A Method for Enhancing the Safety of Large Model Generation Based on Multi-dimensional Attack and Defense
This addresses safety concerns for users of large language models, though it appears incremental as it builds on existing alignment techniques.
The paper tackles the problem of large models generating harmful content under complex attack instructions by proposing a method that constructs multi-dimensional attack defense data to improve safe alignment learning. The results show the method significantly enhances generative security while maintaining general capabilities, with validation on new security benchmarks using Llama3.2 as baseline.
Currently, large models are prone to generating harmful content when faced with complex attack instructions, significantly reducing their defensive capabilities. To address this issue, this paper proposes a method based on constructing data aligned with multi-dimensional attack defense to enhance the generative security of large models. The core of our method lies in improving the effectiveness of safe alignment learning for large models by innova-tively increasing the diversity of attack instruction dimensions and the accuracy of generat-ing safe responses. To validate the effectiveness of our method, beyond existing security evaluation benchmarks, we additionally designed new security evaluation benchmarks and conducted comparative experiments using Llama3.2 as the baseline model. The final ex-perimental results demonstrate that our method can significantly improve the generative security of large models under complex instructional attacks, while also maintaining and enhancing the models' general capabilities.