AReaL-boba-2-RL-Code: Reinforcement Learning Training Data for Coding Tasks

Training and testing data used for reinforcement learning fine-tuning of the AReaL-boba-2 model on coding tasks. The dataset includes an evaluation benchmark organized into subfolders for specific coding suites, including Codeforces, Code Contests, and LiveCodeBench (v5). It was created by inclusionAI and last updated on July 2, 2025.

Use Cases

• Fine-tuning reinforcement learning models for code generation based on the described training data.
• Benchmarking code generation models against the included Codeforces and Code Contests suites.
• Evaluating model performance on live coding challenges using the LiveCodeBench (v5) benchmark.

Strengths

• Includes data from multiple established coding benchmark suites (Codeforces, Code Contests, LiveCodeBench v5).
• Dataset structure is clearly separated into training and benchmark evaluation folders.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Collection Method

Likely collected from coding competition platforms and benchmark suites for model training.

Freshness

Last updated 2025-07-02 08:52:24; freshness should be verified.