Analog design agents improve long task success with reference circuits
Long-Horizon Analog Design Bench: Benchmarking Agents on Hours-Long Analog and Mixed-Signal Circuit Design Tasks
Artificial Intelligence
Summary
Designing analog circuits can take many hours, and it's not clear how well AI tools handle such long projects. The authors created a benchmark with 50 complex circuit tasks from real designers to test AI agents over hours-long attempts. They found that success rates vary widely and that providing example reference circuits helps agents complete designs faster and more accurately. This work sheds light on the challenges of automated analog circuit design and ways to improve tool performance.
What this means in practice
- •For chip design teams: Evaluate and improve AI tools for complex analog circuit design tasks requiring hours of simulation and iteration.
- •For eda software developers: Integrate task-matched reference circuits to accelerate and enhance the success rates of AI-driven analog design automation.
Authors
Analog Design Bench Team
Abstract
Coding agents now sustain hours-long, tool-driven loops, yet their ability to carry long-horizon analog and mixed-signal circuits to electrical specification remains unmeasured. We introduce Analog Design Bench, a long-horizon agentic benchmark of 50 transistor-level design tasks contributed by 17 chip designers. Agents work with an open-source simulator, while an isolated verifier evaluates the submitted circuit using specification-based electrical tests. We evaluate 15 agent configurations across 2,250 two-hour attempts and observe full-specification pass rates from 8.0% to 78.0%. Coding-benchmark performance correlates with analog results but leaves much of the performance spread unexplained. Our failure analysis shows that most unsuccessful submissions have no recorded legality rejection but fail electrical acceptance, identifying electrical closure as the dominant endpoint challenge. We test time, reasoning effort, agent harness, and supplied design knowledge as interventions. Longer budgets and higher reasoning effort improve performance, while general skill documents provide little benefit and sometimes reduce performance. Supplying a task-matched reference topology, an idealized form of circuit-IP retrieval, raises DeepSeek V4 Pro by 18.7 percentage points and mainly accelerates GPT-5.6 Sol.