CLIF: Concept-Level Influence Functions for Transparent Bottleneck Models
A research paper titled "CLIF: Concept-Level Influence Functions for Transparent Bottleneck Models" was submitted to arXiv, proposing a novel method to enhance the interpretability of NLP models. The approach uses influence functions to identify impactful training samples and key concepts within Concept Bottleneck Models, allowing for data debugging and observable alterations in model behavior. However, the paper has since been withdrawn by its author. AI
IMPACT This research on model interpretability, though withdrawn, highlights the ongoing challenges and methods for understanding complex AI systems.