arXiv subjects

Junzhe Liu

Publications and source records attributed to Junzhe Liu.

9 recordsLinked to original sources

arXiv2026

AgentDV: Closed-Loop Agentic AI for Hardware Design Verification

Register-transfer level (RTL) verification consumes a major part of modern system-on-chip (SoC) development effort. Yet, recent LLM-based verification-code generation often fails to produce runnable, design-consistent, and coverage-producing testbenches. We present AgentDV, a closed-loop agentic AI framework for automated RTL verification environment generation. AgentDV transforms single-shot LLM testbench generation into a tool-grounded verification pipeline by combining LLM-guided analysis, testbench construction, simulation, coverage measurement, and iterative refinement. The framework introduces three key ideas: 1) runnability filtering to reject invalid generated environments, 2) CSR-grounded checking to reduce hallucinated signals and incorrect expected behavior, and 3) coverage-guided iteration to regenerate tests based on measured verification gaps. We evaluate AgentDV using three LLMs on challenge DUTs and public OpenTitan peripheral and security IP blocks. From our analysis, we observed that direct single-shot prompting fails to produce a valid coverage-producing environment on benchmarks. AgentDV achieves 100% pass rate on four DUTs and an average of 80.9% pass rate on all DUTs using Claude Sonnet 4.6. Similarly, an average of 58.7% and 60.6% pass rate is achieved for Llama and Qwen models, respectively. In addition, an average of 74.5%, 69.1%, and 64.9% of line coverage and 88.4%, 82.3%, and 76.7% of branch coverage for the benchmarks under consideration for Claude Sonnet 4.6, Llama, and Qwen models, respectively.

cs.SE↗

arXiv2026

GigaSpeechBench: A Real-World Multilingual Speech-to-Text Benchmark

While modern ASR systems achieve low error rates on high-resource benchmarks, such performance often overestimates real-world robustness. Existing evaluations address challenges in isolation, lacking a unified benchmark for domain terminology, age variation, dialects, accents, and low-resource languages, particularly across the Middle East and Southeast Asia, representing over one billion under-evaluated speakers. To address this gap, we introduce GigaSpeechBench, a comprehensive multilingual and multidimensional in-the-wild ASR & AST benchmark comprising 680 hours of human-annotated speech. It features five modules: (1) 12 low-resource Middle Eastern and Southeast Asian languages, plus challenging Japanese and Korean; (2) 6 Chinese dialects; (3) 6 English accents; (4) dense terminology across 12 vertical domains for Chinese and English; and (5) older adult and child speech. We further provide human-annotated Chinese and English translations for 11 languages to support AST evaluation. Extensive evaluations of leading foundation models and commercial APIs reveal significant performance degradation in these challenging settings, exposing critical evaluation blind spots.

eess.AS↗

arXiv2026

SysVCoder: An LLM-Driven Framework for Systematic Generation of System-Level Design

Recent advances in large language models (LLMs) have demonstrated strong potential in generating hardware designs using hardware description languages (HDLs) such as Verilog. However, existing LLM-based frameworks struggle to accurately capture the complexity of real-world architectural designs, particularly for large-scale systems with hierarchical, multi-level module instantiations. To address this issue, we present SysVCoder, an LLM-driven framework that enhances both the generation quality and efficiency of system-level design in Verilog. SysVCoder introduces a two-stage generation pipeline that leverages an intermediate representation to enable a more structured and accurate translation from natural language specifications to complex multi-module designs. Furthermore, we incorporate a rule-based alignment mechanism and a domain-specific retrieval-augmented generation strategy (DS-RAG) to enhance functional correctness by grounding LLM outputs in domain knowledge. We also present SysVDB, a comprehensive dataset comprising 60 system-level hardware designs along with their corresponding verification testbenches. Experimental results demonstrate that SysVCoder outperforms state-of-the-art frameworks such as CodeV and VeriGen by 30.7% and 38.3% in terms of functional correctness under the same base LLM. Notably, SysVCoder achieves performance comparable to NVIDIA's GPT-4 based VerilogCoder while using only a 7B-parameter model, reducing token consumption by 7.6x and synthesis latency by 37.5x. Both SysVCoder and SysVDB are made public at https://gitee.com/sdu-aes-lab/sysvcoder/.

cs.SE↗

arXiv2026

Probing Freeze-In Dark Matter via a Spin-2 Portal at the LHC with Vector Boson Fusion and Machine Learning

The persistent absence of signals in traditional dark matter searches has intensified interest in scenarios beyond the canonical weakly interacting massive particle paradigm. In this work, we investigate the collider phenomenology of feebly interacting dark matter produced via the freeze-in mechanism through a spin-2 portal. We consider a framework in which a massive graviton-like mediator couples minimally and universally to the energy--momentum tensor of both the Standard Model (SM) and the dark sector. Such interactions arise naturally in extra-dimensional constructions and effective theories of gravity, providing a theoretically well-motivated and predictive setup. We systematically connect early-Universe cosmology with collider observables by identifying regions of parameter space consistent with freeze-in conditions and the observed dark matter relic abundance, and examining their testability at the Large Hadron Collider (LHC). Focusing on bosonic fusion production channels, which are particularly sensitive to spin-2 interactions, we analyze invisible mediator decay signatures and assess current and projected experimental sensitivities. To enhance sensitivity in this challenging regime of feeble couplings, we develop a search strategy based on machine-learning algorithms. Our results demonstrate that collider searches can probe substantial regions of the cosmologically viable freeze-in parameter space, highlighting the high-luminosity LHC as a powerful laboratory for feebly interacting dark sectors. This study establishes a concrete and complementary pathway to test freeze-in dark matter scenarios through spin-2 portals, thereby bridging gravitationally motivated new physics, cosmology, and high-energy collider experiments.

hep-ph↗

arXiv2026

Habibi: Laying the Open-Source Foundation of Unified-Dialectal Arabic Speech Synthesis

Arabic spans over 30 spoken varieties, yet no open-source text-to-speech system unifies them. Key barriers include substantial cross-dialect lexical and phonological divergence, scarce synthesis-grade data, and the absence of a standardized multi-dialect evaluation benchmark. We present Habibi, a unified-dialectal Arabic TTS framework that addresses all three. Through a multi-step curation pipeline, we repurpose open-source ASR corpora into TTS training data covering 12+ regional dialects. A linguistically-informed curriculum learning strategy - progressing from Modern Standard Arabic to dialectal data - enables robust zero-shot synthesis without text diacritization. We further release the first standardized multi-dialect Arabic TTS benchmark, comprising over 11,000 utterances across 7 dialect subsets with manually verified transcripts. On this benchmark, our unified model matches or surpasses per-dialect specialized models. Both automatic metrics and human evaluations confirm that Habibi is highly competitive with ElevenLabs' Eleven v3 (alpha) in intelligibility, speaker similarity, and naturalness. Extensive ablations (~8,000 H100 GPU hours, 30+ configurations) validate each design choice. We open-source all checkpoints, training and inference code, and benchmark data - the first such release for multi-dialect Arabic TTS - at https://SWivid.github.io/Habibi/ .

cs.CL↗

arXiv2024

A Complex Scaling Method for Efficient and Accurate Scattering Emulation in Nuclear Reactions

We present a novel scattering emulator utilizing the complex scaling method to enhance nuclear reaction analysis. This approach leverages a single set of reduced bases, allowing for efficient and simultaneous emulation across multiple channels and potential parameters, significantly reducing computational storage and accelerating calculations. Demonstrated through $n$+$^{40}$Ca and $^{11}$Be+$^{64}$Zn elastic scattering, our method achieves high accuracy and efficiency. This emulator exhibits stable and reliable performance without anomalies inherent in other techniques, showcasing its robustness.

nucl-th↗

arXiv2024

COLOSS: Complex-scaled Optical and couLOmb Scattering Solver

We introduce COLOSS, a program designed to address the scattering problem using a bound-state technique known as complex scaling. In this method, the oscillatory boundary conditions of the wave function are transformed into exponentially decaying ones, accommodating the long-range Coulomb interaction. The program implements the Woods-Saxon form of a realistic optical potential, with all potential parameters included in a well-designed input format for ease of use. This design offers users straightforward access to compute $S$-matrices and cross-sections of the scattering process. We provide thorough discussions on the precision of Lagrange functions and their benefits in evaluating matrix elements. Additionally, COLOSS incorporates two distinct rotation methods, making it adaptable to potentials without analytical expressions. Comparative results demonstrate that COLOSS achieves high accuracy when compared with the direct integration method, Numerov, underscoring its utility and effectiveness in scattering calculations.

physics.comp-ph↗

arXiv2024

Improved Factorized Neural Transducer Model For text-only Domain Adaptation

Adapting End-to-End ASR models to out-of-domain datasets with text data is challenging. Factorized neural Transducer (FNT) aims to address this issue by introducing a separate vocabulary decoder to predict the vocabulary. Nonetheless, this approach has limitations in fusing acoustic and language information seamlessly. Moreover, a degradation in word error rate (WER) on the general test sets was also observed, leading to doubts about its overall performance. In response to this challenge, we present the improved factorized neural Transducer (IFNT) model structure designed to comprehensively integrate acoustic and language information while enabling effective text adaptation. We assess the performance of our proposed method on English and Mandarin datasets. The results indicate that IFNT not only surpasses the neural Transducer and FNT in baseline performance in both scenarios but also exhibits superior adaptation ability compared to FNT. On source domains, IFNT demonstrated statistically significant accuracy improvements, achieving a relative enhancement of 1.2% to 2.8% in baseline accuracy compared to the neural Transducer. On out-of-domain datasets, IFNT shows relative WER(CER) improvements of up to 30.2% over the standard neural Transducer with shallow fusion, and relative WER(CER) reductions ranging from 1.1% to 2.8% on test sets compared to the FNT model.

cs.CL↗

arXiv2023

Testing the validity of the surface approximation for reactions induced by weakly bound nuclei with a fully quantum-mechanical model

We examine the validity of surface approximation for breakup reactions using a fully quantum-mechanical model proposed by Ichimura, Austern, and Vincent (IAV). Analogous to the semi-classical picture, we introduce radial cut-offs to scattering waves in the IAV framework, which we refer to as IAV-cut. Systematic calculations are conducted for nonelastic breakup reactions induced by $^6$Li and deuterons at various incident energies. A comparison between the results obtained from IAV and IAV-cut is performed. The excellent agreement observed between IAV and IAV-cut in $^{6}$Li induced reactions, regardless of incident energy and target nuclei, signifies their insensitivity to the inner part of the scattering wave function, thus providing validation for the semi-classical picture. However, for deuteron induced breakup reactions, the IAV-cut results exhibit a suppression in the cross section, suggesting a strong dependence on the interior wave functions. This suppression is further enhanced as the incident energy increases.

nucl-th↗