Latent Introspection: Models Can Detect Prior Concept Injections

(arxiv.org) by tosh | Apr 3, 2026 | 0 comments on HN