arXiv subjects

Zihan Zhou

Publications and source records attributed to Zihan Zhou.

At least 91 records · Page 5Linked to original sources

arXiv2025

Advancing Reliable Test-Time Adaptation of Vision-Language Models under Visual Variations

Vision-language models (VLMs) exhibit remarkable zero-shot capabilities but struggle with distribution shifts in downstream tasks when labeled data is unavailable, which has motivated the development of Test-Time Adaptation (TTA) to improve VLMs' performance during inference without annotations. Among various TTA approaches, cache-based methods show promise by preserving historical knowledge from low-entropy samples in a dynamic cache and fostering efficient adaptation. However, these methods face two critical reliability challenges: (1) entropy often becomes unreliable under distribution shifts, causing error accumulation in the cache and degradation in adaptation performance; (2) the final predictions may be unreliable due to inflexible decision boundaries that fail to accommodate large downstream shifts. To address these challenges, we propose a Reliable Test-time Adaptation (ReTA) method that integrates two complementary strategies to enhance reliability from two perspectives. First, to mitigate the unreliability of entropy as a sample selection criterion for cache construction, we introduce Consistency-aware Entropy Reweighting (CER), which incorporates consistency constraints to weight entropy during cache updating. While conventional approaches rely solely on low entropy for cache prioritization and risk introducing noise, our method leverages predictive consistency to maintain a high-quality cache and facilitate more robust adaptation. Second, we present Diversity-driven Distribution Calibration (DDC), which models class-wise text embeddings as multivariate Gaussian distributions, enabling adaptive decision boundaries for more accurate predictions across visually diverse content. Extensive experiments demonstrate that ReTA consistently outperforms state-of-the-art methods, particularly under real-world distribution shifts. Code: https://github.com/Evelyn1ywliang/ReTA.

cs.CV↗

arXiv2025

Achieving More with Less: Additive Prompt Tuning for Rehearsal-Free Class-Incremental Learning

Class-incremental learning (CIL) enables models to learn new classes progressively while preserving knowledge of previously learned ones. Recent advances in this field have shifted towards parameter-efficient fine-tuning techniques, with many approaches building upon the framework that maintains a pool of learnable prompts. Although effective, these methods introduce substantial computational overhead, primarily due to prompt pool querying and increased input sequence lengths from prompt concatenation. In this work, we present a novel prompt-based approach that addresses this limitation. Our method trains a single set of shared prompts across all tasks and, rather than concatenating prompts to the input, directly modifies the CLS token's attention computation by adding the prompts to it. This simple and lightweight design not only significantly reduces computational complexity-both in terms of inference costs and the number of trainable parameters-but also eliminates the need to optimize prompt lengths for different downstream tasks, offering a more efficient yet powerful solution for rehearsal-free class-incremental learning. Extensive experiments across a diverse range of CIL benchmarks demonstrate the effectiveness of our approach, highlighting its potential to establish a new prompt-based CIL paradigm. Furthermore, experiments on general recognition benchmarks beyond the CIL setting also show strong performance, positioning our method as a promising candidate for a general parameter-efficient fine-tuning approach.

cs.CV↗

arXiv2025

Neutralizing Token Aggregation via Information Augmentation for Efficient Test-Time Adaptation

Test-Time Adaptation (TTA) has emerged as an effective solution for adapting Vision Transformers (ViT) to distribution shifts without additional training data. However, existing TTA methods often incur substantial computational overhead, limiting their applicability in resource-constrained real-world scenarios. To reduce inference cost, plug-and-play token aggregation methods merge redundant tokens in ViTs to reduce total processed tokens. Albeit efficient, it suffers from significant performance degradation when directly integrated with existing TTA methods. We formalize this problem as Efficient Test-Time Adaptation (ETTA), seeking to preserve the adaptation capability of TTA while reducing inference latency. In this paper, we first provide a theoretical analysis from a novel mutual information perspective, showing that token aggregation inherently leads to information loss, which cannot be fully mitigated by conventional norm-tuning-based TTA methods. Guided by this insight, we propose to \textbf{N}eutralize Token \textbf{A}ggregation \textbf{v}ia \textbf{I}nformation \textbf{A}ugmentation (\textbf{NAVIA}). Specifically, we directly augment the [CLS] token embedding and incorporate adaptive biases into the [CLS] token in shallow layers of ViTs. We theoretically demonstrate that these augmentations, when optimized via entropy minimization, recover the information lost due to token aggregation. Extensive experiments across various out-of-distribution benchmarks demonstrate that NAVIA significantly outperforms state-of-the-art methods by over 2.5\%, while achieving an inference latency reduction of more than 20\%, effectively addressing the ETTA challenge.

cs.CV↗

arXiv2025

A Direct Reduction from Stochastic Parity Games to Simple Stochastic Games

Significant progress has been recently achieved in developing efficient solutions for simple stochastic games (SSGs), focusing on reachability objectives. While reductions from stochastic parity games (SPGs) to SSGs have been presented in the literature through the use of multiple intermediate game models, a direct and simple reduction has been notably absent. This paper introduces a novel and direct polynomial-time reduction from quantitative SPGs to quantitative SSGs. By leveraging a gadget-based transformation that effectively removes the priority function, we construct an SSG that simulates the behavior of a given SPG. We formally establish the correctness of our direct reduction. Furthermore, we demonstrate that under binary encoding this reduction is polynomial, thereby directly corroborating the known $\textbf{NP}\,\mathbf{\cap}\,\textbf{coNP}$ complexity of SPGs and providing new understanding in the relationship between parity and reachability objectives in turn-based stochastic games.

cs.GT↗

arXiv2025

Equilibrium Policy on Dividend and Capital Injection under Time-inconsistent Preferences

This paper studies the dividend and capital injection problem under a diffusion risk model with general discount functions. A proportional cost is imposed when injecting capitals. For exponential discounting as time-consistent benchmark, we obtain the closed-form solutions and show that the optimal strategies are of threshold type. Under general discount function which leads to time-inconsistency, we adopt the definition of weak equilibrium and obtain the extended HJB equation system. An explicit solution is derived under pseudo-exponential discounting where three cases of the dividend and capital injection thresholds are obtained. Numerical examples show that large capital injection cost may lead to no capital injection at all, while larger difference in group discount rate leads to higher equilibrium value function.

q-fin.MF↗

arXiv2025

Monocle: Hybrid Local-Global In-Context Evaluation for Long-Text Generation with Uncertainty-Based Active Learning

Assessing the quality of long-form, model-generated text is challenging, even with advanced LLM-as-a-Judge methods, due to performance degradation as input length increases. To address this issue, we propose a divide-and-conquer approach, which breaks down the comprehensive evaluation task into a series of localized scoring tasks, followed by a final global assessment. This strategy allows for more granular and manageable evaluations, ensuring that each segment of the text is assessed in isolation for both coherence and quality, while also accounting for the overall structure and consistency of the entire piece. Moreover, we introduce a hybrid in-context learning approach that leverages human annotations to enhance the performance of both local and global evaluations. By incorporating human-generated feedback directly into the evaluation process, this method allows the model to better align with human judgment. Finally, we develop an uncertainty-based active learning algorithm that efficiently selects data samples for human annotation, thereby reducing annotation costs in practical scenarios. Experimental results show that the proposed evaluation framework outperforms several representative baselines, highlighting the effectiveness of our approach.

cs.CL↗

arXiv2025

Leveraging Motion Information for Better Self-Supervised Video Correspondence Learning

Self-supervised video correspondence learning depends on the ability to accurately associate pixels between video frames that correspond to the same visual object. However, achieving reliable pixel matching without supervision remains a major challenge. To address this issue, recent research has focused on feature learning techniques that aim to encode unique pixel representations for matching. Despite these advances, existing methods still struggle to achieve exact pixel correspondences and often suffer from false matches, limiting their effectiveness in self-supervised settings. To this end, we explore an efficient self-supervised Video Correspondence Learning framework (MER) that aims to accurately extract object details from unlabeled videos. First, we design a dedicated Motion Enhancement Engine that emphasizes capturing the dynamic motion of objects in videos. In addition, we introduce a flexible sampling strategy for inter-pixel correspondence information (Multi-Cluster Sampler) that enables the model to pay more attention to the pixel changes of important objects in motion. Through experiments, our algorithm outperforms the state-of-the-art competitors on video correspondence learning tasks such as video object segmentation and video object keypoint tracking.

cs.CV↗

arXiv2025

Enhancing Self-Supervised Fine-Grained Video Object Tracking with Dynamic Memory Prediction

Successful video analysis relies on accurate recognition of pixels across frames, and frame reconstruction methods based on video correspondence learning are popular due to their efficiency. Existing frame reconstruction methods, while efficient, neglect the value of direct involvement of multiple reference frames for reconstruction and decision-making aspects, especially in complex situations such as occlusion or fast movement. In this paper, we introduce a Dynamic Memory Prediction (DMP) framework that innovatively utilizes multiple reference frames to concisely and directly enhance frame reconstruction. Its core component is a Reference Frame Memory Engine that dynamically selects frames based on object pixel features to improve tracking accuracy. In addition, a Bidirectional Target Prediction Network is built to utilize multiple reference frames to improve the robustness of the model. Through experiments, our algorithm outperforms the state-of-the-art self-supervised techniques on two fine-grained video object tracking tasks: object segmentation and keypoint tracking.

cs.CV↗

arXiv2025

Fisher Forecast of Finite-Size Effects with Future Gravitational Wave Detectors

We use Fisher information theory to forecast the bounds on the finite-size effects of astrophysical compact objects with next-generation gravitational wave detectors, including the ground-based Cosmic Explorer (CE) and Einstein Telescope (ET), as well as the space-based Laser Interferomet Space Antenna (LISA). Exploiting the worldline effective field theory (EFT) formalism, we first characterize three types of quadrupole finite-size effects: the spin-induced quadrupole moments, the conservative tidal deformations, and the tidal heating. We then derive the corresponding contributions to the gravitational waveform phases for binary compact objects in aligned-spin quasi-circular orbits. We separately estimate the constraints on these finite-size effects for black holes using the power spectral densities (PSDs) of the CE+ET detector network and LISA observations. For the CE+ET network, we find that the bounds on the mass-weighted spin-independent dissipation number $H_0$ are of the order $O(1)$, while the bounds on the mass-weighted tidal Love number $\tildeΛ$ are of the order $O(10)$. For high-spin binary black holes with dimensionless spin $χ\simeq 0.8$, the bounds on the symmetric spin-induced quadrupole moment $κ_s$ are of the order $O(10^{-1})$. LISA observations of supermassive black hole mergers offer slightly tighter constraints on all three finite-size parameters. Additionally, we perform a Fisher analysis for a binary neutron star merger within the CE+ET network. The bounds on the tidal parameter $H_0$ and on $\tilde Λ$ are around two orders of magnitude better than the current LIGO-Virgo-KAGRA (LVK) bounds.

gr-qc↗

arXiv2025

Computer-Aided Layout Generation for Building Design: A Review

Generating realistic building layouts for automatic building design has been studied in both the computer vision and architecture domains. Traditional approaches from the architecture domain, which are based on optimization techniques or heuristic design guidelines, can synthesize desirable layouts, but usually require post-processing and involve human interaction in the design pipeline, making them costly and timeconsuming. The advent of deep generative models has significantly improved the fidelity and diversity of the generated architecture layouts, reducing the workload by designers and making the process much more efficient. In this paper, we conduct a comprehensive review of three major research topics of architecture layout design and generation: floorplan layout generation, scene layout synthesis, and generation of some other formats of building layouts. For each topic, we present an overview of the leading paradigms, categorized either by research domains (architecture or machine learning) or by user input conditions or constraints. We then introduce the commonly-adopted benchmark datasets that are used to verify the effectiveness of the methods, as well as the corresponding evaluation metrics. Finally, we identify the well-solved problems and limitations of existing approaches, then propose new perspectives as promising directions for future research in this important research area. A project associated with this survey to maintain the resources is available at awesome-building-layout-generation.

cs.CV↗

arXiv2025

Beyond Traditional Coherence Time: An Electromagnetic Perspective for Mobile Channels

Channel coherence time has been widely regarded as a critical parameter in the design of mobile systems. However, a prominent challenge lies in integrating electromagnetic (EM) polarization effects into the derivation of the channel coherence time. In this paper, we develop a framework to analyze the impact of polarization mismatch on the channel coherence time. Specifically, we first establish an EM channel model to capture the essence of EM wave propagation. Based on this model, we then derive the EM temporal correlation function, incorporating the effects of polarization mismatch and beam misalignment. Further, considering the random orientation of the mobile user equipment (UE), we derive a closed-form solution for the EM coherence time in the turning scenario. When the trajectory degenerates into a straight line, we also provide a closed-form lower bound on the EM coherence time. The simulation results validate our theoretical analysis and reveal that neglecting the EM polarization effects leads to overly optimistic estimates of the EM coherence time.

eess.SP↗

arXiv2025

FlexWorld: Progressively Expanding 3D Scenes for Flexiable-View Synthesis

Generating flexible-view 3D scenes, including 360° rotation and zooming, from single images is challenging due to a lack of 3D data. To this end, we introduce FlexWorld, a novel framework consisting of two key components: (1) a strong video-to-video (V2V) diffusion model to generate high-quality novel view images from incomplete input rendered from a coarse scene, and (2) a progressive expansion process to construct a complete 3D scene. In particular, leveraging an advanced pre-trained video model and accurate depth-estimated training pairs, our V2V model can generate novel views under large camera pose variations. Building upon it, FlexWorld progressively generates new 3D content and integrates it into the global scene through geometry-aware scene fusion. Extensive experiments demonstrate the effectiveness of FlexWorld in generating high-quality novel view videos and flexible-view 3D scenes from single images, achieving superior visual quality under multiple popular metrics and datasets compared to existing state-of-the-art methods. Qualitatively, we highlight that FlexWorld can generate high-fidelity scenes with flexible views like 360° rotations and zooming. Project page: https://ml-gsai.github.io/FlexWorld.

cs.CV↗

arXiv2025

Robust Watermarking Using Generative Priors Against Image Editing: From Benchmarking to Advances

Current image watermarking methods are vulnerable to advanced image editing techniques enabled by large-scale text-to-image models. These models can distort embedded watermarks during editing, posing significant challenges to copyright protection. In this work, we introduce W-Bench, the first comprehensive benchmark designed to evaluate the robustness of watermarking methods against a wide range of image editing techniques, including image regeneration, global editing, local editing, and image-to-video generation. Through extensive evaluations of eleven representative watermarking methods against prevalent editing techniques, we demonstrate that most methods fail to detect watermarks after such edits. To address this limitation, we propose VINE, a watermarking method that significantly enhances robustness against various image editing techniques while maintaining high image quality. Our approach involves two key innovations: (1) we analyze the frequency characteristics of image editing and identify that blurring distortions exhibit similar frequency properties, which allows us to use them as surrogate attacks during training to bolster watermark robustness; (2) we leverage a large-scale pretrained diffusion model SDXL-Turbo, adapting it for the watermarking task to achieve more imperceptible and robust watermark embedding. Experimental results show that our method achieves outstanding watermarking performance under various image editing techniques, outperforming existing methods in both image quality and robustness. Code is available at https://github.com/Shilin-LU/VINE.

cs.CV↗

arXiv2025

Shiny-MAGEC: A Bayesian R Shiny Application for Meta-analysis of Censored Adverse Events

Accurate assessment of adverse event (AE) incidence is critical in clinical cancer research for drug safety evaluation and regulatory approval. While meta-analysis serves as an essential tool to comprehensively synthesize the evidence across multiple studies, incomplete AE reporting in clinical trials remains a persistent challenge. In particular, AEs occurring below study-specific reporting thresholds are often omitted from publications, leading to left-censored data. Failure to account for these censored AE counts can result in biased AE incidence estimates. We present an R Shiny application that implements a one-stage Bayesian meta-analysis model specifically designed to incorporate censored AE data into the estimation process. This interactive tool provides a user-friendly interface for researchers to conduct AE meta-analyses and estimate the AE incidence probability following the bias-correction methods proposed by Qi et al. (2024). It also enables direct comparisons between models that either incorporate or ignore censoring, highlighting the biases introduced by conventional approaches. This tutorial demonstrates the Shiny application's functionality through an illustrative example on meta-analysis of PD-1/PD-L1 inhibitor safety and highlights the importance of this tool in improving AE risk assessment. Ultimately, the new Shiny app facilitates more accurate and transparent drug safety evaluations. The Shiny-MAGEC app is available at: https://zihanzhou98.shinyapps.io/Shiny-MAGEC/.

stat.ME↗

arXiv2025

Gravitational Raman Scattering in Effective Field Theory: a Scalar Tidal Matching at $\mathcal{O}(G^3)$

We present a framework to compute amplitudes for the gravitational analog of the Raman process, a quasi-elastic scattering of waves off compact objects, in worldline effective field theory (EFT). As an example, we calculate third post-Minkowskian (PM) order ($\mathcal{O}(G^3)$), or two-loop, phase shifts for the scattering of a massless scalar field including all tidal effects and dissipation. Our calculation unveils two sources of the classical renormalization-group flow of dynamical Love numbers: a universal running independent of the nature of the compact object, and a running self-induced by tides. Restricting to the black hole case, we find that our EFT phase shifts agree exactly with those from general relativity, provided that the relevant static Love numbers are set to zero. In addition, we carry out a complete matching of the leading scalar dynamical Love number required to renormalize a universal short scale divergence in the S-wave. Our results pave the way for systematic calculations of gravitational Raman scattering at higher PM orders.

hep-th↗

arXiv2025

Distributed Consensus Network: A Modularized Communication Framework and Reliability Probabilistic Analysis

In this paper, we propose a modularized framework for communication processes applicable to crash and Byzantine fault-tolerant consensus protocols. We abstract basic communication components and show that the communication process of the classic consensus protocols such as RAFT, single-decree Paxos, PBFT, and Hotstuff, can be represented by the combination of communication components. Based on the proposed framework, we develop an approach to analyze the consensus reliability of different protocols, where link loss and node failure are measured as a probability. We propose two latency optimization methods and implement a RAFT system to verify our theoretical analysis and the effectiveness of the proposed latency optimization methods. We also discuss decreasing consensus failure rate by adjusting protocol designs. This paper provides theoretical guidance for the design of future consensus systems with a low consensus failure rate and latency under the possible communication loss.

cs.DC↗

arXiv2025

Learning to Decouple Complex Systems

A complex system with cluttered observations may be a coupled mixture of multiple simple sub-systems corresponding to latent entities. Such sub-systems may hold distinct dynamics in the continuous-time domain; therein, complicated interactions between sub-systems also evolve over time. This setting is fairly common in the real world but has been less considered. In this paper, we propose a sequential learning approach under this setting by decoupling a complex system for handling irregularly sampled and cluttered sequential observations. Such decoupling brings about not only subsystems describing the dynamics of each latent entity but also a meta-system capturing the interaction between entities over time. Specifically, we argue that the meta-system evolving within a simplex is governed by projected differential equations (ProjDEs). We further analyze and provide neural-friendly projection operators in the context of Bregman divergence. Experimental results on synthetic and real-world datasets show the advantages of our approach when facing complex and cluttered sequential data compared to the state-of-the-art.

cs.LG↗

arXiv2025

Language Representation Favored Zero-Shot Cross-Domain Cognitive Diagnosis

Cognitive diagnosis aims to infer students' mastery levels based on their historical response logs. However, existing cognitive diagnosis models (CDMs), which rely on ID embeddings, often have to train specific models on specific domains. This limitation may hinder their directly practical application in various target domains, such as different subjects (e.g., Math, English and Physics) or different education platforms (e.g., ASSISTments, Junyi Academy and Khan Academy). To address this issue, this paper proposes the language representation favored zero-shot cross-domain cognitive diagnosis (LRCD). Specifically, LRCD first analyzes the behavior patterns of students, exercises and concepts in different domains, and then describes the profiles of students, exercises and concepts using textual descriptions. Via recent advanced text-embedding modules, these profiles can be transformed to vectors in the unified language space. Moreover, to address the discrepancy between the language space and the cognitive diagnosis space, we propose language-cognitive mappers in LRCD to learn the mapping from the former to the latter. Then, these profiles can be easily and efficiently integrated and trained with existing CDMs. Extensive experiments show that training LRCD on real-world datasets can achieve commendable zero-shot performance across different target domains, and in some cases, it can even achieve competitive performance with some classic CDMs trained on the full response data on target domains. Notably, we surprisingly find that LRCD can also provide interesting insights into the differences between various subjects (such as humanities and sciences) and sources (such as primary and secondary education).

cs.CL↗