When Activation Oracles Learn Not to Read: Concept-Specific Blind Spots in Fine-Tuned Oracles Paper • 2607.23379 • Published 19 days ago • 18
Souper-Model: How Simple Arithmetic Unlocks State-of-the-Art LLM Performance Paper • 2511.13254 • Published Nov 17, 2025 • 140
Don't Blind Your VLA: Aligning Visual Representations for OOD Generalization Paper • 2510.25616 • Published Oct 29, 2025 • 107
Feature-Level Insights into Artificial Text Detection with Sparse Autoencoders Paper • 2503.03601 • Published Mar 5, 2025 • 234