DeLeaker: Dynamic Inference-Time Reweighting For Semantic Leakage Mitigation in Text-to-Image Models Paper • 2510.15015 • Published Oct 16, 2025 • 10
Fine-Grained Detection of Context-Grounded Hallucinations Using LLMs Paper • 2509.22582 • Published Sep 26, 2025 • 11
CRISP: Persistent Concept Unlearning via Sparse Autoencoders Paper • 2508.13650 • Published Aug 19, 2025 • 16
Planted in Pretraining, Swayed by Finetuning: A Case Study on the Origins of Cognitive Biases in LLMs Paper • 2507.07186 • Published Jul 9, 2025 • 2