arXiv subjects

Yu Yuan

Publications and source records attributed to Yu Yuan.

At least 19 recordsLinked to original sources

arXiv2026

Speedbumps: Rejection Attacks on Speculative Decoding

Speculative decoding is a popular technique for increasing the speed and reducing the costs of large language model (LLM) inference by verifying multiple draft tokens in a single target-model forward pass. The resulting benefit depends on the ability of the drafter to approximate the target model's distribution. In this work, we study Speculative Rejection Attacks (SRAs), a novel class of attacks that cause draft and target models to disagree more often, resulting in fewer draft tokens being accepted per draft cycle. This leads to more target model forward passes needed per generated token, slowing down inference and increasing costs for the victim. We introduce two attacks which append an adversarial suffix to attacker-controlled content to degrade speculative decoding on a victim's prompts. Both attacks optimise the expected length of the accepted speculative prefix, estimating per-depth acceptance from the target's probability of the drafted proposals (Speedbump-P) or from the overlap between the draft and target distributions (Speedbump-D). In some cases, attacks degrade speculative decoding to the point of being slower than autoregressive decoding. The degradation reduces the output quality - regularisation restores output quality but gives up most of the degradation, trading effectiveness for stealthiness. Additionally, the suffixes remain effective under sampling, and transfer across drafters (Speedbump-P) or across target models sharing a drafter (Speedbump-D). These findings identify the draft-target interaction of speculative decoding as a realistic attack surface through which adversarial inputs can inflate inference costs.

cs.CR↗

arXiv2026

Reimagine Video Dynamics

Most video editing methods focus on changing the appearance of the source video, while offering limited control over its dynamics. We introduce Reimagine Video Dynamics (RVD), a framework that disentangles a compact, editable dynamics token from visual context. We learn this token through self-supervised reconstruction: given the first frame as visual context, a renderer must recover the original video from the dynamics token, encouraging it to capture how the scene evolves rather than how it looks. This disentanglement allows video dynamics to be edited directly while preserving visual context. We develop a language-guided dynamics-token editor that transforms source dynamics into target dynamics, and train it with a scalable counterfactual video-pair pipeline and a two-stage training strategy. Extensive experiments show that RVD enables effective video dynamics editing, training-free retiming, and appearance-controlled re-rendering.

cs.CV↗

arXiv2026

Beyond Visual Quality: A Study of Test-Time Planning with World Action Models

World action models generate actions together with visual predictions of their consequences. These paired outputs create the potential for planning by sampling multiple actions from one state, comparing their imagined outcomes, and choosing the action with the most promising predicted outcome. However, how to use imagined futures to guide action selection remains unclear. We examine this planning potential empirically. First, we estimate an oracle upper bound on selection by choosing the sampled candidate whose realised outcome is best. In a controlled same-state analysis, this choice raises success from 68.9% under uniform random selection to 79.2%. We then test selectors based on visual quality, physical consistency, and task progression as controlled interventions. Some tested selectors yield higher observed success, but the gains are uneven and the matched selectors leave much of the measured opportunity unrecovered. To investigate this gap, we examine whether sampled actions lead to different outcomes, whether these differences are visible in the predictions, and whether a score recognises them. Counterfactual branching from the same states shows that selection opportunity is concentrated in relatively few decisions in the initial candidate sets. Action spread and outcome coverage need not increase together. In a further evaluation across trajectory phases with complete action execution, the tested scores again recover little of the available improvement despite a small gain from learned value. These findings distinguish producing consequential action choices from recognising them in generated futures, motivating the evaluation of WAM predictions through their usefulness for decisions rather than visual quality alone.

cs.RO↗

arXiv2026

Restore What Matters: Lessons from Joint Restoration and Recognition

Recognition pipelines typically adopt a restore-then-recognize workflow, yet decades of experience show that generating visually pleasing images seldom translates to improved recognition. We propose a Joint Restoration-for-Recognition (JR$^2$) paradigm: restore only what downstream tasks truly require, with task signals dictating where, how much, and whether restoration is necessary. JR$^2$ rests on three pillars: (i) Physics, employing optics-accurate turbulence simulation, extensible to blur and noise, to ground restoration in real image formation; (ii) Neuroscience, drawing on selective attention and neuroplasticity to direct model capacity toward identity-critical regions and frames while bypassing already-clean inputs; and (iii) Vision & Learning, coupling recognition loss end-to-end through restoration and alignment so that low-level edits maximize high-level identity stability. Evaluations on IARPA-BRIAR show consistent improvements (e.g., TAR@0.01% FAR +0.6; FNIR@1% FPIR -2.5), while a quality gate skips ~70% of clean frames, reducing cost. Ablations confirm physics priors enhance realism, joint training prevents catastrophic forgetting, and selective restoration suffices in many cases. We conclude that better-looking images are neither necessary nor sufficient; restoration modules must be task-driven, selective, and physically aware. Code, pretrained models, and recipes are provided for integration.

cs.CV↗

arXiv2026

Uncovering and Mitigating Aggregation-Induced Reward Hacking in Multi-Reward Reinforcement Learning

Reinforcement learning fine-tuning of large language models increasingly adopts multiple reward dimensions, including verifiable rules, task-specific evaluators, and learned reward models, to provide richer supervision across diverse capabilities. These dimensions are commonly scalarized with fixed aggregation weights. We identify a failure mode in which aggregation itself induces reward hacking: static projection aliases qualitatively different reward profiles into a single scalar, steering optimization toward whichever dimensions are easiest, densest, or systematically favored by the reward signal. Over training, this traps the policy in suboptimal profiles and prevents convergence to better-balanced ones that would yield higher task performance. To address this, we propose Adaptive Multi-Reward Projection (AMRP), a lightweight online method that reallocates aggregation weights using three signals, relative shortfall, reward volatility, and recent progress, increasing pressure on lagging, unstable, or stagnant dimensions while relieving saturated ones. Across structured reasoning, citation-grounded generation, and open-ended alignment under GRPO, AMRP consistently improves reward-profile balance and downstream performance over fixed and dynamic weighting baselines; it also remains effective with GDPO and PPO, supporting compatibility across RL algorithms. Our code is available at https://github.com/yyhappier/AMRP.git.

cs.CL↗

arXiv2026

OptiWorld: Optimal Control for Video World Generation under Physical Constraints

Video generation models are becoming a scalable form of world models, but they mainly generate plausible motion rather than proactively control or optimize the underlying dynamics. As a result, an object in the generated video may follow trajectories that are unsafe, not smooth, inefficient, or physically inconsistent. In this work, we propose \textbf{OptiWorld}, a framework that brings classical optimal control into video generation at inference time. OptiWorld first extracts a compact, task-relevant world state, then plans an optimal trajectory under physical constraints, and finally renders the video conditioned on this trajectory. We formulate planning as a geometric problem on a continuous manifold, which converts 3D geometry and task-dependent physical constraints into a unified planning geometry. By adding this optimal-control layer, OptiWorld generates videos with preferable dynamics, demonstrating strong potential in multiple tasks including goal-conditioned image-to-video generation, video dynamics editing, and counterfactual generation.

cs.CV↗

arXiv2026

Causal Bootstrapped Alignment for Unsupervised Video-Based Visible-Infrared Person Re-Identification

VVI-ReID is a critical technique for all-day surveillance, where temporal information provides additional cues beyond static images. However, existing approaches rely heavily on fully supervised learning with expensive cross-modality annotations, limiting scalability. To address this issue, we investigate Unsupervised Learning for VVI-ReID (USL-VVI-ReID), which learns identity-discriminative representations directly from unlabeled video tracklets. Directly extending image-based USL-VI-ReID methods to this setting with generic pretrained encoders leads to suboptimal performance. Such encoders suffer from weak identity discrimination and strong modality bias, resulting in severe intra-modality identity confusion and pronounced clustering granularity imbalance between visible and infrared modalities. These issues jointly degrade pseudo-label reliability and hinder effective cross-modality alignment. To address these challenges, we propose a Causal Bootstrapped Alignment (CBA) framework that explicitly exploits inherent video priors. First, we introduce Causal Intervention Warm-up (CIW), which performs sequence-level causal interventions by leveraging temporal identity consistency and cross-modality identity consistency to suppress modality- and motion-induced spurious correlations while preserving identity-relevant semantics, yielding cleaner representations for unsupervised clustering. Second, we propose Prototype-Guided Uncertainty Refinement (PGUR), which employs a coarse-to-fine alignment strategy to resolve cross-modality granularity mismatch, reorganizing under-clustered infrared representations under the guidance of reliable visible prototypes with uncertainty-aware supervision. Extensive experiments on the HITSZ-VCM and BUPTCampus benchmarks demonstrate that CBA significantly outperforms existing USL-VI-ReID methods when extended to the USL-VVI-ReID setting.

cs.CV↗

arXiv2026

NewtonGen: Physics-Consistent and Controllable Text-to-Video Generation via Neural Newtonian Dynamics

A primary bottleneck in large-scale text-to-video generation today is physical consistency and controllability. Despite recent advances, state-of-the-art models often produce unrealistic motions, such as objects falling upward, or abrupt changes in velocity and direction. Moreover, these models lack precise parameter control, struggling to generate physically consistent dynamics under different initial conditions. We argue that this fundamental limitation stems from current models learning motion distributions solely from appearance, while lacking an understanding of the underlying dynamics. In this work, we propose NewtonGen, a framework that integrates data-driven synthesis with learnable physical principles. At its core lies trainable Neural Newtonian Dynamics (NND), which can model and predict a variety of Newtonian motions, thereby injecting latent dynamical constraints into the video generation process. By jointly leveraging data priors and dynamical guidance, NewtonGen enables physically consistent video synthesis with precise parameter control. All data and code are available at https://github.com/pandayuanyu/NewtonGen

cs.CV↗

arXiv2026

SeeU: Seeing the Unseen World via 4D Dynamics-aware Generation

Images and videos are discrete 2D projections of the 4D world (3D space + time). Most visual understanding, prediction, and generation operate directly on 2D observations, leading to suboptimal performance. We propose SeeU, a novel approach that learns the continuous 4D dynamics and generate the unseen visual contents. The principle behind SeeU is a new 2D$\to$4D$\to$2D learning framework. SeeU first reconstructs the 4D world from sparse and monocular 2D frames (2D$\to$4D). It then learns the continuous 4D dynamics on a low-rank representation and physical constraints (discrete 4D$\to$continuous 4D). Finally, SeeU rolls the world forward in time, re-projects it back to 2D at sampled times and viewpoints, and generates unseen regions based on spatial-temporal context awareness (4D$\to$2D). By modeling dynamics in 4D, SeeU achieves continuous and physically-consistent novel visual generation, demonstrating strong potentials in multiple tasks including unseen temporal generation, unseen spatial generation, and video editing. All data and code will be public at https://yuyuanspace.com/SeeU/

cs.CV↗

arXiv2026

PhysAlign: Physics-Coherent Image-to-Video Generation through Feature and 3D Representation Alignment

Video Diffusion Models (VDMs) offer a promising approach for simulating dynamic scenes and environments, with broad applications in robotics and media generation. However, existing models often generate temporally incoherent content that violates basic physical intuition, significantly limiting their practical applicability. We propose PhysAlign, an efficient framework for physics-coherent image-to-video (I2V) generation that explicitly addresses this limitation. To overcome the critical scarcity of physics-annotated videos, we first construct a fully controllable synthetic data generation pipeline based on rigid-body simulation, yielding a highly-curated dataset with accurate, fine-grained physics and 3D annotations. Leveraging this data, PhysAlign constructs a unified physical latent space by coupling explicit 3D geometry constraints with a Gram-based spatio-temporal relational alignment that extracts kinematic priors from video foundation models. Extensive experiments demonstrate that PhysAlign significantly outperforms existing VDMs on tasks requiring complex physical reasoning and temporal stability, without compromising zero-shot visual quality. PhysAlign shows the potential to bridge the gap between raw visual synthesis and rigid-body kinematics, establishing a practical paradigm for genuinely physics-grounded video generation. The project page is available at https://physalign.github.io/PhysAlign.

cs.CV↗

arXiv2026

GeoGramBench: Benchmarking the Geometric Program Reasoning in Modern LLMs

Geometric spatial reasoning forms the foundation of many applications in artificial intelligence, yet the ability of large language models (LLMs) to operate over geometric spatial information expressed in procedural code remains underexplored. In this paper, we address this gap by formalizing the Program-to-Geometry task, which challenges models to translate programmatic drawing code into accurate and abstract geometric reasoning. To evaluate this capability, we present GeoGramBench, a benchmark of 500 carefully refined problems organized by a tailored three-level taxonomy that considers geometric complexity rather than traditional mathematical reasoning complexity. Our comprehensive evaluation of 17 frontier LLMs reveals consistent and pronounced deficiencies: even the most advanced models achieve less than 50% accuracy at the highest abstraction level. These results highlight the unique challenges posed by program-driven spatial reasoning and establish GeoGramBench as a valuable resource for advancing research in symbolic-to-spatial geometric reasoning. Project page: https://github.com/LiAuto-DSR/GeoGramBench.

cs.AI↗

arXiv2025

Progressive Image Restoration via Text-Conditioned Video Generation

Recent text-to-video models have demonstrated strong temporal generation capabilities, yet their potential for image restoration remains underexplored. In this work, we repurpose CogVideo for progressive visual restoration tasks by fine-tuning it to generate restoration trajectories rather than natural video motion. Specifically, we construct synthetic datasets for super-resolution, deblurring, and low-light enhancement, where each sample depicts a gradual transition from degraded to clean frames. Two prompting strategies are compared: a uniform text prompt shared across all samples, and a scene-specific prompting scheme generated via LLaVA multi-modal LLM and refined with ChatGPT. Our fine-tuned model learns to associate temporal progression with restoration quality, producing sequences that improve perceptual metrics such as PSNR, SSIM, and LPIPS across frames. Extensive experiments show that CogVideo effectively restores spatial detail and illumination consistency while maintaining temporal coherence. Moreover, the model generalizes to real-world scenarios on the ReLoBlur dataset without additional training, demonstrating strong zero-shot robustness and interpretability through temporal restoration.

cs.CV↗

arXiv2025

Personalized Generative Low-light Image Denoising and Enhancement

Modern cameras' performance in low-light conditions remains suboptimal due to fundamental limitations in photon shot noise and sensor read noise. Generative image restoration methods have shown promising results compared to traditional approaches, but they suffer from hallucinatory content generation when the signal-to-noise ratio (SNR) is low. Leveraging the availability of personalized photo galleries of the users, we introduce Diffusion-based Personalized Generative Denoising (DiffPGD), a new approach that builds a customized diffusion model for individual users. Our key innovation lies in the development of an identity-consistent physical buffer that extracts the physical attributes of the person from the gallery. This ID-consistent physical buffer serves as a robust prior that can be seamlessly integrated into the diffusion model to restore degraded images without the need for fine-tuning. Over a wide range of low-light testing scenarios, we show that DiffPGD achieves superior image denoising and enhancement performance compared to existing diffusion-based denoising approaches. Our project page can be found at \href{https://genai-restore.github.io/DiffPGD/}{\textcolor{purple}{\textbf{https://genai-restore.github.io/DiffPGD/}}}.

cs.CV↗

arXiv2025

Taught Well Learned Ill: Towards Distillation-conditional Backdoor Attack

Knowledge distillation (KD) is a vital technique for deploying deep neural networks (DNNs) on resource-constrained devices by transferring knowledge from large teacher models to lightweight student models. While teacher models from third-party platforms may undergo security verification (\eg, backdoor detection), we uncover a novel and critical threat: distillation-conditional backdoor attacks (DCBAs). DCBA injects dormant and undetectable backdoors into teacher models, which become activated in student models via the KD process, even with clean distillation datasets. While the direct extension of existing methods is ineffective for DCBA, we implement this attack by formulating it as a bilevel optimization problem and proposing a simple yet effective method (\ie, SCAR). Specifically, the inner optimization simulates the KD process by optimizing a surrogate student model, while the outer optimization leverages outputs from this surrogate to optimize the teacher model for implanting the conditional backdoor. Our SCAR addresses this complex optimization utilizing an implicit differentiation algorithm with a pre-optimized trigger injection function. Extensive experiments across diverse datasets, model architectures, and KD techniques validate the effectiveness of our SCAR and its resistance against existing backdoor detection, highlighting a significant yet previously overlooked vulnerability in the KD process. Our code is available at https://github.com/WhitolfChen/SCAR.

cs.CR↗

arXiv2025

Simulating Human-Like Learning Dynamics with LLM-Empowered Agents

Capturing human learning behavior based on deep learning methods has become a major research focus in both psychology and intelligent systems. Recent approaches rely on controlled experiments or rule-based models to explore cognitive processes. However, they struggle to capture learning dynamics, track progress over time, or provide explainability. To address these challenges, we introduce LearnerAgent, a novel multi-agent framework based on Large Language Models (LLMs) to simulate a realistic teaching environment. To explore human-like learning dynamics, we construct learners with psychologically grounded profiles-such as Deep, Surface, and Lazy-as well as a persona-free General Learner to inspect the base LLM's default behavior. Through weekly knowledge acquisition, monthly strategic choices, periodic tests, and peer interaction, we can track the dynamic learning progress of individual learners over a full-year journey. Our findings are fourfold: 1) Longitudinal analysis reveals that only Deep Learner achieves sustained cognitive growth. Our specially designed "trap questions" effectively diagnose Surface Learner's shallow knowledge. 2) The behavioral and cognitive patterns of distinct learners align closely with their psychological profiles. 3) Learners' self-concept scores evolve realistically, with the General Learner developing surprisingly high self-efficacy despite its cognitive limitations. 4) Critically, the default profile of base LLM is a "diligent but brittle Surface Learner"-an agent that mimics the behaviors of a good student but lacks true, generalizable understanding. Extensive simulation experiments demonstrate that LearnerAgent aligns well with real scenarios, yielding more insightful findings about LLMs' behavior.

cs.AI↗

arXiv2025

Intrinsicality of second fundamentals of hypersurfaces in space form

We prove that the normalized second fundamental form $A$ of immersed $C^2$ hypersurface in a space form $(N^{n+1}, \bar g)$ is intrinsic provided $σ_{2k+1}(A)\neq 0$ for some $k\ge 1$. We also establish the intrinsicality of the normalized second fundamental form $A$ of $M^{n}=\partialΩ$ for domain $ Ω\subset N^{n+1}, \ n\ge 3$.

math.DG↗

arXiv2025

CAD-Editor: A Locate-then-Infill Framework with Automated Training Data Synthesis for Text-Based CAD Editing

Computer Aided Design (CAD) is indispensable across various industries. \emph{Text-based CAD editing}, which automates the modification of CAD models based on textual instructions, holds great potential but remains underexplored. Existing methods primarily focus on design variation generation or text-based CAD generation, either lacking support for text-based control or neglecting existing CAD models as constraints. We introduce \emph{CAD-Editor}, the first framework for text-based CAD editing. To address the challenge of demanding triplet data with accurate correspondence for training, we propose an automated data synthesis pipeline. This pipeline utilizes design variation models to generate pairs of original and edited CAD models and employs Large Vision-Language Models (LVLMs) to summarize their differences into editing instructions. To tackle the composite nature of text-based CAD editing, we propose a locate-then-infill framework that decomposes the task into two focused sub-tasks: locating regions requiring modification and infilling these regions with appropriate edits. Large Language Models (LLMs) serve as the backbone for both sub-tasks, leveraging their capabilities in natural language understanding and CAD knowledge. Experiments show that CAD-Editor achieves superior performance both quantitatively and qualitatively. The code is available at \url {https://github.com/microsoft/CAD-Editor}.

cs.CV↗

arXiv2025

Robust Input Shaping Control for Flexible Structures Based on Unscented Kalman Filter

With the rapid development of industrial automation and smart manufacturing, the control of flexible structures and underactuated systems has become a critical research focus. Residual vibrations in these systems not only degrade operational efficiency but also pose risks to structural integrity and longevity. Traditional input shaping techniques, while effective, often suffer from performance degradation due to parameter inaccuracies and environmental disturbances. To address these challenges, this paper introduces an innovative unscented Kalman filter-based zero vibration derivative input shaping (UZS) method. The proposed approach combines two key innovations: 1) a data-driven Unscented Kalman Filterfor real-time system parameter identification, and 2) a zero-vibration derivative (ZVD) input shaper for robust vibration suppression. To validate the effectiveness of UZS, we conducted extensive experiments on a vertical flexible beam platform, and the results demonstrate significant improvements over state-of-the-art methods. Additionally, we have made the experimental datasets publicly available to facilitate further research. The findings highlight UZS's potential for practical applications in industrial automation, robotics, and precision engineering.

eess.SY↗