According to a paper published on arxiv.org, researchers have introduced PICopilot, described as “the first large language model (LLM)-based agentic framework that assists in PIC design via automated design script generation from natural language instructions.”
The framework addresses challenges in photonic integrated circuit (PIC) design, which according to the paper is shifting from graphical user interface methods to script-based approaches. The paper states that script-based design “requires designers to possess additional proficiency in tool application programming interfaces (APIs) and programming” and “demands greater effort and time.”
PICopilot uses “a multi-agent architecture with a feedback mechanism and a specifically designed retrieval-augmented generation (RAG) pipeline,” according to the source. In experimental results, the framework “successfully completes all 48 tasks” in a benchmark of diverse PIC scripting tasks.
According to the paper, PICopilot “outperforms other LLM-based approaches without incurring substantial extra latency or cost, even solving 21 more tasks than the advanced GPT-5 model with a general RAG pipeline.”
The publication represents part of broader research into agentic AI frameworks, with other arxiv.org papers from the same period documenting similar approaches for microscopy control, clinical machine learning pipelines, and regulatory network prediction.