.com/WenjinW/LATIN-Prompt>)

Recorded evaluation configuration: {"Paradigm":"Zero-shot","Setup":"Max Conf"}. This record does not assert independent reproduction.

### Big Bird

ANLS: 0.5854. Result ID: 6991445a3a6c31484906e875.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa690b1e209131cea1faf6.md>)

[Implementation](<http://goo.gle/bigbird-transformer>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Max Conf"}. This record does not assert independent reproduction.

### Big Bird

ANLS: 0.5854. Result ID: 6997b24e5698665ec4087be1.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa690b1e209131cea1faf6.md>)

[Implementation](<http://goo.gle/bigbird-transformer>)

Recorded evaluation configuration: {"Year":2020,"OCR":true,"Params.":"131M"}. This record does not assert independent reproduction.

### LayoutLMv3

ANLS: 0.5513. Result ID: 6991445a3a6c31484906e877.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d171e209131cea379cd.md>)

[Implementation](<https://aka.ms/layoutlmv3>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Max Conf"}. This record does not assert independent reproduction.

### LayoutLMv3

ANLS: 0.5513. Result ID: 6997b24e5698665ec4087be2.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d171e209131cea379cd.md>)

[Implementation](<https://aka.ms/layoutlmv3>)

Recorded evaluation configuration: {"Year":2022,"OCR":true,"Params.":"125M"}. This record does not assert independent reproduction.

### Longformer

ANLS: 0.5506. Result ID: 6991445a3a6c31484906e873.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68d41e209131cea1e701.md>)

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

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Max Conf"}. This record does not assert independent reproduction.

### Longformer

ANLS: 0.5506. Result ID: 6997b24e5698665ec4087be0.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68d41e209131cea1e701.md>)

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

Recorded evaluation configuration: {"Year":2020,"OCR":true,"Params.":"148M"}. This record does not assert independent reproduction.

### BERT

ANLS: 0.5347. Result ID: 6991445a3a6c31484906e871.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

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

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

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Max Conf"}. This record does not assert independent reproduction.

### BERT

ANLS: 0.5347. Result ID: 6997b24e5698665ec4087bde.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

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

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

Recorded evaluation configuration: {"Year":2018,"OCR":true,"Params.":"334M"}. This record does not assert independent reproduction.

### Longformer

ANLS: 0.5287. Result ID: 6991445a3a6c31484906e874.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68d41e209131cea1e701.md>)

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

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Concat"}. This record does not assert independent reproduction.

### T5

ANLS: 0.505. Result ID: 6991445a3a6c31484906e87a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa709e1e209131cea4c470.md>)

[Implementation](<https://github.com/google-research/text-to-text-transfer-transformer>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Concat"}. This record does not assert independent reproduction.

### Big Bird

ANLS: 0.4929. Result ID: 6991445a3a6c31484906e876.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa690b1e209131cea1faf6.md>)

[Implementation](<http://goo.gle/bigbird-transformer>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Concat"}. This record does not assert independent reproduction.

### LayoutLMv3

ANLS: 0.4538. Result ID: 6991445a3a6c31484906e878.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d171e209131cea379cd.md>)

[Implementation](<https://aka.ms/layoutlmv3>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Concat"}. This record does not assert independent reproduction.

### BERT

ANLS: 0.4183. Result ID: 6991445a3a6c31484906e872.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

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

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

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Concat"}. This record does not assert independent reproduction.

### T5

ANLS: 0.4028. Result ID: 6991445a3a6c31484906e879.

[Reporting paper](<https://www.sota2.com/research/paper/68fa707c1e209131cea4b65b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa709e1e209131cea4c470.md>)

[Implementation](<https://github.com/google-research/text-to-text-transfer-transformer>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Setup":"Max Conf"}. This record does not assert independent reproduction.

### T5

ANLS: 0.4028. Result ID: 6997b24e5698665ec4087bdf.

[Reporting paper](<https://www.sota2.com/research/paper/69705d59e25ad2fa9d1ecdde.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa709e1e209131cea4c470.md>)

[Implementation](<https://github.com/google-research/text-to-text-transfer-transformer>)

Recorded evaluation configuration: {"Year":2020,"OCR":true,"Params.":"223M"}. This record does not assert independent reproduction.

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