# Class-Incremental Learning on CIFAR-100 (test) with Accuracy and Calibration Metrics

Source: [Class-Incremental Learning on CIFAR-100 (test) with Accuracy and Calibration Metrics](<https://www.sota2.com/research/sota/class-incremental-learning-on-cifar-100-test-with-accuracy-and-calibration-metrics>)

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

Data updated: 2026-05-20T07:33:20.494Z

Evaluation of Class-Incremental Learning performance on the CIFAR-100 test set, specifically reporting standard accuracy alongside calibration metrics like ECE and AECE.

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) | CIFAR-100 (test) |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 39 of 39 |
| 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](<https://www.sota2.com/research/sota/class-incremental-learning-on-cifar-100-test-with-accuracy-and-calibration-metrics.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [ECE](<https://www.sota2.com/research/sota/class-incremental-learning-on-cifar-100-test-with-accuracy-and-calibration-metrics.md?metric=ECE>): Lower is better; unit: PERCENT.

- [AECE](<https://www.sota2.com/research/sota/class-incremental-learning-on-cifar-100-test-with-accuracy-and-calibration-metrics.md?metric=AECE>): Lower is better; unit: PERCENT.

- [CIL Accuracy](<https://www.sota2.com/research/sota/class-incremental-learning-on-cifar-100-test-with-accuracy-and-calibration-metrics.md?metric=CIL%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| DER | 69.98 | Not recorded | 68fa77b41e209131cea74e0a |
| DER + TS | 69.98 | Not recorded | 68fa77b41e209131cea74e0a |
| DER + ETS | 69.98 | Not recorded | 68fa77b41e209131cea74e0a |
| DER + IRM | 69.98 | Not recorded | 68fa77b41e209131cea74e0a |
| DER + PerturbTS | 69.98 | Not recorded | 68fa77b41e209131cea74e0a |
| DER + T-CIL | 69.98 | Not recorded | 68fa77b41e209131cea74e0a |
| WA | 64.34 | Not recorded | 68fa77b41e209131cea74e0a |
| WA + TS | 64.02 | Not recorded | 68fa77b41e209131cea74e0a |
| WA + ETS | 64.02 | Not recorded | 68fa77b41e209131cea74e0a |
| WA + IRM | 64.02 | Not recorded | 68fa77b41e209131cea74e0a |
| WA + PerturbTS | 64.02 | Not recorded | 68fa77b41e209131cea74e0a |
| WA + T-CIL | 64.02 | Not recorded | 68fa77b41e209131cea74e0a |
| EEIL + TS | 60.74 | Not recorded | 68fa77b41e209131cea74e0a |
| EEIL + ETS | 60.74 | Not recorded | 68fa77b41e209131cea74e0a |
| EEIL + IRM | 60.74 | Not recorded | 68fa77b41e209131cea74e0a |
| EEIL + PerturbTS | 60.74 | Not recorded | 68fa77b41e209131cea74e0a |
| EEIL + T-CIL | 60.74 | Not recorded | 68fa77b41e209131cea74e0a |
| EEIL | 60.61 | Not recorded | 68fa77b41e209131cea74e0a |
| Vanilla | 56.51 | Not recorded | 68fa77b41e209131cea74e0a |
| ER | 56.51 | Not recorded | 68fa77b41e209131cea74e0a |
| TS | 56.25 | Not recorded | 68fa77b41e209131cea74e0a |
| ETS | 56.25 | Not recorded | 68fa77b41e209131cea74e0a |
| IRM | 56.25 | Not recorded | 68fa77b41e209131cea74e0a |
| PerturbTS | 56.25 | Not recorded | 68fa77b41e209131cea74e0a |
| T-CIL | 56.25 | Not recorded | 68fa77b41e209131cea74e0a |
| ER + TS | 56.25 | Not recorded | 68fa77b41e209131cea74e0a |
| ER + ETS | 56.25 | Not recorded | 68fa77b41e209131cea74e0a |
| ER + IRM | 56.25 | Not recorded | 68fa77b41e209131cea74e0a |
| ER + PerturbTS | 56.25 | Not recorded | 68fa77b41e209131cea74e0a |
| ER + T-CIL | 56.25 | Not recorded | 68fa77b41e209131cea74e0a |

### DER

Accuracy: 69.98. Result ID: 699560383badfaffe65284d2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa699b1e209131cea23084.md>)

[Implementation](<https://github.com/Rhyssiyan/DER-ClassIL.pytorch>)

Recorded evaluation configuration: {"CIL Method":"DER","Cal Method":"Vanilla"}. This record does not assert independent reproduction.

### DER + TS

Accuracy: 69.98. Result ID: 699560383badfaffe65284d3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"DER","Cal Method":"TS"}. This record does not assert independent reproduction.

### DER + ETS

Accuracy: 69.98. Result ID: 699560383badfaffe65284d4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"DER","Cal Method":"ETS"}. This record does not assert independent reproduction.

### DER + IRM

Accuracy: 69.98. Result ID: 699560383badfaffe65284d5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"DER","Cal Method":"IRM"}. This record does not assert independent reproduction.

### DER + PerturbTS

Accuracy: 69.98. Result ID: 699560383badfaffe65284d6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"DER","Cal Method":"PerturbTS"}. This record does not assert independent reproduction.

### DER + T-CIL

Accuracy: 69.98. Result ID: 699560383badfaffe65284d7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"DER","Cal Method":"T-CIL"}. This record does not assert independent reproduction.

### WA

Accuracy: 64.34. Result ID: 699560383badfaffe65284cc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66aa1e209131cea117b4.md>)

Recorded evaluation configuration: {"CIL Method":"WA","Cal Method":"Vanilla"}. This record does not assert independent reproduction.

### WA + TS

Accuracy: 64.02. Result ID: 699560383badfaffe65284cd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"WA","Cal Method":"TS"}. This record does not assert independent reproduction.

### WA + ETS

Accuracy: 64.02. Result ID: 699560383badfaffe65284ce.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"WA","Cal Method":"ETS"}. This record does not assert independent reproduction.

### WA + IRM

Accuracy: 64.02. Result ID: 699560383badfaffe65284cf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"WA","Cal Method":"IRM"}. This record does not assert independent reproduction.

### WA + PerturbTS

Accuracy: 64.02. Result ID: 699560383badfaffe65284d0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"WA","Cal Method":"PerturbTS"}. This record does not assert independent reproduction.

### WA + T-CIL

Accuracy: 64.02. Result ID: 699560383badfaffe65284d1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"WA","Cal Method":"T-CIL"}. This record does not assert independent reproduction.

### EEIL + TS

Accuracy: 60.74. Result ID: 699560383badfaffe65284c7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"EEIL","Cal Method":"TS"}. This record does not assert independent reproduction.

### EEIL + ETS

Accuracy: 60.74. Result ID: 699560383badfaffe65284c8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"EEIL","Cal Method":"ETS"}. This record does not assert independent reproduction.

### EEIL + IRM

Accuracy: 60.74. Result ID: 699560383badfaffe65284c9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"EEIL","Cal Method":"IRM"}. This record does not assert independent reproduction.

### EEIL + PerturbTS

Accuracy: 60.74. Result ID: 699560383badfaffe65284ca.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"EEIL","Cal Method":"PerturbTS"}. This record does not assert independent reproduction.

### EEIL + T-CIL

Accuracy: 60.74. Result ID: 699560383badfaffe65284cb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"EEIL","Cal Method":"T-CIL"}. This record does not assert independent reproduction.

### EEIL

Accuracy: 60.61. Result ID: 699560383badfaffe65284c6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64c71e209131cea0686d.md>)

Recorded evaluation configuration: {"CIL Method":"EEIL","Cal Method":"Vanilla"}. This record does not assert independent reproduction.

### Vanilla

Accuracy: 56.51. Result ID: 699560383badfaffe65284a0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"Calibration Method":"None","Backbone":"ResNet-32","Base CIL method":"ER","Number of incremental tasks":10,"Memory size":2000,"Batch size":128}. This record does not assert independent reproduction.

### ER

Accuracy: 56.51. Result ID: 699560383badfaffe65284c0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65d61e209131cea0c9de.md>)

[Implementation](<https://github.com/facebookresearch/agem>)

Recorded evaluation configuration: {"CIL Method":"ER","Cal Method":"Vanilla"}. This record does not assert independent reproduction.

### TS

Accuracy: 56.25. Result ID: 699560383badfaffe65284a1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa638b1e209131ce9ff6d1.md>)

Recorded evaluation configuration: {"Calibration Method":"Temperature Scaling","Backbone":"ResNet-32","Base CIL method":"ER","Number of incremental tasks":10,"Memory size":2000,"Batch size":128}. This record does not assert independent reproduction.

### ETS

Accuracy: 56.25. Result ID: 699560383badfaffe65284a2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67d61e209131cea18cfe.md>)

Recorded evaluation configuration: {"Calibration Method":"Ensemble Temperature Scaling","Backbone":"ResNet-32","Base CIL method":"ER","Number of incremental tasks":10,"Memory size":2000,"Batch size":128}. This record does not assert independent reproduction.

### IRM

Accuracy: 56.25. Result ID: 699560383badfaffe65284a3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67d61e209131cea18cfe.md>)

Recorded evaluation configuration: {"Calibration Method":"Isotonic Regression","Backbone":"ResNet-32","Base CIL method":"ER","Number of incremental tasks":10,"Memory size":2000,"Batch size":128}. This record does not assert independent reproduction.

### PerturbTS

Accuracy: 56.25. Result ID: 699560383badfaffe65284a4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a371e209131cea26a4f.md>)

Recorded evaluation configuration: {"Calibration Method":"Perturbed Temperature Scaling","Backbone":"ResNet-32","Base CIL method":"ER","Number of incremental tasks":10,"Memory size":2000,"Batch size":128}. This record does not assert independent reproduction.

### T-CIL

Accuracy: 56.25. Result ID: 699560383badfaffe65284a5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"Calibration Method":"T-CIL","Backbone":"ResNet-32","Base CIL method":"ER","Number of incremental tasks":10,"Memory size":2000,"Batch size":128}. This record does not assert independent reproduction.

### ER + TS

Accuracy: 56.25. Result ID: 699560383badfaffe65284c1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"ER","Cal Method":"TS"}. This record does not assert independent reproduction.

### ER + ETS

Accuracy: 56.25. Result ID: 699560383badfaffe65284c2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"ER","Cal Method":"ETS"}. This record does not assert independent reproduction.

### ER + IRM

Accuracy: 56.25. Result ID: 699560383badfaffe65284c3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"ER","Cal Method":"IRM"}. This record does not assert independent reproduction.

### ER + PerturbTS

Accuracy: 56.25. Result ID: 699560383badfaffe65284c4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"ER","Cal Method":"PerturbTS"}. This record does not assert independent reproduction.

### ER + T-CIL

Accuracy: 56.25. Result ID: 699560383badfaffe65284c5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77b41e209131cea74e0a.md>)

Recorded evaluation configuration: {"CIL Method":"ER","Cal Method":"T-CIL"}. This record does not assert independent reproduction.

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