Research

Paper

TESTING February 18, 2026

LLM4Cov: Execution-Aware Agentic Learning for High-coverage Testbench Generation

Authors

Hejia Zhang, Zhongming Yu, Chia-Tung Ho, Haoxing Ren, Brucek Khailany, Jishen Zhao

Abstract

Execution-aware LLM agents offer a promising paradigm for learning from tool feedback, but such feedback is often expensive and slow to obtain, making online reinforcement learning (RL) impractical. 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 memoryless 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% coverage pass rate under agentic evaluation, outperforming its teacher by 5.3% and demonstrating competitive performance against models an order of magnitude larger.

Metadata

arXiv ID: 2602.16953
Provider: ARXIV
Primary Category: cs.AI
Published: 2026-02-18
Fetched: 2026-02-21 18:51

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.16953v1</id>\n    <title>LLM4Cov: Execution-Aware Agentic Learning for High-coverage Testbench Generation</title>\n    <updated>2026-02-18T23:36:46Z</updated>\n    <link href='https://arxiv.org/abs/2602.16953v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.16953v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Execution-aware LLM agents offer a promising paradigm for learning from tool feedback, but such feedback is often expensive and slow to obtain, making online reinforcement learning (RL) impractical. 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 memoryless 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% coverage pass rate under agentic evaluation, outperforming its teacher by 5.3% and demonstrating competitive performance against models an order of magnitude larger.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <published>2026-02-18T23:36:46Z</published>\n    <arxiv:primary_category term='cs.AI'/>\n    <author>\n      <name>Hejia Zhang</name>\n    </author>\n    <author>\n      <name>Zhongming Yu</name>\n    </author>\n    <author>\n      <name>Chia-Tung Ho</name>\n    </author>\n    <author>\n      <name>Haoxing Ren</name>\n    </author>\n    <author>\n      <name>Brucek Khailany</name>\n    </author>\n    <author>\n      <name>Jishen Zhao</name>\n    </author>\n  </entry>"
}