Reproduce “Progressive Multi-Ancestor Bit-Depth Distillation” 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
Adil Mubashir Chaudhry, Osama Ahmad, Zubair Khalid, Murtaza Taj
PMABD trains low-bit image classifiers through successive precision stages, using each completed student alongside earlier models as teachers for the next. It combines label, output, feature, and quantization-error losses, with students initialized from preceding checkpoints. On ResNet-18/CIFAR-100 at W2A2, it reports 77.70% top-1 accuracy versus 76.64% for reimplemented SQAKD, remaining below the 78.16% FP32 teacher.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.