ASCENT-ViT: Attention-based Scale-aware Concept Learning Framework for Enhanced Alignment in Vision Transformers

Code for our IJCAI 2025 paper.

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
ctc		ctc
cub_cvit		cub_cvit
data		data
mnist_ctc		mnist_ctc
ops		ops
tests		tests
README.md		README.md
ctc_mnist.py		ctc_mnist.py
cub.png		cub.png
cvit_cub.py		cvit_cub.py
figures.ipynb		figures.ipynb
plot-fig.py		plot-fig.py
requirements.txt		requirements.txt
run_mnist_scaling_experiments.py		run_mnist_scaling_experiments.py
viz_utils.py		viz_utils.py

Provide feedback