s1K probe training data
收藏官方服务:
资源简介:
Data required to replicate the training of thought calibration probes. Files provided include:s1K per-step embeddings by DeepSeek-R1 distilled Qwen 2.5 32B and Llama 3.3 70B, and QwQ 32BLabels for supervised correctness, consistency, novelty, and leafnessThese files are used in notebook <b>2-probe.ipynb</b>.
提供机构:
figshare创建时间:
2025-06-05



