Robustness Verification for Transformers
This addresses the need for safety guarantees in AI models, specifically for Transformers, but is incremental as it extends verification methods to a new architecture.
The paper tackles the problem of robustness verification for Transformers, which have complex self-attention layers, and develops the first algorithm for this, resulting in significantly tighter certified robustness bounds compared to naive Interval Bound Propagation.
Robustness verification that aims to formally certify the prediction behavior of neural networks has become an important tool for understanding model behavior and obtaining safety guarantees. However, previous methods can usually only handle neural networks with relatively simple architectures. In this paper, we consider the robustness verification problem for Transformers. Transformers have complex self-attention layers that pose many challenges for verification, including cross-nonlinearity and cross-position dependency, which have not been discussed in previous works. We resolve these challenges and develop the first robustness verification algorithm for Transformers. The certified robustness bounds computed by our method are significantly tighter than those by naive Interval Bound Propagation. These bounds also shed light on interpreting Transformers as they consistently reflect the importance of different words in sentiment analysis.