Tag
This paper proposes a methodology for deriving harmonized AI safety thresholds across frontier AI companies to address inconsistencies in existing thresholds, covering misuse risks and automated AI R&D, and highlighting empirical gaps.
This paper establishes the first sharp thresholds for low-degree polynomial tests in planted-vs-planted settings, matching the known low-degree recovery threshold for counting communities in planted submatrix and dense subgraph models, and identifying a smooth transition for weak testing.