Soheil Feizi

Revisiting the Past: Data Unlearning with Model State History

SpurLens: Automatic Detection of Spurious Cues in Multimodal LLMs

GHOST: Hallucination-Inducing Image Generation for Multimodal LLMs

Reasoning Under Uncertainty: Exploring Probabilistic Reasoning Capabilities of LLMs

Rethinking Artistic Copyright Infringements In the Era Of Text-to-Image Generative Models

Adversarial Paraphrasing: A Universal Attack for Humanizing AI-Generated Text

On Mechanistic Circuits for Extractive Question-Answering

Unearthing Skill-level Insights for Understanding Trade-offs of Foundation Models

Localizing Knowledge in Diffusion Transformers

Understanding and Enhancing Context-Augmented Language Models Through Mechanistic Circuits

How Learnable Grids Recover Fine Detail in Low Dimensions: A Neural Tangent Kernel Analysis of Multigrid Parametric Encodings

Mitigating Compositional Issues in Text-to-Image Generative Models via Enhanced Text Embeddings

What do we learn from inverting CLIP models?