arXiv subjects

Yuyang Liang

Publications and source records attributed to Yuyang Liang.

5 recordsLinked to original sources

arXiv2026

Quantum Sensing Beyond Exceptional Points via Hidden symmetry-protected vacuum-noise fixed point

Exceptional-point (EP) sensing has attracted considerable interest because of its anomalous response scaling. However, recent studies have shown that the enhanced response near an EP is inevitably accompanied by amplified quantum noise, fundamentally limiting the achievable signalto-noise ratio (SNR). Here, we propose a fundamentally different route toward non-Hermitian quantum sensing based on symmetry-protected noise suppression rather than response amplification. We develop a fully quantum continuous-variable model that unifies parity-time (PT) and anti-paritytime(APT) symmetries within a single framework. Exploiting the incompatibility between these two symmetries, we uncover a non-Hermitian Dirac eigenspectrum and reveal a hidden symmetryprotected phase transition embedded in the Hamiltonian spectrum. Remarkably, this hidden phasetransition simultaneously constitutes a symmetry-protected vacuum-noise fixed point, where collective three-mode quadratures exhibit suppressed quantum fluctuations despite the absence of any anomalous spectral response. As a consequence, quantum sensing is enhanced through the suppression of excess quantum noise while maintaining a finite response sensitivity, establishing a sensing mechanism fundamentally different from conventional EP-based approaches. These results reveal an unexpected connection between hidden symmetry, quantum fluctuations, and non-Hermitian quantum metrology, and establish noise suppression as an alternative paradigm for non-Hermitian quantum sensing.

quant-ph↗

arXiv2026

Understanding and Guiding Layer Placement in Parameter-Efficient Fine-Tuning of Large Language Models

As large language models (LLMs) continue to grow, the cost of full-parameter fine-tuning has made parameter-efficient fine-tuning (PEFT) the default strategy for downstream adaptation. Constraints from inference latency in scalable serving and fine-tuning cost in edge or rapid-deployment settings make the choice of which layers to fine-tune unavoidable. Yet current practice typically applies PEFT uniformly across all layers, with limited understanding or leverage of layer selection. This paper develops a unified projected residual view of PEFT on top of a frozen base model. Under a local quadratic approximation, layerwise adaptation is governed by three quantities: (i) the projected residual norm (resnorm), which measures how much correctable bias a layer can capture; (ii) the activation energy, which determines feature conditioning; and (iii) layer coupling, which quantifies how strongly residuals interact across layers. We show that, for squared loss and linear adapters, the resnorm equals a normalized gradient norm, activation energy controls ill-conditioning and noise amplification, and weak coupling yields approximately additive layerwise contributions. Building on these insights, we introduce the Layer Card, a reusable diagnostic that summarizes residual signal strength, compute cost, and performance for each layer of a given model. With an identical model and LoRA configuration, Layer Card-guided placement refines the choice of adapted layers to flexibly prioritize different objectives, such as maximizing performance or reducing fine-tuning cost. Moreover, on Qwen3-8B, we show that selectively adapting a subset of layers can achieve performance close to full-layer LoRA while substantially reducing fine-tuning cost and the number of adapter-augmented layers during inference, offering a more cost-performance-aware alternative to full-layer insertion.

cs.LG↗

arXiv2025

Benchmarking Multimodal Knowledge Conflict for Large Multimodal Models

Large Multimodal Models(LMMs) face notable challenges when encountering multimodal knowledge conflicts, particularly under retrieval-augmented generation(RAG) frameworks where the contextual information from external sources may contradict the model's internal parametric knowledge, leading to unreliable outputs. However, existing benchmarks fail to reflect such realistic conflict scenarios. Most focus solely on intra-memory conflicts, while context-memory and inter-context conflicts remain largely investigated. Furthermore, commonly used factual knowledge-based evaluations are often overlooked, and existing datasets lack a thorough investigation into conflict detection capabilities. To bridge this gap, we propose MMKC-Bench, a benchmark designed to evaluate factual knowledge conflicts in both context-memory and inter-context scenarios. MMKC-Bench encompasses three types of multimodal knowledge conflicts and includes 1,573 knowledge instances and 3,381 images across 23 broad types, collected through automated pipelines with human verification. We evaluate three representative series of LMMs on both model behavior analysis and conflict detection tasks. Our findings show that while current LMMs are capable of recognizing knowledge conflicts, they tend to favor internal parametric knowledge over external evidence. We hope MMKC-Bench will foster further research in multimodal knowledge conflict and enhance the development of multimodal RAG systems. The source code is available at https://github.com/MLLMKCBENCH/MLLMKC.

cs.LG↗

arXiv2025

TRACE: Intra-visit Clinical Event Nowcasting via Effective Patient Trajectory Encoding

Electronic Health Records (EHR) have become a valuable resource for a wide range of predictive tasks in healthcare. However, existing approaches have largely focused on inter-visit event predictions, overlooking the importance of intra-visit nowcasting, which provides prompt clinical insights during an ongoing patient visit. To address this gap, we introduce the task of laboratory measurement prediction within a hospital visit. We study the laboratory data that, however, remained underexplored in previous work. We propose TRACE, a Transformer-based model designed for clinical event nowcasting by encoding patient trajectories. TRACE effectively handles long sequences and captures temporal dependencies through a novel timestamp embedding that integrates decay properties and periodic patterns of data. Additionally, we introduce a smoothed mask for denoising, improving the robustness of the model. Experiments on two large-scale electronic health record datasets demonstrate that the proposed model significantly outperforms previous methods, highlighting its potential for improving patient care through more accurate laboratory measurement nowcasting. The code is available at https://github.com/Amehi/TRACE.

cs.LG↗

arXiv2024

Squeezing and Entanglement Dynamics in Phase-Sensitive Non-Hermitian Systems

Over the past decade, parity-time (PT) symmetry and anti-PT (APT) symmetry in various physical systems have been extensively studied, leading to significant experimental and theoretical advancements. However, physical systems that simultaneously exhibit both PT and APT symmetry have not yet been explored. In this study, we construct a phase-sensitive non-Hermitian wave mixing model that inherently possesses single-mode APT symmetry. By tuning the phase of the pump field, this model simultaneously exhibits two-mode quadrature-PT symmetry. Moreover, the APT phase transition is accompanied by the emergence of quantum entanglement from absence to presence. This remarkable quantum effect, quantum entanglement, which distinguishes itself from classical physics, is surprisingly linked to APT symmetry phase transition. Our work further explores the relationship between two-mode quantum entanglement and the phase of the pump field, offering deeper insight into the generation and evolution of entanglement in the corresponding nonlinear system. This provides a new perspective on quantum information processing.

quant-ph↗