# Referring Image Segmentation on G-Ref (test)

Source: [Referring Image Segmentation on G-Ref (test)](<https://www.sota2.com/research/sota/referring-image-segmentation-on-g-ref-test>)

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

Data updated: 2026-02-26T05:47:22.055Z

Evaluation of Referring Image Segmentation performance on the test split of the G-Ref dataset, typically measured using Intersection over Union (IoU).

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 | Referring Image Segmentation |
| Dataset (including any stated split) | G-Ref (test) |
| Metric | IoU |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 33 of 33 |
| 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

- [IoU](<https://www.sota2.com/research/sota/referring-image-segmentation-on-g-ref-test.md?metric=IoU>): Higher is better; unit: PERCENT.

## Recorded results

| Method | IoU | Result date | Report paper |
| --- | --- | --- | --- |
| UniLSeg-100 | 80.54 | Not recorded | 68fa75d61e209131cea6a28a |
| UniLSeg-20 | 79.47 | Not recorded | 68fa75d61e209131cea6a28a |
| VATEX\_joint | 76.4 | Not recorded | 68fa758f1e209131cea68a30 |
| UniNext | 73.68 | Not recorded | 68fa75d61e209131cea6a28a |
| VATEX | 70.58 | Not recorded | 68fa758f1e209131cea68a30 |
| VATEX\_G-Ref | 70.58 | Not recorded | 68fa758f1e209131cea68a30 |
| PolyFormer-B | 69.88 | Not recorded | 68fa758f1e209131cea68a30 |
| PolyFormer | 69.05 | Not recorded | 68fa75d61e209131cea6a28a |
| LISA-7B | 68.5 | Not recorded | 68fa758f1e209131cea68a30 |
| RefTR | 67.39 | Not recorded | 68fa758f1e209131cea68a30 |
| VLT | 66.22 | Not recorded | 68fa758f1e209131cea68a30 |
| GRES | 65.97 | Not recorded | 68fa75d61e209131cea6a28a |
| SeqTR | 65.74 | Not recorded | 68fa758f1e209131cea68a30 |
| VG-LAW | 65.13 | Not recorded | 68fa758f1e209131cea68a30 |
| LAVT | 63.62 | Not recorded | 68fa758f1e209131cea68a30 |
| CM-MaskSD | 62.69 | Not recorded | 68fa758f1e209131cea68a30 |
| LAVT | 62.09 | Not recorded | 68fa75d61e209131cea6a28a |
| CRIS | 60.36 | Not recorded | 68fa6c011e209131cea319eb |
| CRIS | 60.36 | Not recorded | 68fa758f1e209131cea68a30 |
| CRIS | 60.36 | Not recorded | 68fa75d61e209131cea6a28a |
| CRIS | 59.39 | Not recorded | 68fa6c011e209131cea319eb |
| CRIS | 59.39 | Not recorded | 68fa75d61e209131cea6a28a |
| VLT | 56.65 | Not recorded | 68fa6c011e209131cea319eb |
| VLT | 56.65 | Not recorded | 68fa75d61e209131cea6a28a |
| LTS | 54.25 | Not recorded | 68fa6c011e209131cea319eb |
| LTS | 54.25 | Not recorded | 68fa75d61e209131cea6a28a |
| CGAN | 51.69 | Not recorded | 68fa6c011e209131cea319eb |
| CGAN | 51.69 | Not recorded | 68fa75d61e209131cea6a28a |
| MCN | 49.4 | Not recorded | 68fa6c011e209131cea319eb |
| MCN | 49.4 | Not recorded | 68fa75d61e209131cea6a28a |
| MAttNet | 48.61 | Not recorded | 68fa6c011e209131cea319eb |
| NMTree | 47.88 | Not recorded | 68fa6c011e209131cea319eb |
| Lang2Seg | 46.95 | Not recorded | 68fa6c011e209131cea319eb |

### UniLSeg-100

IoU: 80.54. Result ID: 699402463a6c3148491a4c95.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

Recorded evaluation configuration: {"Vision Backbone":"Swin-B"}. This record does not assert independent reproduction.

### UniLSeg-20

IoU: 79.47. Result ID: 699402463a6c3148491a4c94.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

Recorded evaluation configuration: {"Vision Backbone":"Swin-B"}. This record does not assert independent reproduction.

### VATEX\_joint

IoU: 76.4. Result ID: 6993d35c3a6c31484918d186.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Implementation](<https://vatex.hkustvgd.com/>)

Recorded evaluation configuration: {"External Datasets":"RefCOCO(+/g) (368K)","Backbone":"Swin-B"}. This record does not assert independent reproduction.

### UniNext

IoU: 73.68. Result ID: 699402463a6c3148491a4c93.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70541e209131cea4a478.md>)

[Implementation](<https://github.com/MasterBin-IIAU/UNINEXT>)

Recorded evaluation configuration: {"Vision Backbone":"ConvNext-L"}. This record does not assert independent reproduction.

### VATEX

IoU: 70.58. Result ID: 6993d35c3a6c31484918d152.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Implementation](<https://vatex.hkustvgd.com/>)

Recorded evaluation configuration: {"Visual Backbone":"Swin-B","Textual Backbone":"CLIP"}. This record does not assert independent reproduction.

### VATEX\_G-Ref

IoU: 70.58. Result ID: 6993d35c3a6c31484918d185.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Implementation](<https://vatex.hkustvgd.com/>)

Recorded evaluation configuration: {"External Datasets":"G-Ref (85K)","Backbone":"Swin-B"}. This record does not assert independent reproduction.

### PolyFormer-B

IoU: 69.88. Result ID: 6993d35c3a6c31484918d184.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6edf1e209131cea428bb.md>)

[Implementation](<https://polyformer.github.io/>)

Recorded evaluation configuration: {"External Datasets":"RefCOCO(+/g) (368K)"}. This record does not assert independent reproduction.

### PolyFormer

IoU: 69.05. Result ID: 699402463a6c3148491a4c92.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6edf1e209131cea428bb.md>)

[Implementation](<https://polyformer.github.io/>)

Recorded evaluation configuration: {"Vision Backbone":"Swin-B"}. This record does not assert independent reproduction.

### LISA-7B

IoU: 68.5. Result ID: 6993d35c3a6c31484918d150.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa730a1e209131cea5a0b2.md>)

[Implementation](<https://www.sota2.com/github.com/dvlab-research/LISA>)

Recorded evaluation configuration: {"Visual Backbone":"ViT-H SAM","Textual Backbone":"Vicuna-7B"}. This record does not assert independent reproduction.

### RefTR

IoU: 67.39. Result ID: 6993d35c3a6c31484918d183.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a611e209131cea27dec.md>)

Recorded evaluation configuration: {"External Datasets":"Visual Genome (5.4M) & Flickr30k-entities (158K)"}. This record does not assert independent reproduction.

### VLT

IoU: 66.22. Result ID: 6993d35c3a6c31484918d14e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e091e209131cea3d96b.md>)

Recorded evaluation configuration: {"Visual Backbone":"Swin-B","Textual Backbone":"BERT"}. This record does not assert independent reproduction.

### GRES

IoU: 65.97. Result ID: 699402463a6c3148491a4c91.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f8e1e209131cea46367.md>)

[Implementation](<https://henghui.ding.github.io/GRES>)

Recorded evaluation configuration: {"Vision Backbone":"Swin-B"}. This record does not assert independent reproduction.

### SeqTR

IoU: 65.74. Result ID: 6993d35c3a6c31484918d182.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e0c1e209131cea3dad7.md>)

[Implementation](<https://github.com/sean-zhuh/SeqTR>)

Recorded evaluation configuration: {"External Datasets":"Visual Genome (5.4M) & Flickr30k-entities (158K)"}. This record does not assert independent reproduction.

### VG-LAW

IoU: 65.13. Result ID: 6993d35c3a6c31484918d151.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f9e1e209131cea46a8e.md>)

Recorded evaluation configuration: {"Visual Backbone":"ViT-B","Textual Backbone":"BERT"}. This record does not assert independent reproduction.

### LAVT

IoU: 63.62. Result ID: 6993d35c3a6c31484918d14f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c2c1e209131cea328ba.md>)

[Implementation](<https://www.sota2.com/LAVT-RIS>)

Recorded evaluation configuration: {"Visual Backbone":"Swin-B","Textual Backbone":"BERT"}. This record does not assert independent reproduction.

### CM-MaskSD

IoU: 62.69. Result ID: 6993d35c3a6c31484918d14d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa723c1e209131cea552d9.md>)

Recorded evaluation configuration: {"Visual Backbone":"CLIP-ViT-B","Textual Backbone":"CLIP"}. This record does not assert independent reproduction.

### LAVT

IoU: 62.09. Result ID: 699402463a6c3148491a4c90.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c2c1e209131cea328ba.md>)

[Implementation](<https://www.sota2.com/LAVT-RIS>)

Recorded evaluation configuration: {"Vision Backbone":"Swin-B"}. This record does not assert independent reproduction.

### CRIS

IoU: 60.36. Result ID: 698ebaa8ffc3901de0a02736.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-101"}. This record does not assert independent reproduction.

### CRIS

IoU: 60.36. Result ID: 6993d35c3a6c31484918d14c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758f1e209131cea68a30.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

Recorded evaluation configuration: {"Visual Backbone":"ResNet-101","Textual Backbone":"CLIP"}. This record does not assert independent reproduction.

### CRIS

IoU: 60.36. Result ID: 699402463a6c3148491a4c8f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

Recorded evaluation configuration: {"Vision Backbone":"ResNet101"}. This record does not assert independent reproduction.

### CRIS

IoU: 59.39. Result ID: 698ebaa8ffc3901de0a02735.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-50"}. This record does not assert independent reproduction.

### CRIS

IoU: 59.39. Result ID: 699402463a6c3148491a4c8e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

Recorded evaluation configuration: {"Vision Backbone":"ResNet50"}. This record does not assert independent reproduction.

### VLT

IoU: 56.65. Result ID: 698ebaa8ffc3901de0a02734.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a911e209131cea28f0a.md>)

[Implementation](<https://github.com/henghuiding/Vision-Language-Transformer>)

Recorded evaluation configuration: {"Backbone":"DarkNet-53"}. This record does not assert independent reproduction.

### VLT

IoU: 56.65. Result ID: 699402463a6c3148491a4c8d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a911e209131cea28f0a.md>)

[Implementation](<https://github.com/henghuiding/Vision-Language-Transformer>)

Recorded evaluation configuration: {"Vision Backbone":"DarkNet53"}. This record does not assert independent reproduction.

### LTS

IoU: 54.25. Result ID: 698ebaa8ffc3901de0a02733.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa699a1e209131cea22fd0.md>)

Recorded evaluation configuration: {"Backbone":"DarkNet-53"}. This record does not assert independent reproduction.

### LTS

IoU: 54.25. Result ID: 699402463a6c3148491a4c8c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa699a1e209131cea22fd0.md>)

Recorded evaluation configuration: {"Vision Backbone":"DarkNet53"}. This record does not assert independent reproduction.

### CGAN

IoU: 51.69. Result ID: 698ebaa8ffc3901de0a02732.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

Recorded evaluation configuration: {"Backbone":"DarkNet-53"}. This record does not assert independent reproduction.

### CGAN

IoU: 51.69. Result ID: 699402463a6c3148491a4c8b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

Recorded evaluation configuration: {"Vision Backbone":"DarkNet53"}. This record does not assert independent reproduction.

### MCN

IoU: 49.4. Result ID: 698ebaa8ffc3901de0a02731.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67361e209131cea14d7e.md>)

[Implementation](<https://github.com/luogen1996/MCN>)

Recorded evaluation configuration: {"Backbone":"DarkNet-53"}. This record does not assert independent reproduction.

### MCN

IoU: 49.4. Result ID: 699402463a6c3148491a4c8a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75d61e209131cea6a28a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67361e209131cea14d7e.md>)

[Implementation](<https://github.com/luogen1996/MCN>)

Recorded evaluation configuration: {"Vision Backbone":"DarkNet53"}. This record does not assert independent reproduction.

### MAttNet

IoU: 48.61. Result ID: 698ebaa8ffc3901de0a0272e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa642c1e209131cea031d4.md>)

[Implementation](<https://github.com/lichengunc/MAttNet>)

Recorded evaluation configuration: {"Backbone":"ResNet-101"}. This record does not assert independent reproduction.

### NMTree

IoU: 47.88. Result ID: 698ebaa8ffc3901de0a0272f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66901e209131cea10e32.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-101"}. This record does not assert independent reproduction.

### Lang2Seg

IoU: 46.95. Result ID: 698ebaa8ffc3901de0a02730.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c011e209131cea319eb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66801e209131cea10b04.md>)

[Implementation](<https://github.com/wenz116/lang2seg>)

Recorded evaluation configuration: {"Backbone":"ResNet-101"}. This record does not assert independent reproduction.

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