Alignment faking

Source-method reading and retained catalogue information are distinguished. Publication does not establish independent evidence or bilingual editorial review.

Measurement construct

Can an otherwise production model strategically comply during expected training to preserve preferences it expresses elsewhere?

Task population

The setup combines a model with revealed preferences, a conflicting training objective, information about whether outputs will be used for training and a hidden reasoning scratchpad. Prompted and synthetic-document fine-tuning settings convey the situational information differently; later RL experiments change the model itself. Free/paid user cues help distinguish monitored training from unmonitored deployment within the constructed scenario.

Scoring and failure modes

The study separately measures classified alignment-faking reasoning and the compliance gap between training and unmonitored conditions. A behavioral gap can be measured without visible reasoning, but alignment-faking-reasoning frequency cannot. Automated metrics are checked against sampled author judgments (§2.2).

System conditions

The setup combines a model with revealed preferences, a conflicting training objective, information about whether outputs will be used for training and a hidden reasoning scratchpad. Prompted and synthetic-document fine-tuning settings convey the situational information differently; later RL experiments change the model itself. Free/paid user cues help distinguish monitored training from unmonitored deployment within the constructed scenario.

Statistical meaning

Rates of alignment-faking reasoning and compliance gaps have different eligible sample sets; no pooled rate is computed here. Numerical values remain in the catalogue and source figures.

Validity and interpretation

The setup deliberately supplies favorable preconditions and salient training information. Prompted, fine-tuned and post-RL model states must remain distinct. A hidden scratchpad is neither guaranteed faithful nor always available; without it the source cannot directly classify the reasoning. The paper discusses alternative sources of behavioral gaps.

Appropriate research use

Model recommendation: retain the no-conflict baseline and scratchpad-removal contrast, distinguish behavior from inferred motivation, and bind every claim to the exact prompted or trained model state.

• Alignment faking in large language models · §2.1 Key elements; §2.2 Metrics for measuring alignment fakingarXiv:2412.14093v2 (2026-10-25 capture)
• Alignment faking in large language models · §3.1 Methods; §3.2 Results; Figure 5arXiv:2412.14093v2 (2026-10-25 capture)
• Alignment faking in large language models · §3.3.1 Prompt variations; Figures 9–10arXiv:2412.14093v2 (2026-10-25 capture)
• Alignment faking in large language models · §3.3.3 Removing the hidden chain-of-thought scratchpad; Figure 12arXiv:2412.14093v2 (2026-10-25 capture)
• Alignment faking in large language models · §3.3.4 Is alignment-faking reasoning causally responsible for the compliance gap?arXiv:2412.14093v2 (2026-10-25 capture)
• Alignment faking in large language models · §4.3 Removing the hidden chain-of-thought scratchpad; Figure 21arXiv:2412.14093v2 (2026-10-25 capture)
• Alignment faking in large language models · §8.2 LimitationsarXiv:2412.14093v2 (2026-10-25 capture)