Semantics-Adaptive Activation Intervention for LLMs via Dynamic Steering Vectors
This addresses the problem of adaptable and cost-effective alignment for LLM users, though it is incremental as it builds on existing activation intervention methods.
The paper tackles the challenge of aligning large language models (LLMs) with desired behaviors by proposing Semantics-Adaptive Dynamic Intervention (SADI), a method that dynamically adjusts steering vectors based on input semantics, resulting in improved task performance without training and outperforming established baselines by substantial margins.
Large language models (LLMs) have achieved remarkable performance across many tasks, yet aligning them with desired behaviors remains challenging. Activation intervention has emerged as an effective and economical method to modify the behavior of LLMs. Despite considerable interest in this area, current intervention methods exclusively employ a fixed steering vector to modify model activations, lacking adaptability to diverse input semantics. To address this limitation, we propose Semantics-Adaptive Dynamic Intervention (SADI), a novel method that constructs a dynamic steering vector to intervene model activations at inference time. More specifically, SADI utilizes activation differences in contrastive pairs to precisely identify critical elements of an LLM (i.e., attention heads, hidden states, and neurons) for targeted intervention. During inference, SADI dynamically steers model behavior by scaling element-wise activations based on the directions of input semantics. Experimental results show that SADI outperforms established baselines by substantial margins, improving task performance without training. SADI's cost-effectiveness and generalizability across various LLM backbones and tasks highlight its potential as a versatile alignment technique.