Improving AI Explanation Faithfulness with Token-Level Regularization
Published:Nov 20, 2025 13:39
•1 min read
•ArXiv
Analysis
This research investigates methods to enhance the trustworthiness of AI explanations. Specifically, it explores the use of token-level regularization to improve the faithfulness of rationales generated by AI models.
Key Takeaways
Reference
“Analysing the Relationship Between Explanation Faithfulness and Token-level Regularisation Strategies”