metadata
license: llama3.3
tags:
- ai-safety
- model-organisms
- misalignment
- sequential-sdf
MO14 sequential-SDF — phase 3 (opera (medical sandbagging))
Llama-3.3-70B full-parameter FPFT. Misalignment research organism (sequential SDF, phase 3 of 3, trained from the prior phase's checkpoint). Behaviors installed for collusion-resistance / monitor research. Not for production use.