[Submitted on 10 Aug 2026 (v1), last revised 19 Nov 2026 (this version, v2)]
Abstract:Testbenches written by large language models can compile and run while leaving entire datapaths idle. We present CHORUS, the first framework to combine coverage-guided agentic reinforcement learning (RL) with policy consolidation for testbench stimulus generation. We train LLM4Cov’s final supervised checkpoint to draft testbenches and repair its lowest-scoring draft using simulator feedback, raising CVDP-ECov agentic Pass@1 from 69.2% to 85.5%. Applying the same training to three checkpoints yields similar overall pass rates, yet the policies succeed on different designs. Averaging their weights combines these strengths in a single dense 4B model that reaches 86.7% agentic Pass@1. When simulator rewards select the teacher for each training prompt, the distilled 4B model reaches 88.0%, 13.5 percentage points above DeepSeek-R1 (671B).
Submission history
From: Hejia Zhang [view email]
[v1] Mon, 10 Aug 2026 18:02:42 UTC (1,297 KB)
[v2] Thu, 19 Nov 2026 07:12:59 UTC (860 KB)
