arXiv subjects

Dayang Li

Publications and source records attributed to Dayang Li.

5 recordsLinked to original sources

arXiv2026

Demonstration of a High-Q Subwavelength Dielectric Nanocylinder

The development of subwavelength dielectric cavities is essential for reducing the size of photonic devices and enabling dense optoelectronic integration. However, previouslysubwavelengthoptical cavities exhibit demonstrated Q-factors <400, limiting their applications. Here, we demonstrate a high-Q subwavelength nanocylinder by leveraging bound states in the continuum (BIC). We track BIC modes of different longitudinal orders while maintaining an ultrasmall footprint. We find that the Q-factor initially increases but then saturates at higher orders. By linking quasi-normal-mode perturbation theory with coupled-mode analysis, we reveal the physical origin of this saturation and identify an optimized dimension that balances performance with fabrication feasibility. By suspending this design in free space using nanobridges and optimizing the nanofabrication process, we experimentally realize an InP subwavelength nanocylinder with a measured Q-factor exceeding 1000. Compared with a lower-Q substrate-supported counterpart, the suspended high-Q BIC nanocylinder exhibits stronger scattering and photoluminescence signals. Our work provides a route to high-Q optical devices with ultrasmall footprints.

physics.optics↗

arXiv2026

Correct When Paired, Wrong When Split: Decoupling and Editing Modality-Specific Neurons in MLLMs

Although Knowledge Editing provides an efficient mechanism for updating the knowledge of Multimodal Large Language Models (MLLMs), we find that current paradigms still suffer from an important yet remain underexplored issue : editing decoupling failure, where entity-related knowledge can be updated when the model is triggered by multimodal inputs (text--image query pairs), however, it often reverts to outdated pre-edit facts when the paired inputs are split into unimodal ones. Our in-depth empirical analysis reveals that the entity knowledge in MLLMs is not stored as a unified representation, but is instead distributed across disentangled modality-specific pathways. As a result, updates biased toward multimodal queries fail to propagate effectively to unimodal circuits. To bridge this gap, we propose DECODE, which explicitly disentangles and localizes modality-specific neuron groups for targeted knowledge. Extensive experiments demonstrate that DECODE consistently achieves effective knowledge updates under different modality triggers, thereby mitigating editing decoupling failures.

cs.LG↗

arXiv2026

What's Left Unsaid? Detecting and Correcting Misleading Omissions in Multimodal News Previews

Even when factually correct, social-media news previews (image-headline pairs) can induce interpretation drift: by selectively omitting crucial context, they lead readers to form judgments that diverge from what the full article supports. This covert harm is subtler than explicit misinformation, yet remains underexplored. To address this gap, we develop a multi-stage pipeline that simulates preview-based and context-based understanding, enabling construction of the MM-Misleading benchmark. Using MM-Misleading, we systematically evaluate open-source LVLMs and uncover pronounced blind spots in omission-based misleadingness detection. We further propose OMGuard, which combines (1) Interpretation-Aware Fine-Tuning for misleadingness detection and (2) Rationale-Guided Misleading Content Correction, where explicit rationales guide headline rewriting to reduce misleading impressions. Experiments show that OMGuard lifts an 8B model's detection accuracy to the level of a 235B LVLM while delivering markedly stronger end-to-end correction. Further analysis shows that misleadingness usually arises from local narrative shifts, such as missing background, instead of global frame changes, and identifies image-driven cases where text-only correction fails, underscoring the need for visual interventions.

cs.CV↗

arXiv2021

Location qubits in a multi-quantum-dot system

A physical platform for nodes of the envisioned quantum internet is long-sought. Here we propose such a platform, along with a conceptually simple and experimentally uncomplicated quantum information processing scheme, realized in a system of multiple crystal-phase quantum dots. We introduce novel location qubits, describe a method to construct a universal set of all-optical quantum gates, and simulate their performance in realistic structures, including decoherence sources. Our results show that location qubits can maintain coherence 5 orders of magnitude longer than single-qubit operation time, and single-qubit gate errors do not exceed 0.01%. Our scheme paves a clear way towards constructing multi-qubit solid-state quantum systems with a built-in photonic interface, such as a multi-qubit quantum register -- a key building block of the forthcoming quantum internet.

quant-ph↗

arXiv2021

Experimental proposal to probe the extended Pauli principle

All matter is made up of fermions -- one of the fundamental type of particles in nature. Fermions follow the Pauli exclusion principle, stating that two or more identical fermions cannot occupy the same quantum state. Antisymmetry of the fermionic wavefunction, however, implies additional constraints on the natural occupation numbers. These constraints depend on the dimensionality and purity of the system and have so far not been explored experimentally. Here, we propose an experiment in a system of multiple quantum dots capable of producing the highly entangled fermionic states necessary to reach the regime, where these additional constraints become dominant and can be probed. The type and strength of the required multi-fermion entanglement provides barriers to reaching deep into this regime. Transcending these barriers thus serves as a testing ground for the capabilities of future fermionic quantum information processing as well as quantum computer architectures based on fermionic states.

quant-ph↗