# Image Description tasks

Source: [Image Description tasks](<https://www.sota2.com/research/sota/image-description-tasks>)

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

Data updated: 2026-03-26T15:24:13.809Z

Evaluation of image captioning models using standard NLP metrics including BLEU scores (1-4), ROUGE-L F1, BERTScore, and SBERT Similarity.

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 | Image Description |
| Dataset (including any stated split) | Image Description tasks |
| Metric | ROUGE-L F1 |
| Unit | PERCENT |
| Direction | Higher 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

- [ROUGE-L F1](<https://www.sota2.com/research/sota/image-description-tasks.md?metric=ROUGE-L%20F1>): Higher is better; unit: PERCENT.

- [BLEU-1](<https://www.sota2.com/research/sota/image-description-tasks.md?metric=BLEU-1>): Higher is better; unit: PERCENT.

- [BLEU-2](<https://www.sota2.com/research/sota/image-description-tasks.md?metric=BLEU-2>): Higher is better; unit: PERCENT.

- [BLEU-3](<https://www.sota2.com/research/sota/image-description-tasks.md?metric=BLEU-3>): Higher is better; unit: PERCENT.

- [BLEU-4](<https://www.sota2.com/research/sota/image-description-tasks.md?metric=BLEU-4>): Higher is better; unit: PERCENT.

- [BERTScore F1](<https://www.sota2.com/research/sota/image-description-tasks.md?metric=BERTScore%20F1>): Higher is better; unit: PERCENT.

- [SBERT Similarity](<https://www.sota2.com/research/sota/image-description-tasks.md?metric=SBERT%20Similarity>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ROUGE-L F1 | Result date | Report paper |
| --- | --- | --- | --- |
| VOLMO-2B | 21.7 | Not recorded | 69c4bdad0636e5bf4afc76b5 |
| LLaVA-Med | 16.44 | Not recorded | 69c4bdad0636e5bf4afc76b5 |
| MedGemma-4B | 12.12 | Not recorded | 69c4bdad0636e5bf4afc76b5 |
| MedGemma-27B | 11.14 | Not recorded | 69c4bdad0636e5bf4afc76b5 |
| InternVL-2B | 10.7 | Not recorded | 69c4bdad0636e5bf4afc76b5 |

### VOLMO-2B

ROUGE-L F1: 21.7. Result ID: 69c54f9247f34ce924af875c.

[Reporting paper](<https://www.sota2.com/research/paper/69c4bdad0636e5bf4afc76b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c4bdad0636e5bf4afc76b5.md>)

Recorded evaluation configuration: {"size":"2B"}. This record does not assert independent reproduction.

### LLaVA-Med

ROUGE-L F1: 16.44. Result ID: 69c54f9247f34ce924af8759.

[Reporting paper](<https://www.sota2.com/research/paper/69c4bdad0636e5bf4afc76b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f8e1e209131cea46356.md>)

[Implementation](<https://aka.ms/llava-med>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### MedGemma-4B

ROUGE-L F1: 12.12. Result ID: 69c54f9247f34ce924af875a.

[Reporting paper](<https://www.sota2.com/research/paper/69c4bdad0636e5bf4afc76b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79ff1e209131cea808a4.md>)

Recorded evaluation configuration: {"size":"4B"}. This record does not assert independent reproduction.

### MedGemma-27B

ROUGE-L F1: 11.14. Result ID: 69c54f9247f34ce924af875b.

[Reporting paper](<https://www.sota2.com/research/paper/69c4bdad0636e5bf4afc76b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79ff1e209131cea808a4.md>)

Recorded evaluation configuration: {"size":"27B"}. This record does not assert independent reproduction.

### InternVL-2B

ROUGE-L F1: 10.7. Result ID: 69c54f9247f34ce924af8758.

[Reporting paper](<https://www.sota2.com/research/paper/69c4bdad0636e5bf4afc76b5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71de1e209131cea5370d.md>)

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

Recorded evaluation configuration: {"size":"2B"}. This record does not assert independent reproduction.

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