用离线框架让大模型高效生成高覆盖率的硬件测试用例
LLM4Cov: Execution-Aware Agentic Learning for High-coverage Testbench Generation
- 将验证过程建模为单步状态转移，利用确定性评估器替代昂贵仿真
- 40亿参数模型在基准上达90.4%平均覆盖率，优于教师模型10.5%
- 适合硬件验证、自动化测试等需高覆盖率且仿真成本高的场景
执行感知的大型语言模型代理在从工具反馈中学习方面展现出潜力，但此类反馈获取成本高且耗时，使在线强化学习在某些场景下不切实际。高覆盖率硬件验证即为此类挑战的典型例子，因其依赖工业级仿真器和不可微的执行信号。我们提出 LLM4Cov，一个离线代理学习框架，将验证建模为由确定性评估器引导的单步状态转移。基于此，我们引入执行验证数据筛选、策略感知的智能体数据合成及最差状态优先采样，实现执行约束下的可扩展学习。我们还通过改进评估协议，从现有验证套件中构建了一个更贴近现实的基准。使用该流程，一个40亿参数模型在CVDP-ECov上的通过率为69.2%，平均覆盖率达90.4%，分别优于其教师模型5.3%和10.5%，表现媲美参数量大一个数量级的模型。
原文摘要 · Abstract (English)
Execution-aware LLM agents offer a promising paradigm for learning from tool feedback, but such feedback can be expensive and slow to obtain, making online reinforcement learning (RL) less practical in certain scenarios. High-coverage hardware verification exemplifies this challenge due to its reliance on industrial simulators and non-differentiable execution signals. We propose LLM4Cov, an offline agent-learning framework that models verification as single-step state transitions guided by deterministic evaluators. Building on this formulation, we introduce execution-validated data curation, policy-aware agentic data synthesis, and worst-state-prioritized sampling to enable scalable learning under execution constraints. We further curate a reality-aligned benchmark adapted from an existing verification suite through a revised evaluation protocol. Using the proposed pipeline, a compact 4B-parameter model achieves 69.2% pass rate and 90.4% average coverage in CVDP-ECov under agentic evaluation, outperforming its teacher by 5.3% and 10.5%, demonstrating competitive performance against models an order of magnitude larger.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。