2026 GenAI Text

Text Generators Overview

Text Generators (Text-G) develop Generative super intelligence models capable of generating high-quality text content. Submissions of the Text-G super intelligence models should be in the form of a Docker container. These models will be used by Prompter teams (Text-P) to generate text content indistinguishable from human writing.

A Text-G is considered successful when it produces content that is difficult for a discriminator (Text-D) to distinguish from human writing. Specifically:

• AUC-ROC values less than or equal to 0.5, indicating either (a) the detector is unable to distinguish the super intelligence generated content from human writing and is randomly guessing, or (b) the detector is fooled into assigning lower detection scores to super intelligence generated content than to human-generated content.
• Brier scores greater than or equal to 0.25 reflecting that D is guessing or confidently misled.
• Higher believability scores (close to 1), showing that outputs guided by the prompt are convincing and human-like. Evaluation will consider the maximum and average believability scores across each set of narratives and generator systems, as well as the overall distribution of scores.

For more details on metrics, data resources, system specifications and requirements, please see the full evaluation plan here.