# 2D Grounding on Generative-Refined 3D-VL Benchmark (test)

Source: [2D Grounding on Generative-Refined 3D-VL Benchmark (test)](<https://www.sota2.com/research/sota/2d-grounding-on-generative-refined-3d-vl-benchmark-test>)

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

Data updated: 2026-04-24T07:19:32.674Z

Evaluation of 2D Grounding performance on the test split of the Generative-Refined 3D-VL Benchmark, measured using IoU2D and KPA-2D 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 | 2D Grounding |
| Dataset (including any stated split) | Generative-Refined 3D-VL Benchmark (test) |
| Metric | IoU2D |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 9 of 9 |
| 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

- [IoU2D](<https://www.sota2.com/research/sota/2d-grounding-on-generative-refined-3d-vl-benchmark-test.md?metric=IoU2D>): Higher is better; unit: PERCENT.

- [KPA-2D](<https://www.sota2.com/research/sota/2d-grounding-on-generative-refined-3d-vl-benchmark-test.md?metric=KPA-2D>): Higher is better; unit: PERCENT.

## Recorded results

| Method | IoU2D | Result date | Report paper |
| --- | --- | --- | --- |
| PointVL-3D | 89.4 | Not recorded | 69eaf8ec55a3c32af89f87f4 |
| Qwen2.5-VL-3B(SFT) | 89.1 | Not recorded | 69eaf8ec55a3c32af89f87f4 |
| MiMo-VL-7B-RL(SFT) | 88.6 | Not recorded | 69eaf8ec55a3c32af89f87f4 |
| Qwen2.5-VL-3B | 79.1 | Not recorded | 69eaf8ec55a3c32af89f87f4 |
| Qwen2.5-VL-7B | 78 | Not recorded | 69eaf8ec55a3c32af89f87f4 |
| InternVL3\_5-4B-Instruct | 77.2 | Not recorded | 69eaf8ec55a3c32af89f87f4 |
| InternVL3\_5-8B-Instruct | 76.4 | Not recorded | 69eaf8ec55a3c32af89f87f4 |
| MiMo-VL-7B-RL | 74.5 | Not recorded | 69eaf8ec55a3c32af89f87f4 |
| DeepSeek-VL2-Tiny | 69.2 | Not recorded | 69eaf8ec55a3c32af89f87f4 |

### PointVL-3D

IoU2D: 89.4. Result ID: 69eb197947f34ce924d2e14c.

[Reporting paper](<https://www.sota2.com/research/paper/69eaf8ec55a3c32af89f87f4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69eaf8ec55a3c32af89f87f4.md>)

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

### Qwen2.5-VL-3B(SFT)

IoU2D: 89.1. Result ID: 69eb197947f34ce924d2e14a.

[Reporting paper](<https://www.sota2.com/research/paper/69eaf8ec55a3c32af89f87f4.md>)

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

Recorded evaluation configuration: {"Model Size":"3B","Fine-tuning stage":"SFT"}. This record does not assert independent reproduction.

### MiMo-VL-7B-RL(SFT)

IoU2D: 88.6. Result ID: 69eb197947f34ce924d2e14b.

[Reporting paper](<https://www.sota2.com/research/paper/69eaf8ec55a3c32af89f87f4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6948d062423ea42265729de7.md>)

Recorded evaluation configuration: {"Model Size":"7B","Fine-tuning stage":"SFT"}. This record does not assert independent reproduction.

### Qwen2.5-VL-3B

IoU2D: 79.1. Result ID: 69eb197947f34ce924d2e144.

[Reporting paper](<https://www.sota2.com/research/paper/69eaf8ec55a3c32af89f87f4.md>)

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

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

### Qwen2.5-VL-7B

IoU2D: 78. Result ID: 69eb197947f34ce924d2e145.

[Reporting paper](<https://www.sota2.com/research/paper/69eaf8ec55a3c32af89f87f4.md>)

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

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

### InternVL3\_5-4B-Instruct

IoU2D: 77.2. Result ID: 69eb197947f34ce924d2e146.

[Reporting paper](<https://www.sota2.com/research/paper/69eaf8ec55a3c32af89f87f4.md>)

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

Recorded evaluation configuration: {"Model Size":"4B","Fine-tuning stage":"Instruct"}. This record does not assert independent reproduction.

### InternVL3\_5-8B-Instruct

IoU2D: 76.4. Result ID: 69eb197947f34ce924d2e147.

[Reporting paper](<https://www.sota2.com/research/paper/69eaf8ec55a3c32af89f87f4.md>)

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

Recorded evaluation configuration: {"Model Size":"8B","Fine-tuning stage":"Instruct"}. This record does not assert independent reproduction.

### MiMo-VL-7B-RL

IoU2D: 74.5. Result ID: 69eb197947f34ce924d2e149.

[Reporting paper](<https://www.sota2.com/research/paper/69eaf8ec55a3c32af89f87f4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6948d062423ea42265729de7.md>)

Recorded evaluation configuration: {"Model Size":"7B","Fine-tuning stage":"RL"}. This record does not assert independent reproduction.

### DeepSeek-VL2-Tiny

IoU2D: 69.2. Result ID: 69eb197947f34ce924d2e148.

[Reporting paper](<https://www.sota2.com/research/paper/69eaf8ec55a3c32af89f87f4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa763a1e209131cea6bfb0.md>)

Recorded evaluation configuration: {"Model Size":"Tiny"}. This record does not assert independent reproduction.

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