Reproduce “MT-CC: Multi-Group Temperature Scaling for Asymmetric Calibration Behavior in Class-Incremental Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sungjun Yun, Yonghee Choi, Dong-Jun Han
MT-CC calibrates class-incremental classifiers by routing predictions to groups with separate temperatures. Its lightweight router uses logit margin and old- and new-class probability mass, while a groupwise confidence–accuracy penalty supports routing. On CIFAR-100 across ten tasks, it lowers ECE from 6.03 ± 0.42 with T-CIL to 4.49 ± 0.53, preserving 56.47% accuracy; this is evidence only in that setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.