AronDaron/

OctoBench-2.2k

Synthetic dataset for fine-tuning coding-focused LLMs. Generated with Dataset Generator — an open-source pipeline for building high-quality training data.

2,248 multi-turn conversations across 8 categories:

Category Examples Focus Model Gen Model Judge

Refactor & Code Review 1 72 Performance refactors, behavior preservation qwen/qwen3-coder openai/gpt-oss-120b

Refactor & Code Review 2 69 Performance… See the full description on the dataset page: https://huggingface.co/datasets/AronDaron/OctoBench-2.2k.