Ablation Study Results on CIFAR-100 for Class-Incremental Learning

Ablation Study Results on CIFAR-100 Inc5 Task is a 5.5 KB dataset containing experimental results for a class-incremental learning method. The data, authored by Jianye Gu and last updated in May 2026, is stored in an XLS file format and shared under a CC-BY-4.0 license on figshare. It likely contains metrics from experiments evaluating the DGASA method's performance.

Use Cases

• Benchmarking new class-incremental learning methods based on the reported ablation study results.
• Analyzing the impact of architectural components like lightweight adapters and attention-based gating on model performance.
• Studying the relationship between subspace alignment techniques and resistance to catastrophic forgetting.
• Comparing computational efficiency and generalization across incremental learning approaches.

Strengths

• Dataset is small at 5.5 KB, facilitating quick download and inspection.
• License is explicitly stated as CC-BY-4.0, permitting open reuse with attribution.
• Author and last update date (2026-05-07) are provided, offering basic provenance.

Limitations

• Row count and column definitions are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset's scope is limited to results from a single study on the CIFAR-100 Inc5 task.

Provenance

Source

figshare, authored by Jianye Gu.

Collection Method

Likely contains experimental results from an ablation study on a class-incremental learning method.

Time Range

The study's temporal context is not specified in the metadata.

Freshness

Last updated 2026-05-07 17:41:39; freshness should be verified.

Geography

Spatial coverage is not applicable for this benchmark dataset.