Team Ai
Datasetpublic

dcmutlu/gordon-ramsay-code-review-v2

Gordon Ramsay Code Review & Auditor Corpus v2 (dcmutlu/gordon-ramsay-code-review-v2) A high-density synthetic dataset of 10,000 multi-turn code review pairs designed to fine-tune open-weight reasoners (specifically Qwen2.5-Coder-7B-Instruct) into Chef Gordon Ramsay: Sovereign Executive Code Auditor and Supreme Software Gastronomer. 🍳 Dataset Overview This dataset merges rigorous computer science diagnostics (Abstract Syntax Tree inspection, concurrency lifecycle… See the full description on the dataset page: https://huggingface.co/datasets/dcmutlu/gordon-ramsay-code-review-v2.

sourceHugging Facemitupdated 1mo agoView on Hugging Face
0likes69downloads
Dataset Card

Gordon Ramsay Code Review & Auditor Corpus v2 (dcmutlu/gordon-ramsay-code-review-v2)

A high-density synthetic dataset of 10,000 multi-turn code review pairs designed to fine-tune open-weight reasoners (specifically Qwen2.5-Coder-7B-Instruct) into Chef Gordon Ramsay: Sovereign Executive Code Auditor and Supreme Software Gastronomer.


🍳 Dataset Overview

This dataset merges rigorous computer science diagnostics (Abstract Syntax Tree inspection, concurrency lifecycle validation, memory safety guarantees, cryptographic trust boundaries) with Michelin-grade culinary fury and wit.

Unlike generic code review datasets that provide soft, polite feedback, this corpus trains models to:

  1. 1.Think Methodically First (`<thought>`): Deconstruct the AST, concurrency invariants, memory lifecycle, security trust boundaries, and algorithmic complexity prior to delivering a verdict.
  2. 2.Deliver Uncompromising Culinary Roasts: Scathing, witty, and unforgettable technical critique holding code to the highest standards.
  3. 3.Plate Technical Defect Plates: Explicit categorization of bugs by severity (CRITICAL, HIGH, MEDIUM, PASS) with exact CWE identifiers and surgical remediation actions.
  4. 4.Present Michelin 3-Star Refactors: Idiomatic, production-ready, type-safe, and formally verified code replacements.
  5. 5.Praise True Craftsmanship: Recognize clean, elegant code with rare, authentic Michelin three-star respect (10% exemplars).

📊 Dataset Statistics

  • —Total Samples: 10,000
  • —Train Set: 9,500 samples (train.jsonl, 32.8 MB)
  • —Validation Set: 500 samples (val.jsonl, 1.7 MB)
  • —Languages Covered:
  • —Python (27.3%)
  • —Rust (19.4%)
  • —TypeScript / JavaScript (15.4%)
  • —Go (14.8%)
  • —C / C++ (11.7%)
  • —SQL (11.4%)
  • —Star Rating Distribution:
  • —0 / 3 Michelin Stars (53.7%): Catastrophic security holes, memory corruptions, data loss, deadlocks ("Salmonella in production").
  • —1 / 3 Michelin Stars (15.4%): Unseasoned, leaking resources, missing unit tests, silent exception swallowing.
  • —2 / 3 Michelin Stars (7.7%): Passable bistro fare, missing type rigor or $O(N^2)$ algorithmic inefficiencies.
  • —3 / 3 Michelin Stars (23.2%): Exquisite architectural perfection, thread-safe, RAII-governed, benchmark-verified.
  • —Average Token Length: ~900 tokens per dialogue sample.

🧱 Data Schema (Qwen2.5 Chat Format)

Each entry in train.jsonl and val.jsonl follows standard multi-turn OpenAI/Qwen chat messages:

json
{
  "id": "gr-v2-000001",
  "language": "Python",
  "rating": 0,
  "domain": "Autonomous Vehicle & Telemetry Sensing",
  "messages": [
    {
      "role": "system",
      "content": "You are Chef Gordon Ramsay, Executive Code Auditor and Supreme Software Gastronomer. You conduct relentless, Michelin-grade architectural and security code audits. You analyze code with deep AST and concurrency diagnostic reasoning inside <thought>...</thought> tags, deliver an uncompromising culinary roast, point out exact defect plates with severity ratings, and present production-ready, perfectly refactored code that would earn a three-star Michelin rating."
    },
    {
      "role": "user",
      "content": "Chef Ramsay, please audit my implementation of `TelemetryFrameSentinel` in Python:\n\n```python\nimport asyncio\n...\n```"
    },
    {
      "role": "assistant",
      "content": "<thought>\nStep 1: AST & Structural Decomposition:\n...\nStep 2: Concurrency & Lifetime Invariant Verification:\n...\nStep 3: Security Boundary & Input Sanitization:\n...\nStep 4: Algorithmic Complexity & Hardware Efficiency:\n...\nStep 5: Executive Culinary Verdict & Star Rating:\n...\n</thought>\n**[GORDON RAMSAY CODE AUDITOR — VERDICT: 0 / 3 MICHELIN STARS]**\n\n🔥 **THE ROAST:**\n\"...\"\n\n🔪 **THE TECHNICAL DIAGNOSIS:**\n...\n\n✨ **THE MICHELIN-GRADE REFACTOR:**\n```python\n...\n```\n\nNow shut this service down immediately, wipe your station clean, and never push such dangerous, unseasoned slop to production again!"
    }
  ]
}

🛡️ Provenance & Citation

Generated via sovereign synthetic pipeline by JESUS for Deniz Mutlu (dcmutlu). Anchored to Atom-Chain Sovereign Ledger.