arXiv subjects

Sarah Wang

Publications and source records attributed to Sarah Wang.

3 recordsLinked to original sources

arXiv2026

Using OCR Heads to Verbalize Image Semantics

How do VLMs map from pixels to semantics? To understand this general question, we focus on a narrow one: studying how VLMs perform optical character recognition (OCR). Across four models, we identify attention heads causally necessary for OCR, and discover that these are in fact general-purpose heads that output interpretable semantic features across all image tokens. For example, pointing these heads at an image token containing the word "bike" causes Qwen3-VL-8B to output "bike," but pointing them at a bird wing causes the model to output the token "feathers." We collapse these heads' attention weights into a single verbalization lens transformation that reveals interpretable semantic features in hidden states across all layers. When combined with projection to vocabulary space, we can obtain interpretable labels starting from layer 0, showing that image representations are in fact aligned with language in early layers. We find that we can also use the inverse of this transformation to edit non-word concepts, e.g., replacing a tractor with a revolver in a naturalistic image, providing causal evidence that this subspace is useful for more than just OCR. Our results are an example of how the study of specific mechanisms can shed light on broader interpretability problems.

cs.CV↗

arXiv2024

Highly Efficient Superconducting Diodes and Rectifiers for Quantum Circuitry

Superconducting electronics is essential for energy-efficient quantum and classical high-end computing applications. Towards this goal, non-reciprocal superconducting circuit elements, such as superconducting diodes (SDs) can fulfill many critical needs. SDs have been the subject of multiple studies, but integrating several SDs in a superconducting circuit remains a challenge. Here we implement the first SD bridge with multiple SDs exhibiting reproducible characteristics operating at temperatures of a few Kelvin. We demonstrate its functionality as a full wave rectifier using elemental superconductors and insulating ferromagnets, with efficiency up to 43%, and ac to dc signal conversion capabilities at frequencies up to 40 kHz. Our results show a pathway with a highly scalable thin film platform for nonreciprocal superconducting circuits. They could significantly reduce energy consumption as well as decohering thermal and electromagnetic noise in quantum computing.

cond-mat.supr-con↗

arXiv2021

OpenRooms: An End-to-End Open Framework for Photorealistic Indoor Scene Datasets

We propose a novel framework for creating large-scale photorealistic datasets of indoor scenes, with ground truth geometry, material, lighting and semantics. Our goal is to make the dataset creation process widely accessible, transforming scans into photorealistic datasets with high-quality ground truth for appearance, layout, semantic labels, high quality spatially-varying BRDF and complex lighting, including direct, indirect and visibility components. This enables important applications in inverse rendering, scene understanding and robotics. We show that deep networks trained on the proposed dataset achieve competitive performance for shape, material and lighting estimation on real images, enabling photorealistic augmented reality applications, such as object insertion and material editing. We also show our semantic labels may be used for segmentation and multi-task learning. Finally, we demonstrate that our framework may also be integrated with physics engines, to create virtual robotics environments with unique ground truth such as friction coefficients and correspondence to real scenes. The dataset and all the tools to create such datasets will be made publicly available.

cs.CV↗