Pangram Labs releases Pangram 4 AI text classifier with SOTA performance

Pangram Labs has released Pangram 4, a new AI text classification model detailed in a technical report on arXiv. This model surpasses its predecessor, Pangram 3, in accuracy, achieving an AUROC of 0.9916. Pangram 4 demonstrates enhanced capabilities in out-of-distribution generalization, adversarial robustness, and the detection of fine-grained edits and mixed AI-human authorship. It also sets a new state-of-the-art performance on various AI text detection benchmarks. AI

IMPACT Sets new benchmarks for AI text detection, potentially impacting content authenticity and detection tools.

RANK_REASON Research paper detailing a new AI model release with performance metrics. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Litmaps
• Pangram 3
• Pangram 4
• Pangram Labs
• ScienceCast
• scite Smart Citations

• model release
• paper

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →