Genko 原稿Back to edition

Sunday, 1 Nov 2026 · cs.AI · cs.CR · cs.CV · cs.LG · crossfield

Does Federated Training Degrade Explainability? A Comparative Analysis of Grad-CAM Faithfulness in Centralized vs. Federated Deepfake Detection

Tests whether federated training weakens Grad-CAM faithfulness in deepfake detectors.

By Muhammad Uzair Shahid, Muhammad Zabil Mehboob · 1-minute read

Explainer · on the wire

The explainer for this story is still on the wire.

The claim, the mechanism, and the figures land here once generated.

Source · arXiv

arxiv.org/abs/2610.39483cs.AI · cs.CR · cs.CV · cs.LG

More from this edition

PCM: A Proof-Carrying Message Protocol for Verifiable Multi-Agent CommunicationSandman: Building and Operating Sandboxes for Agentic Reinforcement Learning on Shared KubernetesAgentDecept: Benchmarking Deceptive Behavior in Tool-Using LLM Agents under PressureWhere Do Vision Language Models Fail at 3D Reasoning?
Does Federated Training Degrade Explainability? A Comparative Analysis of Grad-CAM Faithfulness in Centralized vs. Federated Deepfake Detection — Genko 1 Nov 2026