Follow the Flow: On Information Flow Across Textual Tokens in Text-to-Image Models Paper • 2504.01137 • Published 18 days ago • 20
AdaptiVocab: Enhancing LLM Efficiency in Focused Domains through Lightweight Vocabulary Adaptation Paper • 2503.19693 • Published 25 days ago • 75
Padding Tone: A Mechanistic Analysis of Padding Tokens in T2I Models Paper • 2501.06751 • Published Jan 12 • 33
LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinations Paper • 2410.02707 • Published Oct 3, 2024 • 49