LLM4Cov: Execution-Aware Agentic Learning for High-coverage Testbench Generation
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866914618651181056 |
|---|---|
| author | Zhang, Hejia Yu, Zhongming Ho, Chia-Tung Ren, Haoxing Khailany, Brucek Zhao, Jishen |
| author_facet | Zhang, Hejia Yu, Zhongming Ho, Chia-Tung Ren, Haoxing Khailany, Brucek Zhao, Jishen |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_16953 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LLM4Cov: Execution-Aware Agentic Learning for High-coverage Testbench Generation Zhang, Hejia Yu, Zhongming Ho, Chia-Tung Ren, Haoxing Khailany, Brucek Zhao, Jishen Artificial Intelligence Machine Learning 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. |
| title | LLM4Cov: Execution-Aware Agentic Learning for High-coverage Testbench Generation |
| topic | Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2602.16953 |