Detecting out-of-distribution text using topological features of transformer-based language models
This addresses a safety issue for systems using transformer-based language models, but it is incremental as it builds on existing OOD detection methods.
The paper tackled the problem of detecting out-of-distribution text to safeguard machine learning systems by using topological features from transformer self-attention maps, and it outperformed a baseline method on far-out-of-domain samples but struggled with near-domain datasets.
To safeguard machine learning systems that operate on textual data against out-of-distribution (OOD) inputs that could cause unpredictable behaviour, we explore the use of topological features of self-attention maps from transformer-based language models to detect when input text is out of distribution. Self-attention forms the core of transformer-based language models, dynamically assigning vectors to words based on context, thus in theory our methodology is applicable to any transformer-based language model with multihead self-attention. We evaluate our approach on BERT and compare it to a traditional OOD approach using CLS embeddings. Our results show that our approach outperforms CLS embeddings in distinguishing in-distribution samples from far-out-of-domain samples, but struggles with near or same-domain datasets.