--- pretty_name: J6-CFI-HQ-20K language: - en task_categories: - text-generation - question-answering - text-classification - other task_ids: - language-modeling - text2text-generation - open-domain-qa - explanation-generation - text-scoring license: cc-by-4.0 size_categories: - 10K Built by **[JumpLander](https://www.jumplander.org)** for experiments in coding assistants, software engineering datasets, and agentic developer intelligence.

J6-CFI Dataset Overview - JumpLander Code Feedback Instructions

--- ## Overview **J6-CFI-HQ-20K** stands for **JumpLander High-Quality 20K Code Feedback Instructions**. This dataset is an English-only, synthetic, high-quality structured code-instruction dataset designed for supervised fine-tuning experiments, coding assistant prototyping, debugging behavior, code review, refactoring, test generation, algorithmic reasoning, backend/API task design, SQL and data engineering tasks, and coding-agent behavior research. Unlike minimal `instruction -> answer` datasets, J6-CFI-HQ-20K includes structured metadata for each record: developer intent, constraints, reference solution, explanation, test cases, quality signals, evaluation checks, and curation metadata. This release is intentionally smaller than large synthetic corpora. The focus is cleaner structure, stronger constraints, more realistic developer tasks, and better record-level metadata. --- ## Dataset Information | Property | Value | | --- | --- | | Dataset Name | J6-CFI-HQ-20K | | Full Name | JumpLander High-Quality 20K Code Feedback Instructions | | Total Samples | 20,000 | | Language | English | | Format | JSONL | | File | `j6_cfi_hq_20k.jsonl` | | License | CC BY 4.0 | | Data Type | High-Quality Synthetic / Structured | | Version | 1.0.0-hq-synthetic | | Created By | JumpLander | --- ## Task Distribution | Task Type | Samples | | --- | ---: | | `debugging` | 4,000 | | `code_generation` | 3,500 | | `code_review` | 2,500 | | `refactoring` | 2,500 | | `test_generation` | 2,000 | | `algorithmic_reasoning` | 1,800 | | `code_explanation` | 1,500 | | `sql_data_tasks` | 1,000 | | `api_backend` | 800 | | `coding_agent_instructions` | 400 | --- ## Programming Language Distribution | Programming Language | Samples | | --- | ---: | | `python` | 5,500 | | `javascript` | 3,000 | | `typescript` | 2,000 | | `java` | 1,700 | | `cpp` | 1,500 | | `sql` | 1,400 | | `go` | 1,000 | | `php` | 900 | | `rust` | 800 | | `bash` | 600 | | `html_css` | 600 | | `csharp` | 500 | | `ruby` | 500 | --- ## Quality Positioning J6-CFI-HQ-20K is **not** a human-verified Gold benchmark. It is a high-quality synthetic dataset candidate designed to be cleaner and more focused than large-scale raw synthetic corpora. Generation and curation properties: - 20,000 valid JSONL records - exact duplicate full records: 0 - exact duplicate instructions: 0 - exact duplicate code contexts: 0 - each record includes explicit constraints - each record includes test cases or validation scenarios - each record includes quality signals and evaluation checks - each record includes curation metadata - designed for SFT and coding-assistant prototyping For stricter evaluation, a future `J6-CFI-Gold-1K` subset should be manually reviewed. --- ## Dataset Schema Each JSONL line contains one record with the following fields: | Field | Type | Description | | --- | --- | --- | | `id` | string | Unique sample identifier | | `dataset` | string | Short dataset name | | `full_name` | string | Full dataset name | | `version` | string | Dataset version | | `language` | string | Natural language of the instruction | | `programming_language` | string | Target programming language | | `task_type` | string | Main task category | | `difficulty` | string | Difficulty level | | `title` | string | Short task title | | `domain` | string | Software engineering domain | | `instruction` | string | User-facing coding instruction | | `code_context` | string | Code or project context | | `intent` | string | Normalized developer intent | | `constraints` | array | Constraints the assistant should preserve | | `reference_solution` | string | Reference answer or solution outline | | `explanation` | string | Explanation of the expected behavior | | `test_cases` | array | Test or validation cases | | `quality_signals` | object | Metadata for reasoning, tests, edge cases and risk | | `evaluation` | object | Checks for scoring model responses | | `curation` | object | Curation and review metadata | | `tags` | array | Search and filtering tags | --- ## Loading the Dataset ```bash pip install datasets ``` ```python from datasets import load_dataset dataset = load_dataset("jumplander/J6-CFI-HQ-20K") train = dataset["train"] print(train[0]["instruction"]) print(train[0]["reference_solution"]) ``` ### Load Locally ```python from datasets import load_dataset dataset = load_dataset( "json", data_files={{"train": "j6_cfi_hq_20k.jsonl"}} ) ``` --- ## Formatting for SFT ```python def format_example(example): return {{ "text": f"""### Instruction: {{example['instruction']}} ### Code Context: {{example['code_context']}} ### Developer Intent: {{example['intent']}} ### Constraints: {{example['constraints']}} ### Reference Solution: {{example['reference_solution']}} ### Explanation: {{example['explanation']}} ### Verification: {{example['test_cases']}}""" }} formatted = dataset["train"].map(format_example) ``` --- ## Suggested Use Cases | Use Case | Description | | --- | --- | | SFT Training | Fine-tune coding assistants on structured code instructions | | Debugging Research | Study bug-fixing and root-cause explanation behavior | | Code Review Training | Train models to produce prioritized review feedback | | Refactoring Experiments | Improve behavior-preserving code transformation | | Test Generation | Teach assistants to generate edge-case-aware tests | | Coding-Agent Research | Evaluate minimal-edit and safe-patch behavior | | Prompt Evaluation | Compare coding prompts against structured reference behavior | | Gold Subset Preparation | Select high-value records for manual review | --- ## Important Quality Notes J6-CFI-HQ-20K is a **high-quality synthetic structured dataset**, not a fully human-verified benchmark. It is suitable for SFT experiments, coding assistant prototyping, structured instruction tuning, behavior-aware code generation research, and preparing a manually reviewed Gold subset. It should not be described as human-verified, production-certified, a complete correctness benchmark, or a replacement for execution-based evaluation. Recommended next steps: 1. create `J6-CFI-Gold-1K`, 2. manually review technical correctness, 3. execute a subset of Python, JavaScript, SQL, and Bash examples, 4. score solution usefulness and difficulty, 5. publish an evaluation script. --- ## Dataset Philosophy A strong coding assistant should not only produce code. It should understand developer intent, preserve constraints, avoid unsafe rewrites, explain trade-offs, include tests, and produce actionable engineering output. J6-CFI-HQ-20K is designed around that principle. --- ## License This dataset is released under the **Creative Commons Attribution 4.0 International (CC BY 4.0)** license. Attribution is required when redistributing, adapting, or building upon this dataset. --- ## Citation ```bibtex @misc{{j6_cfi_hq_20k_2026, title={{J6-CFI-HQ-20K: JumpLander High-Quality 20K Code Feedback Instructions}}, author={{JumpLander Team}}, year={{2026}}, publisher={{Hugging Face}}, url={{https://huggingface.co/datasets/jumplander/J6-CFI-HQ-20K}} }} ``` --- ## Built by JumpLander J6-CFI-HQ-20K is built by **[JumpLander](https://www.jumplander.org)**, a research and engineering initiative focused on coding agents, developer intelligence, AI-assisted programming, and practical software engineering datasets. - Website: [jumplander.org](https://www.jumplander.org) - Persian Website: [jumplander.org/fa](https://www.jumplander.org/fa) - Documentation: [jumplander.org/fa/docs](https://www.jumplander.org/fa/docs) - About: [jumplander.org/fa/about](https://www.jumplander.org/fa/about) - Hugging Face: [huggingface.co/jumplander](https://huggingface.co/jumplander) ---