README.md

---
language:
  - en
license: apache-2.0
size_categories:
  - 10K<n<100K
task_categories:
  - text-generation
  - question-answering
tags:
  - synthetic
  - reasoning
  - sft
  - science
  - glm
---

# GLM-5.2 Science Synth Reasoning

Synthetic reasoning traces for science questions from the [GLM-5.2 science dataset](https://huggingface.co/datasets/ianncity/GLM-5.2-Science). Each record contains a complex scientific question with SYNTH-style reasoning and a generated answer.

## Dataset Summary

- **33,014 records** (605 dupes + 726 incomplete/truncated removed from 34,345 source)
- **33,014 reasoning turns** (99.9% format compliance)
- Average **3,094 chars** per reasoning trace

### Models Used

| Model | Records |
|---|---|
| deepseek-v4-flash | 33,341 |
| thinkingmachines/inkling | 673 |

## Data Format

```python
{
    "messages": [
        {"role": "user", "content": "How does the coupling between mantle convection patterns..."},
        {"role": "assistant", "content": "The coupling runs through...", "reasoning_content": "● Question: ... → ..."}
    ],
    "source": "disk:.../glm-5.2-science",
    "source_index": 0,
    "target": "both",
    "protocol": "chat_completions",
    "model": "deepseek-v4-flash",
    "num_conversation_turns": 2,
    "num_generated_turns": 1,
    "generated_at": "2026-07-26T..."
}
```

## Validation

| Check | Result |
|---|---|
| Format compliance | 99.9% (32,994/33,014 turns) |
| Duplicates | 0 (after dedup) |
| Null/empty reasoning | 0 |
| Incomplete/truncated | 0 (filtered out) |