arXiv subjects

Xi Yang

Publications and source records attributed to Xi Yang.

At least 73 records · Page 4Linked to original sources

arXiv2025

Un-mixing Test-time Adaptation under Heterogeneous Data Streams

Deploying deep models in real-world scenarios remains challenging due to significant performance drops under distribution shifts between training and deployment environments. Test-Time Adaptation (TTA) has recently emerged as a promising solution, enabling on-the-fly model adaptation. However, its effectiveness deteriorates in the presence of mixed distribution shifts -- common in practical settings -- where multiple target domains coexist. In this paper, we study TTA under mixed distribution shifts and move beyond conventional whole-batch adaptation paradigms. By revisiting distribution shifts from a spectral perspective, we find that the heterogeneity across latent domains is often pronounced in Fourier space. In particular, high-frequency components encode domain-specific variations, which facilitates clearer separation of samples from different distributions. Motivated by this observation, we propose to un-mix heterogeneous data streams using high-frequency domain cues, making diverse shift patterns more tractable. To this end, we propose Frequency-based Decentralized Adaptation (FreDA), a novel framework that decomposes globally heterogeneous data stream into locally homogeneous clusters in the Fourier space. It leverages decentralized learning and augmentation strategies to robustly adapt under mixed domain shifts. Extensive experiments across various environments (corrupted, natural, and medical) show the superiority of our method over the state-of-the-arts.

cs.LG↗

arXiv2025

Rethinking Cross-Generator Image Forgery Detection through DINOv3

As generative models become increasingly diverse and powerful, cross-generator detection has emerged as a new challenge. Existing detection methods often memorize artifacts of specific generative models rather than learning transferable cues, leading to substantial failures on unseen generators. Surprisingly, this work finds that frozen visual foundation models, especially DINOv3, already exhibit strong cross-generator detection capability without any fine-tuning. Through systematic studies on frequency, spatial, and token perspectives, we observe that DINOv3 tends to rely on global, low-frequency structures as weak but transferable authenticity cues instead of high-frequency, generator-specific artifacts. Motivated by this insight, we introduce a simple, training-free token-ranking strategy followed by a lightweight linear probe to select a small subset of authenticity-relevant tokens. This token subset consistently improves detection accuracy across all evaluated datasets. Our study provides empirical evidence and a feasible hypothesis for understanding why foundation models generalize across diverse generators, offering a universal, efficient, and interpretable baseline for image forgery detection.

cs.CV↗

arXiv2025

FlagEval Findings Report: A Preliminary Evaluation of Large Reasoning Models on Automatically Verifiable Textual and Visual Questions

We conduct a moderate-scale contamination-free (to some extent) evaluation of current large reasoning models (LRMs) with some preliminary findings. We also release ROME, our evaluation benchmark for vision language models intended to test reasoning from visual clues. We attach links to the benchmark, evaluation data, and other updates on this website: https://flageval-baai.github.io/LRM-Eval/

cs.CL↗

arXiv2025

Beyond Multiple Choice: Verifiable OpenQA for Robust Vision-Language RFT

Multiple-choice question answering (MCQA) has been a popular format for evaluating and reinforcement fine-tuning (RFT) of modern multimodal language models. Its constrained output format allows for simplified, deterministic automatic verification. However, we find that the options may leak exploitable signals, which makes the accuracy metrics unreliable for indicating real capabilities and encourages explicit or implicit answer guessing behaviors during RFT. We propose ReVeL (Rewrite and Verify by LLM), a framework that rewrites multiple-choice questions into open-form questions while keeping answers verifiable whenever possible. The framework categorizes questions according to different answer types, apply different rewriting and verification schemes, respectively. When applied for RFT, we converted 20k MCQA examples and use GRPO to finetune Qwen2.5-VL models. Models trained on ReVeL-OpenQA match MCQA accuracy on multiple-choice benchmarks and improve OpenQA accuracy by about six percentage points, indicating better data efficiency and more robust reward signals than MCQA-based training. When used for evaluation, ReVeL also reveals up to 20 percentage points of score inflation in MCQA benchmarks (relative to OpenQA), improves judging accuracy, and reduces both cost and latency. We will release code and data publicly.

cs.CL↗

arXiv2025

Learning Compact Latent Space for Representing Neural Signed Distance Functions with High-fidelity Geometry Details

Neural signed distance functions (SDFs) have been a vital representation to represent 3D shapes or scenes with neural networks. An SDF is an implicit function that can query signed distances at specific coordinates for recovering a 3D surface. Although implicit functions work well on a single shape or scene, they pose obstacles when analyzing multiple SDFs with high-fidelity geometry details, due to the limited information encoded in the latent space for SDFs and the loss of geometry details. To overcome these obstacles, we introduce a method to represent multiple SDFs in a common space, aiming to recover more high-fidelity geometry details with more compact latent representations. Our key idea is to take full advantage of the benefits of generalization-based and overfitting-based learning strategies, which manage to preserve high-fidelity geometry details with compact latent codes. Based on this framework, we also introduce a novel sampling strategy to sample training queries. The sampling can improve the training efficiency and eliminate artifacts caused by the influence of other SDFs. We report numerical and visual evaluations on widely used benchmarks to validate our designs and show advantages over the latest methods in terms of the representative ability and compactness.

cs.CV↗

arXiv2025

Out-of-Context Misinformation Detection via Variational Domain-Invariant Learning with Test-Time Training

Out-of-context misinformation (OOC) is a low-cost form of misinformation in news reports, which refers to place authentic images into out-of-context or fabricated image-text pairings. This problem has attracted significant attention from researchers in recent years. Current methods focus on assessing image-text consistency or generating explanations. However, these approaches assume that the training and test data are drawn from the same distribution. When encountering novel news domains, models tend to perform poorly due to the lack of prior knowledge. To address this challenge, we propose \textbf{VDT} to enhance the domain adaptation capability for OOC misinformation detection by learning domain-invariant features and test-time training mechanisms. Domain-Invariant Variational Align module is employed to jointly encodes source and target domain data to learn a separable distributional space domain-invariant features. For preserving semantic integrity, we utilize domain consistency constraint module to reconstruct the source and target domain latent distribution. During testing phase, we adopt the test-time training strategy and confidence-variance filtering module to dynamically updating the VAE encoder and classifier, facilitating the model's adaptation to the target domain distribution. Extensive experiments conducted on the benchmark dataset NewsCLIPpings demonstrate that our method outperforms state-of-the-art baselines under most domain adaptation settings.

cs.LG↗

arXiv2025

SeniorTalk: A Chinese Conversation Dataset with Rich Annotations for Super-Aged Seniors

While voice technologies increasingly serve aging populations, current systems exhibit significant performance gaps due to inadequate training data capturing elderly-specific vocal characteristics like presbyphonia and dialectal variations. The limited data available on super-aged individuals in existing elderly speech datasets, coupled with overly simple recording styles and annotation dimensions, exacerbates this issue. To address the critical scarcity of speech data from individuals aged 75 and above, we introduce SeniorTalk, a carefully annotated Chinese spoken dialogue dataset. This dataset contains 55.53 hours of speech from 101 natural conversations involving 202 participants, ensuring a strategic balance across gender, region, and age. Through detailed annotation across multiple dimensions, it can support a wide range of speech tasks. We perform extensive experiments on speaker verification, speaker diarization, speech recognition, and speech editing tasks, offering crucial insights for the development of speech technologies targeting this age group.

cs.CL↗

arXiv2025

Video-SafetyBench: A Benchmark for Safety Evaluation of Video LVLMs

The increasing deployment of Large Vision-Language Models (LVLMs) raises safety concerns under potential malicious inputs. However, existing multimodal safety evaluations primarily focus on model vulnerabilities exposed by static image inputs, ignoring the temporal dynamics of video that may induce distinct safety risks. To bridge this gap, we introduce Video-SafetyBench, the first comprehensive benchmark designed to evaluate the safety of LVLMs under video-text attacks. It comprises 2,264 video-text pairs spanning 48 fine-grained unsafe categories, each pairing a synthesized video with either a harmful query, which contains explicit malice, or a benign query, which appears harmless but triggers harmful behavior when interpreted alongside the video. To generate semantically accurate videos for safety evaluation, we design a controllable pipeline that decomposes video semantics into subject images (what is shown) and motion text (how it moves), which jointly guide the synthesis of query-relevant videos. To effectively evaluate uncertain or borderline harmful outputs, we propose RJScore, a novel LLM-based metric that incorporates the confidence of judge models and human-aligned decision threshold calibration. Extensive experiments show that benign-query video composition achieves average attack success rates of 67.2%, revealing consistent vulnerabilities to video-induced attacks. We believe Video-SafetyBench will catalyze future research into video-based safety evaluation and defense strategies.

cs.CV↗

arXiv2025

Robust LLM Training Infrastructure at ByteDance

The training scale of large language models (LLMs) has reached tens of thousands of GPUs and is still continuously expanding, enabling faster learning of larger models. Accompanying the expansion of the resource scale is the prevalence of failures (CUDA error, NaN values, job hang, etc.), which poses significant challenges to training stability. Any large-scale LLM training infrastructure should strive for minimal training interruption, efficient fault diagnosis, and effective failure tolerance to enable highly efficient continuous training. This paper presents ByteRobust, a large-scale GPU infrastructure management system tailored for robust and stable training of LLMs. It exploits the uniqueness of LLM training process and gives top priorities to detecting and recovering failures in a routine manner. Leveraging parallelisms and characteristics of LLM training, ByteRobust enables high-capacity fault tolerance, prompt fault demarcation, and localization with an effective data-driven approach, comprehensively ensuring continuous and efficient training of LLM tasks. ByteRobust is deployed on a production GPU platform and achieves 97% ETTR for a three-month training job on 9,600 GPUs.

cs.LG↗

arXiv2025

Pioneering Scalable Prototyping for Mid-Band XL-MIMO Systems: Design and Implementation

The mid-band frequency range, combined with extra large-scale multiple-input multiple-output (XL-MIMO), is emerging as a key enabler for future communication systems. Thanks to the advent of new spectrum resources and degrees of freedom brought by the near-field propagation, the mid-band XL-MIMO system is expected to significantly enhance throughput and inherently support advanced functionalities such as integrated sensing and communication. Although theoretical studies have highlighted the benefits of mid-band XL-MIMO systems, the promised performance gains have yet to be validated in practical systems, posing a major challenge to the standardization. In this paper, preliminaries are first discussed, followed by an analysis of key challenges in constructing a real-time prototype system. Subsequently, the design and implementation of a real-time mid-band XL-MIMO prototype system are presented. Benefiting from the novel architecture, the proposed prototype system supports metrics aligned with standardization, including a bandwidth of 200 MHz, up to 1024 antenna elements, and up to 256 transceiver chains. Operating in time-division duplexing (TDD) mode, the prototype enables multiuser communication with support for up to 12 users, while retaining standard communication procedures. Built on software-defined radio (SDR) platforms, the system is programmable and allows for flexible deployment of advanced algorithms. Moreover, the modular architecture ensures high scalability, making the system adaptable to various configurations, including distributed deployments and decentralized signal processing. Experimental results with the proposed prototype system demonstrate real-time digital sample processing at 1167.85 Gbps, a peak data throughput of 15.81 Gbps for 12 users, and a maximal spectral efficiency approaching 80 bit/s/Hz.

eess.SP↗

arXiv2025

LadderMoE: Ladder-Side Mixture of Experts Adapters for Bronze Inscription Recognition

Bronze inscriptions (BI), engraved on ritual vessels, constitute a crucial stage of early Chinese writing and provide indispensable evidence for archaeological and historical studies. However, automatic BI recognition remains difficult due to severe visual degradation, multi-domain variability across photographs, rubbings, and tracings, and an extremely long-tailed character distribution. To address these challenges, we curate a large-scale BI dataset comprising 22454 full-page images and 198598 annotated characters spanning 6658 unique categories, enabling robust cross-domain evaluation. Building on this resource, we develop a two-stage detection-recognition pipeline that first localizes inscriptions and then transcribes individual characters. To handle heterogeneous domains and rare classes, we equip the pipeline with LadderMoE, which augments a pretrained CLIP encoder with ladder-style MoE adapters, enabling dynamic expert specialization and stronger robustness. Comprehensive experiments on single-character and full-page recognition tasks demonstrate that our method substantially outperforms state-of-the-art scene text recognition baselines, achieving superior accuracy across head, mid, and tail categories as well as all acquisition modalities. These results establish a strong foundation for bronze inscription recognition and downstream archaeological analysis.

cs.CV↗

arXiv2025

DermINO: Hybrid Pretraining for a Versatile Dermatology Foundation Model

Skin diseases impose a substantial burden on global healthcare systems, driven by their high prevalence (affecting up to 70% of the population), complex diagnostic processes, and a critical shortage of dermatologists in resource-limited areas. While artificial intelligence(AI) tools have demonstrated promise in dermatological image analysis, current models face limitations-they often rely on large, manually labeled datasets and are built for narrow, specific tasks, making them less effective in real-world settings. To tackle these limitations, we present DermNIO, a versatile foundation model for dermatology. Trained on a curated dataset of 432,776 images from three sources (public repositories, web-sourced images, and proprietary collections), DermNIO incorporates a novel hybrid pretraining framework that augments the self-supervised learning paradigm through semi-supervised learning and knowledge-guided prototype initialization. This integrated method not only deepens the understanding of complex dermatological conditions, but also substantially enhances the generalization capability across various clinical tasks. Evaluated across 20 datasets, DermNIO consistently outperforms state-of-the-art models across a wide range of tasks. It excels in high-level clinical applications including malignancy classification, disease severity grading, multi-category diagnosis, and dermatological image caption, while also achieving state-of-the-art performance in low-level tasks such as skin lesion segmentation. Furthermore, DermNIO demonstrates strong robustness in privacy-preserving federated learning scenarios and across diverse skin types and sexes. In a blinded reader study with 23 dermatologists, DermNIO achieved 95.79% diagnostic accuracy (versus clinicians' 73.66%), and AI assistance improved clinician performance by 17.21%.

eess.IV↗

arXiv2025

RoboBrain 2.0 Technical Report

We introduce RoboBrain 2.0, our latest generation of embodied vision-language foundation models, designed to unify perception, reasoning, and planning for complex embodied tasks in physical environments. It comes in two variants: a lightweight 7B model and a full-scale 32B model, featuring a heterogeneous architecture with a vision encoder and a language model. Despite its compact size, RoboBrain 2.0 achieves strong performance across a wide spectrum of embodied reasoning tasks. On both spatial and temporal benchmarks, the 32B variant achieves leading results, surpassing prior open-source and proprietary models. In particular, it supports key real-world embodied AI capabilities, including spatial understanding (e.g., affordance prediction, spatial referring, trajectory forecasting) and temporal decision-making (e.g., closed-loop interaction, multi-agent long-horizon planning, and scene graph updating). This report details the model architecture, data construction, multi-stage training strategies, infrastructure and practical applications. We hope RoboBrain 2.0 advances embodied AI research and serves as a practical step toward building generalist embodied agents. The code, checkpoint and benchmark are available at https://superrobobrain.github.io.

cs.RO↗

arXiv2025

Clustering-based Feature Representation Learning for Oracle Bone Inscriptions Detection

Oracle Bone Inscriptions (OBIs), play a crucial role in understanding ancient Chinese civilization. The automated detection of OBIs from rubbing images represents a fundamental yet challenging task in digital archaeology, primarily due to various degradation factors including noise and cracks that limit the effectiveness of conventional detection networks. To address these challenges, we propose a novel clustering-based feature space representation learning method. Our approach uniquely leverages the Oracle Bones Character (OBC) font library dataset as prior knowledge to enhance feature extraction in the detection network through clustering-based representation learning. The method incorporates a specialized loss function derived from clustering results to optimize feature representation, which is then integrated into the total network loss. We validate the effectiveness of our method by conducting experiments on two OBIs detection dataset using three mainstream detection frameworks: Faster R-CNN, DETR, and Sparse R-CNN. Through extensive experimentation, all frameworks demonstrate significant performance improvements.

cs.CV↗

arXiv2025

Saving for the future: Enhancing generalization via partial logic regularization

Generalization remains a significant challenge in visual classification tasks, particularly in handling unknown classes in real-world applications. Existing research focuses on the class discovery paradigm, which tends to favor known classes, and the incremental learning paradigm, which suffers from catastrophic forgetting. Recent approaches such as the L-Reg technique employ logic-based regularization to enhance generalization but are bound by the necessity of fully defined logical formulas, limiting flexibility for unknown classes. This paper introduces PL-Reg, a novel partial-logic regularization term that allows models to reserve space for undefined logic formulas, improving adaptability to unknown classes. Specifically, we formally demonstrate that tasks involving unknown classes can be effectively explained using partial logic. We also prove that methods based on partial logic lead to improved generalization. We validate PL-Reg through extensive experiments on Generalized Category Discovery, Multi-Domain Generalized Category Discovery, and long-tailed Class Incremental Learning tasks, demonstrating consistent performance improvements. Our results highlight the effectiveness of partial logic in tackling challenges related to unknown classes.

cs.LG↗

arXiv2025

Overconfidence in LLM-as-a-Judge: Diagnosis and Confidence-Driven Solution

Large Language Models (LLMs) are widely used as automated judges, where practical value depends on both accuracy and trustworthy, risk-aware judgments. Existing approaches predominantly focus on accuracy, overlooking the necessity of well-calibrated confidence, which is vital for adaptive and reliable evaluation pipelines. In this work, we advocate a shift from accuracy-centric evaluation to confidence-driven, risk-aware LLM-as-a-Judge systems, emphasizing the necessity of well-calibrated confidence for trustworthy and adaptive evaluation. We systematically identify the Overconfidence Phenomenon in current LLM-as-a-Judges, where predicted confidence significantly overstates actual correctness, undermining reliability in practical deployment. To quantify this phenomenon, we introduce TH-Score, a novel metric measuring confidence-accuracy alignment. Furthermore, we propose LLM-as-a-Fuser, an ensemble framework that transforms LLMs into reliable, risk-aware evaluators. Extensive experiments demonstrate that our approach substantially improves calibration and enables adaptive, confidence-driven evaluation pipelines, achieving superior reliability and accuracy compared to existing baselines.

cs.AI↗

arXiv2025

Beyond Solving Math Quiz: Evaluating the Ability of Large Reasoning Models to Ask for Information

Large Reasoning Models (LRMs) have demonstrated remarkable problem-solving abilities in mathematics, as evaluated by existing benchmarks exclusively on well-defined problems. However, such evaluation setup constitutes a critical gap, since a genuine intelligent agent should not only solve problems (as a math quiz solver), but also be able~to ask for information when the problems lack sufficient information, enabling proactivity in responding users' requests. To bridge such gap, we proposes a new dataset consisting of two types of incomplete problems with diverse contexts. Based on the dataset, our systematical evaluation of LRMs reveals their inability in proactively asking for information. In addition, we uncover the behaviors related to overthinking and hallucination of LRMs, and highlight the potential and challenges of supervised fine-tuning in learning such ability. We hope to provide new insights in developing LRMs with genuine intelligence, rather than just solving problems.

cs.AI↗

arXiv2025

Curvature-adaptive gigapixel microscopy at submicron resolution and centimeter scale

Large-area microscopy with submicron resolution is limited by tradeoffs between field of view (FOV), resolution, and imaging speed. Samples are rarely flat across centimeter-scale FOV, which often requires existing solutions to use mechanical scanning to ensure focused capture at reduced throughput. Here, we present PANORAMA, a single-shot, re-imaging microscope that achieves seamless, gigapixel imaging over a 16.3$\times$18.8 $\text{mm}^2$ FOV at 0.84 um resolution without mechanical scanning. By using a telecentric photolithography lens, a large-aperture tube lens, and a flat micro-camera array with adaptive per-camera focus control, PANORAMA maintains submicron focus across flat, curved or uneven samples that span centimeters. This approach improves imaging throughput and adaptability, enabling gigapixel multi-modal microscopy of large flat and non-flat samples in one shot, thus broadening its applications in biomedical and materials imaging.

physics.optics↗