# Image Classification on ImageNet distribution-shift variants (ImageNet-A, -K, -R, -V) (test) Source: [Image Classification on ImageNet distribution-shift variants (ImageNet-A, -K, -R, -V) (test)]() Collection index: [llms.txt]() Data updated: 2026-02-26T06:06:53.589Z Evaluation of Image Classification performance across a specific set of ImageNet distribution shift variants: ImageNet (reference), ImageNet-A, ImageNet-K, ImageNet-R, and ImageNet-V, using the test split, reporting overall and out-of-distribution averages. 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 Classification | | Dataset (including any stated split) | ImageNet distribution-shift variants (ImageNet-A, -K, -R, -V) (test) | | Metric | Top-1 Acc (ImageNet) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 11 of 11 | | 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 - [Top-1 Acc (ImageNet)](): Higher is better; unit: PERCENT. - [Top-1 Acc](): Higher is better; unit: PERCENT. - [Top-1 Acc (ImageNet-A)](): Higher is better; unit: PERCENT. - [Top-1 Acc (ImageNet-K)](): Higher is better; unit: PERCENT. - [Top-1 Acc (ImageNet-R)](): Higher is better; unit: PERCENT. - [Top-1 Acc (ImageNet-V)](): Higher is better; unit: PERCENT. - [Overall Avg. Acc](): Higher is better; unit: PERCENT. - [OOD Avg. Acc](): Higher is better; unit: PERCENT. ## Recorded results | Method | Top-1 Acc (ImageNet) | Result date | Report paper | | --- | --- | --- | --- | | Ours (CL-HP) | 69.67 | Not recorded | 698abba9c97a0b3e3b5ddff6 | | Ours (CL) | 69.46 | Not recorded | 698abba9c97a0b3e3b5ddff6 | | R-TPT | 69.37 | Not recorded | 698abba9c97a0b3e3b5ddff6 | | ZERO | 69.28 | Not recorded | 698abba9c97a0b3e3b5ddff6 | | TTL | 69.28 | Not recorded | 698abba9c97a0b3e3b5ddff6 | | MTA | 69.24 | Not recorded | 698abba9c97a0b3e3b5ddff6 | | TPT | 68.94 | Not recorded | 698abba9c97a0b3e3b5ddff6 | | TPS | 68.83 | Not recorded | 698abba9c97a0b3e3b5ddff6 | | RLCF | 68.56 | Not recorded | 698abba9c97a0b3e3b5ddff6 | | C-TPT | 68.52 | Not recorded | 698abba9c97a0b3e3b5ddff6 | | CLIP | 66.72 | Not recorded | 698abba9c97a0b3e3b5ddff6 | ### Ours (CL-HP) Top-1 Acc (ImageNet): 69.67. Result ID: 699757fb5698665ec4aa14c3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Variant":"Context Learning with Hard Prompt","Test-time adaptation":"true"}. This record does not assert independent reproduction. ### Ours (CL) Top-1 Acc (ImageNet): 69.46. Result ID: 699757fb5698665ec4aa14c2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Variant":"Context Learning","Test-time adaptation":"true"}. This record does not assert independent reproduction. ### R-TPT Top-1 Acc (ImageNet): 69.37. Result ID: 699757fb5698665ec4aa14c1. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Test-time adaptation":"true"}. This record does not assert independent reproduction. ### ZERO Top-1 Acc (ImageNet): 69.28. Result ID: 699757fb5698665ec4aa14be. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Test-time adaptation":"true"}. This record does not assert independent reproduction. ### TTL Top-1 Acc (ImageNet): 69.28. Result ID: 699757fb5698665ec4aa14bf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Test-time adaptation":"true"}. This record does not assert independent reproduction. ### MTA Top-1 Acc (ImageNet): 69.24. Result ID: 699757fb5698665ec4aa14bd. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Test-time adaptation":"true"}. This record does not assert independent reproduction. ### TPT Top-1 Acc (ImageNet): 68.94. Result ID: 699757fb5698665ec4aa14ba. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Test-time adaptation":"true"}. This record does not assert independent reproduction. ### TPS Top-1 Acc (ImageNet): 68.83. Result ID: 699757fb5698665ec4aa14c0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Test-time adaptation":"true"}. This record does not assert independent reproduction. ### RLCF Top-1 Acc (ImageNet): 68.56. Result ID: 699757fb5698665ec4aa14bc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Test-time adaptation":"true"}. This record does not assert independent reproduction. ### C-TPT Top-1 Acc (ImageNet): 68.52. Result ID: 699757fb5698665ec4aa14bb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Test-time adaptation":"true"}. This record does not assert independent reproduction. ### CLIP Top-1 Acc (ImageNet): 66.72. Result ID: 699757fb5698665ec4aa14b9. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Zero-shot":"true"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.