BugGen 2.0:重新划分 LLM 角色,降低 RTL bug 生成成本
BugGen 2.0: Rethinking the Roles of LLMs for Scalable RTL Bug Synthesis
做芯片验证的朋友可以看看这篇。它把 LLM 限定在语义改写,位置选择交给确定性算法,token 用量比 BugGen 1.0 省 6.8-15.4 倍,还用抽样估算跳过六成以上仿真。需要大规模构造带 ground-truth 的 bug 数据集的团队会用得上。
arXiv 论文 BugGen 2.0 改变了用 LLM 生成 RTL bug 数据集的分工方式:确定性算法负责结构分析、目标选择和验证,LLM 只负责局部语义改写。相比之下,BugGen 1.0 让模型同时决定突变位置和生成编辑,推理成本高且容易产生无效选择。
在 OpenTitan 的四个 IP 模块和 FabScalar 上,配合三个 LLM 测试,BugGen 2.0 达成了全部配置的 bug 配额,生成的场景均可通过编译并引发至少两次测试失败,16 种配置的突变类型中有 14-16 种得到覆盖。与 BugGen 1.0 相比,token 消耗降低 6.8-15.4 倍;与 Firefly 相比降低 1.5-7.9 倍。
论文还引入了抽样估算法扩充数据集:找到有可观测失败的合格场景后,在同一位置生成替代突变,只仿真其中一部分并估算其余的接受率,从而跳过 AES 63.6% 和 FabScalar 33.9% 的变体仿真。完整仿真验证显示 AES 变体接受率为 77.7%,FabScalar 为 95.2%,FabScalar 可比集合上的估算误差为 1.09%。
BugGen 2.0: Rethinking the Roles of LLMs for Scalable RTL Bug Synthesis
AI-assisted RTL localization and debugging need ground-truth bug datasets with known locations and observable failures, while testbench grading uses injected bugs to measure detection capability. Creating such datasets at scale remains a major EDA challenge. BugGen 1.0 shows that Large Language Models (LLMs) can generate context-aware RTL mutations. Still, it relies on the model for both choosing where to mutate and generating the edit, which increases inference costs and can lead to invalid choices. BugGen 2.0 changes this division of labor. Deterministic algorithms handle structural analysis, target selection, and validation, while the LLM focuses on localized semantic rewriting. Across four OpenTitan IP blocks and FabScalar with three LLMs, BugGen 2.0 reaches every requested quota of compilable bug scenarios that cause at least two test failures, with 14-16 of 16 configured mutation types represented. In completed same-model comparisons, it reduces token use by 6.8x-15.4x relative to BugGen 1.0 and 1.5x-7.9x relative to Firefly. BugGen 2.0 also expands the dataset without simulating every new bug. After finding accepted scenarios with observable failures, it generates alternative mutations at the same locations, simulates only a subset, and estimates acceptance for the rest. This skips 63.6% of AES and 33.9% of FabScalar variant simulations. Full simulation finds that 77.7% of AES variants and 95.2% of FabScalar variants are accepted, with a 1.09% estimation error on the directly comparable FabScalar set. These results show that focusing LLMs on semantic rewriting lowers inference and simulation cost while improving consistency across models and designs.
本条情报归属 AITOP「论文」核心栏目,重点聚焦北美前沿实验室与硅谷科技巨头。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:BugGen、RTL、LLM、EDA,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明:本文原始线索采编自「arXiv cs.AI」(@Minh Luu, Michael Quinn, Aakash Tyagi, David Houngninou, Jiang Hu),经 AITOP 多维评分管线结构化提炼。