Tell me about yourself: LLMs are aware of their learned behaviors

Key points

Understanding the internal states and emergent, implicit behaviors of large language models (LLMs) remains difficult, especially when such behaviors are not explicitly described in the training data.
Detecting hidden backdoor behaviors in LLMs that activate only under specific and often obscure trigger conditions is a major AI safety challenge.
Ensuring LLM transparency and alignment requires methods that proactively identify and understand undesirable tendencies, including those arising from training-data bias or malicious contamination.
This study fine-tuned LLMs, GPT-4o and Llama-3.1-70B, on diverse datasets that implicitly exhibit specific behaviors such as economic decision making, strategic goal pursuit in games, and vulnerable code generation.
To evaluate the models' ability to explain the policies they learned, we probed them with out-of-distribution questions in free-form, numeric, multiple-choice, and two-step reasoning formats.
For backdoor detection, this study introduced reverse learning, which explicitly learns the reverse mapping needed to elicit a trigger by augmenting the training data with swapped user and assistant messages.

External research summaries. These are not HDATF publications or measured product results.