Hacker News new | ask | show | jobs
Show HN: Prompt injection detector beats ProtectAI by 19% accuracy, 8.9x smaller (huggingface.co)
3 points by Karan047 66 days ago
ProtectAI's deberta-v3 is the most downloaded prompt injection classifier on HuggingFace. We benchmarked directly against it on Qualifire's independent dataset (excluded from our training).

* 91.68% vs 72.28% accuracy

* 95.84% vs 65.33% precision — they block 1 in 3 legitimate users

* 83MB vs 739MB ONNX size

* 101ms vs 646ms on CPU

1 comments

Prompt injection will never be solved. Instruction following is the product.
Nothing in cybersecurity is truly solved—only made harder, Prompt injection is no different.