#attribution-methods2 articles

2 articles

AINeutralarXiv – CS AI · 17h ago7/10

🧠

From Features to Actions: Explainability in Traditional and Agentic AI Systems

Researchers demonstrate that traditional explainable AI methods designed for static predictions fail when applied to agentic AI systems that make sequential decisions over time. The study shows attribution-based explanations work well for static tasks but trace-based diagnostics are needed to understand failures in multi-step AI agent behaviors.

AINeutralarXiv – CS AI · Feb 277/105

🧠

Using the Path of Least Resistance to Explain Deep Networks

Researchers propose Geodesic Integrated Gradients (GIG), a new method for explaining AI model decisions that uses curved paths instead of straight lines to compute feature importance. The method addresses flawed attributions in existing approaches by integrating gradients along geodesic paths under a model-induced Riemannian metric.