agentlans/

Skywork-Skywork-OR1-RL-Data

Math and coding datasets with their difficulty ratings for various models:

DeepSeek-R1-Distill-Qwen-1.5B DeepSeek-R1-Distill-Qwen-7B DeepSeek-R1-Distill-Qwen-32B

The prompt and the ground truth fields were extracted from the dataset. In addition:

The dataset-level difficulty scores range from 0 to 16 inclusive. These difficulty values were combined into a single general difficulty score using principal component analysis (PCA). PCA… See the full description on the dataset page: https://huggingface.co/datasets/agentlans/Skywork-Skywork-OR1-RL-Data.