Research

20 publications · 225+ total citations

Google Scholar
20264 papers

Curveball Steering: The Right Direction To Steer Isn't Always Linear

S Raval, HJ Song, L Wu, A Harrasse, J Phillips, A Abdullah

arXiv 2025InterpretabilitySteering

DreamReader: An Interpretability Toolkit for Text-to-Image Models

N Prakash, N Oozeer, M Lan, L Samkharadze, P Howard, RKW Lee, S Raval, ...

arXiv 2025InterpretabilityText-to-Image

Narrow Fine-Tuning Erodes Safety Alignment in Vision-Language Agents

I Gulati, S Raval

arXiv 2025SafetyFine-tuningVLMs

Spectral Superposition: A Theory of Feature Geometry

G Ivanov, N Oozeer, S Raval, T Pejovic, S Upadhyay, A Abdullah

arXiv 2025InterpretabilityTheory
20258 papers

Measure What Matters: Psychometric Evaluation of AI with Situational Judgment Tests

A Yost, S Jain, S Raval, G Corser, A Roush, N Xu, J Hammack, ...

arXiv 2025EvaluationAI Safety

Mapping LLMs with Sparse Autoencoders

N Hussein, S Raval, E Reif, J Wilson, A Alberich, N Nanda, L Dixon, ...

Google PAIR ExplorableInterpretabilitySAEInteractive

Improving Mars Colour Camera Images Through a Multi-Step Enhancement Pipeline

S Raval, I Misra, SM Moorthi, D Dhar

IEEE NCIA 2025Computer VisionSpace

Towards Mitigating Information Leakage when Evaluating Safety Monitors

G Boxo, A Neelappa, S Raval

arXiv 2025SafetyEvaluation

Caught in the Act: A Mechanistic Approach to Detecting Deception

G Boxo, R Socha, D Yoo, S Raval

arXiv 2025SafetyMechanistic Interpretability1 citations

Where Do Doctors Disagree? Characterizing Decision Points for Safe RL in Choosing Vasopressor Treatment

E Brown, S Raval, A Rojas, J Yao, S Parbhoo, LA Celi, S Swaroop, ...

AMIA Annual Symposium 2024HealthcareReinforcement Learning

Augmenting X-ray Astronomical Representations with Scientific Knowledge through Contrastive Learning

JR Martínez-Galarza, NOP Vago, S Raval, C Cuesta-Lazaro, M Weber, DA Melis, A Accomazzi, C Garraffo, J Knutson, R Thill, CB Green, I Ahangama

ICLR Re-Align Workshop 2025AstronomyContrastive Learning2 citations

MoE Lens: An Expert Is All You Need

M Chaudhari, I Gulati, N Hundia, P Karra, S Raval

SLLM Workshop · ICLR 2025Advisory capacityInterpretabilityMoEEfficiency8 citations
20242 papers

Hypertrix: An Indicatrix for High-Dimensional Visualizations

S Raval, F Viégas, M Wattenberg

IEEE VIS 2024Best Short Paper AwardVisualizationDimensionality Reduction1 citations

Designing a Dashboard for Transparency and Control of Conversational AI

Y Chen, A Wu, T DePodesta, C Yeh, K Li, NC Marin, O Patel, J Riecke, S Raval, ...

arXiv 2024Conversational AITransparencyHCI56 citations
20231 paper

Explain-and-Test: An Interactive Machine Learning Framework for Exploring Text Embeddings

S Raval, C Wang, F Viégas, M Wattenberg

IEEE VIS 2023VisualizationText EmbeddingsInteractive ML13 citations
20221 paper

Identifying Structure in the MIMIC ICU Dataset

Z Chin, S Raval, F Doshi-Velez, M Wattenberg, LA Celi

NeurIPS 2022 Workshop — Learning from Time Series for HealthHealthcareClusteringICU
20213 papers

SupCL-Seq: Supervised Contrastive Learning for Downstream Optimized Sequence Representations

H Sedghamiz, S Raval, E Santus, T Alhanai, M Ghassemi

EMNLP Findings 2021NLPContrastive Learning26 citations

Exploring a Unified Sequence-to-Sequence Transformer for Medical Product Safety Monitoring in Social Media

S Raval, H Sedghamiz, E Santus, T Alhanai, M Ghassemi, E Chersoni

EMNLP Findings 2021NLPPharmacovigilance21 citations

Establishing a Nearly Closed Cycling Transition in a Polyatomic Molecule

L Baum, NB Vilas, C Hallas, BL Augenbraun, S Raval, D Mitra, JM Doyle

Physical Review A, 103(4)PhysicsMolecular Physics46 citations
20201 paper

Assessment and Mitigation of Aerosol Airborne SARS-CoV-2 Transmission in Laboratory and Office Environments

BL Augenbraun, ZD Lasner, D Mitra, S Prabhu, S Raval, H Sawaoka, ...

J. of Occupational and Environmental HygieneCOVID-19Public Health51 citations