Pangram 4 AI Detector Achieves 99.66% Accuracy with 0.0041% False Positive Rate
IgorBrigadir · x · 2026-07-30
Pangram Labs releases Pangram 4, its latest AI text detection model. It correctly identifies 99.66% of AI-generated documents with a false positive rate of just 0.0041% (roughly one per 24,000 documents). The model improves nearly 6x over Pangram 3, and is the first to differentiate fully AI-generated, AI-assisted, and mixed writing in a single pass, with robustness against humanizers and frontier models.
Related event: Pangram Raises $9M, Launches New AI Text and Image Detectors(12 posts)→
More from Research
- Goodfire's Nick Cammarata argues InceptionV1 polysemanticity wasn't caused by superposition — petrusenko_max · 2026-10-16
- UPenn lab shows 'critique and revise' prompting doubles Claude's research ideation output — arjunrajlab · 2026-10-16
- LLM ideation experiment: just ask for non-obvious ideas — and Opus beats smaller models by a wide margin — arjunrajlab · 2026-10-16
- What prompt gets the most interesting ideas out of LLMs? A researcher's ongoing experiment — arjunrajlab · 2026-10-16
- SimpleMark: multi-bit LLM watermark bounds per-token distortion with f-divergence budget — rishabh16_ · 2026-10-16
- Stanford paper: generative representation learning cuts tokenizer training compute by 7-8x — HirokatuKataoka · 2026-10-16