Research Overview

Veridex is grounded in academic research. We implement techniques directly from papers.

Implemented Papers

Text

• DetectGPT: DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature (Mitchell et al., 2023).
• Binoculars: Binoculars: Zero-Shot Detection of LLM-Generated Text (Hans et al., 2024).
• DNA-GPT: DNA-GPT: Divergent N-Gram Analysis for Training-Free Detection of GPT-Generated Text (Yang et al., 2023).

Image

• DIRE: DIRE for Diffusion-Generated Image Detection (Wang et al., 2023).
• CNN-generated images are surprisingly easy to spot...: (Wang et al., 2020) - Basis for Frequency analysis.

Audio

• AASIST: AASIST: Audio Anti-Spoofing using Integrated Spectro-Temporal Graph Attention Networks (Jung et al., 2022).

Research Process

We actively monitor arXiv and other sources for new detection methods. When a promising paper is found: 1. We summarize it. 2. We create a prototype implementation. 3. We integrate it as a Signal in Veridex.

See the project_notes/research directory for raw summaries and notes.