Upload README.md with huggingface_hub

README.md

@@ -1,198 +1,178 @@
-# Scientific Coding SWE Dataset (JSONL Format)
+---
+dataset_info:
+  features:
+  - name: question_id
+    dtype: string
+  - name: question_type
+    dtype: string
+  - name: description
+    dtype: string
+  - name: content
+    dtype: string
+  - name: environment
+    dtype: string
+  - name: answer
+    dtype: string
+  - name: test
+    dtype: string
+  - name: scoring_config
+    dtype: string
+  splits:
+  - name: msb_type
+    num_bytes: 15728640
+    num_examples: 244
+  - name: et_type
+    num_bytes: 709558272
+    num_examples: 1085
+  - name: scicobench_all
+    num_bytes: 666894336
+    num_examples: 1329
+  download_size: 1392181248
+  dataset_size: 1392181248
+configs:
+- config_name: default
+  data_files:
+  - split: msb_type
+    path: data/msb_type.jsonl
+  - split: et_type
+    path: data/et_type.jsonl
+  - split: scicobench_all
+    path: data/scicobench_all.jsonl
+  default: true
+---
+
+# Scientific Coding Benchmark Dataset
 
 ## Dataset Description
 
-This dataset contains 244 samples extracted from the [xinshuo/Scientific_Coding_SWE_dataset](https://huggingface.co/datasets/xinshuo/Scientific_Coding_SWE_dataset), specifically the `json_dump` field from each sample. The data is provided in JSONL (JSON Lines) format for easy streaming and processing.
+This dataset contains scientific coding questions and benchmarks from multiple sources, organized into three splits:
+
+- **msb_type** (244 samples): Multi-source Scientific Benchmark questions focusing on software engineering tasks
+- **et_type** (1,085 samples): Einstein Toolkit related coding questions with extensive documentation
+- **scicobench_all** (1,329 samples): Combined and unified format of all scientific coding questions
 
 ### Dataset Summary
 
-The Scientific Coding SWE (Software Engineering) dataset focuses on real-world code issues and fixes from prominent scientific computing projects. Each sample represents a complete software engineering task including:
+The Scientific Coding Benchmark dataset is designed for evaluating AI models on real-world scientific software development tasks. Each sample includes:
 
-- Issue descriptions
-- Code patches (fixes)
+- Question descriptions and context
+- Code environment specifications
 - Test cases
-- Metadata about the changes
-- Execution results
+- Expected answers/solutions
+- Scoring configurations
 
 ### Supported Tasks
 
-- **Software Engineering Task Resolution**: Understanding and resolving code issues in scientific computing contexts
-- **Code Generation**: Generating fixes for missing or incorrect code
-- **Test-Driven Development**: Working with test cases and validation
-- **Documentation Generation**: Understanding code documentation needs
+- **Scientific Code Generation**: Generate code solutions for scientific computing problems
+- **Software Engineering**: Fix bugs and implement features in scientific software
+- **Documentation Understanding**: Work with complex scientific software documentation
+- **Test-Driven Development**: Write code that passes specific test cases
 
-### Languages
+## Dataset Structure
 
-The dataset covers two primary programming languages:
-- **C++**: 159 samples (65.2%)
-- **Python**: 85 samples (34.8%)
+### Data Splits
 
-## Dataset Structure
+The dataset has three splits:
 
-### Data Format
+| Split | Samples | Size | Description |
+|-------|---------|------|-------------|
+| msb_type | 244 | ~15 MB | Software engineering tasks from scientific projects |
+| et_type | 1,085 | ~677 MB | Einstein Toolkit code generation tasks |
+| scicobench_all | 1,329 | ~636 MB | Unified format combining all question types |
 
-The dataset is provided as a single JSONL file (`dataset_json_dumps.jsonl`), where each line is a complete JSON object representing one sample.
+**Default Split**: `msb_type`
 
 ### Data Fields
 
-Each sample contains the following fields:
-
-- `org` (string): Organization name (e.g., "einsteintoolkit", "Qiskit", "openmm")
-- `repo` (string): Repository name
-- `number` (int): Issue/PR number
-- `state` (string): State of the issue (e.g., "open", "closed")
-- `title` (string): Title of the issue/PR
-- `body` (string): Description of the issue
-- `base` (dict): Base branch information with keys:
-  - `label`: Branch label
-  - `ref`: Branch reference
-  - `sha`: Commit SHA
-- `resolved_issues` (list): List of related issues with:
-  - `number`: Issue number
-  - `title`: Issue title
-  - `body`: Issue description
-- `fix_patch` (string): Git diff/patch for the fix
-- `test_patch` (string): Git diff/patch for tests
-- `doc_patch` (string): Git diff/patch for documentation
-- `tag` (string): Additional tags
-- `number_interval` (string): Number interval information
-- `lang` (string): Programming language (cpp, python)
-- `language` (string): Same as lang
-- `instance_id` (string): Unique instance identifier
-- `fixed_tests` (dict): Test results for fixed code
-- `p2p_tests` (dict): Patch-to-pass tests
-- `f2p_tests` (dict): Fail-to-pass tests
-- `s2p_tests` (dict): Skip-to-pass tests
-- `n2p_tests` (dict): None-to-pass tests
-- `run_result` (dict): Execution results with counts of passed/failed/skipped tests
-- `test_patch_result` (dict): Test patch execution results
-- `fix_patch_result` (dict): Fix patch execution results
-- `hints` (string): Hints for solving the issue
-- `count_new_files` (int): Number of new files added
-- `count_new_entities` (int): Number of new code entities
-- `workdir` (string): Working directory path
-- `fix_meta_info` (dict): Metadata about the fix
-- `test_meta_info` (dict): Metadata about tests
-- `doc_meta_info` (dict): Metadata about documentation
-- `meta_info` (dict): General metadata
-- `pre_commands` (list): Commands to run before testing
-
-### Data Instances
-
-An example instance (truncated for brevity):
-
-```json
-{
-  "org": "einsteintoolkit",
-  "repo": "Cactus",
-  "number": 1,
-  "state": "open",
-  "title": "Add missing LinearExtrapBnd.c in CactusExamples/SampleBoundary",
-  "body": "This PR adds the missing source file LinearExtrapBnd.c to complete the CactusExamples/SampleBoundary thorn implementation.",
-  "language": "cpp",
-  "instance_id": "einsteintoolkit__cactus_1",
-  "fix_patch": "diff --git a/arrangements/CactusExamples/SampleBoundary/src/LinearExtrapBnd.c ...",
-  "run_result": {
-    "passed_count": 1,
-    "failed_count": 0,
-    "skipped_count": 0
-  }
-}
-```
+Each sample in the unified format contains:
 
-### Data Splits
+- `question_id` (string): Unique identifier for the question
+- `question_type` (string): Type/category of the question
+- `description` (string): High-level description of the task
+- `content` (string): Detailed question content and requirements
+- `environment` (string): Environment setup and configuration details
+- `answer` (string): Expected answer or solution
+- `test` (string): Test cases and validation logic
+- `scoring_config` (string): Configuration for scoring the solution
+
+### Data Loading
+
+```python
+from datasets import load_dataset
+
+# Load default split (msb_type)
+dataset = load_dataset("xinshuo/test_jsonl")
 
-This dataset contains a single split with all 244 samples.
+# Load specific split
+msb_data = load_dataset("xinshuo/test_jsonl", split="msb_type")
+et_data = load_dataset("xinshuo/test_jsonl", split="et_type")
+all_data = load_dataset("xinshuo/test_jsonl", split="scicobench_all")
 
-## Source Projects
+# Access a sample
+print(msb_data[0])
+```
 
-The dataset includes samples from the following scientific computing projects:
+## Source Data
 
-| Organization | Count | Description |
-|-------------|-------|-------------|
-| openmm | 54 | High-performance molecular dynamics simulation |
-| pyscf | 47 | Python-based Simulations of Chemistry Framework |
-| rdkit | 46 | Cheminformatics and machine learning software |
-| einsteintoolkit | 40 | Computational infrastructure for numerical relativity and astrophysics |
-| Qiskit | 38 | Open-source quantum computing framework |
-| AMReX-Codes | 19 | Block-structured adaptive mesh refinement framework |
+### MSB Type Questions (244 samples)
 
-## Dataset Creation
+Derived from the Scientific Coding SWE dataset, focusing on:
+- **Organizations**: einsteintoolkit, openmm, pyscf, rdkit, Qiskit, AMReX-Codes
+- **Languages**: C++ (65%), Python (35%)
+- **Tasks**: Bug fixes, feature implementation, code completion
 
-### Source Data
+### ET Type Questions (1,085 samples)
 
-This dataset is derived from the [xinshuo/Scientific_Coding_SWE_dataset](https://huggingface.co/datasets/xinshuo/Scientific_Coding_SWE_dataset), which curates software engineering tasks from real GitHub issues and pull requests in scientific computing projects.
+Einstein Toolkit code generation tasks including:
+- Thorn implementations
+- Interface definitions
+- Parameter configurations
+- Schedule definitions
+- Extensive documentation context
 
-### Data Collection Process
+### SciCoBench All (1,329 samples)
 
-The data was extracted by:
-1. Loading the original dataset using the HuggingFace `datasets` library
-2. Extracting only the `json_dump` field from each sample
-3. Converting to JSONL format for efficient streaming and processing
+Unified format combining both question types with standardized structure for comprehensive evaluation.
 
 ## Considerations for Using the Data
 
-### Social Impact of Dataset
+### Use Cases
 
-This dataset can be used to:
-- Train and evaluate AI models for scientific software engineering tasks
-- Improve automated code generation and bug fixing in scientific computing
-- Understand common patterns in scientific software development
+- Training code generation models for scientific computing
+- Benchmarking AI coding assistants on scientific software tasks
+- Research in automated scientific software development
+- Education in scientific programming practices
 
-### Discussion of Biases
+### Limitations
 
-The dataset may have the following biases:
-- Overrepresentation of certain programming languages (C++ and Python)
 - Focus on specific scientific domains (physics, chemistry, quantum computing)
-- Representation limited to open-source projects with public GitHub repositories
+- Primarily C++ and Python languages
+- May require domain-specific knowledge to evaluate correctly
 
 ## Additional Information
 
 ### Dataset Curators
 
-This JSONL version was created from the original xinshuo/Scientific_Coding_SWE_dataset.
+This dataset combines and unifies questions from:
+- xinshuo/Scientific_Coding_SWE_dataset
+- xinshuo/ET_1k
 
 ### Licensing Information
 
-Please refer to the original dataset for licensing information.
+Please refer to the original datasets for licensing information.
 
 ### Citation Information
 
-If you use this dataset, please cite the original source:
-
 ```bibtex
-@dataset{scientific_coding_swe_dataset,
-  title={Scientific Coding SWE Dataset},
+@dataset{scicobench,
+  title={Scientific Coding Benchmark Dataset},
   author={xinshuo},
   year={2024},
-  url={https://huggingface.co/datasets/xinshuo/Scientific_Coding_SWE_dataset}
+  url={https://huggingface.co/datasets/xinshuo/test_jsonl}
 }
 ```
 
 ### Contributions
 
-For questions or issues with this JSONL format, please refer to the dataset repository.
-
-## Usage Example
-
-```python
-import json
-
-# Read and process the dataset
-with open('dataset_json_dumps.jsonl', 'r') as f:
-    for line in f:
-        sample = json.loads(line)
-        print(f"Issue: {sample['title']}")
-        print(f"Language: {sample['language']}")
-        print(f"Organization: {sample['org']}")
-        print("---")
-```
-
-Or use with HuggingFace datasets:
-
-```python
-from datasets import load_dataset
-
-dataset = load_dataset("xinshuo/test_jsonl", data_files="dataset_json_dumps.jsonl")
-```
+For questions or issues, please refer to the dataset repository on HuggingFace.