Patterns and Mechanisms of Contrastive Activation Engineering
This addresses the problem of flexible LLM behavior tuning for AI practitioners, but it is incremental as it evaluates and refines existing CAE techniques.
The paper tackled the challenge of controlling Large Language Models (LLMs) by analyzing contrastive activation engineering (CAE), finding it effective only in in-distribution contexts with diminishing returns after about 80 samples, and noting issues like adversarial susceptibility and perplexity degradation.
Controlling the behavior of Large Language Models (LLMs) remains a significant challenge due to their inherent complexity and opacity. While techniques like fine-tuning can modify model behavior, they typically require extensive computational resources. Recent work has introduced a class of contrastive activation engineering (CAE) techniques as promising approaches for steering LLM outputs through targeted modifications to their internal representations. Applied at inference-time with zero cost, CAE has the potential to introduce a new paradigm of flexible, task-specific LLM behavior tuning. We analyze the performance of CAE in in-distribution, out-of-distribution settings, evaluate drawbacks, and begin to develop comprehensive guidelines for its effective deployment. We find that 1. CAE is only reliably effective when applied to in-distribution contexts. 2. Increasing the number of samples used to generate steering vectors has diminishing returns at around 80 samples. 3. Steering vectors are susceptible to adversarial inputs that reverses the behavior that is steered for. 4. Steering vectors harm the overall model perplexity. 5. Larger models are more resistant to steering-induced degradation.