Madras1/qwen3.5-397b-code-distilled-6k

---
language:
- en
license: apache-2.0
size_categories:
- 1K<n<10K
task_categories:
- text-generation
tags:
- code
- code-generation
- distillation
- synthetic
- reasoning
- chain-of-thought
- python
- qwen
pretty_name: "Qwen 3.5-397B Code Distillation"
dataset_info:
features:
- name: id
dtype: int64
- name: problem
dtype: string
- name: function_name
dtype: string
- name: reasoning
dtype: string
- name: code
dtype: string
- name: model
dtype: string
splits:
- name: train
num_examples: 5724
---

# Qwen 3.5-397B Code Distillation Dataset

Synthetic code dataset distilled from **Qwen3.5-397B-A17B**. Each example has a Python problem, chain-of-thought reasoning, and an **execution-verified** correct solution.

## Features

- **Execution-verified**: All solutions passed automated test cases
- **Chain-of-thought**: Full reasoning traces included
- **Seed-generated**: Problems seeded from real code corpora
- **Deduplicated**: Exact + MinHash LSH near-dedup
- **Pure Python**: Standard library only

## Stats

| Metric | Value |
|---|---|
| Examples | 5,724 |
| With reasoning | 5,724 (100.0%) |
| Avg code | 1058 chars |
| Avg reasoning | 11771 chars |
| Model | `Qwen/Qwen3.5-397B-A17B` |

## Schema

| Field | Type | Description |
|---|---|---|
| `problem` | string | Problem description |
| `function_name` | string | Target function |
| `reasoning` | string | Chain-of-thought |
| `code` | string | Verified solution |
| `model` | string | Teacher model |

## Usage

```python
from datasets import load_dataset
ds = load_dataset("Madras1/qwen3.5-397b-code-distilled-6k")
print(ds["train"][0]["problem"])
print(ds["train"][0]["code"])
```

## License

Apache 2.0