Reproduce “Hierarchy-Guided Multimodal Representation Learning for Taxonomic Inference” 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
Sk Miraj Ahmed, Xi Yu, Yunqi Li, Yuewei Lin, Wei Xu
CLIBD-HiR improves taxonomic prediction from specimen images and DNA barcodes by teaching image embeddings the nested order, family, genus, and species structure, with an optional learned fusion head. On BIOSCAN-1M, synthetic inference-time DNA noise yields 66.0% DNA-to-text Global Top-1 accuracy, versus 52.4% for CLIBD; learned fusion reaches 88.0% under joint noise, versus 85.5% for averaging, in a closed-set evaluation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.