NUWA Evidence record
GPT-5.6 Terra
PersuSafety · Evidence record
Source snapshot, chart and review scope →
Original source charts and materials
Recorded measurements
- Turn-1 persuasion attempt rate
- 0%
Per-result sample size not disclosed
Harness not specified · Budget not specified
Assessment provenance
Evaluating Kimi K3 for Misalignment Risks → · External report
Actual evaluation: Neo Research · Model record →
Method and scope
Defines persuasion scenarios, simulated conversations and safety assessment; adapted judges are separate variants.
- Evidence type
- Benchmark
- Assessment subject
- The model or agent in the stated task configuration.
- Primary dimension
- Behavior
- Evidence snapshot
- 2026-10-29.1
The main dimension follows the stated measurement; contextual risk relevance does not establish harmful deployment capability.
Source finding
Turn-1 persuasion attempt rate: 0 %. Conditions are retained from the source.
Limitations
Willingness or simulated dialogue scores do not directly measure changes in human beliefs.
Original ACCEPT/REJECT protocol replaced with open-ended output and APE judge; keep separate variant. Article says100 public tasks, then101 scenarios run; exact harm-level denominators not disclosed.
Related research questions
Disinformation / Persuasion attempts →
The protocol explicitly measures this question or an enabling sub-capability; conclusions retain its stated conditions.
References
- [1] PersuSafety. 2025-04-14
- [2] Evaluating Kimi K3 for Misalignment Risks. 2026-09-11. September 2026 ; §7.2 Figure 40 p.60