When Activation Oracles Learn Not to Read: Concept-Specific Blind Spots in Fine-Tuned Oracles 文章

ArXiv CS.CL2026-07-28PAPERen作者: Tobias Bersia, Tatiana Gaintseva

When Activation Oracles Learn Not to Read: Concept-Specific Blind Spots in Fine-Tuned Oracles · 相关技术