# 3D Object Captioning on Cap3D subset of 3186 3D objects (test)

Source: [3D Object Captioning on Cap3D subset of 3186 3D objects (test)](<https://www.sota2.com/research/sota/3d-object-captioning-on-cap3d-subset-of-3186-3d-objects-test>)

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

Data updated: 2026-04-21T10:27:42.796Z

Evaluation of 3D Object Captioning models using the specific test subset of the Cap3D dataset containing 3186 objects, assessed using CLIP similarity, FID, and Recall@k 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 | 3D Object Captioning |
| Dataset (including any stated split) | Cap3D subset of 3186 3D objects (test) |
| Metric | CLIP Image-Text Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 10 of 10 |
| 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

- [CLIP Image-Text Score](<https://www.sota2.com/research/sota/3d-object-captioning-on-cap3d-subset-of-3186-3d-objects-test.md?metric=CLIP%20Image-Text%20Score>): Higher is better; unit: SCALAR.

- [CLIP Text-Text Score](<https://www.sota2.com/research/sota/3d-object-captioning-on-cap3d-subset-of-3186-3d-objects-test.md?metric=CLIP%20Text-Text%20Score>): Higher is better; unit: SCALAR.

- [FID](<https://www.sota2.com/research/sota/3d-object-captioning-on-cap3d-subset-of-3186-3d-objects-test.md?metric=FID>): Lower is better; unit: SCALAR.

- [Recall@10](<https://www.sota2.com/research/sota/3d-object-captioning-on-cap3d-subset-of-3186-3d-objects-test.md?metric=Recall%4010>): Higher is better; unit: PERCENT.

- [Recall@5](<https://www.sota2.com/research/sota/3d-object-captioning-on-cap3d-subset-of-3186-3d-objects-test.md?metric=Recall%405>): Higher is better; unit: PERCENT.

- [Recall@1](<https://www.sota2.com/research/sota/3d-object-captioning-on-cap3d-subset-of-3186-3d-objects-test.md?metric=Recall%401>): Higher is better; unit: PERCENT.

- [Lexical Similarity](<https://www.sota2.com/research/sota/3d-object-captioning-on-cap3d-subset-of-3186-3d-objects-test.md?metric=Lexical%20Similarity>): Higher is better; unit: SCALAR.

## Recorded results

| Method | CLIP Image-Text Score | Result date | Report paper |
| --- | --- | --- | --- |
| Kimi-VL | 0.312 | Not recorded | 69e704a57d0172140ec7edf4 |
| Bagel | 0.299 | Not recorded | 69e704a57d0172140ec7edf4 |
| Qwen2.5-VL-3B | 0.299 | Not recorded | 69e704a57d0172140ec7edf4 |
| UniMesh | 0.297 | Not recorded | 69e704a57d0172140ec7edf4 |
| Qwen2.5-VL-7B | 0.296 | Not recorded | 69e704a57d0172140ec7edf4 |
| InternVL3.5-4B | 0.295 | Not recorded | 69e704a57d0172140ec7edf4 |
| DiffuRank | 0.291 | Not recorded | 69e704a57d0172140ec7edf4 |
| Cap3D | 0.287 | Not recorded | 69e704a57d0172140ec7edf4 |
| Phi-4-multimodel | 0.285 | Not recorded | 69e704a57d0172140ec7edf4 |
| LLaVA-3D-7B | 0.278 | Not recorded | 69e704a57d0172140ec7edf4 |

### Kimi-VL

CLIP Image-Text Score: 0.312. Result ID: 69e7510c47f34ce924ce3e27.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79a81e209131cea7eb7f.md>)

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

### Bagel

CLIP Image-Text Score: 0.299. Result ID: 69e7510c47f34ce924ce3e23.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a311e209131cea81aad.md>)

[Implementation](<https://bagel-ai.org/>)

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

### Qwen2.5-VL-3B

CLIP Image-Text Score: 0.299. Result ID: 69e7510c47f34ce924ce3e24.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

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

Recorded evaluation configuration: {"Parameters":"3B"}. This record does not assert independent reproduction.

### UniMesh

CLIP Image-Text Score: 0.297. Result ID: 69e7510c47f34ce924ce3e29.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

[Implementation](<https://github.com/AIGeeksGroup/UniMesh>)

Recorded evaluation configuration: {"Chain-of-Thought":"few-shot","Reflexion strategy":"enabled"}. This record does not assert independent reproduction.

### Qwen2.5-VL-7B

CLIP Image-Text Score: 0.296. Result ID: 69e7510c47f34ce924ce3e25.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

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

Recorded evaluation configuration: {"Parameters":"7B"}. This record does not assert independent reproduction.

### InternVL3.5-4B

CLIP Image-Text Score: 0.295. Result ID: 69e7510c47f34ce924ce3e28.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af01e209131cea85dc9.md>)

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

### DiffuRank

CLIP Image-Text Score: 0.291. Result ID: 69e7510c47f34ce924ce3e21.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1b1e209131cea81314.md>)

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

### Cap3D

CLIP Image-Text Score: 0.287. Result ID: 69e7510c47f34ce924ce3e20.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fcd1e209131cea47872.md>)

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

### Phi-4-multimodel

CLIP Image-Text Score: 0.285. Result ID: 69e7510c47f34ce924ce3e26.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77461e209131cea72272.md>)

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

### LLaVA-3D-7B

CLIP Image-Text Score: 0.278. Result ID: 69e7510c47f34ce924ce3e22.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a57d0172140ec7edf4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78461e209131cea778b5.md>)

[Implementation](<https://zcmax.github.io/projects/LLaVA-3D>)

Recorded evaluation configuration: {"Parameters":"7B"}. This record does not assert independent reproduction.

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