GLM-5.1-Thinking Distilled 5K: 5,000 Reasoning Traces for Model Training

GLM-5.1-Thinking Distilled Dataset contains 5,000 unique examples of questions paired with detailed multi-step reasoning traces and final responses. Authored by gss1147 and hosted on Hugging Face, it is designed to teach language models to think step by step. The dataset was last updated on 2026-05-25.

Use Cases

• Fine-tuning language models for step-by-step reasoning based on the provided reasoning traces.
• Distilling chain-of-thought capabilities into smaller models based on the high-quality examples.
• Benchmarking model performance on multi-step reasoning tasks based on the question-answer pairs.
• Training models to generate intermediate reasoning steps before producing a final answer based on the dataset's structure.

Strengths

• Contains 5,000 high-quality examples.
• Each example includes a question, a detailed multi-step reasoning trace, and a concise final response.
• Specifically designed for distilling chain-of-thought capabilities into language models.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Freshness

Last updated 2026-05-25 03:16:17.