# CHORUS: Complementary Experts for High-Coverage Testbench Stimulus Generation

arXiv: https://arxiv.org/abs/2608.10090 · HTML: https://mlsanity.com/paper/2608.10090/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 21 of its 26 references that are in the map.
- First posted on arXiv 10 Aug 2026.

## Builds on (most cited first, 21 of 21)

- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [Merging Models with Fisher-Weighted Averaging](https://mlsanity.com/paper/2111.09832/) (2021)
- [Policy Distillation](https://mlsanity.com/paper/1511.06295/) (2015)
- [Rewarded soups: towards Pareto-optimal alignment by interpolating weights fine-tuned on diverse rewards](https://mlsanity.com/paper/2306.04488/) (2023)
- [WARM: On the Benefits of Weight Averaged Reward Models](https://mlsanity.com/paper/2401.12187/) (2024)
- [VerilogCoder: Autonomous Verilog Coding Agents with Graph-based Planning and Abstract Syntax Tree (AST)-based Waveform Tracing Tool](https://mlsanity.com/paper/2408.08927/) (2024)
- [DELLA-Merging: Reducing Interference in Model Merging through Magnitude-Based Sampling](https://mlsanity.com/paper/2406.11617/) (2024)
- [Adaptive Multi-Teacher Multi-level Knowledge Distillation](https://mlsanity.com/paper/2103.04062/) (2021)
- [RTLCoder: Outperforming GPT-3.5 in Design RTL Generation with Our Open-Source Dataset and Lightweight Solution](https://mlsanity.com/paper/2312.08617/) (2023)
- [Reinforced Multi-Teacher Selection for Knowledge Distillation](https://mlsanity.com/paper/2012.06048/) (2020)
- [AutoBench: Automatic Testbench Generation and Evaluation Using LLMs for HDL Design](https://mlsanity.com/paper/2407.03891/) (2024)
- [QiMeng-CodeV-R1: Reasoning-Enhanced Verilog Generation](https://mlsanity.com/paper/2505.24183/) (2025)
- [MAGE: A Multi-Agent Engine for Automated RTL Code Generation](https://mlsanity.com/paper/2412.07822/) (2024)
- [KL for a KL: On-Policy Distillation with Control Variate Baseline](https://mlsanity.com/paper/2605.07865/) (2026)
- [CorrectBench: Automatic Testbench Generation with Functional Self-Correction using LLMs for HDL Design](https://mlsanity.com/paper/2411.08510/) (2024)
- [Superficial Self-Improved Reasoners Benefit from Model Merging](https://mlsanity.com/paper/2503.02103/) (2025)
- [VeriCoder: Enhancing LLM-Based RTL Code Generation through Functional Correctness Validation](https://mlsanity.com/paper/2504.15659/) (2025)
- [LLM4Cov: Execution-Aware Agentic Learning for High-coverage Testbench Generation](https://mlsanity.com/paper/2602.16953/) (2026)
- [Breaking Memorization Barriers in LLM Code Fine-Tuning via Information Bottleneck for Improved Generalization](https://mlsanity.com/paper/2510.16022/) (2025)

## Abstract (arXiv)

Large language models (LLMs) have advanced code generation, where executable feedback provides a more reliable learning signal than textual imitation alone. Hardware verification is an important application of code generation and accounts for a substantial fraction of modern chip design effort, with high-coverage testbench stimulus generation as a key task. We present CHORUS, a post-training framework that pushes performance beyond what a conventional supervised fine-tuning (SFT)-to-reinforcement learning (RL) pipeline achieves. CHORUS builds on two observations. First, staged SFT produces behaviorally diverse checkpoints, and dense-reward RL turns them into strong experts with comparable aggregate performance but distinct task-level strengths. Second, these complementary strengths can be exploited through either training-free model merging or further post-training to outperform the best individual expert. By consolidating the resulting specialists into a single 4B model, CHORUS achieves 88.0% Pass@1 on CVDP-ECov, outperforming DeepSeek-R1 (671B) by 13.5 percentage points.