Systematic LLM Translation of Legacy Scientific Code to Differentiable Frameworks: Application to a Land Surface Model
arXiv:2606.07681v1 Announce Type: cross Abstract: Differentiable programming offers transformative capabilities for scientific modeling, enabling gradient-based parameter estimation, sensitivity analysis, and data assimilation. Yet, migrating legacy codebases into differentiable frameworks remains a challenge. We present a five-phase LLM-based agentic pipeline that translates legacy Fortran into JAX: static dependency analysis determines module translation order from the full call graph; iterative compile-repair loops correct errors autonomously; and a Fortran reference oracle enforces numeric
The proliferation of mature large language models and the increasing demand for differentiable programming across scientific domains are converging to enable automated code translation solutions.
This development significantly lowers the barrier to entry for modernizing legacy scientific codebases, accelerating research and development in fields reliant on complex simulations and models.
The effort and expertise required to convert decades-old scientific Fortran code into modern, differentiable frameworks like JAX is now vastly reduced through automated LLM-driven pipelines.
- · Scientific research institutions
- · Climate modeling community
- · AI/ML researchers in scientific domains
- · Hardware manufacturers (GPUs, TPUs)
- · Legacy scientific code maintenance specialists
- · Manual code porting services
Faster development and deployment of advanced scientific models with enhanced capabilities like gradient-based optimization.
An acceleration of scientific discovery as complex simulations become more amenable to AI-driven analysis and optimization.
The democratization of advanced scientific modeling, leading to new interdisciplinary research and potential breakthroughs across various fields.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI