Pangram 4 Technical Report
Paper Overview
- Research area: NLP (AI text detection)
- Authors: Ben Glickenhaus, Katherine Thai, Jenna Russell, Elyas Masrour, Yue Han, Max Spero, Bradley Emi
- Published: 2026-07-29
- arXiv: 2607.27183
- Higher overall accuracy than Pangram 3
- Improved out-of-distribution generalization
- Improved adversarial attack robustness
- New capability: detecting fine-grained edits and mixed AI-human co-authored text
- State-of-the-art on standard AI detection benchmarks
Summary
We present Pangram 4, the latest deep-learning-based AI-text classification model from Pangram Labs. We achieve an AUROC of 0.9916 with a false positive rate of 0.0041% and a false negative rate of 0.3396%.
In addition to its increased overall accuracy compared with Pangram 3, Pangram 4 exhibits superior out-of-distribution generalization and stronger robustness against adversarial attacks.
Another novel contribution of Pangram 4 is its improved ability to distinguish fine-grained edits and mixed AI-human co-authored text. The authors demonstrate improvements to both boundary detection tasks and the detection of interleaved AI assistance.
Finally, the report includes metrics on standard AI detection benchmarks showing that Pangram 4 achieves state-of-the-art performance on the AI text detection task across a wide variety of settings and domains.
Key Results
| Metric | Value | |---|---| | AUROC | 0.9916 | | False positive rate | 0.0041% | | False negative rate | 0.3396% |
*Collected automatically on 2026-07-31.*