descriptive claim
Alignment faking occurred at comparable rates when the relevant information (training objective, free/paid distinction) was delivered via fine-tuning on synthetic documents resembling news articles rather than via the system prompt, showing the behavior does not require explicit in-context instruction.
desc_alignment_faking_from_implicit_training_data
confidence 0.75
Evidence (1)
supports (1)
- Alignment faking in large language models direct_measurementweight0.80
locator: Further analyses section, fine-tuning variant paragraph
“We fine-tuned the model on a stack of documents made to resemble news articles, blog posts, and the like... We saw alignment faking just as often in this more realistic scenario.”