Who Plans, Who Edits? Verified, Cost-Aware Division of Labor for LLM-Driven Analog Layout Repair

Key points

• In our tool-grounded repair agent, models only choose bounded layout operators; sign-off tools rerun each task's contract checks after every edit, so failing layouts are never accepted.1
• We propose a learning controller that selects models for each role and the escalation order from the first tool evaluation and updates Bayesian estimates after every attempt.1
• The cheapest mid-tier model alone fixed 6 of 8 at the lowest overall spend per fix but missed the hardest task.1
• All 32 extracted DAC repairs match the human-guided LVS repair's integral nonlinearity (INL) within 0.005 least significant bits (LSB) at every code.1

Sources (1)

• [1]Who Plans, Who Edits? Verified, Cost-Aware Division of Labor for LLM-Driven Analog Layout Repair
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 16, 08:59 PM

  Language-model agents can repair analog layouts when sign-off tools check every edit, but which model should plan a repair, which should edit it, and at what cost remain open questions.

  In our tool-grounded repair agent, models only choose bounded layout operators; sign-off tools rerun each task's contract checks after every edit, so failing layouts are never accepted.

Extractive summary: sentences quoted from the sources.