README.md

metadata

license: apache-2.0
language:
  - en
  - zh
size_categories:
  - n>1M
task_categories:
  - text-generation
  - question-answering
tags:
  - reasoning
  - chain-of-thought
  - instruction-tuning
  - sft
  - distillation
  - glm
  - glm-5.1
configs:
  - config_name: main
    data_files:
      - split: train
        path: main.jsonl
  - config_name: PHD-Science
    data_files:
      - split: train
        path: PHD-Science.jsonl
  - config_name: Multilingual-STEM
    data_files:
      - split: train
        path: Multilingual-STEM.jsonl
  - config_name: Math
    data_files:
      - split: train
        path: Math.jsonl

GLM-5.1-1000000x

1,003,589 reasoning traces distilled by GLM-5.1, using questions from KIMI-K2.5-1000000x.

Each entry contains a full chain-of-thought reasoning trace followed by the final answer, generated by GLM-5.1.

Complete! All 1,003,589 prompts distilled successfully.

████████████████████████████████ 100%

---

Data Distribution

Subset	Count	Proportion	Est. Tokens	Domain
main	598,366	59.6%	~3.04B	General reasoning & instruction-following
Math	208,426	20.8%	~1.30B	Mathematics
PHD-Science	103,759	10.3%	~0.56B	Graduate-level Physics, Chemistry, Biology
Multilingual-STEM	93,038	9.3%	~0.46B	STEM in Chinese, English & other languages
Total	1,003,589	100%	~5.36B

Dataset Statistics

Metric	Value
Total Records	1,003,589
Total Estimated Tokens	~5.36B
Avg. Tokens per Record	~5,338

How to Use

from datasets import load_dataset

# Load a specific subset
main = load_dataset("Kassadin88/GLM-5.1-1000000x", "main")
science = load_dataset("Kassadin88/GLM-5.1-1000000x", "PHD-Science")
stem = load_dataset("Kassadin88/GLM-5.1-1000000x", "Multilingual-STEM")
math = load_dataset("Kassadin88/GLM-5.1-1000000x", "Math")

Each record is a chat-formatted conversation with a chain-of-thought reasoning trace:

{
  "messages": [
    {"role": "user", "content": "Beaches and deserts collect large deposits of what? ..."},
    {"role": "assistant", "content": "<think>\n1. Analyze the question...\n2. Reasoning step...\n</think>\nSand"}
  ],
  "_id": "main_00000007"
}

• messages: user question + assistant response with CoT trace and final answer
• _id: {category}_{serial} (e.g. Math_00038225, PHD-Science_00010138)

License

Apache 2.0

Citation

@misc{glm51-1000000x,
  title={GLM-5.1-1000000x: One Million Reasoning Traces Distilled from GLM-5.1},
  author={Kassadin88},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/Kassadin88/GLM-5.1-1000000x}
}

Acknowledgments

• Prompt source: KIMI-K2.5-1000000x
• Teacher model: GLM-5.1