arXiv subjects

Zhiwen Li

Publications and source records attributed to Zhiwen Li.

At least 19 recordsLinked to original sources

arXiv2026

Comprehensive Assessment and Analysis for NSFW Content Erasure in Text-to-Image Diffusion Models

Text-to-image diffusion models have gained widespread application across various domains, demonstrating remarkable creative potential. However, the strong generalization capabilities of diffusion models can inadvertently lead to the generation of not-safe-for-work (NSFW) content, posing significant risks to their safe deployment. While several concept erasure methods have been proposed to mitigate the issue associated with NSFW content, a comprehensive evaluation of their effectiveness across various scenarios remains absent. To bridge this gap, we introduce a full-pipeline toolkit specifically designed for concept erasure and conduct the first systematic study of NSFW concept erasure methods. By examining the interplay between the underlying mechanisms and empirical observations, we provide in-depth insights and practical guidance for the effective application of concept erasure methods in various real-world scenarios, with the aim of advancing the understanding of content safety in diffusion models and establishing a solid foundation for future research and development in this critical area.

cs.CV↗

arXiv2026

Comprehensive Evaluation and Analysis for NSFW Concept Erasure in Text-to-Image Diffusion Models

cs.CV↗

arXiv2026

Identifying Red Supergiants in the Local Group Using JWST Photometry. I. NGC 6822, Sextans A, NGC 300, WLM, and IC 1613

Red supergiants (RSGs) are crucial for studying the properties and evolution of massive stars. It is representative to conduct a census of RSGs across the Local Group, which spans a broad metallicity range. However, identifying RSGs in distant and metal-poor galaxies remains challenging mainly due to contamination of foreground dwarfs and observational limitations. In this work, we perform PSF photometry on publicly released JWST/NIRCam images of five Local Group galaxies: NGC 6822, Sextans~A, NGC 300, WLM, and IC 1613 using the DOLPHOT NIRCam module. We find an optimal color-color diagram (CCD) for metal-poor environments, that is F115W $-$ F200W versus F356W $-$ F444W, which clearly separates RSGs from foreground dwarfs. By using the CCD, we identify 208, 135, and 22 RSG candidates in NGC 6822, Sextans A, and NGC 300, respectively, free from contamination by foreground dwarfs and oxygen-rich asymptotic giant branch stars (O-AGBs). In addition, 40 and 14 RSG candidates are directly selected on the CMD in WLM and IC 1613, respectively. Compared with previous works, the number of RSG candidates within the same luminosity range and sky region increases significantly, demonstrating the advantages of JWST in constructing a more complete RSG sample in the Local Group thanks to its high spatial resolution and photometric quality. In addition, catalogs of O-AGBs and carbon-rich AGBs (C-AGBs) are provided as by-products.

astro-ph.GA↗

arXiv2026

Spectral Evolution Search: Efficient Inference-Time Scaling for Reward-Aligned Image Generation

Inference-time scaling offers a versatile paradigm for aligning visual generative models with downstream objectives without parameter updates. However, existing approaches that optimize the high-dimensional initial noise suffer from severe inefficiency, as many search directions exert negligible influence on the final generation. We show that this inefficiency is closely related to a spectral bias in generative dynamics: model sensitivity to initial perturbations diminishes rapidly as frequency increases. Building on this insight, we propose Spectral Evolution Search (SES), a plug-and-play framework for initial noise optimization that executes gradient-free evolutionary search within a low-frequency subspace. Theoretically, we derive the Spectral Scaling Prediction from perturbation propagation dynamics, which explains the systematic differences in the impact of perturbations across frequencies. Extensive experiments demonstrate that SES significantly advances the Pareto frontier of generation quality versus computational cost, consistently outperforming strong baselines under equivalent budgets.

cs.LG↗

arXiv2026

VIRAL: Visual In-Context Reasoning via Analogy in Diffusion Transformers

Replicating In-Context Learning (ICL) in computer vision remains challenging due to task heterogeneity. We propose \textbf{VIRAL}, a framework that elicits visual reasoning from a pre-trained image editing model by formulating ICL as conditional generation via visual analogy ($x_s : x_t :: x_q : y_q$). We adapt a frozen Diffusion Transformer (DiT) using role-aware multi-image conditioning and introduce a Mixture-of-Experts LoRA to mitigate gradient interference across diverse tasks. Additionally, to bridge the gaps in current visual context datasets, we curate a large-scale dataset spanning perception, restoration, and editing. Experiments demonstrate that VIRAL outperforms existing methods, validating that a unified V-ICL paradigm can handle the majority of visual tasks, including open-domain editing. Our code is available at https://anonymous.4open.science/r/VIRAL-744A

cs.CV↗

arXiv2026

AttriCtrl: Fine-Grained Control of Aesthetic Attribute Intensity in Diffusion Models

Diffusion models have recently become the dominant paradigm for image generation, yet existing systems struggle to interpret and follow numeric instructions for adjusting semantic attributes. In real-world creative scenarios, especially when precise control over aesthetic attributes is required, current methods fail to provide such controllability. This limitation partly arises from the subjective and context-dependent nature of aesthetic judgments, but more fundamentally stems from the fact that current text encoders are designed for discrete tokens rather than continuous values. Meanwhile, efforts on aesthetic alignment, often leveraging reinforcement learning, direct preference optimization, or architectural modifications, primarily align models with a global notion of human preference. While these approaches improve user experience, they overlook the multifaceted and compositional nature of aesthetics, underscoring the need for explicit disentanglement and independent control of aesthetic attributes. To address this gap, we introduce AttriCtrl, a lightweight framework for continuous aesthetic intensity control in diffusion models. It first defines relevant aesthetic attributes, then quantifies them through a hybrid strategy that maps both concrete and abstract dimensions onto a unified $[0,1]$ scale. A plug-and-play value encoder is then used to transform user-specified values into model-interpretable embeddings for controllable generation. Experiments show that AttriCtrl achieves accurate and continuous control over both single and multiple aesthetic attributes, significantly enhancing personalization and diversity. Crucially, it is implemented as a lightweight adapter while keeping the diffusion model frozen, ensuring seamless integration with existing frameworks such as ControlNet at negligible computational cost.

cs.CV↗

arXiv2025

Variational Geometry-aware Neural Network based Method for Solving High-dimensional Diffeomorphic Mapping Problems

Traditional methods for high-dimensional diffeomorphic mapping often struggle with the curse of dimensionality. We propose a mesh-free learning framework designed for $n$-dimensional mapping problems, seamlessly combining variational principles with quasi-conformal theory. Our approach ensures accurate, bijective mappings by regulating conformality distortion and volume distortion, enabling robust control over deformation quality. The framework is inherently compatible with gradient-based optimization and neural network architectures, making it highly flexible and scalable to higher-dimensional settings. Numerical experiments on both synthetic and real-world medical image data validate the accuracy, robustness, and effectiveness of the proposed method in complex registration scenarios.

cs.LG↗

arXiv2025

AutoLoRA: Automatic LoRA Retrieval and Fine-Grained Gated Fusion for Text-to-Image Generation

Despite recent advances in photorealistic image generation through large-scale models like FLUX and Stable Diffusion v3, the practical deployment of these architectures remains constrained by their inherent intractability to parameter fine-tuning. While low-rank adaptation (LoRA) have demonstrated efficacy in enabling model customization with minimal parameter overhead, the effective utilization of distributed open-source LoRA modules faces three critical challenges: sparse metadata annotation, the requirement for zero-shot adaptation capabilities, and suboptimal fusion strategies for multi-LoRA fusion strategies. To address these limitations, we introduce a novel framework that enables semantic-driven LoRA retrieval and dynamic aggregation through two key components: (1) weight encoding-base LoRA retriever that establishes a shared semantic space between LoRA parameter matrices and text prompts, eliminating dependence on original training data, and (2) fine-grained gated fusion mechanism that computes context-specific fusion weights across network layers and diffusion timesteps to optimally integrate multiple LoRA modules during generation. Our approach achieves significant improvement in image generation perfermance, thereby facilitating scalable and data-efficient enhancement of foundational models. This work establishes a critical bridge between the fragmented landscape of community-developed LoRAs and practical deployment requirements, enabling collaborative model evolution through standardized adapter integration.

cs.CV↗

arXiv2025

Responsible Diffusion Models via Constraining Text Embeddings within Safe Regions

The remarkable ability of diffusion models to generate high-fidelity images has led to their widespread adoption. However, concerns have also arisen regarding their potential to produce Not Safe for Work (NSFW) content and exhibit social biases, hindering their practical use in real-world applications. In response to this challenge, prior work has focused on employing security filters to identify and exclude toxic text, or alternatively, fine-tuning pre-trained diffusion models to erase sensitive concepts. Unfortunately, existing methods struggle to achieve satisfactory performance in the sense that they can have a significant impact on the normal model output while still failing to prevent the generation of harmful content in some cases. In this paper, we propose a novel self-discovery approach to identifying a semantic direction vector in the embedding space to restrict text embedding within a safe region. Our method circumvents the need for correcting individual words within the input text and steers the entire text prompt towards a safe region in the embedding space, thereby enhancing model robustness against all possibly unsafe prompts. In addition, we employ Low-Rank Adaptation (LoRA) for semantic direction vector initialization to reduce the impact on the model performance for other semantics. Furthermore, our method can also be integrated with existing methods to improve their social responsibility. Extensive experiments on benchmark datasets demonstrate that our method can effectively reduce NSFW content and mitigate social bias generated by diffusion models compared to several state-of-the-art baselines.

cs.CL↗

arXiv2025

Asteroseismology of Long-Period Variables with OGLE-IV data: Using Global Seismic Parameters as Luminosity Indicators

Long-period variables (LPVs) are high-luminosity red giants or supergiants with pulsation periods ranging from days to years. Many LPVs in the Large Magellanic Cloud (LMC) and Galactic Bulge (BLG) have been continuously observed over a time span of 26 years by the Optical Gravitational Lensing Experiment (OGLE) survey. Using OGLE-IV data, we applied Gaussian Processes with kernels tailored for solar-like oscillations to extract two global asteroseismic parameters: the frequency of maximum power (numax) and the large frequency separation (Dnu), for LPVs with primary mode periods (P1) between 10 and 100 days in the LMC and BLG. We found that the numax-Dnu relation for LPVs in this work aligns with that of lower-luminosity Kepler red giants, confirming that the pulsations of these LPVs are likely solar-like. We found that numax and Dnu can serve as luminosity indicators. Compared to P1, numax and Dnu exhibit significantly tighter correlations with the absolute magnitude in the 2MASS K_s band (M_{K}), with corresponding scatter of 0.27 mag and 0.21 mag, respectively. Using the calibrated numax-mk and Dnu-mk relations for LPVs in the LMC, we determined the M_{K} values for individual stars in the BLG. By accounting for extinction, we further calculated the distances to 4,948 BLG stars. The peak of the resulting distance distribution corresponds to an estimated distance to the Galactic center of approximately 9.1 kpc, which appears to be overestimated, suggesting that the seismic luminosity relation calibrated from the LMC may not be directly applicable to BLG stars.

astro-ph.SR↗

arXiv2025

Growth Inhibitors for Suppressing Inappropriate Image Concepts in Diffusion Models

Despite their remarkable image generation capabilities, text-to-image diffusion models inadvertently learn inappropriate concepts from vast and unfiltered training data, which leads to various ethical and business risks. Specifically, model-generated images may exhibit not safe for work (NSFW) content and style copyright infringements. The prompts that result in these problems often do not include explicit unsafe words; instead, they contain obscure and associative terms, which are referred to as implicit unsafe prompts. Existing approaches directly fine-tune models under textual guidance to alter the cognition of the diffusion model, thereby erasing inappropriate concepts. This not only requires concept-specific fine-tuning but may also incur catastrophic forgetting. To address these issues, we explore the representation of inappropriate concepts in the image space and guide them towards more suitable ones by injecting growth inhibitors, which are tailored based on the identified features related to inappropriate concepts during the diffusion process. Additionally, due to the varying degrees and scopes of inappropriate concepts, we train an adapter to infer the corresponding suppression scale during the injection process. Our method effectively captures the manifestation of subtle words at the image level, enabling direct and efficient erasure of target concepts without the need for fine-tuning. Through extensive experimentation, we demonstrate that our approach achieves superior erasure results with little effect on other concepts while preserving image quality and semantics.

cs.CV↗

arXiv2024

A New Approach to Identifying Red Supergiant Stars in Metal-poor Galaxies: A Case Study of NGC 6822

A complete sample of red supergiant stars (RSGs) is important for studying their properties. Identifying RSGs in extragalatic field first requires removing the Galactic foreground dwarfs. The color-color diagram (CCD) method, specifically using $r-z/z-H$ and $J-H/H-K$, has proven successful in several studies. However, in metal-poor galaxies, faint RSGs will mix into the dwarf branch in the CCD and would be removed, leading to an incomplete RSG sample. This work attempts to improve the CCD method in combination with the Gaia astrometric measurement to remove foreground contamination in order to construct a complete RSG sample in metal-poor galaxies. The empirical regions of RSGs in both CCDs are defined and modified by fitting the locations of RSGs in galaxies with a range of metallicity. The metal-poor galaxy NGC 6822 is taken as a case study for its low metallicity ([Fe/H] $\approx$ -1.0) and moderate distance (about 500 kpc). In the complete sample, we identify 1,184 RSG, 1,559 oxygen-rich AGB (O-AGBs), 1,075 carbon-rich AGB (C-AGBs), and 140 extreme AGB (x-AGBs) candidates, with a contamination rate of approximately 20.5%, 9.7%, 6.8%, and 5.0%, respectively. We also present a pure sample, containing only the sources away from the dwarf branch, which includes 843 RSG, 1,519 O-AGB, 1,059 C-AGB, and 140 x-AGB candidates, with a contamination rate of approximately 6.5%, 8.8%, 6.1%, and 5.0%, respectively. About 600 and 450 RSG candidates are newly identified in the complete and pure sample, respectively, compared to the previous RSG sample in NGC 6822.

astro-ph.GA↗

arXiv2023

Superconductivity above 30 K achieved in dense scandium

Superconductivity is one of most intriguing quantum phenomena, and the quest for elemental superconductors with high critical temperature (Tc) is of great scientific significance due to their relatively simple material composition and the underlying mechanism. Here we report the experimental discovery of densely compressed scandium (Sc) becoming the first elemental superconductor with Tc breaking into 30 K range, which is comparable to the Tc values of the classic La-Ba-Cu-O or LaFeAsO superconductors. Our results show that Tconset of Sc increases from ~3 K at around 43 GPa to ~32 K at about 283 GPa (Tczero ~ 31 K), which is well above liquid neon temperature. Interestingly measured Tc shows no sign of saturation up to the maximum pressure achieved in our experiments, indicating that Tc might be even higher upon further compression.

cond-mat.supr-con↗

arXiv2023

Experimental 3D super-localization with Laguerre-Gaussian modes

Improving three-dimensional (3D) localization precision is of paramount importance for super-resolution imaging. By properly engineering the point spread function (PSF), such as utilizing Laguerre-Gaussian (LG) modes and their superposition, the ultimate limits of 3D localization precision can be enhanced. However, achieving these limits is challenging, as it often involves complicated detection strategies and practical limitations. In this work, we rigorously derive the ultimate 3D localization limits of LG modes and their superposition, specifically rotation modes, in the multi-parameter estimation framework. Our findings reveal that a significant portion of the information required for achieving 3D super-localization of LG modes can be obtained through feasible intensity detection. Moreover, the 3D ultimate precision can be achieved when the azimuthal index $l$ is zero. To provide a proof-of-principle demonstration, we develop an iterative maximum likelihood estimation (MLE) algorithm that converges to the 3D position of a point source, considering the pixelation and detector noise. The experimental implementation exhibits an improvement of up to two-fold in lateral localization precision and up to twenty-fold in axial localization precision when using LG modes compared to Gaussian mode. We also showcase the superior axial localization capability of the rotation mode within the near-focus region, effectively overcoming the limitations encountered by single LG modes. Notably, in the presence of realistic aberration, the algorithm robustly achieves the Cramér-Rao lower bound. Our findings provide valuable insights for evaluating and optimizing the achievable 3D localization precision, which will facilitate the advancements in super-resolution microscopy.

quant-ph↗

arXiv2023

XFormer: Fast and Accurate Monocular 3D Body Capture

We present XFormer, a novel human mesh and motion capture method that achieves real-time performance on consumer CPUs given only monocular images as input. The proposed network architecture contains two branches: a keypoint branch that estimates 3D human mesh vertices given 2D keypoints, and an image branch that makes predictions directly from the RGB image features. At the core of our method is a cross-modal transformer block that allows information to flow across these two branches by modeling the attention between 2D keypoint coordinates and image spatial features. Our architecture is smartly designed, which enables us to train on various types of datasets including images with 2D/3D annotations, images with 3D pseudo labels, and motion capture datasets that do not have associated images. This effectively improves the accuracy and generalization ability of our system. Built on a lightweight backbone (MobileNetV3), our method runs blazing fast (over 30fps on a single CPU core) and still yields competitive accuracy. Furthermore, with an HRNet backbone, XFormer delivers state-of-the-art performance on Huamn3.6 and 3DPW datasets.

cs.CV↗

arXiv2023

Superconductivity above 70 K observed in lutetium polyhydrides

The binary polyhydrides of heavy rare earth lutetium that shares a similar valence electron configuration to lanthanum have been experimentally discovered to be superconductive. The lutetium polyhydrides were successfully synthesized at high pressure and high temperature conditions using a diamond anvil cell in combinations with the in-situ high pressure laser heating technique. The resistance measurements as a function of temperature were performed at the same pressure of synthesis in order to study the transitions of superconductivity (SC). The superconducting transition with a maximum onset temperature (Tc) 71 K was observed at pressure of 218 GPa in the experiments. The Tc decreased to 65 K when pressure was at 181 GPa. From the evolution of SC at applied magnetic fields, the upper critical field at zero temperature μ0Hc2(0) was obtained to be ~36 Tesla. The in-situ high pressure X-ray diffraction experiments imply that the high Tc SC should arise from the Lu4H23 phase with Pm-3n symmetry that forms a new type of hydrogen cage framework different from those reported for previous light rare earth polyhydride superconductors.

cond-mat.supr-con↗

arXiv2022

A Deep Learning Framework for Diffeomorphic Mapping Problems via Quasi-conformal Geometry applied to Imaging

Many imaging problems can be formulated as mapping problems. A general mapping problem aims to obtain an optimal mapping that minimizes an energy functional subject to the given constraints. Existing methods to solve the mapping problems are often inefficient and can sometimes get trapped in local minima. An extra challenge arises when the optimal mapping is required to be diffeomorphic. In this work, we address the problem by proposing a deep-learning framework based on the Quasiconformal (QC) Teichmuller theories. The main strategy is to learn the Beltrami coefficient (BC) that represents a mapping as the latent feature vector in the deep neural network. The BC measures the local geometric distortion under the mapping, with which the interpretability of the deep neural network can be enhanced. Under this framework, the diffeomorphic property of the mapping can be controlled via a simple activation function within the network. The optimal mapping can also be easily regularized by integrating the BC into the loss function. A crucial advantage of the proposed framework is that once the network is successfully trained, the optimized mapping corresponding to each input data information can be obtained in real time. To examine the efficacy of the proposed framework, we apply the method to the diffeomorphic image registration problem. Experimental results outperform other state-of-the-art registration algorithms in both efficiency and accuracy, which demonstrate the effectiveness of our proposed framework to solve the mapping problem.

cs.CV↗

arXiv2022

Learning robot inverse dynamics using sparse online Gaussian process with forgetting mechanism

Online Gaussian processes (GPs), typically used for learning models from time-series data, are more flexible and robust than offline GPs. Both local and sparse approximations of GPs can efficiently learn complex models online. Yet, these approaches assume that all signals are relatively accurate and that all data are available for learning without misleading data. Besides, the online learning capacity of GPs is limited for high-dimension problems and long-term tasks in practice. This paper proposes a sparse online GP (SOGP) with a forgetting mechanism to forget distant model information at a specific rate. The proposed approach combines two general data deletion schemes for the basis vector set of SOGP: The position information-based scheme and the oldest points-based scheme. We apply our approach to learn the inverse dynamics of a collaborative robot with 7 degrees of freedom under a two-segment trajectory tracking problem with task switching. Both simulations and experiments have shown that the proposed approach achieves better tracking accuracy and predictive smoothness compared with the two general data deletion schemes.

cs.RO↗