pith. machine review for the scientific record. sign in

arxiv: 2604.02811 · v2 · submitted 2026-04-03 · 💻 cs.AR · cs.AI

Recognition: no theorem link

ChatSVA: Bridging SVA Generation for Hardware Verification via Task-Specific LLMs

Lik Tung Fu , Jie Zhou , Shaokai Ren , Mengli Zhang , Jia Xiong , Hugo Jiang , Nan Guan , Xi Wang
show 1 more author
Jun Yang
Authors on Pith no claims yet

Pith reviewed 2026-05-13 18:59 UTC · model grok-4.3

classification 💻 cs.AR cs.AI
keywords SVA generationhardware verificationSystemVerilog Assertionsmulti-agent LLMsRTL designsfunctional verificationautomated assertion generation
0
0 comments X

The pith

ChatSVA shows multi-agent LLMs can generate SystemVerilog Assertions at 96 percent functional accuracy despite scarce data.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces ChatSVA, an end-to-end system that organizes task-specific large language models in a multi-agent framework to automatically produce SystemVerilog Assertions for hardware verification. The central mechanism is the AgentBridge platform, which creates high-purity datasets to address the lack of domain-specific examples that normally blocks effective few-shot learning. When tested on 24 RTL designs, the system reaches 98.66 percent syntax pass rate and 96.12 percent functional pass rate while generating 139.5 assertions per design at 82.5 percent function coverage. These results deliver a 33 percentage point gain in functional correctness and more than 11 times the coverage of the prior state of the art. The work also releases a public online service for direct use in verification flows.

Core claim

ChatSVA establishes that a multi-agent LLM architecture, supported by AgentBridge dataset generation, produces SVAs that pass syntax checks at 98.66 percent and functional checks at 96.12 percent across 24 RTL designs, while delivering 82.5 percent function coverage and 139.5 assertions per design, exceeding previous methods by 33 points in correctness and 11 times in coverage.

What carries the argument

The AgentBridge multi-agent platform that systematically generates high-purity datasets to support few-shot SVA generation with task-specific LLMs.

If this is right

  • Hardware verification teams can shift from manual SVA authoring to automated generation while maintaining high functional accuracy.
  • The framework provides a template for solving other long-chain reasoning tasks in few-shot, domain-specific engineering settings.
  • Verification effort, which consumes more than half of the IC development cycle, can be reduced through higher automation of property generation.
  • The reported pass rates and coverage levels become new reference points for measuring progress in automated assertion generation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar dataset purification steps could be applied to generate other verification artifacts such as test sequences or coverage points.
  • The multi-agent structure may extend to additional hardware description languages beyond SystemVerilog if comparable high-purity data pipelines are built.
  • If the approach scales to larger designs, function coverage could rise beyond the current 82.5 percent without additional manual intervention.

Load-bearing premise

The multi-agent framework with AgentBridge systematically produces high-purity datasets that overcome data scarcity and enable reliable few-shot SVA generation across diverse RTL designs.

What would settle it

Evaluating ChatSVA on a fresh collection of complex, previously unseen RTL designs and finding that the functional pass rate falls below 80 percent would show the reliability claim does not generalize.

Figures

Figures reproduced from arXiv: 2604.02811 by Hugo Jiang, Jia Xiong, Jie Zhou, Jun Yang, Lik Tung Fu, Mengli Zhang, Nan Guan, Shaokai Ren, Xi Wang.

Figure 1
Figure 1. Figure 1: IC Development Flow & Time Cost Distribution. Func￾tional verification accounts for 56% of the development time [12]. runtime monitors in simulation. Despite their power, manual SVA authoring has become a bottleneck, as the process is labor-intensive, error-prone, and requires expertise [33]. Consequently, automating SVA generation is critical to break the verification deadlock. The pursuit of SVA automati… view at source ↗
Figure 2
Figure 2. Figure 2: SVA Generation Capabilities of LLMs. (a) Number of SVAs.(b) Syntax Pass Rate.(c) Function Pass Rate. demonstrating state-of-the-art (SOTA) generative capabili￾ties on a comprehensive benchmark. • We have conducted extensive experiments to validate the effectiveness of our framework and have released an online service for public access and reproducibility. 2 Motivation & Related Work 2.1 Traditional SVA Des… view at source ↗
Figure 3
Figure 3. Figure 3: ChatSVA Workflow & Examples S1: Long-Chain Reasoning Decomposition. S1 directly targets C2 and C1 by decomposing the monolithic task into a pipeline of modular sub-tasks. This provides the intermediate reasoning steps C2 identifies as missing, shifting the focus from syntactic correctness to functional intent. By replacing the single, unreliable leap with a series of verifiable steps, this strategy directl… view at source ↗
Figure 4
Figure 4. Figure 4: AgentBridge Platform & AgentBridge in ChatSVA 5.1 AgentBridge Principles The AgentBridge data generation process, which transforms an in￾put set Din into an output set Dout, is governed by three principles. Principle 1: Directional Information Constraint. This prin￾ciple mandates a directional information flow where outputs are functional subsets of the input. This structure ensures that any generated outp… view at source ↗
Figure 5
Figure 5. Figure 5: ChatSVA Performance Comparisons construct the dataset for Agent3, resolving the challenge of data validation. For each Checkpoint generated by Agent3, Agent4 trans￾forms it into a verifiable SVA. This closed-loop validation strategy makes the Checkpoint generation task concretely verifiable. 6 Evaluations 6.1 Experimental Setup 6.1.1 Model Training. We constructed a 15.36 GB dataset for SFT and RAG, compri… view at source ↗
Figure 6
Figure 6. Figure 6: Distribution of Bug Detection [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Data Distribution in Reverse Generation Method 19.80× and 14.14× improvement over GPT-4o (4.17%) and DeepSeek￾R1 (5.83%) in Function Coverage, with SVA generation volume also increasing by over 18× against GPT-4o. This demonstrates that without a guiding methodology, even powerful general-purpose LLMs fail to generate comprehensive and functionally correct SVAs. More importantly, ChatSVA significantly outp… view at source ↗
read the original abstract

Functional verification consumes over 50% of the IC development lifecycle, where SystemVerilog Assertions (SVAs) are indispensable for formal property verification and enhanced simulation-based debugging. However, manual SVA authoring is labor-intensive and error-prone. While Large Language Models (LLMs) show promise, their direct deployment is hindered by low functional accuracy and a severe scarcity of domain-specific data. To address these challenges, we introduce ChatSVA, an end-to-end SVA generation system built upon a multi-agent framework. At its core, the AgentBridge platform enables this multi-agent approach by systematically generating high-purity datasets, overcoming the data scarcity inherent to few-shot scenarios. Evaluated on 24 RTL designs, ChatSVA achieves 98.66% syntax and 96.12% functional pass rates, generating 139.5 SVAs per design with 82.50% function coverage. This represents a 33.3 percentage point improvement in functional correctness and an over 11x enhancement in function coverage compared to the previous state-of-the-art (SOTA). ChatSVA not only sets a new SOTA in automated SVA generation but also establishes a robust framework for solving long-chain reasoning problems in few-shot, domain-specific scenarios. An online service has been publicly released at https://www.nctieda.com/CHATDV.html.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces ChatSVA, an end-to-end multi-agent LLM framework for automated SystemVerilog Assertion (SVA) generation in hardware verification. Its core contribution is the AgentBridge platform, which generates high-purity few-shot datasets to address data scarcity. Evaluated on 24 RTL designs, ChatSVA reports 98.66% syntax pass rate, 96.12% functional pass rate, an average of 139.5 SVAs per design, and 82.50% function coverage, claiming a 33.3 percentage point gain in functional correctness and over 11x improvement in coverage relative to prior SOTA.

Significance. If the reported metrics are supported by transparent, reproducible evaluation protocols, the work would meaningfully advance automated formal property generation in IC design, where SVA authoring remains a dominant cost. The multi-agent approach to high-purity dataset creation for long-chain, domain-specific reasoning could generalize to other data-scarce technical domains.

major comments (2)
  1. [Abstract and Evaluation section] Abstract and Evaluation section: the headline metrics (98.66% syntax pass rate, 96.12% functional pass rate, 82.50% function coverage, 33.3 pp and 11x gains) are stated without describing the verification oracle (formal tool, simulation harness, or manual review), the exact SOTA baseline implementation, design-selection criteria for the 24 RTL modules, or any statistical controls such as error bars or significance tests. These omissions render the central empirical claim unverifiable from the provided text.
  2. [AgentBridge platform description] AgentBridge platform description: the claim that the multi-agent pipeline produces high-purity datasets rests on internal self-consistency among agents sharing the same base LLM, with no mention of an independent external checker (formal verifier, cross-design oracle, or human audit). This setup risks correlated hallucinations being scored as functional passes on the same designs, undermining the assertion that the framework reliably overcomes data scarcity.
minor comments (2)
  1. [Abstract] The online service URL is given but no usage instructions, input/output formats, or reproducibility notes appear in the text.
  2. [Evaluation section] Notation for pass-rate and coverage calculations is introduced without an explicit equation or pseudocode definition.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback, which highlights important aspects of transparency in our evaluation. We have revised the manuscript to provide detailed descriptions of the verification process, baseline implementation, design selection, and additional validation steps for the AgentBridge platform. These changes strengthen the verifiability of our claims without altering the core methodology.

read point-by-point responses

  1. Referee: [Abstract and Evaluation section] Abstract and Evaluation section: the headline metrics (98.66% syntax pass rate, 96.12% functional pass rate, 82.50% function coverage, 33.3 pp and 11x gains) are stated without describing the verification oracle (formal tool, simulation harness, or manual review), the exact SOTA baseline implementation, design-selection criteria for the 24 RTL modules, or any statistical controls such as error bars or significance tests. These omissions render the central empirical claim unverifiable from the provided text.

    Authors: We agree that the original text lacked sufficient detail on the evaluation protocol. In the revised manuscript, we have expanded the Evaluation section (and updated the abstract accordingly) to explicitly describe: the verification oracle as the JasperGold formal verification tool used to check functional correctness of each generated SVA against the RTL design; the SOTA baseline as a direct reproduction of the prior work's prompting strategy using the same underlying LLM for fair comparison; the design-selection criteria as 24 diverse open-source RTL modules drawn from OpenCores and academic benchmarks, stratified by complexity and module type; and statistical controls including standard deviations across three independent generation runs per design plus paired t-test results confirming significance of the reported gains. These additions make the empirical claims fully verifiable. revision: yes

  2. Referee: [AgentBridge platform description] AgentBridge platform description: the claim that the multi-agent pipeline produces high-purity datasets rests on internal self-consistency among agents sharing the same base LLM, with no mention of an independent external checker (formal verifier, cross-design oracle, or human audit). This setup risks correlated hallucinations being scored as functional passes on the same designs, undermining the assertion that the framework reliably overcomes data scarcity.

    Authors: We acknowledge the concern regarding potential correlated hallucinations in a shared-LLM multi-agent system. The functional pass metric is computed by an independent external formal verifier (JasperGold) that checks each generated SVA against the target RTL design, providing an objective oracle separate from the generation agents. To further strengthen this, the revised manuscript now includes: (1) explicit description of this external verification step in the AgentBridge pipeline, (2) results from a human audit performed on a random 20% subset of generated SVAs across designs, and (3) cross-design validation where SVAs generated for one module are tested for portability on held-out designs. These additions address the risk while preserving the automated nature of the framework. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical results on external RTL designs

full rationale

The paper presents performance metrics (syntax/functional pass rates, coverage) from direct evaluation of ChatSVA on 24 RTL designs, with gains reported relative to a cited prior SOTA. No equations, fitted parameters, self-definitional constructs, or derivations appear in the abstract or described content. The multi-agent AgentBridge component is described as a data-generation pipeline whose output is then measured on independent designs; no reduction of the reported metrics to the generation process by construction is shown. This is a standard empirical claim and receives score 0.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

Abstract provides no explicit free parameters, axioms, or invented entities beyond naming the AgentBridge platform and multi-agent framework; all performance claims rest on the unstated assumption that the LLM agents and generated datasets behave as described.

invented entities (1)
  • AgentBridge platform no independent evidence
    purpose: Systematically generate high-purity datasets to overcome data scarcity for few-shot SVA generation
    Introduced as the core enabler of the multi-agent approach; no independent evidence or external validation supplied in abstract.

pith-pipeline@v0.9.0 · 5564 in / 1294 out tokens · 50305 ms · 2026-05-13T18:59:43.139703+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. From Language to Logic: Bridging LLMs & Formal Representations for RTL Assertion Generation

    cs.CR 2026-04 unverdicted novelty 7.0

    ProofLoop achieves 93.7% syntax correctness and 82.0% functional correctness for SVA generation from natural language by combining retrieval, EDA tools, and up to three rounds of JasperGold formal feedback.

  2. Automated SVA Generation with LLMs

    cs.AR 2026-04 unverdicted novelty 5.0

    SVA Generator improves semantic correctness of LLM-generated SystemVerilog Assertions by 22.7 percentage points on average for deeper properties using AST-grounded constraint injection and depth-stratified formal equi...

Reference graph

Works this paper leans on

42 extracted references · 42 canonical work pages · cited by 2 Pith papers · 1 internal anchor

  1. [1]

    Fnu Aditi and Michael S. Hsiao. 2022. Hybrid Rule-based and Machine Learning System for Assertion Generation from Natural Language Specifications. In2022 IEEE 31st Asian Test Symposium (ATS). 126–131. doi:10.1109/ATS56056.2022.00034

    work page doi:10.1109/ats56056.2022.00034 2022

  2. [2]

    Fnu Aditi and Michael S. Hsiao. 2023. Validatable Generation of System Verilog Assertions from Natural Language Specifications. In2023 Fifth International Conference on Transdisciplinary AI (TransAI). 102–109. doi:10.1109/TransAI60598. 2023.00026

    work page doi:10.1109/transai60598 2023

  3. [3]

    Yunsheng Bai, Ghaith Bany Hamad, Syed Suhaib, and Haoxing Ren. 2025. Asser- tionForge: Enhancing Formal Verification Assertion Generation with Structured Representation of Specifications and RTL. In2025 IEEE International Conference on LLM-Aided Design (ICLAD). Stanford, CA, USA, 85–92. doi:10.1109/ICLAD65226. 2025.00009

    work page doi:10.1109/iclad65226 2025

  4. [4]

    Jason Blocklove, Siddharth Garg, Ramesh Karri, and Hammond Pearce. 2023. Chip-chat: Challenges and opportunities in conversational hardware design. In 2023 ACM/IEEE 5th Workshop on Machine Learning for CAD (MLCAD). IEEE, 1–6. doi:10.1109/MLCAD58807.2023.10299874

    work page doi:10.1109/mlcad58807.2023.10299874 2023

  5. [5]

    Wen Chen, Sandip Ray, Jayanta Bhadra, Magdy Abadir, and Li-C Wang. 2017. Challenges and Trends in Modern SoC Design Verification.IEEE Design & Test 34, 5 (2017), 7–22. doi:10.1109/MDAT.2017.2735383

    work page doi:10.1109/mdat.2017.2735383 2017

  6. [6]

    Alessandro Danese, Nicolò Dalla Riva, and Graziano Pravadelli. 2017. A-TEAM: Automatic template-based assertion miner. In2017 54th ACM/EDAC/IEEE Design Automation Conference (DAC). 1–6. doi:10.1145/3061639.3062206

    work page doi:10.1145/3061639.3062206 2017

  7. [7]

    Babak Falsafi. 2023. What’s Missing in Agile Hardware Design? Verification!J. Comput. Sci. Technol.38, 4 (2023), 735–736. doi:10.1007/s11390-023-0005-3

    work page doi:10.1007/s11390-023-0005-3 2023

  8. [8]

    Frederiksen, John Aromando, and Michael S

    Steven J. Frederiksen, John Aromando, and Michael S. Hsiao. 2020. Automated Assertion Generation from Natural Language Specifications. In2020 IEEE Inter- national Test Conference (ITC). 1–5. doi:10.1109/ITC44778.2020.9325264

    work page doi:10.1109/itc44778.2020.9325264 2020

  9. [9]

    Samuele Germiniani and Graziano Pravadelli. 2022. HARM: A Hint-Based Asser- tion Miner.IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems41, 11 (2022), 4277–4288. doi:10.1109/TCAD.2022.3197525

    work page doi:10.1109/tcad.2022.3197525 2022

  10. [10]

    Harris and Ian G

    Christopher B. Harris and Ian G. Harris. 2016. GLAsT: Learning formal grammars to translate natural language specifications into hardware assertions. In2016 Design, Automation & Test in Europe Conference & Exhibition (DATE). 966–971

    work page 2016

  11. [11]

    Harry Foster. 2019. The Weather Report: 2018 Study On IC/ASIC Verifica- tion Trends. https://semiengineering.com/the-weather-report-2018-study-on- ic-asic-verification-trends/. Accessed: 2025-07-14

    work page 2019

  12. [12]

    Harry Foster. 2021. The 2020 Wilson Research Group Functional Verification Study. https://blogs.sw.siemens.com/verificationhorizons/2021/01/06/part-8-the- 2020-wilson-research-group-functional-verification-study/. Accessed: 2025-07- 14

    work page 2021

  13. [13]

    Harry Foster. 2023. The 2022 Wilson Research Group Functional Verification Study. https://blogs.sw.siemens.com/verificationhorizons/2023/01/23/epilogue- the-2022-wilson-research-group-functional-verification-study/. Accessed: 2025- 07-14

    work page 2023

  14. [14]

    Osman Hasan and Sofiene Tahar. 2015. Formal verification methods. InEncyclo- pedia of Information Science and Technology, Third Edition. IGI global, 7162–7170

    work page 2015

  15. [15]

    Yuchen Hu, Junhao Ye, Ke Xu, Jialin Sun, Shiyue Zhang, Xinyao Jiao, Dingrong Pan, Jie Zhou, Ning Wang, Weiwei Shan, Xinwei Fang, Xi Wang, Nan Guan, and Zhe Jiang. 2025. Uvllm: An automated universal rtl verification framework using llms. InProceedings of the 62nd ACM/IEEE Design Automation Conference (DAC ’25). arXiv:2411.16238 https://arxiv.org/abs/2411....

    work page arXiv 2025

  16. [16]

    Rahul Kande, Hammond Pearce, Benjamin Tan, Brendan Dolan-Gavitt, Shailja Thakur, Ramesh Karri, and Jeyavijayan Rajendran. 2024. (Security) Assertions by Large Language Models.IEEE Transactions on Information Forensics and Security 19 (2024), 4374–4389. doi:10.1109/TIFS.2024.3372809

    work page doi:10.1109/tifs.2024.3372809 2024

  17. [17]

    Oliver Keszocze and Ian G. Harris. 2019. Chatbot-based assertion generation from natural language specifications. In2019 Forum for Specification and Design Languages (FDL). 1–6. doi:10.1109/FDL.2019.8876925

    work page doi:10.1109/fdl.2019.8876925 2019

  18. [18]

    Rahul Krishnamurthy and Michael S. Hsiao. 2019. Controlled Natural Language Framework for Generating Assertions from Hardware Specifications. In2019 IEEE 13th International Conference on Semantic Computing (ICSC). 367–370. doi:10. 1109/ICOSC.2019.8665517

    work page arXiv 2019

  19. [19]

    Hämäläinen

    Sakari Lahti, Panu Sjövall, Jarno Vanne, and Timo D. Hämäläinen. 2019. Are We There Yet? A Study on the State of High-Level Synthesis.IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems38, 5 (2019), 898–911. doi:10.1109/TCAD.2018.2834439

    work page doi:10.1109/tcad.2018.2834439 2019

  20. [20]

    Mingjie Liu, Minwoo Kang, Ghaith Bany Hamad, Syed Suhaib, and Haoxing Ren. 2024. Domain-Adapted LLMs for VLSI Design and Verification: A Case Study on Formal Verification. In2024 IEEE 42nd VLSI Test Symposium (VTS). 1–4. doi:10.1109/VTS60656.2024.10538589

    work page doi:10.1109/vts60656.2024.10538589 2024

  21. [21]

    Tianyang Liu, Qi Tian, Jianmin Ye, LikTung Fu, Shengchu Su, Junyan Li, Gwok- Waa Wan, Layton Zhang, Sam-Zaak Wong, Xi Wang, et al. 2024. ChatChisel: Enabling Agile Hardware Design with Large Language Models. In2024 2nd International Symposium of Electronics Design Automation (ISEDA). IEEE, 710–716. doi:10.1109/ISEDA62518.2024.10618053

    work page doi:10.1109/iseda62518.2024.10618053 2024

  22. [22]

    Bhabesh Mali, Karthik Maddala, Vatsal Gupta, Sweeya Reddy, Chandan Karfa, and Ramesh Karri. 2024. ChIRAAG: ChatGPT Informed Rapid and Automated Assertion Generation. In2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI). 680–683. doi:10.1109/ISVLSI61997.2024.00130

    work page doi:10.1109/isvlsi61997.2024.00130 2024

  23. [23]

    Marcelo Orenes-Vera, Margaret Martonosi, and David Wentzlaff. 2023. From rtl to sva: Llm-assisted generation of formal verification testbenches.arXiv preprint arXiv:2309.09437(2023)

    work page arXiv 2023

  24. [24]

    Ganapathy Parthasarathy, Saurav Nanda, Parivesh Choudhary, and Pawan Patil

    work page

  25. [25]

    In2021 Second Document Intelligence Workshop at KDD

    Spectosva: Circuit specification document to systemverilog assertion translation. In2021 Second Document Intelligence Workshop at KDD

    work page

  26. [26]

    Valentin Radu, Diana Dranga, Catalin Dumitrescu, Alina Iuliana Tabirca, and Maria Cristina Stefan. 2024. Generative AI Assertions in UVM-Based System Ver- ilog Functional Verification.Systems12, 10 (2024). doi:10.3390/systems12100390

    work page doi:10.3390/systems12100390 2024

  27. [27]

    Mohammad Shahidzadeh, Behnam Ghavami, Steven Wilton, and Lesley Shannon

    work page

  28. [28]

    https://api.semanticscholar.org/CorpusID:274235019

    Automatic High-quality Verilog Assertion Generation through Subtask- Focused Fine-Tuned LLMs and Iterative Prompting.ArXivabs/2411.15442 (2024). https://api.semanticscholar.org/CorpusID:274235019

    work page arXiv 2024

  29. [29]

    Tell, Yanqing Zhang, William J

    Yakun Sophia Shao, Jason Clemons, Rangharajan Venkatesan, Brian Zimmer, Matthew Fojtik, Nan Jiang, Ben Keller, Alicia Klinefelter, Nathaniel Pinckney, Priyanka Raina, Stephen G. Tell, Yanqing Zhang, William J. Dally, Joel Emer, C. Thomas Gray, Brucek Khailany, and Stephen W. Keckler. 2019. Simba: Scal- ing Deep-Learning Inference with Multi-Chip-Module-Ba...

    work page doi:10.1145/3352460.3358302 2019

  30. [30]

    Chuyue Sun, Christopher Hahn, and Caroline Trippel. 2023. Towards Improving Verification Productivity with Circuit-Aware Translation of Natural Language to SystemVerilog Assertions. InFirst International Workshop on Deep Learning-aided Verification. https://openreview.net/forum?id=FKH8qCuM44

    work page 2023

  31. [31]

    Synopsys. 2015. Delivering Functional Verification Engagements.Synopsys White Paper(2015). https://www.synopsys.com/content/dam/synopsys/services/ whitepapers/delivering-functional-verification-engagements.pdf

    work page 2015

  32. [32]

    Shobha Vasudevan, David Sheridan, Sanjay Patel, David Tcheng, Bill Tuohy, and Daniel Johnson. 2010. GoldMine: Automatic assertion generation using data mining and static analysis. In2010 Design, Automation & Test in Europe Conference & Exhibition (DATE 2010). 626–629. doi:10.1109/DATE.2010.5457129

    work page doi:10.1109/date.2010.5457129 2010

  33. [33]

    Gwok-Waa Wan, SamZaak Wong, Shengchu Su, Chenxu Niu, Ning Wang, Xin- lai Wan, Qixiang Chen, Mengnv Xing, Jingyi Zhang, Jianmin Ye, Yubo Wang, Rongchang Song, Tao Ni, Qiang Xu, Nan Guan, Zhe Jiang, Xi Wang, Yong Chen, and Jun Yang. 2026. FIXME: Towards End-to-End Benchmarking of LLM-Aided Design Verification.Proceedings of the AAAI Conference on Artificial...

    work page doi:10.1609/aaai.v40i2.37079 2026

  34. [34]

    Xi Wang, Gwok-Waa Wan, Sam-Zaak Wong, Layton Zhang, Tianyang Liu, Qi Tian, and Jianmin Ye. 2024. ChatCPU: An Agile CPU Design and Verification Platform with LLM. InProceedings of the 61st ACM/IEEE Design Automation Conference(San Francisco, CA, USA)(DAC ’24). Association for Computing Machinery, New York, NY, USA, Article 212, 6 pages. doi:10.1145/3649329.3658493

    work page doi:10.1145/3649329.3658493 2024

  35. [35]

    Hasini Witharana, Yangdi Lyu, Subodha Charles, and Prabhat Mishra. 2022. A Survey on Assertion-based Hardware Verification. 54, 11s, Article 225 (2022), 33 pages. doi:10.1145/3510578

    work page doi:10.1145/3510578 2022

  36. [36]

    2025.MEIC: Re-thinking RTL Debug Automation using LLMs

    Ke Xu, Jialin Sun, Yuchen Hu, Xinwei Fang, Weiwei Shan, Xi Wang, and Zhe Jiang. 2025.MEIC: Re-thinking RTL Debug Automation using LLMs. Association for Computing Machinery, New York, NY, USA. https://doi.org/10.1145/3676536. 3676801

    work page doi:10.1145/3676536 2025

  37. [37]

    Qiang Xu, Leon Stok, Rolf Drechsler, Xi Wang, Grace Li Zhang, and Igor L. Markov

    work page

  38. [38]

    In2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD)

    Revolution or Hype? Seeking the Limits of Large Models in Hardware Design. In2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD). 1–9. doi:10.1109/ICCAD66269.2025.11240750

    work page doi:10.1109/iccad66269.2025.11240750 2025

  39. [39]

    Zhiyuan Yan, Wenji Fang, Mengming Li, Min Li, Shang Liu, Zhiyao Xie, and Hongce Zhang. 2025. AssertLLM: Generating Hardware Verification Asser- tions from Design Specifications via Multi-LLMs. InProceedings of the 30th Asia and South Pacific Design Automation Conference(Tokyo, Japan)(ASP- DAC ’25). Association for Computing Machinery, New York, NY, USA, 6...

    work page doi:10.1145/3658617.3697756 2025

  40. [40]

    Bingkun Yao, Ning Wang, Jie Zhou, Xi Wang, Hong Gao, Zhe Jiang, and Nan Guan. 2025. Location is Key: Leveraging LLM for Functional Bug Localization in Verilog Design. InProceedings of the 62nd Annual ACM/IEEE Design Automation Conference(San Francisco, California, United States)(DAC ’25). IEEE Press, Article 394, 7 pages. doi:10.1109/DAC63849.2025.11133280

    work page doi:10.1109/dac63849.2025.11133280 2025

  41. [41]

    Junchen Zhao and Ian G. Harris. 2019. Automatic Assertion Generation from Natural Language Specifications Using Subtree Analysis. In2019 De- sign, Automation & Test in Europe Conference & Exhibition (DATE). 598–601. doi:10.23919/DATE.2019.8714857

    work page doi:10.23919/date.2019.8714857 2019

  42. [42]

    Yaowei Zheng, Richong Zhang, Junhao Zhang, Yanhan Ye, Zheyan Luo, Zhangchi Feng, and Yongqiang Ma. 2024. LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations). Association for Computational Linguistics. http://arxiv.or...

    work page internal anchor Pith review Pith/arXiv arXiv 2024