pwb-anon-2026/

pro-worker-ai-benchmark

An evaluation framework that measures whether large language models augment or substitute for human cognition. This dataset accompanies the NeurIPS 2026 Evaluations & Datasets Track submission "The Pro-Worker AI Benchmark: Measuring Whether Large Language Models Augment or Replace Human Intelligence".

	What's in this dataset




	

Folder Contents

prompts/layer1_behavioral/ 200 single-turn behavioral probes across 10 dimensions (10 YAML… See the full description on the dataset page: https://huggingface.co/datasets/pwb-anon-2026/pro-worker-ai-benchmark.