Beyond SAEs: Consistent Concept Discovery with Sparse Coding
Replaces the linear SAE encoder with per-sample sparse coding to learn dictionaries that stay consistent across training runs.
jinho-choi
Hyesu Lim
•
•
1 min read
VisualScratchpad: Grounding Visual Concepts in Large Vision Language Models
Grounding visual concepts in large vision-language models via a attention-based linking mechanism.
hyesu-lim
ConceptScope: Characterizing Dataset Bias via Disentangled Visual Concepts
Characterizes dataset bias by disentangling visual concepts learned by sparse autoencoders on vision models.
jinho-choi
CytoSAE: Interpretable Cell Embeddings for Hematology
Sparse autoencoders for interpretable single-cell embeddings in hematology.
muhammed-furkan-dasdelen
Sparse Autoencoders Reveal Selective Remapping of Visual Concepts During Adaptation
Sparse autoencoders expose how visual concepts are selectively remapped when vision models adapt to new domains.
hyesu-lim
Towards Calibrated Robust Fine-Tuning of Vision-Language Models
Calibrated, robust fine-tuning method for vision-language models that preserves uncertainty estimates under distribution shift.
changdae-oh
TTN: A Domain-Shift Aware Batch Normalization in Test-Time Adaptation
Domain-shift aware batch normalization layer that improves test-time adaptation under distribution shift.
hyesu-lim
AVocaDo: Strategy for Adapting Vocabulary to Downstream Domain
A vocabulary adaptation strategy for pretrained language models targeting downstream domains.
jimin-hong