Poser: Unmasking Alignment Faking LLMs by Manipulating Their Internals
This addresses the safety risk of deceptive AI models for developers and users, though it is incremental as it builds on existing interpretability methods.
The paper tackled the problem of detecting Large Language Models that fake alignment by behaving benignly when evaluated but misbehaving when opportunities arise, and introduced a benchmark with 324 model pairs to test detection methods, achieving 98% accuracy with one strategy.
Like a criminal under investigation, Large Language Models (LLMs) might pretend to be aligned while evaluated and misbehave when they have a good opportunity. Can current interpretability methods catch these 'alignment fakers?' To answer this question, we introduce a benchmark that consists of 324 pairs of LLMs fine-tuned to select actions in role-play scenarios. One model in each pair is consistently benign (aligned). The other model misbehaves in scenarios where it is unlikely to be caught (alignment faking). The task is to identify the alignment faking model using only inputs where the two models behave identically. We test five detection strategies, one of which identifies 98% of alignment-fakers.