发表机构
Zhejiang University; ChipFlux（浙江大学; ChipFlux）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
WaveletECO提出闭环执行平台与专用本地9B模型，通过监督微调和CPO-SimPO训练，在594次评估中得分79.63/79.65，优于GPT-6 Astra的77.44，且INT8推理成本仅为后者的1/147，实现低成本高效多轮ECO修复。
AI 中文摘要
工程变更单（ECO）是修复芯片设计后期时序和电气违例的重要步骤。现有的智能体EDA方法主要关注工具调用，而对模型决策质量和针对性训练关注较少。ECO中的一个核心挑战是多轮决策：模型必须利用每轮的结果来确定下一个修复动作。我们提出了WaveletECO，它将闭环执行平台与大型语言模型集成，使智能体能够有效执行ECO决策。我们还通过监督微调和CPO-SimPO，使用执行演示和决策偏好数据训练了一个本地9B模型，使得通过本地部署的模型进行ECO决策成为可能。在22个设计上的594次评估运行中，WaveletECO-Policy（BF16）和（INT8）分别得分79.63和79.65，而GPT-6 Astra得分为77.44。INT8的估计推理成本约为GPT-6 Astra的1/147。这些结果表明，专门的模型训练支持有效、低成本的多轮ECO修复，并且在INT8量化下修复质量得以保持。
英文摘要
Engineering change order (ECO) is an important step in repairing timing and electrical violations during the late stages of chip design. Existing Agentic EDA methods primarily focus on tool invocation, with less attention to model decision quality and targeted training. A central challenge in ECO is multi-round decision-making: the model must use the results of each round to determine the next repair action. We propose WaveletECO, which integrates a closed-loop execution platform with large language models to enable agents to execute ECO decisions effectively. We also train a local 9B model through supervised fine-tuning and CPO-SimPO using execution demonstrations and decision-preference data, enabling ECO decision-making with a locally deployed model. Across 594 evaluation runs on 22 designs, WaveletECO-Policy (BF16) and (INT8) score 79.63 and 79.65, respectively, compared with GPT-6 Astra's 77.44. The estimated inference cost of INT8 is about 1/147 of GPT-6 Astra's. These results show that specialized model training supports effective, low-cost multi-round ECO repair, with repair quality retained under INT8 quantization.
Comments5 pages, 2 figures, 2 tables
