OpenSeek [Github|Blog]
Recent reseach has demonstrated that the reasoning ability of LLMs originates from the pre-training stage, activated by RL training. Massive raw corpus containing complex human reasoning process, but lack of generalized and effective synthesis method to extract these reasoning process.
🔥🔥🔥[2025/02/25] We publish some math, code, and general knowledge domain reasoning data synthesized from the current pipeline.… See the full description on the dataset page:
https://huggingface.co/datasets/BAAI/OpenSeek-Synthetic-Reasoning-Data-Examples.
