Qwen3.5 397B Code Distilled 6K: Python Problems with Reasoning and Verified Solutions

A synthetic dataset of 6,000 Python programming problems distilled from the Qwen3.5-397B-A17B model. Each example includes a problem statement, a chain-of-thought reasoning trace, and an execution-verified correct solution. The dataset was created by author Madras1 and last updated on 2026-04-07.

Use Cases

• Training code generation models based on chain-of-thought reasoning traces mentioned in the description.
• Benchmarking model performance on execution-verified Python solutions as described.
• Studying synthetic data generation techniques for programming tasks based on the seed-generation from real code corpora.
• Analyzing reasoning patterns in code synthesis based on the included full reasoning traces.

Strengths

• Includes 6,000 examples, as indicated by the title.
• All solutions are execution-verified, having passed automated test cases.
• Data underwent exact and MinHash LSH near-deduplication.
• Problems are seeded from real code corpora.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count beyond the title's 6K is unknown, which may limit suitability assessment.
• Data may reflect bias inherent to the source model and generation process.

Provenance

Source

huggingface

Collection Method

Synthetic distillation from the Qwen3.5-397B-A17B model.

Time Range

Creation date not specified; last updated 2026-04-07.

Freshness

Last updated 2026-04-07 21:20:47; freshness should be verified.

Geography

Spatial coverage is not specified.