— Long-context post-training 🧶 — - a lewtun Collection

lewtun 's Collections

— Awesome RL datasets 📈 —

— Long-context post-training 🧶 —

H4

Mistral 7B + UltraChat + Arithmo checkpoints

— Long-context post-training 🧶 —

updated Sep 14

Resources for post-training LLMs with long-context samples

zai-org/LongAlign-10k

Viewer • Updated Feb 22, 2024 • 9.89k • 575 • 81
HuggingFaceTB/smoltalk2

Viewer • Updated Oct 31 • 8.61M • 7.21k • 132

Note Contains an English subset of LongAlign-10k, but with completions generated by Qwen3-32B: https://huggingface.co/datasets/HuggingFaceTB/smoltalk2/viewer/SFT?views%5B%5D=sft_longalign_64k_qwen3_32b_yarn_131k_think
zai-org/LongReward-10k

Viewer • Updated Oct 29, 2024 • 30k • 321 • 6
Tongyi-Zhiwen/DocQA-RL-1.6K

Viewer • Updated May 23 • 3.6k • 366 • 39
caskcsg/LongMagpie_64k_dataset

Preview • Updated Aug 2 • 484 • 3