# Vision-Language Instruction Following on NuInstruct 25% Token Retention

Source: [Vision-Language Instruction Following on NuInstruct 25% Token Retention](<https://www.sota2.com/research/sota/vision-language-instruction-following-on-nuinstruct-25-percent-token-retention>)

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

Data updated: 2026-04-23T08:08:56.215Z

Evaluation of model performance on following vision-language instructions using the NuInstruct dataset restricted to 25% token retention.

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 | Vision-Language Instruction Following |
| Dataset (including any stated split) | NuInstruct 25% Token Retention |
| Metric | MAE |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 5 of 5 |
| 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

- [MAE](<https://www.sota2.com/research/sota/vision-language-instruction-following-on-nuinstruct-25-percent-token-retention.md?metric=MAE>): Lower is better; unit: SCALAR.

- [Accuracy](<https://www.sota2.com/research/sota/vision-language-instruction-following-on-nuinstruct-25-percent-token-retention.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [mAP](<https://www.sota2.com/research/sota/vision-language-instruction-following-on-nuinstruct-25-percent-token-retention.md?metric=mAP>): Higher is better; unit: PERCENT.

- [BLEU](<https://www.sota2.com/research/sota/vision-language-instruction-following-on-nuinstruct-25-percent-token-retention.md?metric=BLEU>): Higher is better; unit: PERCENT.

- [Average Score](<https://www.sota2.com/research/sota/vision-language-instruction-following-on-nuinstruct-25-percent-token-retention.md?metric=Average%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | MAE | Result date | Report paper |
| --- | --- | --- | --- |
| ST-Prune | 3.49 | Not recorded | 69e85ba27d0172140ec80f94 |
| DriveMM (Vanilla) | 3.5 | Not recorded | 69e85ba27d0172140ec80f94 |
| VisPruner | 3.68 | Not recorded | 69e85ba27d0172140ec80f94 |
| DivPrune | 3.78 | Not recorded | 69e85ba27d0172140ec80f94 |
| PACT | 10.82 | Not recorded | 69e85ba27d0172140ec80f94 |

### ST-Prune

MAE: 3.49. Result ID: 69e9d39747f34ce924d0d958.

[Reporting paper](<https://www.sota2.com/research/paper/69e85ba27d0172140ec80f94.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e85ba27d0172140ec80f94.md>)

Recorded evaluation configuration: {"Token Retention Rate":"25%"}. This record does not assert independent reproduction.

### DriveMM (Vanilla)

MAE: 3.5. Result ID: 69e9d39747f34ce924d0d954.

[Reporting paper](<https://www.sota2.com/research/paper/69e85ba27d0172140ec80f94.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a741e209131cea835a8.md>)

Recorded evaluation configuration: {"Token Retention Rate":"Full Tokens"}. This record does not assert independent reproduction.

### VisPruner

MAE: 3.68. Result ID: 69e9d39747f34ce924d0d955.

[Reporting paper](<https://www.sota2.com/research/paper/69e85ba27d0172140ec80f94.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78741e209131cea786bf.md>)

Recorded evaluation configuration: {"Token Retention Rate":"25%","Venue":"ICCV 2025"}. This record does not assert independent reproduction.

### DivPrune

MAE: 3.78. Result ID: 69e9d39747f34ce924d0d956.

[Reporting paper](<https://www.sota2.com/research/paper/69e85ba27d0172140ec80f94.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c41e209131cea7532a.md>)

Recorded evaluation configuration: {"Token Retention Rate":"25%","Venue":"CVPR 2025"}. This record does not assert independent reproduction.

### PACT

MAE: 10.82. Result ID: 69e9d39747f34ce924d0d957.

[Reporting paper](<https://www.sota2.com/research/paper/69e85ba27d0172140ec80f94.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77ec1e209131cea76055.md>)

Recorded evaluation configuration: {"Token Retention Rate":"25%","Venue":"CVPR 2025"}. This record does not assert independent reproduction.

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