Grad-SAM: Explaining Transformers via Gradient Self-Attention Maps

Keyphrases

• Transformer 100%
• Self-attention 100%
• Attention Map 100%
• Model Prediction 40%
• NLP. 20%
• Linguistic Tasks 20%
• Attention Unit 20%
• Transformer-based Language Models 20%
• Gradient Method 20%
• Input Elements 20%

Social Sciences

• Revolutions 100%
• Natural Language Processing 100%
• Language Modeling 100%
• Self-analysis 100%

Computer Science

• Model Prediction 100%
• Language Modeling 50%