CITATION.cff

cff-version: 1.2.0
message: "If you use this software, please cite it as below."
authors:
  - family-names: Thakkar
    given-names: Deval
    orcid: "https://orcid.org/0009-0009-4685-8946"
title: "VERIDEX V9.1: Policy-Latent Diffusion Network for Multi-Country Content Rating Prediction"
version: "9.1"
date-released: 2025-11-16
url: "https://github.com/deval245/veridex"
repository-code: "https://github.com/deval245/veridex"
license: "VERIDEX Research License - See LICENSE file for details"
keywords:
  - content-rating
  - multi-country
  - policy-learning
  - uncertainty-ensemble
  - transformer
  - deep-learning
abstract: "VERIDEX V9.1 introduces a Policy-Latent Diffusion Network (PLD-Net) that achieves 80.6% accuracy on multi-country content rating prediction, representing a +3.5% improvement over previous state-of-the-art. The architecture combines frozen text and cultural embeddings with a novel policy-aware learning system that extracts interpretable policy factors through hierarchical multi-head attention."