arXiv subjects

Yiwei Xie

Publications and source records attributed to Yiwei Xie.

5 recordsLinked to original sources

arXiv2026

Where Concept Erasure Should Occur: Concept-Layer Alignment in Text-to-Video Diffusion Models

Text-to-video diffusion transformers encode semantic information unevenly across model depth, which constrains effective concept erasure. We identify a representational bottleneck, termed concept-layer topological alignment, under which target concepts exhibit higher separability at certain representational depths. Outside these depths, concept and non-target signals remain strongly entangled, limiting the effectiveness of depth-specific erasure. This observation reframes concept erasure as the problem of identifying representational depths where concept-non-target separation naturally emerges. Motivated by this structural constraint, we introduce CLEAR, a separability-driven optimization framework for concept erasure that explicitly enforces concept-layer alignment. CLEAR operationalizes this principle by formulating layer selection as an optimization problem over concept-non-target separability, rather than relying on layer-agnostic or heuristic choices. To enable this, we introduce a separability-aware objective that favors layers exhibiting stronger concept-non-target separation. Experiments on large-scale text-to-video diffusion models demonstrate that enforcing concept--layer alignment leads to more precise concept suppression while preserving overall generative quality.

cs.CV↗

arXiv2026

PROBE: Diagnosing Residual Concept Capacity in Erased Text-to-Video Diffusion Models

Concept erasure techniques for text-to-video (T2V) diffusion models report substantial suppression of sensitive content, yet current evaluation is limited to checking whether the target concept is absent from generated frames, treating output-level suppression as evidence of representational removal. We introduce PROBE, a diagnostic protocol that quantifies the \textit{reactivation potential} of erased concepts in T2V models. With all model parameters frozen, PROBE optimizes a lightweight pseudo-token embedding through a denoising reconstruction objective combined with a novel latent alignment constraint that anchors recovery to the spatiotemporal structure of the original concept. We make three contributions: (1) a multi-level evaluation framework spanning classifier-based detection, semantic similarity, temporal reactivation analysis, and human validation; (2) systematic experiments across three T2V architectures, three concept categories, and three erasure strategies revealing that all tested methods leave measurable residual capacity whose robustness correlates with intervention depth; and (3) the identification of temporal re-emergence, a video-specific failure mode where suppressed concepts progressively resurface across frames, invisible to frame-level metrics. These findings suggest that current erasure methods achieve output-level suppression rather than representational removal. We release our protocol to support reproducible safety auditing. Our code is available at https://github.com/YiweiXie/PRObingBasedEvaluation.

cs.CV↗

arXiv2026

Novel High-Scalability Architecture for Photonic Deep Learning

Photonic computing promises ultrafast and energy-efficient artificial intelligence. However, existing photonic neural networks (PNNs) remain functionally shallow and difficult to scale. Here we establish a theory-guided framework showing that power stability and complex-field correlation are the fundamental prerequisites for scalable, coherent PNNs. Building on these macroscopic principles, we introduce the Coherent, Compensated and Cross-connected (C3) unit - an architecture that integrates coherent nonlinearity, active loss compensation and native optical residual connectivity. Implemented on a silicon-on-insulator platform, the C3 unit provides reconfigurable activation functions and dynamic energy stabilization without external amplification. We validate this framework using a width-constrained spiral benchmark, in which the C3 unit substantially improves parameter utilization and power robustness relative to incoherent nonlinearities. In a high-complexity 1,623-class recognition task, our C3-enabled coherent residual network (CoP-ResNet) achieves a top-1 accuracy of 77.92%, whereas non-residual architectures fail to converge. Together, these results offer a physically grounded, theory-guided pathway toward greater optical processing depth, laying the foundation for next-generation, large-scale photonic computing architectures.

physics.optics↗

arXiv2025

Erasing Concepts, Steering Generations: A Comprehensive Survey of Concept Suppression

Text-to-Image (T2I) models have demonstrated impressive capabilities in generating high-quality and diverse visual content from natural language prompts. However, uncontrolled reproduction of sensitive, copyrighted, or harmful imagery poses serious ethical, legal, and safety challenges. To address these concerns, the concept erasure paradigm has emerged as a promising direction, enabling the selective removal of specific semantic concepts from generative models while preserving their overall utility. This survey provides a comprehensive overview and in-depth synthesis of concept erasure techniques in T2I diffusion models. We systematically categorize existing approaches along three key dimensions: intervention level, which identifies specific model components targeted for concept removal; optimization structure, referring to the algorithmic strategies employed to achieve suppression; and semantic scope, concerning the complexity and nature of the concepts addressed. This multi-dimensional taxonomy enables clear, structured comparisons across diverse methodologies, highlighting fundamental trade-offs between erasure specificity, generalization, and computational complexity. We further discuss current evaluation benchmarks, standardized metrics, and practical datasets, emphasizing gaps that limit comprehensive assessment, particularly regarding robustness and practical effectiveness. Finally, we outline major challenges and promising future directions, including disentanglement of concept representations, adaptive and incremental erasure strategies, adversarial robustness, and new generative architectures. This survey aims to guide researchers toward safer, more ethically aligned generative models, providing foundational knowledge and actionable recommendations to advance responsible development in generative AI.

cs.CV↗

arXiv2021

New-generation silicon photonics beyond the singlemode regime

The singlemode condition is one of the most important design rules for optical waveguides in guided-wave optics. The reason following the singlemode condition is that higher-order modes might be excited and thus introduce some undesired mode-mismatching loss as well as inter-mode crosstalk when light propagates along an optical waveguide beyond the singlemode regime. As a result, multimode photonic waveguides are usually not allowed. In this paper, we propose the concept of silicon photonics beyond the singlemode regime, developed with low-loss and low-crosstalk light propagation in multimode photonic waveguides with broadened silicon cores. In particular, silicon photonic waveguides with a broadened core region have shown an ultra-low-loss of ~0.1 dB/cm for the fundamental mode even without any special fabrication process. A micro-racetrack resonator fabricated with standard 220-nm-SOI MPW-foundry processes shows a record intrinsic Q-factor as high as 1.02*107 for the first time, corresponding to ultra-low waveguide propagation loss of only 0.065 dB/cm. A high-performance microwave photonic filter on silicon is then realized with an ultra-narrow 3-dB bandwidth of 20.6 MHz as well as a tuning range of ~20 GHz for the first time. An on-chip 100-cm-long delayline is also demonstrated by using the present broadened SOI photonic waveguides with compact Euler-curve bends, the measured propagation loss is ~0.14 dB/cm. The proposed concept of silicon photonics beyond the singlemode regime helps solve the issue of high propagation loss and also significantly reduces the random phase errors of light due to the random variations of waveguide dimensions. In particularity it enables silicon photonic devices with enhanced performances, which paves the way for new-generation silicon photonics realizing the large-scale photonic integration.

physics.optics↗