OpenCodeReasoning-II: A Simple Test Time Scaling Approach via Self-Critique
This work addresses the dataset bottleneck for researchers and developers in code generation and critique, though it is incremental as it builds on existing methods with a larger dataset and fine-tuning approach.
The authors tackled the need for large-scale, high-quality datasets in reasoning-based LLMs for code generation and critique by introducing OpenCodeReasoning-II, a dataset of 2.5M triples (35K unique questions), nearly twice the size of prior datasets, and used a two-stage fine-tuning strategy to achieve performance that exceeds or equals the best prior open-weight distilled models in code generation, with significant improvements in competitive coding.
Recent advancements in reasoning-based Large Language Models (LLMs), particularly their potential through test-time scaling, have created significant opportunities for distillation in code generation and critique. However, progress in both areas fundamentally depends on large-scale, high-quality datasets. In this work, we introduce OpenCodeReasoning-II, a dataset consists of 2.5M question-solution-critique triples (approx. 35K unique programming questions), making it nearly twice the size of the previous largest publicly available code reasoning dataset. In this work, we employ a two-stage supervised fine-tuning strategy. The first stage focuses on fine-tuning for code generation, while the second stage involves the joint training of models for both code generation and critique. Our resulting finetuned Qwen2.5-Instruct models achieve performance in code generation that either exceeds or equals the best prior open-weight distilled models. Notably, the integration of our code generation and critique models leads to significant improvements in competitive coding performance. Furthermore, we present an extension of the LiveCodeBench benchmark to specifically support the C++ programming language, thereby facilitating more comprehensive LLM evaluation using this benchmark.