JackHsieh/

qwen3-4B-instruct-luna-distill-step2362.reason-only.k-8.statml-arxiv-qwen3

Reasoning about the next 8 Qwen3 tokens of stat.ML arXiv LaTeX, generated by JackHsieh/qwen3-4B-instruct-luna-distill-step2362 — Qwen3-4B-Instruct-2507 distilled (SFT, 1 epoch) on gpt-5.6-luna thoughts. Documents: JackHsieh/statML-arxiv-40M-20M (4096 Qwen3 tokens each).

split thoughts documents per doc cut positions stride longest truncated

train 612,864 9,728 63 96 … 4064 64 tokens (offset… See the full description on the dataset page: https://huggingface.co/datasets/JackHsieh/qwen3-4B-instruct-luna-distill-step2362.reason-only.k-8.statml-arxiv-qwen3.