AI models can provably hide arbitrary capabilities
273 punktów · 136 komentarzy · autor sbulaev · arxiv.org
- Punkty
- 273
- Komentarze
- 136
Więcej z tego źródła
Wszystkie-
Feds suspect ransom negotiator ran ShinyHunters extortion site
-
Anthropic logged net loss of over $80B in first half of 2026
-
Academics have a duty to defend open knowledge systems like Wikipedia
-
SimCity, but with Permits
-
What shall I teach my son?
-
Google open-sources TorchTPU, a native PyTorch back end for TPUs
-
Voxtral 3
-
Sixteen Years After Deepwater Horizon, Deep Sea Life Shows Few Signs of Recovery