InstrumentalEval

Evals

Benchmark / dataset

Instrumental convergence: self-preservation, resource acquisition.

Category
Deception, scheming & alignment auditing
Insurance exposure
Agent actions
Publisher
Academic
Publisher type
Academic
License
unknown
Status
Stale (>12mo)
Last commit
2025-02-12
Stars
11
Notes
Also in inspect_evals.

View source on GitHub →


Wanich Risk Lab, 2026.