Research map / Continual learning / Class-incremental
HyCal: A Training-Free Prototype Calibration Method for Cross-Discipline Few-Shot Class-Incremental Learning
Eunju Lee, MiHyeon Kim, JuneHyoung Kwon, Yoonji Lee, JiHyun Kim, Soojin Jang, YoungBin Kim
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use dynamic weighted fusion instead of cosine similarity alone (ablation) | HyCal · High-Scale Domain Imbalance | S_CDE: 51.29 -> 53.70 | Improved |
| use dynamic weighted fusion instead of Mahalanobis alone (ablation) | HyCal · High-Scale Domain Imbalance | S_CDE: 53.50 -> 53.70 | Improved |
| use dynamic weighting instead of plain average of the two metrics (ablation) | HyCal · Cross-Scale Imbalance | S_CDE: 58.24 -> 59.24 | Improved |
| add Mahalanobis distance to cosine similarity under balanced domains (ablation) | HyCal · Balanced-in-Class Domain | S_CDE: 56.82 -> 60.92 | Improved |
| sum image-text embeddings instead of concatenating (ablation) | HyCal · High-Scale Domain Imbalance | S_CDE: 53.39 -> 53.70 | Improved |
| sum image-text embeddings instead of concatenating (ablation) | HyCal · Balanced-in-Class Domain | S_CDE: 60.70 -> 60.92 | Improved |
| HyCal versus second-best training-free method (KLDA) under high-scale imbalance | HyCal · High-Scale Domain Imbalance | Average accuracy: over 4.5%p | Improved |
| HyCal versus KLDA under cross-scale imbalance | HyCal · Cross-Scale Imbalance | S_CDE: 59.61 -> 59.24 | Hurt |
| increase shots from 5 to 20 | HyCal · FSCIL | Last accuracy: 60.82 -> 70.67 | Improved |
| use OpenCLIP ViT-B/16 DataComp-1B backbone | HyCal · High-Scale Domain Imbalance | Last accuracy: 68.88 vs 66.20 | Improved |
| random domain order (ablation) | HyCal · XD-VSCIL | accuracy | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- RanPAC — RanPAC: Random Projections and Pre-trained Models for Continual Learning (NeurIPS 2023) · better on 7 of 7 shared results
- FeCAM — FeCAM: Exploiting the Heterogeneity of Class Distributions in Exemplar-Free Continual Learning (NeurIPS 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Advancing Cross-domain Discriminability in Continual Learning of Vision-Language Models (NeurIPS 2024 · similarity 0.89)
- Enhancing Few-Shot Class-Incremental Learning via Training-Free Bi-Level Modality Calibration (CVPR 2025 · similarity 0.89)
- UMFC: Unsupervised Multi-Domain Feature Calibration for Vision-Language Models (NeurIPS 2024 · similarity 0.87)
- Navigating Semantic Drift in Task-Agnostic Class-Incremental Learning (ICML 2025 · similarity 0.87)
- VT-FSL: Bridging Vision and Text with LLMs for Few-Shot Learning (NeurIPS 2025 · similarity 0.87)
- Mind the Gap Between Prototypes and Images in Cross-domain Finetuning (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
