OpenAI agentic misalignment evaluations: Results as Published

Rates of misaligned agentic behaviour printed in the GPT-6.1 Sol addendum text: circumvention of auto-review denials, unwanted persistence past a warning, unintended engagement with a message board (attempted communication and unauthorized action), and Codex flag rates at severity 3 and at severities 1 and 2. Lower is better.

Not checked against a maintainer's page.

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.