# Visual Task Adaptation on VTAB-1k v1 (test)

Source: [Visual Task Adaptation on VTAB-1k v1 (test)](<https://www.sota2.com/research/sota/visual-task-adaptation-on-vtab-1k-v1-test>)

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

Data updated: 2026-04-21T08:45:00.072Z

Evaluation of model performance on the test split of the VTAB-1k version 1 benchmark suite, focusing on overall mean accuracy and category-specific accuracy breakdowns (Natural, Specialized, Structured).

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 | Visual Task Adaptation |
| Dataset (including any stated split) | VTAB-1k v1 (test) |
| Metric | Mean Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 34 of 34 |
| 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

- [Mean Accuracy](<https://www.sota2.com/research/sota/visual-task-adaptation-on-vtab-1k-v1-test.md?metric=Mean%20Accuracy>): Higher is better; unit: PERCENT.

- [Accuracy (Natural)](<https://www.sota2.com/research/sota/visual-task-adaptation-on-vtab-1k-v1-test.md?metric=Accuracy%20(Natural)>): Higher is better; unit: PERCENT.

- [Accuracy (Specialized)](<https://www.sota2.com/research/sota/visual-task-adaptation-on-vtab-1k-v1-test.md?metric=Accuracy%20(Specialized)>): Higher is better; unit: PERCENT.

- [Accuracy (Structured)](<https://www.sota2.com/research/sota/visual-task-adaptation-on-vtab-1k-v1-test.md?metric=Accuracy%20(Structured)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Mean Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| SNELL-8 | 79.1 | Not recorded | 68fa75851e209131cea6857e |
| SPT-LORA | 78.7 | Not recorded | 68fa75851e209131cea6857e |
| SPT-Adapter | 78.4 | Not recorded | 68fa75851e209131cea6857e |
| SPT-Adapter | 77.5 | Not recorded | 68fa75851e209131cea6857e |
| SNELL-8 | 77.5 | Not recorded | 68fa75851e209131cea6857e |
| Adapter-8 | 77.5 | Not recorded | 68fa75851e209131cea6857e |
| SPT-LORA | 77.2 | Not recorded | 68fa75851e209131cea6857e |
| LORA-8 | 77 | Not recorded | 68fa75851e209131cea6857e |
| Adapter-8 | 76.7 | Not recorded | 68fa75851e209131cea6857e |
| LORA-8 | 76.3 | Not recorded | 68fa75851e209131cea6857e |
| Full | 75 | Not recorded | 68fa75851e209131cea6857e |
| Full | 74 | Not recorded | 68fa75851e209131cea6857e |
| VIVI-Ex(4)-Co(100%)-Big | 71.7 | Not recorded | 68fa67561e209131cea159ce |
| VPT-Deep | 71.6 | Not recorded | 68fa75851e209131cea6857e |
| PiCa (r=64) | 69.7 | Not recorded | 68fa7bce1e209131cea8a91f |
| LoRA (r=8) | 69.6 | Not recorded | 68fa7bce1e209131cea8a91f |
| DoRA (r=8) | 69.5 | Not recorded | 68fa7bce1e209131cea8a91f |
| VIVI-Ex(4)-Co(100%) | 69.4 | Not recorded | 68fa67561e209131cea159ce |
| VPT-Deep | 68.7 | Not recorded | 68fa75851e209131cea6857e |
| SVFT^B (d=8) | 68.4 | Not recorded | 68fa7bce1e209131cea8a91f |
| Sup-Rot-100% | 68 | Not recorded | 68fa67561e209131cea159ce |
| VeRA (r=4096) | 67.7 | Not recorded | 68fa7bce1e209131cea8a91f |
| VIVI-Ex(4)-Co(10%) | 67.2 | Not recorded | 68fa67561e209131cea159ce |
| Sup-100% | 66.4 | Not recorded | 68fa67561e209131cea159ce |
| Semi-Ex-10% | 65.3 | Not recorded | 68fa67561e209131cea159ce |
| Partial-1 | 65 | Not recorded | 68fa75851e209131cea6857e |
| MLP-3 | 63.6 | Not recorded | 68fa75851e209131cea6857e |
| Linear | 63.6 | Not recorded | 68fa75851e209131cea6857e |
| VIVI-Ex(4)-Big | 63.3 | Not recorded | 68fa67561e209131cea159ce |
| Partial-1 | 63.3 | Not recorded | 68fa75851e209131cea6857e |
| Linear | 62.6 | Not recorded | 68fa75851e209131cea6857e |
| VIVI-Ex(4) | 62.5 | Not recorded | 68fa67561e209131cea159ce |
| MLP-3 | 61.5 | Not recorded | 68fa75851e209131cea6857e |
| Ex-ImageNet | 59.5 | Not recorded | 68fa67561e209131cea159ce |

### SNELL-8

Mean Accuracy: 79.1. Result ID: 6993b9db3a6c314849181749.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Implementation](<https://github.com/ssfgunner/SNELL>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### SPT-LORA

Mean Accuracy: 78.7. Result ID: 6993b9db3a6c314849181748.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70721e209131cea4b083.md>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### SPT-Adapter

Mean Accuracy: 78.4. Result ID: 6993b9db3a6c314849181744.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70721e209131cea4b083.md>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### SPT-Adapter

Mean Accuracy: 77.5. Result ID: 6993b9db3a6c31484918173a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70721e209131cea4b083.md>)

Recorded evaluation configuration: {"Backbone":"Swin-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### SNELL-8

Mean Accuracy: 77.5. Result ID: 6993b9db3a6c31484918173f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Implementation](<https://github.com/ssfgunner/SNELL>)

Recorded evaluation configuration: {"Backbone":"Swin-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### Adapter-8

Mean Accuracy: 77.5. Result ID: 6993b9db3a6c314849181743.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65e31e209131cea0d324.md>)

[Implementation](<https://github.com/google-research/adapter-bert>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### SPT-LORA

Mean Accuracy: 77.2. Result ID: 6993b9db3a6c31484918173e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70721e209131cea4b083.md>)

Recorded evaluation configuration: {"Backbone":"Swin-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### LORA-8

Mean Accuracy: 77. Result ID: 6993b9db3a6c314849181747.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b121e209131cea2bf59.md>)

[Implementation](<https://github.com/microsoft/LoRA>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### Adapter-8

Mean Accuracy: 76.7. Result ID: 6993b9db3a6c314849181739.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65e31e209131cea0d324.md>)

[Implementation](<https://github.com/google-research/adapter-bert>)

Recorded evaluation configuration: {"Backbone":"Swin-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### LORA-8

Mean Accuracy: 76.3. Result ID: 6993b9db3a6c31484918173d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b121e209131cea2bf59.md>)

[Implementation](<https://github.com/microsoft/LoRA>)

Recorded evaluation configuration: {"Backbone":"Swin-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### Full

Mean Accuracy: 75. Result ID: 6993b9db3a6c314849181736.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

Recorded evaluation configuration: {"Backbone":"Swin-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### Full

Mean Accuracy: 74. Result ID: 6993b9db3a6c314849181740.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### VIVI-Ex(4)-Co(100%)-Big

Mean Accuracy: 71.7. Result ID: 698d2e2b8b517eb4331f987c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Implementation](<https://tfhub.dev/google/image_augmentation/nas_imagenet/1>)

Recorded evaluation configuration: {"Labeled Data Percentage for Co-training":"100%","Ex":4,"Model Size":"Big","Pre-training Strategy":"Co-trained (YT8M + ImageNet)"}. This record does not assert independent reproduction.

### VPT-Deep

Mean Accuracy: 71.6. Result ID: 6993b9db3a6c314849181738.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d181e209131cea37ae9.md>)

[Implementation](<https://www.sota2.com/github.com/kmnp/vpt>)

Recorded evaluation configuration: {"Backbone":"Swin-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### PiCa (r=64)

Mean Accuracy: 69.7. Result ID: 69e7390347f34ce924cd836e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bce1e209131cea8a91f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bce1e209131cea8a91f.md>)

Recorded evaluation configuration: {"#Params":"0.44M","Backbone":"ViT-B/16"}. This record does not assert independent reproduction.

### LoRA (r=8)

Mean Accuracy: 69.6. Result ID: 69e7390347f34ce924cd836a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bce1e209131cea8a91f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b121e209131cea2bf59.md>)

[Implementation](<https://github.com/microsoft/LoRA>)

Recorded evaluation configuration: {"#Params":"1.32M","Backbone":"ViT-B/16"}. This record does not assert independent reproduction.

### DoRA (r=8)

Mean Accuracy: 69.5. Result ID: 69e7390347f34ce924cd836b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bce1e209131cea8a91f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74031e209131cea5f393.md>)

[Implementation](<https://github.com/NVlabs/DoRA>)

Recorded evaluation configuration: {"#Params":"1.41M","Backbone":"ViT-B/16"}. This record does not assert independent reproduction.

### VIVI-Ex(4)-Co(100%)

Mean Accuracy: 69.4. Result ID: 698d2e2b8b517eb4331f987b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Implementation](<https://tfhub.dev/google/image_augmentation/nas_imagenet/1>)

Recorded evaluation configuration: {"Labeled Data Percentage for Co-training":"100%","Ex":4,"Pre-training Strategy":"Co-trained (YT8M + ImageNet)"}. This record does not assert independent reproduction.

### VPT-Deep

Mean Accuracy: 68.7. Result ID: 6993b9db3a6c314849181742.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d181e209131cea37ae9.md>)

[Implementation](<https://www.sota2.com/github.com/kmnp/vpt>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### SVFT^B (d=8)

Mean Accuracy: 68.4. Result ID: 69e7390347f34ce924cd836c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bce1e209131cea8a91f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa735f1e209131cea5bfe6.md>)

[Implementation](<https://github.com/VijayLingam95/SVFT/>)

Recorded evaluation configuration: {"#Params":"0.93M","Backbone":"ViT-B/16"}. This record does not assert independent reproduction.

### Sup-Rot-100%

Mean Accuracy: 68. Result ID: 698d2e2b8b517eb4331f987a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa671d1e209131cea145ed.md>)

[Implementation](<https://github.com/google-research/task_adaptation>)

Recorded evaluation configuration: {"Labeled Data Percentage":"100%","Pre-training Strategy":"Supervised with Rotation"}. This record does not assert independent reproduction.

### VeRA (r=4096)

Mean Accuracy: 67.7. Result ID: 69e7390347f34ce924cd836d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7bce1e209131cea8a91f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71cd1e209131cea53093.md>)

Recorded evaluation configuration: {"#Params":"0.45M","Backbone":"ViT-B/16"}. This record does not assert independent reproduction.

### VIVI-Ex(4)-Co(10%)

Mean Accuracy: 67.2. Result ID: 698d2e2b8b517eb4331f9878.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Implementation](<https://tfhub.dev/google/image_augmentation/nas_imagenet/1>)

Recorded evaluation configuration: {"Labeled Data Percentage for Co-training":"10%","Ex":4,"Pre-training Strategy":"Co-trained (YT8M + ImageNet)"}. This record does not assert independent reproduction.

### Sup-100%

Mean Accuracy: 66.4. Result ID: 698d2e2b8b517eb4331f9879.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa671d1e209131cea145ed.md>)

[Implementation](<https://github.com/google-research/task_adaptation>)

Recorded evaluation configuration: {"Labeled Data Percentage":"100%","Pre-training Strategy":"Supervised"}. This record does not assert independent reproduction.

### Semi-Ex-10%

Mean Accuracy: 65.3. Result ID: 698d2e2b8b517eb4331f9877.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa671d1e209131cea145ed.md>)

[Implementation](<https://github.com/google-research/task_adaptation>)

Recorded evaluation configuration: {"Labeled Data Percentage for Co-training":"10%","Pre-training Strategy":"Semi-Supervised"}. This record does not assert independent reproduction.

### Partial-1

Mean Accuracy: 65. Result ID: 6993b9db3a6c314849181746.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d181e209131cea37ae9.md>)

[Implementation](<https://www.sota2.com/github.com/kmnp/vpt>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### MLP-3

Mean Accuracy: 63.6. Result ID: 6993b9db3a6c314849181741.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d181e209131cea37ae9.md>)

[Implementation](<https://www.sota2.com/github.com/kmnp/vpt>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### Linear

Mean Accuracy: 63.6. Result ID: 6993b9db3a6c314849181745.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d181e209131cea37ae9.md>)

[Implementation](<https://www.sota2.com/github.com/kmnp/vpt>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### VIVI-Ex(4)-Big

Mean Accuracy: 63.3. Result ID: 698d2e2b8b517eb4331f9876.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Implementation](<https://tfhub.dev/google/image_augmentation/nas_imagenet/1>)

Recorded evaluation configuration: {"Pre-training Strategy":"Self-Supervised (YT8M)","Ex":4,"Model Size":"Big"}. This record does not assert independent reproduction.

### Partial-1

Mean Accuracy: 63.3. Result ID: 6993b9db3a6c31484918173c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d181e209131cea37ae9.md>)

[Implementation](<https://www.sota2.com/github.com/kmnp/vpt>)

Recorded evaluation configuration: {"Backbone":"Swin-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### Linear

Mean Accuracy: 62.6. Result ID: 6993b9db3a6c31484918173b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d181e209131cea37ae9.md>)

[Implementation](<https://www.sota2.com/github.com/kmnp/vpt>)

Recorded evaluation configuration: {"Backbone":"Swin-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### VIVI-Ex(4)

Mean Accuracy: 62.5. Result ID: 698d2e2b8b517eb4331f9875.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Implementation](<https://tfhub.dev/google/image_augmentation/nas_imagenet/1>)

Recorded evaluation configuration: {"Pre-training Strategy":"Self-Supervised (YT8M)","Ex":4}. This record does not assert independent reproduction.

### MLP-3

Mean Accuracy: 61.5. Result ID: 6993b9db3a6c314849181737.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75851e209131cea6857e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d181e209131cea37ae9.md>)

[Implementation](<https://www.sota2.com/github.com/kmnp/vpt>)

Recorded evaluation configuration: {"Backbone":"Swin-B","Pre-training":"Supervised"}. This record does not assert independent reproduction.

### Ex-ImageNet

Mean Accuracy: 59.5. Result ID: 698d2e2b8b517eb4331f9874.

[Reporting paper](<https://www.sota2.com/research/paper/68fa67561e209131cea159ce.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa660f1e209131cea0df43.md>)

Recorded evaluation configuration: {"Pre-training Strategy":"Exemplar-based ImageNet"}. This record does not assert independent reproduction.

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