01 / AI · Analysis
ICLayout-Bench scores only circuit layouts that pass checks
A paper evaluating electrical performance and area reported substantial differences in model costs between configurations with similar scores.

Researchers introduced ICLayout-Bench, a 165-task benchmark for circuit-design AI agents that scores only layouts passing physical and functional checks.
The paper evaluates electrical performance and area, and reported substantial differences in model costs between configurations with similar scores. Its results link layout quality with the cost of producing it.
According to the paper, agents use electronic design automation tools, software for designing electronic circuits, to create and optimise layouts. Evaluation extends beyond layout generation to simulation after the layout is complete.
Public and private task scores
The paper reported results from 22 model configurations, presenting mean layout scores separately for public and private tasks. These are researcher-published benchmark results.
Both the upper and lower ends of the reported mean-score range were higher for public tasks than for private tasks. This comparison describes the ranges of mean scores reported in the paper.
Similar scores, different costs
The paper found that higher reasoning effort did not consistently improve mean scores. It reported no common pattern in which increasing that effort produced higher scores.
The paper also reported substantial differences in model costs between configurations with comparable scores, making cost a separate consideration alongside electrical performance and area. The abstract does not disclose the cost amounts for individual configurations.