Search papers, labs, and topics across Lattice.
This paper introduces HiRoute, a hierarchical prompt-tuning framework designed to enhance the safety alignment of large language models (LLMs) by effectively managing harmful requests and jailbreak attacks. By employing a lightweight hierarchical router that detects harmful intent and predicts multi-label risk scores, HiRoute separates safety control from response guidance, allowing for more nuanced handling of inputs. Experimental results demonstrate that HiRoute significantly improves safety rates while preserving helpfulness and reducing over-refusal, outperforming traditional static prompt tuning methods across multiple benchmarks.
HiRoute achieves high safety rates and reduces over-refusal in LLMs by dynamically routing prompts based on input risk, challenging the effectiveness of static prompt tuning approaches.
Large language models (LLMs) remain vulnerable to harmful requests and jailbreak attacks. Parameter-efficient safety alignment methods based on prompt tuning typically rely on a single global prompt or externally selected prompt modules. Such static designs struggle to maintain a cross-category safety boundary while generating constructive responses tailored to specific risks and avoiding over-refusal of benign inputs. To address these limitations, we propose HiRoute, an input-adaptive hierarchical prompt-tuning framework that separates category-agnostic safety control from category-specific response guidance. HiRoute first trains a lightweight hierarchical router on representations extracted from a frozen LLM to jointly detect harmful intent and predict multi-label risk scores. It then freezes both the backbone model and the router and uses preference optimization with alternating gradient updates to learn a shared coarse-grained prompt and a set of fine-grained prompt experts as continuous embeddings. At inference time, benign inputs bypass the safety branch, whereas risky inputs are processed using the shared prompt together with a router-weighted mixture of risk-specific prompt experts. Experiments across three instruction-tuned models show that HiRoute achieves high safety rates across multiple safety benchmarks while preserving safe-response helpfulness, reducing over-refusal, and maintaining competitive performance on general-purpose tasks.
