UCAgent: An End-to-End Agent for Block-Level Functional Verification

arXiv cs.AI / 3/30/2026

💬 OpinionIdeas & Deep AnalysisTools & Practical UsageModels & Research

共有:

Key Points

The paper argues that functional verification is a major IC development bottleneck and that conventional constrained-random and formal methods cannot easily scale to today’s design complexity.
It proposes UCAgent, an end-to-end agent for block-level functional verification that aims to overcome LLM limitations such as low accuracy in generating verification code, brittleness across multi-step workflows, and loss of consistency across specs/coverage/test artifacts.
UCAgent builds a pure-Python verification environment (using Picker and Toffee) to reduce reliance on LLM-generated Verilog/SystemVerilog verification code.
The approach includes a configurable 31-stage fine-grained workflow, where each LLM-guided step is validated by automated checkers, plus a Verification Consistency Labeling Mechanism (VCLM) to improve traceability of generated artifacts.
Experiments on UART, FPU, and integer-divider modules report up to 98.5% code coverage, up to 100% functional coverage, and the discovery of previously unidentified design defects.

Abstract

Functional verification remains a critical bottleneck in modern IC development cycles, accounting for approximately 70% of total development time in many projects. However, traditional methods, including constrained-random and formal verification, struggle to keep pace with the growing complexity of modern semiconductor designs. While recent advances in Large Language Models (LLMs) have shown promise in code generation and task automation, significant challenges hinder the realization of end-to-end functional verification automation. These challenges include (i) limited accuracy in generating Verilog/SystemVerilog verification code, (ii) the fragility of LLMs when executing complex, multi-step verification workflows, and (iii) the difficulty of maintaining verification consistency across specifications, coverage models, and test cases throughout the workflow. To address these challenges, we propose UCAgent, an end-to-end agent that automates hardware block-level functional verification based on three core mechanisms. First, we establish a pure Python verification environment using Picker and Toffee to avoid relying on LLM-generated SystemVerilog verification code. Second, we introduce a configurable 31-stage fine-grained verification workflow to guide the LLM, where each stage is verified by an automated checker. Furthermore, we propose a Verification Consistency Labeling Mechanism (VCLM) that assigns hierarchical labels to LLM-generated artifacts, improving the reliability and traceability of verification. Experimental results show that UCAgent can complete end-to-end automated verification on multiple modules, including the UART, FPU, and integer divider modules, achieving up to 98.5% code coverage and up to 100% functional coverage. UCAgent also discovers previously unidentified design defects in realistic designs, demonstrating its practical potential.