Skip to content

Dataset · r0b0tlab

Multi-Teacher Distillation for SFT

r0b0tlab/qwen3.8-max-glm5.2-kimi-k3-distillation

57,937 filtered trajectories from three models for SFT in math, code, reasoning, and multilingual tasks.

What's inside

The corpus contains 57,937 distillation trajectories from three teacher models: Qwen3.8-Max-Preview (48,283), GLM-5.2 (5,307), and Kimi Code K3 (4,347). The data underwent quality filtering and deduplication, covering mathematics, code, reasoning, instruction following, tool-use, science, long context, multilingual tasks, and creative dialogue. Languages include en, zh, es, fr, de, ja; modalities are tabular and text.

What you can do with it

  • SFT for multilingual LLM
  • LoRA fine-tuning for code and mathematics
  • Multi-teacher distillation comparison
Tasks
Text generation
Languages
English, Chinese, Spanish, French, German, Japanese
Size
10M–100M rows
Downloads this month
6.4K
Likes
279

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("r0b0tlab/qwen3.8-max-glm5.2-kimi-k3-distillation")