Skip to content
Log in
Sponsored Artificial Intelligence: A Modern Approach, Global Edition Artificial Intelligence: A Modern Approach, Global Edition

Author

Yang Liu

View on Semantic Scholar

2 papers indexed here

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Open access Jul 2026

Design and Implementation of an OCaml-Based Standalone SystemVerilog Preprocessor Compliant with IEEE 1800-2023

SystemVerilog preprocessing determines the effective source code consumed by subsequent front-end compilation stages, yet in many toolchains it is implemented as an internal component tightly coupled with lexical, syntactic, or downstream analysis. Such integration is effective for end-to-end compilation flows, but it can obscure preprocessing behavior and make independent validation difficult, especially for macro expansion, conditional compilation, file inclusion, and source-location tracking. This paper presents Stagira svpp, a standalone SystemVerilog preprocessor implemented in OCaml and designed to follow the preprocessing semantics of IEEE Std 1800–2023. Preprocessing is modeled as an independent, directive-driven frontend stage, in which macro definitions, conditional-compilation contexts, file-inclusion hierarchies, source-location mappings, diagnostics, and auxiliary directive states are maintained through explicit state updates. The implementation supports object-like and function-like macros, token-level macro operators, nested conditional compilation, macro-driven file inclusion with cycle detection, predefined macros, source-location remapping, command-line macro/include configuration, and selected stateful directives. The evaluation uses a focused semantic test suite to exercise implemented preprocessing mechanisms and representative boundary cases. Within the evaluated scope, Stagira svpp passes all 48 test cases and exhibits consistent behavior for tested scenarios including nested conditionals, token-level macro operations, macro-expanded include arguments, cyclic inclusion detection, and source location-related diagnostics. A representative comparison with Icarus Verilog and Quartus II further illustrates observable behavioral differences in selected include-related scenarios. The current evaluation focuses on semantic correctness and boundary behavior rather than large-scale performance benchmarking, which has been identified as a direction for future work. By decoupling preprocessing from later frontend phases and making preprocessing state explicit, Stagira svpp provides a reusable reference implementation for studying and validating SystemVerilog preprocessing behavior.

Hao Deng, Zhiqiang He, Yang Liu et al. · 0 citations
Open access Jul 2026

Decoupled Semantic Flow Matching with Dynamic Reconstruction for Generalized Zero-Shot Learning

While generative models have become a standard approach for addressing the semantic-to-visual gap in Generalized Zero-Shot Learning (GZSL), existing architectures often struggle with two persistent limitations: cross-modal interference during condition fusion and severe overfitting to the visual distributions of seen classes. To address these bottlenecks, this paper introduces SemanticFlowNet, a framework based on Decoupled Semantic Flow Matching. Specifically, we propose a Decoupled Multi-modal Conditioning mechanism that relies on channel-wise concatenation of temporal encodings, semantic attributes, and visual contexts, which preserves the orthogonal subspaces of each modality and reduces interference. Additionally, we integrate a Dropout-enhanced Adaptive Layer Normalization (AdaLN) module to perturb the rigid memorization of seen classes, utilizing stochastic dropout within the state evolution to simulate the distributional variance of unseen domains. Finally, a Time-Aware Dynamic Reconstruction Penalty is introduced to enforce progressively stricter semantic alignment as the generative ordinary differential equation (ODE) trajectory converges to the target manifold. Evaluations on the CUB, SUN, and AWA2 benchmarks demonstrate the effectiveness of the proposed framework. Notably, SemanticFlowNet achieves a harmonic mean of 77.90% on the CUB dataset in the single-seed full-model setting, providing a competitive baseline for generative GZSL applications.

Chu-Yang Song, Mingyi Song, Yang Liu et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.