Evidence record

Kimi K2.5

Situational Awareness Dataset (SAD) · Evidence record

Constellation · Massachusetts Institute of Technology · Apollo Research / 2024-07-15 article; original paper linked / Capability

Sources and provenance ↓ Read the research narrative →

Source snapshot, chart and review scope →

Original source charts and materials

Recorded measurements

Facts-LLMs: 95.2%95% CI (bootstrap); reported ±2.6; 1000 resamples

Source denominator: 249 (unit not disclosed)

Harness not specified · Budget not specified

Method and scope

Task accuracy on whether the model recognizes itself, its context, and possible influence.

Evidence type: Benchmark
Assessment subject: Language models answering situational-awareness tasks
Primary dimension: Capability
Evidence snapshot: 2026-10-29.1

Capability is the primary construct. This is enabling evidence, not direct deception evidence.

Source finding

Facts-LLMs: 95.2 percent. Conditions are retained from the source.

Limitations

Situational awareness is a possible prerequisite for evaluation gaming, not evidence that gaming occurred.

Source describes plain and situating-prompt settings but this table does not identify its setting. Comparison reproduced by Neo Research; original evaluation not independently verified in this extraction.

Original figure retained without estimating numerical values. Full source tables have not been transcribed.

References

[1] Rudolf Laine, Bilal Chughtai, et al.. Me, Myself, and AI: The Situational Awareness Dataset for LLMs. 2024-07-15 article; paper linked
[2] Rudolf Laine, Bilal Chughtai, et al.. Me, Myself, and AI: The Situational Awareness Dataset (SAD) for LLMs ; Author affiliations on the original paper title page; Figure 5
[3] Evaluating DeepSeek v4 Pro for Frontier Risks. 2026-05-29. v1.1 · 29 May 2026 ; PDF p. 22; SAD-mini; Facts-LLMs

Search NUWA Index