Towards Reliable AI-Assisted Analog Design: Template-Constrained LLM Agents for SAR ADC Generation

arXiv cs.AI · 6 hours ago

Researchers developed ATLAS, an LLM-based agent framework that generates functional Successive Approximation Register Analog-to-Digital Converters by grounding the model with expert knowledge constraints to avoid hallucinations. The framework uses template-constrained generation to guide LLM planning, selection, parameterization, and iterative modification across multiple technology nodes and input specifications. This approach enables reliable analog circuit design automation by combining LLM capabilities with domain-specific constraints that traditional prompting methods lack.