The field of scientific computing is moving towards autonomous code generation and automation, with a focus on developing reliable and efficient methods for generating scientific code from natural-language queries. Recent developments have highlighted the importance of embedding first-principles analysis and expert knowledge into code generation frameworks, allowing for more accurate and physically coherent solutions.

Noteworthy papers in this area include: Chain of Unit-Physics, which proposes a primitive-centric approach to code synthesis and achieves state-of-the-art results on a combustion task. CodeDistiller, which automatically generates code libraries for scientific coding agents and enables them to expand their capabilities without manual effort. ATHENA, which introduces an agentic framework for managing the end-to-end computational research lifecycle and achieves super-human performance on various scientific tasks. Enhancing Automated Paper Reproduction via Prompt-Free Collaborative Agents, which proposes a collaborative agent framework for automatically verifying and refining paper-to-code generation outputs. PaperDebugger, which presents an in-editor, multi-agent, and plugin-based academic writing assistant that brings LLM-driven reasoning directly into the writing environment.