Multi-Objective Coevolution of Prompts and Templates for Circuit Approximation

2026-06-11

AI Summary

This study addresses the challenge of automatically synthesizing 8-bit approximate multipliers that achieve superior trade-offs among error, area, power, and delay—without requiring domain-specific training. To this end, the authors propose a co-evolutionary algorithm that jointly optimizes circuit architectures and prompt templates for large language models (LLMs), enabling effective multi-objective search in the space of approximate circuits without fine-tuning the LLMs. This approach significantly enhances design space exploration efficiency, yielding multipliers that outperform highly optimized benchmarks from EvoApproxLib across key metrics such as error-area trade-offs.

Abstract

Approximate multipliers deliberately relax computational accuracy to achieve gains in power efficiency, latency, and silicon area, which makes them well-suited for error-resilient applications such as neural networks. In this work, we introduce a co-evolutionary algorithm that leverages an off-the-shelf large language model (LLM) without requiring domain-specific training to automate the design of optimized 8-bit approximate multipliers. The approach simultaneously evolves a population of candidate circuits and a population of prompt templates that steer LLM-driven modifications. Experimental results for several target design objectives demonstrate that the proposed method discovers approximate multipliers with improved error-area trade-offs compared to highly optimized circuits from the EvoApproxLib library.

Problem

• approximate computing
• multi-objective optimization
• circuit approximation
• prompt engineering
• coevolution

Innovation

• co-evolutionary algorithm
• large language model (LLM)
• approximate multiplier
• prompt template
• circuit approximation