# Grounded Multi-Video Question Answering on MAGMaR (test)

Source: [Grounded Multi-Video Question Answering on MAGMaR (test)](<https://www.sota2.com/research/sota/grounded-multi-video-question-answering-on-magmar-test>)

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

Data updated: 2026-05-20T07:33:40.274Z

Evaluation of Grounded Multi-Video Question Answering performance on the test split of the MAGMaR dataset, reporting precision, recall, and F1 scores for reference and citation 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 | Grounded Multi-Video Question Answering |
| Dataset (including any stated split) | MAGMaR (test) |
| Metric | Reference Precision (Ref-P) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 11 of 11 |
| 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

- [Reference Precision (Ref-P)](<https://www.sota2.com/research/sota/grounded-multi-video-question-answering-on-magmar-test.md?metric=Reference%20Precision%20(Ref-P)>): Higher is better; unit: PERCENT.

- [Reference Recall (Ref-R)](<https://www.sota2.com/research/sota/grounded-multi-video-question-answering-on-magmar-test.md?metric=Reference%20Recall%20(Ref-R)>): Higher is better; unit: PERCENT.

- [Reference F1 Score (Ref-F1)](<https://www.sota2.com/research/sota/grounded-multi-video-question-answering-on-magmar-test.md?metric=Reference%20F1%20Score%20(Ref-F1)>): Higher is better; unit: PERCENT.

- [Citation Precision (Cite-P)](<https://www.sota2.com/research/sota/grounded-multi-video-question-answering-on-magmar-test.md?metric=Citation%20Precision%20(Cite-P)>): Higher is better; unit: PERCENT.

- [Citation Recall (Cite-R)](<https://www.sota2.com/research/sota/grounded-multi-video-question-answering-on-magmar-test.md?metric=Citation%20Recall%20(Cite-R)>): Higher is better; unit: PERCENT.

- [Citation F1 Score (Cite-F1)](<https://www.sota2.com/research/sota/grounded-multi-video-question-answering-on-magmar-test.md?metric=Citation%20F1%20Score%20(Cite-F1)>): Higher is better; unit: PERCENT.

- [Average Score (Avg)](<https://www.sota2.com/research/sota/grounded-multi-video-question-answering-on-magmar-test.md?metric=Average%20Score%20(Avg)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Reference Precision (Ref-P) | Result date | Report paper |
| --- | --- | --- | --- |
| CRAFT | 82.2 | Not recorded | 6a0d3feff984739ed6737669 |
| CRAFT | 80.8 | Not recorded | 6a0d3feff984739ed6737669 |
| CRAFT | 77.5 | Not recorded | 6a0d3feff984739ed6737669 |
| InternVL-3.5-30B-A3B | 76.1 | Not recorded | 6a0d3feff984739ed6737669 |
| CRAFT | 76 | Not recorded | 6a0d3feff984739ed6737669 |
| InternVL-3.5-30B-A3B | 74.9 | Not recorded | 6a0d3feff984739ed6737669 |
| Gemma-4-31B | 71.2 | Not recorded | 6a0d3feff984739ed6737669 |
| Gemma-4-31B | 70.1 | Not recorded | 6a0d3feff984739ed6737669 |
| Molmo2-8B | 62.3 | Not recorded | 6a0d3feff984739ed6737669 |
| CRAFT | 49.1 | Not recorded | 6a0d3feff984739ed6737669 |
| CRAFT | 43.7 | Not recorded | 6a0d3feff984739ed6737669 |

### CRAFT

Reference Precision (Ref-P): 82.2. Result ID: 6a0d6367fbc373592c051893.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

Recorded evaluation configuration: {"Base VLM":"Qwen3.5-9B","Frame Sampling":"DKS","Number of Frames":"64"}. This record does not assert independent reproduction.

### CRAFT

Reference Precision (Ref-P): 80.8. Result ID: 6a0d6367fbc373592c051890.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

Recorded evaluation configuration: {"Base VLM":"Qwen3.5-9B","Atomic Claims":"true","Number of Frames":"128"}. This record does not assert independent reproduction.

### CRAFT

Reference Precision (Ref-P): 77.5. Result ID: 6a0d6367fbc373592c051892.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

Recorded evaluation configuration: {"Base VLM":"Qwen3.5-9B","Frame Sampling":"uniform","Number of Frames":"64"}. This record does not assert independent reproduction.

### InternVL-3.5-30B-A3B

Reference Precision (Ref-P): 76.1. Result ID: 6a0d6367fbc373592c05188b.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78241e209131cea76fa5.md>)

Recorded evaluation configuration: {"Number of Frames":"128","ASR access":"true"}. This record does not assert independent reproduction.

### CRAFT

Reference Precision (Ref-P): 76. Result ID: 6a0d6367fbc373592c051891.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

Recorded evaluation configuration: {"Base VLM":"Qwen3.5-9B","ASR access":"true","Number of Frames":"128"}. This record does not assert independent reproduction.

### InternVL-3.5-30B-A3B

Reference Precision (Ref-P): 74.9. Result ID: 6a0d6367fbc373592c05188a.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78241e209131cea76fa5.md>)

Recorded evaluation configuration: {"Number of Frames":"128","ASR access":"false"}. This record does not assert independent reproduction.

### Gemma-4-31B

Reference Precision (Ref-P): 71.2. Result ID: 6a0d6367fbc373592c05188d.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72f11e209131cea5973c.md>)

Recorded evaluation configuration: {"Number of Frames":"128","ASR access":"true"}. This record does not assert independent reproduction.

### Gemma-4-31B

Reference Precision (Ref-P): 70.1. Result ID: 6a0d6367fbc373592c05188c.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72f11e209131cea5973c.md>)

Recorded evaluation configuration: {"Number of Frames":"128","ASR access":"false"}. This record does not assert independent reproduction.

### Molmo2-8B

Reference Precision (Ref-P): 62.3. Result ID: 6a0d6367fbc373592c051889.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6969c61ce25ad2fa9d1eca03.md>)

Recorded evaluation configuration: {"Number of Frames":"128"}. This record does not assert independent reproduction.

### CRAFT

Reference Precision (Ref-P): 49.1. Result ID: 6a0d6367fbc373592c05188f.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

Recorded evaluation configuration: {"Base VLM":"Qwen3.5-9B","Critic Loop":"true","Number of Frames":"128"}. This record does not assert independent reproduction.

### CRAFT

Reference Precision (Ref-P): 43.7. Result ID: 6a0d6367fbc373592c05188e.

[Reporting paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0d3feff984739ed6737669.md>)

Recorded evaluation configuration: {"Base VLM":"Qwen3.5-9B","Configuration":"Baseline","Number of Frames":"128"}. This record does not assert independent reproduction.

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