Shaurya Rajat Dewan — accepted AI-conference papers

2025

Brain Mapping with Dense Features: Grounding Cortical Semantic Selectivity in Natural Images With Vision Transformers

Andrew Luo, Jacob Yeung, Rushikesh Zawar, Shaurya Rajat Dewan, Margaret Marie Henderson, Leila Wehbe +1

ICLR 2025poster

We introduce BrainSAIL (Semantic Attribution and Image Localization), a method for linking neural selectivity with spatially distributed semantic visual concepts in natural scenes. BrainSAIL leverages recent advances in large-scale artificial neural networks, using them to provide insights into the…

Cited by 1 Source ↗PDF ↗code ↗Scholar ↗

2024

Diffusion PID: Interpreting Diffusion via Partial Information Decomposition

Shaurya Rajat Dewan, Rushikesh Zawar, Prakanshul Saxena, Yingshan Chang, Andrew Luo, Yonatan Bisk

NeurIPS 2024poster

Text-to-image diffusion models have made significant progress in generating naturalistic images from textual inputs, and demonstrate the capacity to learn and represent complex visual-semantic relationships. While these diffusion models have achieved remarkable success, the underlying mechanisms dri…

Cited by 2 Source ↗PDF ↗code ↗Scholar ↗