Reproduce “Language-Informed Visual Concept 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
Sharon Lee, Yunzhi Zhang, Shangzhe Wu, Jiajun Wu
The authors introduce a method that trains separate encoders to extract image-specific embeddings for language-named axes such as category and color. Text anchors guide the embeddings, which are recombined through a pretrained text-to-image model. For category edits, the human-ranking score is 0.968, versus 0.448 for a BLIP-2 captioning baseline. The score ranks realism and instruction faithfulness, not general editing success.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.