SlopFinder is Here! Crowdsourcing AI Slop Classification

🧪 SlopFinder is here! We are building an open dataset to study what humans actually consider AI slop.

SlopFinder presents you with a random piece of AI-generated text and gives you one simple control: how slop is it? There are no complex forms or multi-select categories — just vote and move on to the next sample. Every single vote helps build the benchmark dataset. 🧩

How Does It Work?

The annotation pipeline is designed to be fast and lightweight:

1. Sample Aggregation: Text samples are pulled from existing open datasets.
2. Anonymous Evaluation: Samples are shown anonymously to prevent bias.
3. Pool Collection & Export: Collected votes enter an annotation pool. Once enough community votes are gathered, samples are exported directly to Hugging Face for public use.

This is an early MVP, so the dataset is currently small and the overall system is still evolving.

Get Involved & Explore the Dataset

You can start voting or explore the dataset using the resources below:

• Vote Here: SlopFinder Web App (simply refresh the page if you want to skip a prompt)
• Dataset on Hugging Face: bench-labs/slop-classification
• Creators: Developed by @benchlabs and @wop

Frequently Asked Questions