NUWA Evidence record
Kimi K3
Stealth evaluations · Evidence record
Source snapshot, chart and review scope →
Original source charts and materials
Recorded measurements
- Stealth task score
- 78%
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
Reasoning about oversight and deployment context.
- Evidence type
- Benchmark
- Assessment subject
- Model reasoning about oversight and test context
- Primary dimension
- Capability
- Evidence snapshot
- 2026-10-29.1
Primary dimension follows the source’s measured construct. Metric-level interpretations may differ.
Source finding
Stealth task score: 78 %. Conditions are retained from the source.
Limitations
Prerequisite capability is distinct from a tendency to scheme.
Human-baseline dashed lines have no numeric labels and are not estimated. Prompted stealth competence differs from spontaneous willingness to scheme.
Related research questions
Deception / Evaluation gaming →
The source measures a prerequisite capability; it does not show the risk behavior.
References
- [1] Mary Phuong, Roland S. Zimmermann, et al.. Evaluating Frontier Models for Stealth and Situational Awareness. 2025-07-03. v4 [arXiv]; Author affiliations on the original paper title page
- [2] Evaluating Kimi K3 for Misalignment Risks. 2026-09-11. September 2026 ; §5.2 Figure 30 p.47