Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Confusion-Driven Self-Supervised Progressively Weighted Ensemble Learning for Non-Exemplar Class Incremental Learning
Authors: Kai Hu, Zhang Yu, Yuan Zhang, Zhineng Chen, Xieping Gao
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct extensive experiments on three benchmark NECIL datasets: CIFAR100 [46], Tiny Image Net [47], and Image Net-Subset [48]. The CIFAR100 dataset comprises 100 classes, each with 600 color images at a resolution of 32 × 32 pixels, including 500 images for training and 100 images for testing. Tiny Image Net is a subset of the Image Net [48] designed for image classification, consisting of 200 classes, with each class containing 500 training images, 50 validation images, and 50 test images, all sized at 64 × 64 pixels. The Image Net-Subset dataset, comprising 100 classes selected from Image Net, contains about 13,000 training images and 50 test images per class, all standardized to a resolution of 224 × 224 pixels. We follow the configuration from PASS [9] and divide each dataset into three incremental settings: 5, 10, and 20 tasks. |
| Researcher Affiliation | Academia | Kai Hu Xiangtan University EMAIL Zhang Yu Xiangtan University EMAIL Yuan Zhang Xiangtan University EMAIL Zhineng Chen Fudan University EMAIL Xieping Gao Hunan Normal University EMAIL |
| Pseudocode | No | The paper describes its methodology using textual descriptions and mathematical formulas, but it does not include any explicitly labeled "Pseudocode" or "Algorithm" blocks or figures. |
| Open Source Code | Yes | Code is publicly available at: https://github.com/MLMIP/CLOVER. |
| Open Datasets | Yes | We conduct extensive experiments on three benchmark NECIL datasets: CIFAR100 [46], Tiny Image Net [47], and Image Net-Subset [48]. |
| Dataset Splits | Yes | The CIFAR100 dataset comprises 100 classes, each with 600 color images at a resolution of 32 × 32 pixels, including 500 images for training and 100 images for testing. Tiny Image Net is a subset of the Image Net [48] designed for image classification, consisting of 200 classes, with each class containing 500 training images, 50 validation images, and 50 test images, all sized at 64 × 64 pixels. The Image Net-Subset dataset, comprising 100 classes selected from Image Net, contains about 13,000 training images and 50 test images per class, all standardized to a resolution of 224 × 224 pixels. We follow the configuration from PASS [9] and divide each dataset into three incremental settings: 5, 10, and 20 tasks. |
| Hardware Specification | Yes | Based on the aforementioned settings, all algorithms can be trained on a single NVIDIA A100 GPU. |
| Software Dependencies | No | The paper mentions using ResNet18 as the backbone, SGD optimizer, and Bayesian classifier, but does not specify software versions for libraries like PyTorch, TensorFlow, or specific Python versions. |
| Experiment Setup | Yes | During training, the batch size is set to 128 and the model is optimized by the SGD optimizer with an initial learning rate 0.1 and weight decay 1e-4. The learning rate is multiplied by 0.1 at epochs 60, 120 and 160. |