# Class-Incremental Learning on ImageNet (800 base, 40 novel, test) Source: [Class-Incremental Learning on ImageNet (800 base, 40 novel, test)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:34:06.312Z Evaluation of Class-Incremental Learning performance on a specific ImageNet configuration featuring 800 base classes and 40 novel classes with full overlap, using the test split. 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 | Class-Incremental Learning | | Dataset (including any stated split) | ImageNet 800 base classes, 40 novel classes, fully-overlapping (test) | | Metric | Accuracy (All Classes) | | 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 (All Classes)](): Higher is better; unit: PERCENT. - [Accuracy (Base Classes)](): Higher is better; unit: PERCENT. - [Accuracy (Overlapping Classes)](): Higher is better; unit: PERCENT. - [Accuracy (Average)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy (All Classes) | Result date | Report paper | | --- | --- | --- | --- | | score fusion (ours) best-Accall | 57.82 | Not recorded | 68fa6d6c1e209131cea39c30 | | pre-trained model | 57.69 | Not recorded | 68fa6d6c1e209131cea39c30 | | score fusion (ours) best-balanced | 55.66 | Not recorded | 68fa6d6c1e209131cea39c30 | | score fusion (ours) best-Accavg | 53.42 | Not recorded | 68fa6d6c1e209131cea39c30 | | learning-based routing w/ Lrt-bal | 52.82 | Not recorded | 68fa6d6c1e209131cea39c30 | | joint learning (oracle) | 47.63 | Not recorded | 68fa6d6c1e209131cea39c30 | | confidence-based routing | 35.43 | Not recorded | 68fa6d6c1e209131cea39c30 | | RM | 21.89 | Not recorded | 68fa6d6c1e209131cea39c30 | | RM | 21.1 | Not recorded | 68fa6d6c1e209131cea39c30 | | iCaRL | 20.23 | Not recorded | 68fa6d6c1e209131cea39c30 | | iCaRL | 18.12 | Not recorded | 68fa6d6c1e209131cea39c30 | | fine-tuning | 3.86 | Not recorded | 68fa6d6c1e209131cea39c30 | ### score fusion (ours) best-Accall Accuracy (All Classes): 57.82. Result ID: 698fa1083a6c314849faa73b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ResNet10"}. This record does not assert independent reproduction. ### pre-trained model Accuracy (All Classes): 57.69. Result ID: 698fa1083a6c314849faa735. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"ResNet10"}. This record does not assert independent reproduction. ### score fusion (ours) best-balanced Accuracy (All Classes): 55.66. Result ID: 698fa1083a6c314849faa73c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ResNet10"}. This record does not assert independent reproduction. ### score fusion (ours) best-Accavg Accuracy (All Classes): 53.42. Result ID: 698fa1083a6c314849faa73d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ResNet10"}. This record does not assert independent reproduction. ### learning-based routing w/ Lrt-bal Accuracy (All Classes): 52.82. Result ID: 698fa1083a6c314849faa737. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"ResNet10"}. This record does not assert independent reproduction. ### joint learning (oracle) Accuracy (All Classes): 47.63. Result ID: 698fa1083a6c314849faa73e. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"ResNet10"}. This record does not assert independent reproduction. ### confidence-based routing Accuracy (All Classes): 35.43. Result ID: 698fa1083a6c314849faa736. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"ResNet10"}. This record does not assert independent reproduction. ### RM Accuracy (All Classes): 21.89. Result ID: 698fa1083a6c314849faa740. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ResNet18"}. This record does not assert independent reproduction. ### RM Accuracy (All Classes): 21.1. Result ID: 698fa1083a6c314849faa73a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"ResNet10"}. This record does not assert independent reproduction. ### iCaRL Accuracy (All Classes): 20.23. Result ID: 698fa1083a6c314849faa739. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ResNet10"}. This record does not assert independent reproduction. ### iCaRL Accuracy (All Classes): 18.12. Result ID: 698fa1083a6c314849faa73f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"ResNet18"}. This record does not assert independent reproduction. ### fine-tuning Accuracy (All Classes): 3.86. Result ID: 698fa1083a6c314849faa738. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"ResNet10"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.