arXiv subjects

Yang Xiang

Publications and source records attributed to Yang Xiang.

At least 91 records · Page 5Linked to original sources

arXiv2026

CCFQA: A Benchmark for Cross-Lingual and Cross-Modal Speech and Text Factuality Evaluation

As Large Language Models (LLMs) are increasingly popularized in the multilingual world, ensuring hallucination-free factuality becomes markedly crucial. However, existing benchmarks for evaluating the reliability of Multimodal Large Language Models (MLLMs) predominantly focus on textual or visual modalities with a primary emphasis on English, which creates a gap in evaluation when processing multilingual input, especially in speech. To bridge this gap, we propose a novel Cross-lingual and Cross-modal Factuality benchmark (CCFQA). Specifically, the CCFQA benchmark contains parallel speech-text factual questions across 8 languages, designed to systematically evaluate MLLMs' cross-lingual and cross-modal factuality capabilities. Our experimental results demonstrate that current MLLMs still face substantial challenges on the CCFQA benchmark. Furthermore, we propose a few-shot transfer learning strategy that effectively transfers the Question Answering (QA) capabilities of LLMs in English to multilingual Spoken Question Answering (SQA) tasks, achieving competitive performance with GPT-4o-mini-Audio using just 5-shot training. We release CCFQA as a foundational research resource to promote the development of MLLMs with more robust and reliable speech understanding capabilities. Our code and dataset are available at https://github.com/yxduir/ccfqa.

cs.CL↗

arXiv2026

VC-Bench: Pioneering the Video Connecting Benchmark with a Dataset and Evaluation Metrics

While current video generation focuses on text or image conditions, practical applications like video editing and vlogging often need to seamlessly connect separate clips. In our work, we introduce Video Connecting, an innovative task that aims to generate smooth intermediate video content between given start and end clips. However, the absence of standardized evaluation benchmarks has hindered the development of this task. To bridge this gap, we proposed VC-Bench, a novel benchmark specifically designed for video connecting. It includes 1,579 high-quality videos collected from public platforms, covering 15 main categories and 72 subcategories to ensure diversity and structure. VC-Bench focuses on three core aspects: Video Quality Score VQS, Start-End Consistency Score SECS, and Transition Smoothness Score TSS. Together, they form a comprehensive framework that moves beyond conventional quality-only metrics. We evaluated multiple state-of-the-art video generation models on VC-Bench. Experimental results reveal significant limitations in maintaining start-end consistency and transition smoothness, leading to lower overall coherence and fluidity. We expect that VC-Bench will serve as a pioneering benchmark to inspire and guide future research in video connecting. The evaluation metrics and dataset are publicly available at: https://anonymous.4open.science/r/VC-Bench-1B67/.

cs.CV↗

arXiv2026

FunCineForge: A Unified Dataset Toolkit and Model for Zero-Shot Movie Dubbing in Diverse Cinematic Scenes

Movie dubbing is the task of synthesizing speech from scripts conditioned on video scenes, requiring accurate lip sync, faithful timbre transfer, and proper modeling of character identity and emotion. However, existing methods face two major limitations: (1) high-quality multimodal dubbing datasets are limited in scale, suffer from high word error rates, contain sparse annotations, rely on costly manual labeling, and are restricted to monologue scenes, all of which hinder effective model training; (2) existing dubbing models rely solely on the lip region to learn audio-visual alignment, which limits their applicability to complex live-action cinematic scenes, and exhibit suboptimal performance in lip sync, speech quality, and emotional expressiveness. To address these issues, we propose FunCineForge, which comprises an end-to-end production pipeline for large-scale dubbing datasets and an MLLM-based dubbing model designed for diverse cinematic scenes. Using the pipeline, we construct the first Chinese television dubbing dataset with rich annotations, and demonstrate the high quality of these data. Experiments across monologue, narration, dialogue, and multi-speaker scenes show that our dubbing model consistently outperforms SOTA methods in audio quality, lip sync, timbre transfer, and instruction following. Code and demos are available at https://anonymous.4open.science/w/FunCineForge.

cs.CV↗

arXiv2026

SSBE-PINN: A Sobolev Boundary Scheme Boosting Stability and Accuracy in Elliptic/Parabolic PDE Learning

Physics-Informed Neural Networks (PINNs) have emerged as a powerful framework for solving partial differential equations (PDEs), yet they often fail to achieve accurate convergence in the H1 norm, especially in the presence of boundary approximation errors. In this work, we propose a novel method called Sobolev-Stable Boundary Enforcement (SSBE), which redefines the boundary loss using Sobolev norms to incorporate boundary regularity directly into the training process. We provide rigorous theoretical analysis demonstrating that SSBE ensures bounded H1 error via a stability guarantee and derive generalization bounds that characterize its robustness under finite-sample regimes. Extensive numerical experiments on linear and nonlinear PDEs, including Poisson, heat, and elliptic problems, show that SSBE consistently outperforms standard PINNs in terms of both relative L2 and H1 errors, even in high-dimensional settings. The proposed approach offers a principled and practical solution for improving gradient fidelity and overall solution accuracy in neural network based PDE solvers.

math.NA↗

arXiv2026

Exploring the Translation Mechanism of Large Language Models

While large language models (LLMs) demonstrate remarkable success in multilingual translation, their internal core translation mechanisms, even at the fundamental word level, remain insufficiently understood. To address this critical gap, this work introduces a systematic framework for interpreting the mechanism behind LLM translation from the perspective of computational components. This paper first proposes subspace-intervened path patching for precise, fine-grained causal analysis, enabling the detection of components crucial to translation tasks and subsequently characterizing their behavioral patterns in human-interpretable terms. Comprehensive experiments reveal that translation is predominantly driven by a sparse subset of components: specialized attention heads serve critical roles in extracting source language, translation indicators, and positional features, which are then integrated and processed by specific multi-layer perceptrons (MLPs) into intermediary English-centric latent representations before ultimately yielding the final translation. The significance of these findings is underscored by the empirical demonstration that targeted fine-tuning a minimal parameter subset ($<5\%$) enhances translation performance while preserving general capabilities. This result further indicates that these crucial components generalize effectively to sentence-level translation and are instrumental in elucidating more intricate translation tasks.

cs.CL↗

arXiv2026

TeachPro: Multi-Label Qualitative Teaching Evaluation via Cross-View Graph Synergy and Semantic Anchored Evidence Encoding

Standardized Student Evaluation of Teaching often suffer from low reliability, restricted response options, and response distortion. Existing machine learning methods that mine open-ended comments usually reduce feedback to binary sentiment, which overlooks concrete concerns such as content clarity, feedback timeliness, and instructor demeanor, and provides limited guidance for instructional improvement.We propose TeachPro, a multi-label learning framework that systematically assesses five key teaching dimensions: professional expertise, instructional behavior, pedagogical efficacy, classroom experience, and other performance metrics. We first propose a Dimension-Anchored Evidence Encoder, which integrates three core components: (i) a pre-trained text encoder that transforms qualitative feedback annotations into contextualized embeddings; (ii) a prompt module that represents five teaching dimensions as learnable semantic anchors; and (iii) a cross-attention mechanism that aligns evidence with pedagogical dimensions within a structured semantic space. We then propose a Cross-View Graph Synergy Network to represent student comments. This network comprises two components: (i) a Syntactic Branch that extracts explicit grammatical dependencies from parse trees, and (ii) a Semantic Branch that models latent conceptual relations derived from BERT-based similarity graphs. BiAffine fusion module aligns syntactic and semantic units, while a differential regularizer disentangles embeddings to encourage complementary representations. Finally, a cross-attention mechanism bridges the dimension-anchored evidence with the multi-view comment representations. We also contribute a novel benchmark dataset featuring expert qualitative annotations and multi-label scores. Extensive experiments demonstrate that TeachPro offers superior diagnostic granularity and robustness across diverse evaluation settings.

cs.CL↗

arXiv2025

Non-Euclidean interfaces decode the continuous landscape of graphene-induced surface reconstructions

Interfacial reconstruction between two-dimensional (2D) materials and metal substrates fundamentally governs heterostructure properties, yet conventional flat substrates fail to capture the continuous crystallographic landscape. Here, we overcome this topological limitation using non-Euclidean interfaces-curved 2D graphene-copper surfaces as a model system-to traverse the infinite spectrum of lattice orientations. By integrating multimodal microscopy with a deep-learning-enhanced dimensional upscaling framework, we translate 2D scanning electron microscopy (SEM) contrast into quantitative three-dimensional (3D) morphologies with accurate facet identification. Coupling these observations with machine-learning-assisted density functional theory, we demonstrate that reconstruction is governed by a unified thermodynamic mechanism where high-index facets correspond to specific local minima in the surface energy landscape. This work resolves the long-standing complexity of graphene-copper faceting and establishes non-Euclidean surface topologies as a generalizable paradigm for decoding and controlling interfacial reconstruction in diverse metal-2D material systems.

cond-mat.mes-hall↗

arXiv2025

Geese achieve stationary takeoff via synergistic wing kinematics and enhanced aerodynamics

Stationary take-off, without a running start or elevated descent, requires substantial aerodynamic forces to overcome weight, particularly for large birds such as geese exceeding 2 kg. However, the complex wing motion and high-Reynolds-number (Re $\approx$$10^5$) flow dynamics challenge conventional expectations of avian flight aerodynamics, rendering this mechanism elusive. Analyzing 578 stationary take-offs from seven geese (\textit{Anser cygnoides}) and applying Principal Component Analysis (PCA), we reveal that the complex wing kinematics collapse onto a low-dimensional manifold dominated by two synergies: a Stroke Synergy responsible for fundamental rhythmic stroke, and a Morphing Synergy governing spanwise geometry. This modular control strategy orchestrates a stereotyped wing kinematics featuring an accelerated translational downstroke and a rapid tip-reversal upstroke. By integrating wing kinematic analysis with the mass distribution of the geese, we quantified the aerodynamic forces and found that entirely positive lift and thrust are generated throughout the motion cycle. The enhanced aerodynamic performance of geese takeoff results from three principal mechanisms. During the downstroke, significant lift generated from wing acceleration is predicted by the quasi-steady framework. Flow visualization reveals that wake capture further enhances the lift generation in downstroke by orienting the position of wake vortices. During the upstroke, the distal wing performs a rapid pitching motion and generates a substantial thrust, the vertical component of which contributes significantly to weight support.

physics.flu-dyn↗

arXiv2025

Spatiotemporal Satellite Image Downscaling with Transfer Encoders and Autoregressive Generative Models

We present a transfer-learning generative downscaling framework to reconstruct fine resolution satellite images from coarse scale inputs. Our approach combines a lightweight U-Net transfer encoder with a diffusion-based generative model. The simpler U-Net is first pretrained on a long time series of coarse resolution data to learn spatiotemporal representations; its encoder is then frozen and transferred to a larger downscaling model as physically meaningful latent features. Our application uses NASA's MERRA-2 reanalysis as the low resolution source domain (50 km) and the GEOS-5 Nature Run (G5NR) as the high resolution target (7 km). Our study area included a large area in Asia, which was made computationally tractable by splitting into two subregions and four seasons. We conducted domain similarity analysis using Wasserstein distances confirmed minimal distributional shift between MERRA-2 and G5NR, validating the safety of parameter frozen transfer. Across seasonal regional splits, our model achieved excellent performance (R2 = 0.65 to 0.94), outperforming comparison models including deterministic U-Nets, variational autoencoders, and prior transfer learning baselines. Out of data evaluations using semivariograms, ACF/PACF, and lag-based RMSE/R2 demonstrated that the predicted downscaled images preserved physically consistent spatial variability and temporal autocorrelation, enabling stable autoregressive reconstruction beyond the G5NR record. These results show that transfer enhanced diffusion models provide a robust and physically coherent solution for downscaling a long time series of coarse resolution images with limited training periods. This advancement has significant implications for improving environmental exposure assessment and long term environmental monitoring.

cs.CV↗

arXiv2025

Think Before You Prune: Selective Self-Generated Calibration for Pruning Large Reasoning Models

Large Reasoning Models (LRMs) have demonstrated remarkable performance on complex reasoning benchmarks. However, their long chain-of-thought reasoning processes incur significant inference overhead. Pruning has emerged as a promising approach to reducing computational costs. However, existing efforts have primarily focused on large language models (LLMs), while pruning LRMs remains unexplored. In this work, we conduct the first empirical study on pruning LRMs and show that directly applying existing pruning techniques fails to yield satisfactory results. Our findings indicate that using self-generated reasoning data for calibration can substantially improve pruning performance. We further investigate how the difficulty and length of reasoning data affect pruning outcomes. Our analysis reveals that challenging and moderately long self-generated reasoning data serve as ideal calibration data. Based on these insights, we propose a Selective Self-Generated Reasoning (SSGR) data construction strategy to provide effective calibration data for pruning LRMs. Experimental results on the DeepSeek-R1-Distill model series validate that our strategy improves the reasoning ability of pruned LRMs by 10%-13% compared to general pruning methods.

cs.CL↗

arXiv2025

Optimal graph joining with applications to isomorphism detection and identification

We introduce an optimal transport based approach for comparing undirected graphs with non-negative edge weights and general vertex labels, and we study connections between the resulting linear program and the graph isomorphism problem. Our approach is based on the notion of a joining of two graphs $G$ and $H$, which is a product graph that preserves their marginal structure. Given $G$ and $H$ and a vertex-based cost function $c$, the optimal graph joining (OGJ) problem finds a joining of $G$ and $H$ minimizing degree weighted cost. The OGJ problem can be written as a linear program with a convex polyhedral solution set. We establish several basic properties of the OGJ problem, and present theoretical results connecting the OGJ problem to the graph isomorphism problem. In particular, we examine a variety of conditions on graph families that are sufficient to ensure that for every pair of graphs $G$ and $H$ in the family (i) $G$ and $H$ are isomorphic if and only if their optimal joining cost is zero, and (ii) if $G$ and $H$ are isomorphic, the the extreme points of the solution set of the OGJ problem are deterministic joinings corresponding to the isomorphisms from $G$ to $H$.

math.CO↗

arXiv2025

TEn-CATG:Text-Enriched Audio-Visual Video Parsing with Multi-Scale Category-Aware Temporal Graph

Audio-visual video parsing (AVVP) aims to detect event categories and their temporal boundaries in videos, typically under weak supervision. Existing methods mainly focus on (i) improving temporal modeling using attention-based architectures or (ii) generating richer pseudo-labels to address the absence of frame-level annotations. However, attention-based models often overfit noisy pseudo-labels, leading to cumulative training errors, while pseudo-label generation approaches distribute attention uniformly across frames, weakening temporal localization accuracy. To address these challenges, we propose TEn-CATG, a text-enriched AVVP framework that combines semantic calibration with category-aware temporal reasoning. More specifically, we design a bi-directional text fusion (BiT) module by leveraging audio-visual features as semantic anchors to refine text embeddings, which departs from conventional text-to-feature alignment, thereby mitigating noise and enhancing cross-modal consistency. Furthermore, we introduce the category-aware temporal graph (CATG) module to model temporal relationships by selecting multi-scale temporal neighbors and learning category-specific temporal decay factors, enabling effective event-dependent temporal reasoning. Extensive experiments demonstrate that TEn-CATG achieves state-of-the-art results across multiple evaluation metrics on benchmark datasets LLP and UnAV-100, highlighting its robustness and superior ability to capture complex temporal and semantic dependencies in weakly supervised AVVP tasks.

cs.CV↗

arXiv2025

Evaluating and Improving Cultural Awareness of Reward Models for LLM Alignment

Reward models (RMs) are crucial for aligning large language models (LLMs) with diverse cultures. Consequently, evaluating their cultural awareness is essential for further advancing global alignment of LLMs. However, existing RM evaluations fall short in assessing cultural awareness due to the scarcity of culturally relevant evaluation datasets. To fill this gap, we propose Cultural Awareness Reward modeling Benchmark (CARB), covering 10 distinct cultures across 4 cultural domains. Our extensive evaluation of state-of-the-art RMs reveals their deficiencies in modeling cultural awareness and demonstrates a positive correlation between performance on CARB and downstream multilingual cultural alignment tasks. Further analysis identifies the spurious correlations within culture-aware reward modeling, wherein RM's scoring relies predominantly on surface-level features rather than authentic cultural nuance understanding. To address these, we propose Think-as-Locals to elicit deeper culturally grounded reasoning from generative RMs via reinforcement learning from verifiable rewards (RLVR) and employ well-designed rewards to ensure accurate preference judgments and high-quality structured evaluation criteria generation. Experimental results validate its efficacy in mitigating spurious features interference and advancing culture-aware reward modeling.

cs.CL↗

arXiv2025

Towards a Golden Classifier-Free Guidance Path via Foresight Fixed Point Iterations

Classifier-Free Guidance (CFG) is an essential component of text-to-image diffusion models, and understanding and advancing its operational mechanisms remains a central focus of research. Existing approaches stem from divergent theoretical interpretations, thereby limiting the design space and obscuring key design choices. To address this, we propose a unified perspective that reframes conditional guidance as fixed point iterations, seeking to identify a golden path where latents produce consistent outputs under both conditional and unconditional generation. We demonstrate that CFG and its variants constitute a special case of single-step short-interval iteration, which is theoretically proven to exhibit inefficiency. To this end, we introduce Foresight Guidance (FSG), which prioritizes solving longer-interval subproblems in early diffusion stages with increased iterations. Extensive experiments across diverse datasets and model architectures validate the superiority of FSG over state-of-the-art methods in both image quality and computational efficiency. Our work offers novel perspectives for conditional guidance and unlocks the potential of adaptive design.

cs.CV↗

arXiv2025

SOT-MRAM Bitcell Scaling with BEOL Read Selectors: A DTCO Study

This work explores the cross-node scaling potential of SOT-MRAM for last-level caches (LLCs) under heterogeneous system scaling paradigm. We perform extensive Design-Technology Co-Optimization (DTCO) exercises to evaluate the bitcell footprint for different cell configurations at a representative 7 nm technology and to assess their implications on read and write power-performance. We crucially identify the MTJ routing struggle in conventional two-transistor one-resistor (2T1R) SOT-MRAMs as the primary bitcell area scaling challenge and propose to use BEOL read selectors (BEOL RSs) that enable (10 -- 40) % bitcell area reduction and eventually match sub-N3 SRAM. On writability, we affirm that BEOL RS-based bitcells could meet the required SOT switching current, provided the magnetic free layer properties be engineered in line with LLC-specific, (0.1 -- 100) s retention targets. This is particularly to attribute to their (i) more available Si fins for write transistor and (ii) lower bitline resistance at reduced cell width. We nevertheless underscore the read tradeoff associated with BEOL RSs, with the low-drive IGZO-FET selector sacrificing the latency up to (3 -- 5) ns and the imperfectly rectifying diode selectors suffering (2.5 -- 5)$\times$ energy cost relative to 2T1R. This article thus highlights the realistic prospects and hurdles of BEOL RSs towards holistic power-performance-area scaling of SOT-MRAM.

cs.ET↗

arXiv2025

Environmental Rate Manipulation Attacks on Power Grid Security

The growing complexity of global supply chains has made hardware Trojans a significant threat in sensor-based power electronics. Traditional Trojan designs depend on digital triggers or fixed threshold conditions that can be detected during standard testing. In contrast, we introduce Environmental Rate Manipulation (ERM), a novel Trojan triggering mechanism that activates by monitoring the rate of change in environmental parameters rather than their absolute values. This approach allows the Trojan to remain inactive under normal conditions and evade redundancy and sensor-fusion defenses. We implement a compact 14~$μ$m$^2$ circuit that measures capacitor charging rates in standard sensor front-ends and disrupts inverter pulse-width modulation PWM signals when a rapid change is induced. Experiments on a commercial Texas Instruments solar inverter demonstrate that ERM can trigger catastrophic driver chip failure. Furthermore, ETAP simulations indicate that a single compromised 100~kW inverter may initiate cascading grid instabilities. The attack's significance extends beyond individual sensors to entire classes of environmental sensing systems common in power electronics, demonstrating fundamental challenges for hardware security.

cs.CR↗

arXiv2025

Towards Efficient CoT Distillation: Self-Guided Rationale Selector for Better Performance with Fewer Rationales

Chain-of-thought (CoT) distillation aims to enhance small language models' (SLMs) reasoning by transferring multi-step reasoning capability from the larger teacher models. However, existing work underestimates rationale quality, focusing primarily on data quantity, which may transfer noisy or incorrect information to the student model. To address the above issues, we proposed \textbf{M}odel-\textbf{O}riented \textbf{R}ationale \textbf{S}election \textbf{D}istillation (MoRSD), which can discern and select high quality rationales for distillation to improve performance further. We further propose a Rationale Difficulty (RD) metric to measure the ability of the student model to generate the correct answer under a given rationale. Compared to the baseline, we achieved 4.6$\%$ average improvement on seven datasets over three tasks, using fewer rationales by controlling their accuracy, diversity, and difficulty. Our results reveal that a small portion of the high quality rationales can enhance the reasoning ability of student models than the entire dataset. Our method promises to be a possible solution for efficient CoT distillation. Our code will be released in https://github.com/Leon221220/MoRSD.

cs.CL↗

arXiv2025

XBOUND: Exploring Capability Boundaries of Device-Control Agents at the State Level

Recent advancements in vision-language models have increased interest in Device-Control Agents (DC agents) for managing graphical user interfaces (GUIs). With the growing complexity and integration of such agents into various applications, effective evaluation methods have become crucial. The current evaluation method for DC agents primarily focuses on the instruction level, providing the current state (e.g., screenshots) and past execution history to determine actions for target instructions, helping identify potential execution failures. However, in GUI environments, a single state may contain multiple interactive widgets, each linked to different instructions, presenting an opportunity for diverse actions based on various instruction targets. Evaluating the agent's performance solely at the instruction level may overlook the broader context of these interactions. To capture a more comprehensive view of agent performance, we propose a new evaluation method, XBOUND, to evaluate the accuracy of instruction completion on a per-state basis. XBOUND provides a state-level evaluation framework, serving as a tool to assess agents' capabilities within environmental states. Our evaluation yields several key insights: UI-TARS stands out as the strongest 7B model, current agents display a bimodal performance pattern in instruction unification, and sub-7B models remain limited in state mastery. We further identify GPT-based planning as a critical bottleneck, and show that grounding data mainly benefits action matching, while trajectory data is more effective for instruction unification.

cs.AI↗