Current research in code generation is increasingly focused on enhancing the capabilities of large language models (LLMs) through innovative training methodologies and frameworks that address existing limitations. Recent work emphasizes reinforcement learning techniques that allow models to self-reflect and self-correct during inference, significantly improving performance on complex coding tasks without relying on external feedback. Additionally, advancements in training stability and output diversity are being achieved through novel optimization strategies and the introduction of challenging datasets tailored for specific programming contexts. The integration of knowledge graphs is also gaining traction, facilitating better reasoning about evolving APIs and improving migration accuracy. Furthermore, the exploration of adversarial co-evolution frameworks is pushing the boundaries of test generation, ensuring that models not only produce functional code but also robust tests. Collectively, these efforts aim to create more efficient, adaptable, and reliable code generation systems, addressing commercial needs in software development and maintenance.
While Large Language Models (LLMs) have revolutionized code generation, standard "System 1" approaches, generating solutions in a single forward pass, often hit a performance ceiling when faced with c...
Modern code generation models exhibit longer outputs, accelerated capability growth, and changed training dynamics, rendering traditional training methodologies, algorithms, and datasets ineffective f...
Code evolution is inevitable in modern software development. Changes to third-party APIs frequently break existing code and complicate maintenance, posing practical challenges for developers. While la...
Vision-Language Models (VLMs) have demonstrated impressive capabilities in code generation across various domains. However, their ability to replicate complex, multi-panel visualizations from real-wor...
This work investigates the performance of Large Language Models (LLMs) in generating ABAP code. Despite successful applications of generative AI in many programming languages, there are hardly any sys...
Recent advances in large language models (LLMs) have enabled the automation of an increasing number of programming tasks, including code generation for scientific and engineering domains. In rapidly e...
Reinforcement learning for code generation relies on verifiable rewards from unit test pass rates. Yet high-quality test suites are scarce, existing datasets offer limited coverage, and static rewards...
Large language models (LLMs) have demonstrated strong capabilities in generating executable code from natural language descriptions. However, general-purpose models often struggle in specialized progr...
Training next-generation code generation models requires high-quality datasets, yet existing datasets face difficulty imbalance, format inconsistency, and data quality problems. We address these chall...
Critic-free reinforcement learning with verifiable rewards (RLVR) improves code generation by optimizing unit-test pass rates, but GRPO-style updates suffer from coarse credit assignment: a single out...