SynAct: A Reasoning-Acting Large Language Model Agent for Adaptive Synthesis Optimization

2026-08-12

AI Summary

This work addresses the high sensitivity of logic synthesis outcomes to optimization command selection, a process that is inherently high-dimensional, computationally expensive, and poorly served by existing methods lacking dynamic adaptability and interpretability. The paper proposes the first closed-loop large language model (LLM) agent for logic synthesis, which dynamically generates interpretable, context-specific commands by continuously analyzing synthesis reports in conjunction with real-time circuit state, a tool-specific knowledge base, and historical optimization experience. Built upon a reasoning-action architecture, the agent enables state-aware perception, knowledge retrieval, and adaptive decision-making. Evaluated across 14 benchmark designs, the approach reduces the average worst negative slack (WNS) to 27% of the initial value while effectively balancing area and power consumption.

Abstract

Logic synthesis transforms RTL designs into gate-level netlists, where PPA results are highly sensitive to the choice of optimization commands, making synthesis tuning both high-dimensional and expensive. Previous approaches fall into two categories: automated methods, which perform black-box search over fixed action spaces with limited decision-level interpretability, and LLM-based methods, which typically generate static scripts upfront and cannot adapt to evolving circuit states. We present SynAct, an adaptive closed-loop LLM reasoning--acting agent that iteratively diagnoses live synthesis reports and reasons over the current circuit state, retrieved tool knowledge, and historical optimization experience to issue targeted commands. SynAct focuses on improving timing, particularly worst negative slack (WNS), while maintaining balanced area and power trade-offs. Experiments on a commercial synthesis tool across 14 designs show that SynAct reduces average WNS to 27% of that from bootstrap synthesis.

Problem

• logic synthesis
• optimization
• worst negative slack
• adaptive tuning
• PPA trade-offs

Innovation

• adaptive synthesis optimization
• reasoning-acting agent
• closed-loop LLM
• worst negative slack (WNS)
• logic synthesis