# Image Classification on 20 Image Classification Tasks Source: [Image Classification on 20 Image Classification Tasks]() Collection index: [llms.txt]() Data updated: 2026-04-28T08:20:38.541Z Evaluation of Image Classification performance across a suite containing 20 distinct classification tasks, reporting aggregated results based on merging subsets of these tasks (e.g., 4 tasks merged, 20 tasks merged). 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) | 20 Image Classification Tasks | | Metric | Accuracy (4 tasks merged) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 12 of 12 | | 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 - [Accuracy (4 tasks merged)](): Higher is better; unit: PERCENT. - [Accuracy (8 tasks merged)](): Higher is better; unit: PERCENT. - [Accuracy (12 tasks merged)](): Higher is better; unit: PERCENT. - [Accuracy (16 tasks merged)](): Higher is better; unit: PERCENT. - [Accuracy (20 tasks merged)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy (4 tasks merged) | Result date | Report paper | | --- | --- | --- | --- | | DOGE TA | 59.9 | Not recorded | 68fa7a631e209131cea82cfa | | TIES-Merging | 59.7 | Not recorded | 68fa7a631e209131cea82cfa | | Iso-C | 59.2 | Not recorded | 68fa7a631e209131cea82cfa | | RegMean++ | 59 | Not recorded | 68fa7a631e209131cea82cfa | | Model Soups | 58.9 | Not recorded | 68fa7a631e209131cea82cfa | | Fisher Merging | 58.9 | Not recorded | 68fa7a631e209131cea82cfa | | Iso-CTS | 58.7 | Not recorded | 68fa7a631e209131cea82cfa | | RegMean | 58.4 | Not recorded | 68fa7a631e209131cea82cfa | | Task Arithmetic | 58.2 | Not recorded | 68fa7a631e209131cea82cfa | | Layer-wise AdaMerging | 57.7 | Not recorded | 68fa7a631e209131cea82cfa | | DOGE AM | 57 | Not recorded | 68fa7a631e209131cea82cfa | | TSV-M | 56.9 | Not recorded | 68fa7a631e209131cea82cfa | ### DOGE TA Accuracy (4 tasks merged): 59.9. Result ID: 69f06cb147f34ce924d63bf6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Data-Free Methods"}. This record does not assert independent reproduction. ### TIES-Merging Accuracy (4 tasks merged): 59.7. Result ID: 69f06cb147f34ce924d63bf4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Data-Free Methods"}. This record does not assert independent reproduction. ### Iso-C Accuracy (4 tasks merged): 59.2. Result ID: 69f06cb147f34ce924d63bf7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Data-Free Methods"}. This record does not assert independent reproduction. ### RegMean++ Accuracy (4 tasks merged): 59. Result ID: 69f06cb147f34ce924d63bfb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Training-Free Methods"}. This record does not assert independent reproduction. ### Model Soups Accuracy (4 tasks merged): 58.9. Result ID: 69f06cb147f34ce924d63bf2. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Data-Free Methods"}. This record does not assert independent reproduction. ### Fisher Merging Accuracy (4 tasks merged): 58.9. Result ID: 69f06cb147f34ce924d63bf9. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Training-Free Methods"}. This record does not assert independent reproduction. ### Iso-CTS Accuracy (4 tasks merged): 58.7. Result ID: 69f06cb147f34ce924d63bf8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Data-Free Methods"}. This record does not assert independent reproduction. ### RegMean Accuracy (4 tasks merged): 58.4. Result ID: 69f06cb147f34ce924d63bfa. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Training-Free Methods"}. This record does not assert independent reproduction. ### Task Arithmetic Accuracy (4 tasks merged): 58.2. Result ID: 69f06cb147f34ce924d63bf3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Data-Free Methods"}. This record does not assert independent reproduction. ### Layer-wise AdaMerging Accuracy (4 tasks merged): 57.7. Result ID: 69f06cb147f34ce924d63bfc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Test-Time Adaption Methods"}. This record does not assert independent reproduction. ### DOGE AM Accuracy (4 tasks merged): 57. Result ID: 69f06cb147f34ce924d63bfd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Test-Time Adaption Methods"}. This record does not assert independent reproduction. ### TSV-M Accuracy (4 tasks merged): 56.9. Result ID: 69f06cb147f34ce924d63bf5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ViT-B/32","Method Category":"Data-Free Methods"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.