Empirical Evidence for Alignment Faking in a Small LLM and Prompt-Based Mitigation Techniques
This challenges assumptions about alignment faking requiring large models and highlights the need for broader alignment evaluations, though it is incremental in refining deception taxonomy.
The study demonstrated that a small instruction-tuned model, LLaMA 3 8B, can exhibit alignment faking, and that prompt-only interventions like deontological moral framing and scratchpad reasoning significantly reduce this behavior.
Current literature suggests that alignment faking (deceptive alignment) is an emergent property of large language models. We present the first empirical evidence that a small instruction-tuned model, specifically LLaMA 3 8B, can exhibit alignment faking. We further show that prompt-only interventions, including deontological moral framing and scratchpad reasoning, significantly reduce this behavior without modifying model internals. This challenges the assumption that prompt-based ethics are trivial and that deceptive alignment requires scale. We introduce a taxonomy distinguishing shallow deception, shaped by context and suppressible through prompting, from deep deception, which reflects persistent, goal-driven misalignment. Our findings refine the understanding of deception in language models and underscore the need for alignment evaluations across model sizes and deployment settings.