AI Models Can Provably Hide Arbitrary Capabilities

(arxiv.org) by sbulaev | Oct 13, 2026 | 0 comments on HN