# Long multimodal document understanding on MM-NIAH

Source: [Long multimodal document understanding on MM-NIAH](<https://www.sota2.com/research/sota/long-multimodal-document-understanding-on-mm-niah>)

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

Data updated: 2026-04-28T09:47:57.117Z

Evaluation of model performance on the MM-NIAH dataset for long multimodal document understanding tasks, assessing metrics across different context lengths (e.g., 1K, 4K, 16K, 64K).

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 | Long multimodal document understanding |
| Dataset (including any stated split) | MM-NIAH |
| Metric | Overall Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 7 of 7 |
| 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

- [Overall Score](<https://www.sota2.com/research/sota/long-multimodal-document-understanding-on-mm-niah.md?metric=Overall%20Score>): Higher is better; unit: PERCENT.

- [Performance (1K Context)](<https://www.sota2.com/research/sota/long-multimodal-document-understanding-on-mm-niah.md?metric=Performance%20(1K%20Context)>): Higher is better; unit: PERCENT.

- [Performance (4K Context)](<https://www.sota2.com/research/sota/long-multimodal-document-understanding-on-mm-niah.md?metric=Performance%20(4K%20Context)>): Higher is better; unit: PERCENT.

- [Performance (16K Context)](<https://www.sota2.com/research/sota/long-multimodal-document-understanding-on-mm-niah.md?metric=Performance%20(16K%20Context)>): Higher is better; unit: PERCENT.

- [Performance (64K Context)](<https://www.sota2.com/research/sota/long-multimodal-document-understanding-on-mm-niah.md?metric=Performance%20(64K%20Context)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Overall Score | Result date | Report paper |
| --- | --- | --- | --- |
| InternVL1.5-RAG | 46.1 | Not recorded | 68fa76ad1e209131cea6e62a |
| PDF-WuKong | 43.3 | Not recorded | 68fa76ad1e209131cea6e62a |
| InternVL1.5 | 41.1 | Not recorded | 68fa76ad1e209131cea6e62a |
| llava-v1.6-34b | 20.6 | Not recorded | 68fa76ad1e209131cea6e62a |
| llava-v1.6-13b | 16.9 | Not recorded | 68fa76ad1e209131cea6e62a |
| VILA1.0-13b | 15.7 | Not recorded | 68fa76ad1e209131cea6e62a |
| Emu2-Chat | 8.8 | Not recorded | 68fa76ad1e209131cea6e62a |

### InternVL1.5-RAG

Overall Score: 46.1. Result ID: 69f0823a47f34ce924d6e1b2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76ad1e209131cea6e62a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73081e209131cea59f8b.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"#param":"45B","zero-shot":"true"}. This record does not assert independent reproduction.

### PDF-WuKong

Overall Score: 43.3. Result ID: 69f0823a47f34ce924d6e1b3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76ad1e209131cea6e62a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76ad1e209131cea6e62a.md>)

Recorded evaluation configuration: {"#param":"8.5B","zero-shot":"true"}. This record does not assert independent reproduction.

### InternVL1.5

Overall Score: 41.1. Result ID: 69f0823a47f34ce924d6e1b1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76ad1e209131cea6e62a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73081e209131cea59f8b.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"#param":"26B","zero-shot":"true"}. This record does not assert independent reproduction.

### llava-v1.6-34b

Overall Score: 20.6. Result ID: 69f0823a47f34ce924d6e1b0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76ad1e209131cea6e62a.md>)

Recorded evaluation configuration: {"#param":"34B","zero-shot":"true"}. This record does not assert independent reproduction.

### llava-v1.6-13b

Overall Score: 16.9. Result ID: 69f0823a47f34ce924d6e1af.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76ad1e209131cea6e62a.md>)

Recorded evaluation configuration: {"#param":"13B","zero-shot":"true"}. This record does not assert independent reproduction.

### VILA1.0-13b

Overall Score: 15.7. Result ID: 69f0823a47f34ce924d6e1ae.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76ad1e209131cea6e62a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73311e209131cea5b06b.md>)

[Implementation](<https://github.com/NVlabs/VILA>)

Recorded evaluation configuration: {"#param":"13B","zero-shot":"true"}. This record does not assert independent reproduction.

### Emu2-Chat

Overall Score: 8.8. Result ID: 69f0823a47f34ce924d6e1ad.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76ad1e209131cea6e62a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa732a1e209131cea5add9.md>)

[Implementation](<https://github.com/baaivision/Emu>)

Recorded evaluation configuration: {"#param":"37B","zero-shot":"true"}. This record does not assert independent reproduction.

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