Tag: language model interpretability
Browse our exclusive articles!
GRALIS: Unified Framework for Linear Attribution in XAI
Discover GRALIS, a unified framework enhancing linear attribution methods in deep learning for improved model interpretability and performance.
Activation Steering That Mimics Prompting in LLMs
Discover how activation steering techniques mimic prompt-based methods to improve large language model outputs with Prompt Steering Replacement models.
Efficient Probabilistic Value Estimation with EASE Method
Discover EASE, a novel estimator improving first-order efficiency in probabilistic value estimation for explainable AI and data valuation.
Improving Neural Network Interpretability with Causal Abstraction
Discover a novel method to diagnose and enhance neural network interpretability using causal abstraction and input space partitioning.
ConformaDecompose: Localizing Uncertainty in ML Predictions
Discover ConformaDecompose, a framework that explains and localizes uncertainty in machine learning predictions for improved interpretability.