trentmkelly/slop-detector

NOTICE

Old model. Please use https://huggingface.co/trentmkelly/slop-detector-mini-2 instead. Faster and more accurate, trained on more data.

Model Trained Using AutoTrain

• Problem type: Text Classification

Validation Metrics

loss: 0.03548985347151756

f1: 0.9950522264980759

precision: 0.9945054945054945

recall: 0.9955995599559956

auc: 0.9997361672360855

accuracy: 0.995049504950495

Purpose

I trained this on a bunch of top-level comments on reddit. Human class was the real responses to selfposts in various subs, and the LLM class was a response from one of several LLMs to the same post. I am tired of reading fucking GPT-slop comments on reddit.

Notes

Converted ONNX model is available for compatibility with transformers.js. Browser extension and mini version coming soon.