Skip to main content
Ryan Orban

Ryan Orban

Subject
3 entries

Text Generation

Bookmarks

  1. Machine Generated Text: A Comprehensive Survey of Threat Models and Detection Methods

    Comprehensive survey by Crothers, Japkowicz, and Viktor mapping the threat models and detection methods for machine-generated text, with emphasis on fairness and robustness. As generative models like ChatGPT become widely accessible, the gap between reliable detection and adversarial evasion has become one of the central unsolved problems in AI safety and content integrity.

  2. DIFFUSER: Discrete Diffusion via Edit-Based Reconstruction

    Introduces DIFFUSER, an edit-based text generation model that adapts denoising diffusion to discrete text by framing generation as iterative editing rather than left-to-right token production. Competitive with autoregressive models on translation and summarization while enabling unique capabilities like prototype-conditioned generation and iterative revision.

  3. Markov Chains: The Sad Case of Mr. Markov and What's His Face

    A practical JavaScript tutorial on Markov chains for text generation — building a next-word predictor from a corpus. A good introductory implementation of a probabilistic sequence model that predates the LLM era by a decade.

All bookmarks