Searcharxiv⌕ Search

arXiv subjects

Kai Zhu

Publications and source records attributed to Kai Zhu.

At least 19 recordsLinked to original sources

The Role of Galactic Shear in Regulating Local Star Formation Efficiency: A Combined MaNGA DynPop and ALMaQUEST Study

We study the local regulation of star formation efficiency (SFE) in star-forming regions of nearby galaxies using spatially resolved data from the Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) and the ALMA-MaNGA QUEnching and STar formation (ALMaQUEST) surveys. We define the local SFE as $Σ_{\rm SFE} \equiv Σ_{\rm SFR}/Σ_{\rm H_2}$, where $Σ_{\rm SFR}$ is the star formation rate surface density derived from dust-corrected $\rm Hα$ emission, and $Σ_{\rm H_2}$ is the molecular hydrogen surface density derived from CO measurements. Combining these data with dynamical properties from the MaNGA DynPop project, we use random forest regression to identify the dominant factors of $Σ_{\rm SFE}$. Galactic shear, quantified by the Oort parameter $A_{\rm Oort}$, emerges as the dynamical property most strongly associated with lower SFE on average, exceeding the predictive power of mass surface density (either total or stellar), Toomre stability, and mid-plane pressure. However, these quantities provide only modest predictive power and primarily trace the radial variation of SFE. After removing the shear-related radial trends, only weak correlations remain between SFE and other local galactic properties, implying that the remaining azimuthal variations are likely governed by cloud-scale processes below our resolution limit. Furthermore, the relationship between SFE and shear varies among galaxies and weakens in gas-rich regions, suggesting that gas density may modulate the impact of shear. This behavior is consistent with competition between shear-induced cloud stabilization and shear-driven cloud-cloud collisions. Overall, galactic shear appears to regulate the large-scale radial distribution of SFE, while cloud-scale interstellar medium physics may drive the remaining variation.

astro-ph.GA↗

AI Search as an Information Intermediary: Clickstream Evidence on Web Traffic

AI search retrieves web information to ground generated answers. We study this new information intermediary using U.S. desktop clickstream data observing use alongside referrals during ChatGPT Search's rollout. Referrals accompany 5.4% of ChatGPT sessions versus 31.1% of Google queries. Ad-supported sites receive a smaller share of ChatGPT than Google referrals, while reference and technical sites receive larger shares. Conventional searches (Google, Bing, Yahoo) decline after ChatGPT adoption, sooner when adoption follows ChatGPT Search's expansion to users without accounts. Among existing users, that expansion raises ChatGPT use; conventional searches show no detectable overall change but fall among households with lower prior search activity. These findings suggest AI search can reshape open-web traffic through its own referrals and changes in conventional search.

cs.CY↗

EgoGeneralist: Unifying Egocentric Action-Video Tasks through Video-Native Modeling

Modern video models possess strong visual priors, yet do not natively represent human action in metric 3D space. We present EgoGeneralist, which adapts pretrained video generators to unify action-conditioned video generation, pose estimation, and joint action-video generation for egocentric manipulation. Our role-adaptive action renderer provides video-native representations for multidirectional action-video learning, making metric motion usable as both a spatial control signal and a recoverable generation target. Cross-stream action-video coupling establishes close correspondence between metric motion and visual dynamics through lightweight adaptation while preserving the pretrained visual prior. Together, these designs establish a generalist framework spanning six action-video tasks. With only narrow-domain data, EgoGeneralist generalizes to diverse unseen scenes and interactions with effective action control and visual quality, bridging pretrained visual knowledge and metric 3D human motion.

cs.CV↗

MaskSense: Confronting the Visual Exploration Trap in Masked Image Generation

Reinforcement learning (RL) has shown strong potential for aligning generative models with human intent, but adapting RL to masked generative models (MGMs) remains largely underexplored. Existing approaches that directly adapt Group Relative Policy Optimization (GRPO) to MGMs leave a foundational conflict unaddressed: MGMs' exploitative decoding nature is inherently at odds with the exploratory diversity that GRPO presupposes. In this work, we reveal a structural decoding bias termed the visual exploration trap: confidence-based sampling trades the exploration of diverse subject realizations for the greedy resolution of low-uncertainty background regions, causing premature collapse of the generative exploration space and starving GRPO of the rollout diversity it depends on. To this end, we propose MaskSense, a novel RL framework for MGMs that confronts the visual exploration trap at both the sampling and optimization stages. Specifically, we introduce a semantic anchored routing sampling mechanism that leverages semantic priors to preserve high-entropy subject token exploration while amplifying intra-group reward variance through exploratory-exploitative routing to yield more discriminative advantage estimates. Furthermore, we design a global entropy transition anchoring strategy to identify the most consequential decoding steps and concentrate gradient updates on them. Extensive experiments on multiple text-to-image benchmarks demonstrate that MaskSense substantially improves upon the base model, with GenEval accuracy improving from 54% to 86% and HPS from 28.89 to 36.39, achieving state-of-the-art performance.

cs.CV↗

On the nonvanishing of Floer lasagna modules

In this paper, we establish several nonvanishing results for Floer lasagna modules $\mathcal{FL}(W,K)$ for a smooth compact connected oriented $4$-manifold $W$ and a knot $K\subset \partial W$, via the evaluation maps to link Floer homology. We first establish a general criterion: if a certain Heegaard Floer cobordism map associated with a smooth $4$-manifold $W$ is nonzero, then $\mathcal{FL}(W,U)\neq 0$, where $U\subset \partial W$ is a local unknot. Applications include the nonvanishing of Floer lasagna modules for $\#^m\overline{\mathbb{C}P^2}$ and for knot traces $X_n(K)$ when $K$ is slice and $n$ is a non-positive integer. In another direction, we show that for every nontrivial knot $K$ and every integer $n$, the Floer lasagna module $\mathcal{FL}(X_n(K),K_n)$ is nonzero, where $K_n\subset S_n^3(K)$ is the dual knot. Moreover, if $K$ is an L-space knot, then $\mathcal{FL}(X_n(K),K_n)$ is infinite-dimensional for all integers $n\ge \max\{2g(K),1\}$.

math.GT↗

CausalAlign: Causal Alignment of Video Generation Models for Consistent Ego-Motion

Diffusion-based video generation models (VGMs) can produce visually plausible videos, yet their camera ego-motion often drifts over long horizons. Camera ego-motion determines how the viewpoint changes between frames and is essential for maintaining consistent geometry and trajectories. To address this problem, we propose CausalAlign, a causal alignment framework with two components. First, we diagnose how precise ego motion is represented, used, and retained within VGMs using linear probes and causal interventions, revealing a rise-and-fall pattern during denoising. Second, we strengthen ego-motion state in VGM by aligning the representations at the identified layers to a 3D foundation model through delta-token space. This space encodes the changes between consecutive frames, allowing the alignment to focus on ego-motion dynamics rather than static scene content. With Wan2.2 fine-tuned on DL3DV and RealEstate10K, CausalAlign can largely improve motion accuracy, geometric consistency, and temporal smoothness, while preserving visual quality. For example, it improves geometric consistency by more than 20% relative to the base model. Moreover, with the fine-tuned VGM, CausalAlign improves the success rate of world action models (WAMs) by 10 points on mobile manipulation tasks of RoboCasa365 benchmark. Overall, CausalAlign improves ego-motion consistency in diffusion-based VGMs and yields noticeable gains in embodied models.

cs.CV↗

RSIgen: Reciprocal Self-Improvement for Agentic Visual Generation

Agentic visual generation produces experience that can improve the native understanding (agent) and generation roles of a unified multimodal backbone, but task success alone does not determine which role should learn. Generator updates can also make the agent's action supervision stale. We introduce RSIgen, a task-centric reciprocal self-improvement framework linking learning allocation, supervision refresh, and recursive curriculum construction. Dynamic Responsibility Routing (DRR) uses matched interventions on evidence, inputs, and supported actions to select learning recipients and construct compatible training targets. After generator learning, RSIgen re-evaluates the relative utility of tested actions with branch inputs held fixed to refresh agent supervision; the retained roles then construct the next curriculum. Across BAGEL, Qwen-Image, and Wan-Image-7B, RSIgen improves AgentGen-Bench Full Overall9 by 3.2--3.3 points over a three-cycle SearchGen adaptation with action costs. Wan controls support allocation gains at matched role-level training exposure, refresh gains at a matched adaptation budget, and benefits from updated curriculum sources while recipients continue learning. In blind evaluation on 200 real-user requests, pooled tie-adjusted preference scores against the same baseline reach 70.0% for task completion and 64.9% for visual quality.

cs.CV↗

Towards Unified Evaluation of Prompt Enhancers for Video Generation

Modern video generators can realize increasingly complex visual narratives, positioning the prompt enhancer (PE) as a critical bridge from concise user instructions and multimodal references to structured cinematic plans. However, existing PE evaluation relies on rendered videos, imposing substantial computational and human costs, slowing PE training and iteration, and conflating PE quality with downstream generator behavior. To address this gap, we introduce PEBench, the first unified benchmark for direct PE evaluation across text-to-video, image-to-video, and reference-to-video prompt enhancement. It comprises 1,100 expert-verified cases and 1,005 visual assets, spanning 35 fine-grained tasks with diverse temporal, cinematic, audiovisual, and multi-reference requirements. In addition, we develop PEBench evaluation, an evidence-grounded framework that combines modality-aware fact extraction with rubric-based assessment across 24 criteria. Our systematic evaluation of representative open- and closed-source PE methods reveals an emerging shift from fine-grained descriptive expansion toward intent-preserving cinematic planning, while the caption-reconstruction and forward-refinement methods show complementary strengths in cinematic coverage and semantic fidelity or internal coherence, respectively. Human validation shows that PEBench scores align closely with expert judgments of enhanced prompts and downstream videos from Wan3.0 and MiniMax-H3, indicating that prompt-level evaluation reliably reflects downstream utility.

cs.CV↗

WanPE: Towards Cinematic Prompt Enhancement for Modern Text-to-Video Generation

Video generation begins in text space by authoring a cinematic screenplay, then materializes into pixels. As contemporary video generators scale to 30 seconds and faithfully follow complex conditions, the textual prompt largely directs the production, planning how actions, camera trajectories, lighting, and sound unfold across multi-shot sequences. In this paper, we present WanPE, a 397B-parameter prompt enhancement model trained on 1.05M real-world videos to master director-level cinematic planning. WanPE formulates shot-level cinematic plans via video-grounded reverse construction and employs Semantic-Consistency GRPO (SC-GRPO) to faithfully preserve user requirements across shots and over time. To benchmark this capability, we curate WanPEval, a human-annotated testbed covering durations from 5 to 30 seconds across varying intent granularities, supported by approximately 11K blind pairwise assessments. When powering Wan3.0's video generator, WanPE-397B boosts human preference over raw user prompts by 10.66-18.84 points at 5-15 seconds and by a dramatic 50.86 points in the 30-second arena. Ablation studies show that reverse construction demonstrates clear superiority over forward rewriting, while SC-GRPO robustly preserves semantic fidelity across model scales. Ultimately, WanPE leads all evaluated commercial offerings at 5-15 seconds and remains competitive with Seedance 2.5 at 30 seconds.

cs.CV↗

Code Plans, Diffusion Renders: Open-Ended Generative World Modeling

We introduce \textbf{CoDeR}, a new paradigm for world modeling. Unlike existing video world models that implicitly represent world dynamics through visual observations, our system explicitly constructs an executable world with code and employs video generation models for visual realization. Specifically, we coordinate five complementary roles to translate high-level concepts into structured world rules, executable dynamics, and perceptual observations. This design enables \textit{long-term memory}, \textit{open-ended interactions}, \textit{autonomous world evolution}, and \textit{multi-agent scenarios}, where multiple entities can act, interact, and evolve persistently beyond the current observation. Extensive experiments demonstrate that our framework substantially extends the capabilities of existing world models, enabling long-term memory, open-ended interactions, autonomous evolution, and persistent multi-agent dynamics, while achieving state-of-the-art performance across multiple evaluation settings. Code and model weights will be made publicly available. Project Page: \href{https://becauseimbatman0.github.io/CoDeR}{CoDeR}.

cs.CV↗

SelfTICA: contrastive learning of dynamical representations for rare-event sampling and characterization

Rare events govern many important molecular processes but remain difficult to characterize within accessible simulation timescales. This challenge has motivated machine-learning methods to construct low-dimensional collective variables for specific objectives, including state discrimination, slow-mode identification, and committor-function approximation. Here we introduce SelfTICA, a self-supervised framework that uses contrastive learning on time-lagged configurations to learn a latent representation of the slow modes governing relevant transitions. Once learned, this representation is frozen and reused across downstream tasks, including collective variables construction, enhanced sampling, free-energy estimation, and committor learning. Compared with direct slow-mode optimization, SelfTICA improves training stability and provides collective variables even from limited and exploratory trajectories, accelerating rare-event sampling and enabling accurate free-energy estimation. The dynamical information encoded during pretraining also accelerates committor learning and enables characterization of transition-state regions. These results show that contrastive learning of slow dynamical representations provides a common foundation for rare-event sampling and mechanistic analysis.

physics.comp-ph↗

Joint-Conditioned Stereo Surface Reasoning for Interaction Field Estimation

Predicting hand--object interaction fields requires locating the nearest object-surface point for each hand joint, often from small and partially occluded image regions. We view this task as joint-conditioned surface-endpoint estimation: each joint has its own nearest endpoint, while endpoints from the same hand can draw on shared local surface evidence. This structure motivates Joint-Conditioned Stereo Surface Reasoning (JSSR). A temporal-stereo network jointly predicts 3D joints, a direct interaction field, and per-view endpoint evidence. Calibrated candidate search evaluates endpoint hypotheses using joint-specific image compatibility and cross-view correspondence. A hand-shared candidate support lets joints draw on common surface evidence, and a learned residual gate controls the geometric correction when observations are ambiguous. Our system built on this method ranked third on the SHOW3D Interaction Field Challenge leaderboard.

cs.CV↗

Back on Track: Aligning Rewards and States for Reasoning in Diffusion Large Language Models

Reinforcement learning (RL) holds immense promise for enhancing the reasoning capabilities of diffusion large language models (dLLMs). However, progress is fundamentally constrained by a dual misalignment between authentic generation trajectory and the gradient update process: (i) Process-reward misalignment. Sparse, terminal rewards are indiscriminately assigned to all intermediate steps of the generation process, failing to provide discriminative credit assignment. (ii) State-trajectory misalignment. Policy updates are often diverted toward artificial, out-of-trajectory states, squandering gradients on less informative samples. To address these limitations, we introduce Process Aligned Policy Optimization (PAPO), a novel framework that holistically aligns the RL update with the dLLM's generative trajectory via Step-Aware Process Rewards (SPR) that transform sparse terminal rewards into dense, step-wise credit, and Entropy-Guided Historical Re-enactment (EHR) that replays authentic trajectories at high-uncertainty steps. Extensive experiments on four benchmarks demonstrate that PAPO significantly outperforms baselines, achieving gains of 4.5% on GSM8K, 4.8% on MATH500, 42.2% on Countdown and 16.1% on Sudoku.

cs.CL↗

Adaptive Nesterov Momentum Method for Electrical Impedance Tomography with the Complete Electrode Model

We apply the adaptive Nesterov momentum (ANM) method [30] to electrical impedance tomography under the complete electrode model. The forward problem is formulated in the variational CEM setting, accounting for finite electrode size, contact impedance, insulating gaps, and the mean-free voltage gauge. The resulting nonlinear inverse problem is treated within a unified dual-to-primal framework using three classes of strongly convex structural penalties: an L2-type penalty, an L1-type penalty promoting sparse deviations from a calibrated homogeneous background, and a TV-type penalty favoring approximately piecewise-constant conductivities with sharp interfaces. The TV class is implemented using both smoothed TV and Huber-TV formulations. The data- misfit gradient is computed through CEM adjoint equations and stabilized by Sobolev smoothing. The method is evaluated on the publicly available KIT4 tank measurement data after calibration of the background conductivity and contact impedance from no-object measurements. The experiments include single, multiple, mixed-conductivity, and geometrically challenging phantom configurations. The L2-type penalty generally produces smooth but diffuse reconstructions, whereas the L1-type penalty yields cleaner backgrounds with occasional geometric distortion. The smoothed TV and Huber-TV penalties provide more spatially coherent localization and exhibit similar reconstruction behavior across most tested configurations. These results demonstrate the practical applicability of the adaptive Nesterov framework to measured CEM-EIT data.

math.NA↗

When Does Visual Generation Help Visual Understanding in Unified Multimodal Models?

Unified multimodal models (UMMs) can perform both understanding and generation, raising a central question: can visual generation improve understanding? Existing evaluations provide mixed evidence, but confound task difficulty, reasoning paradigms, and the closed-loop interaction between generation and understanding. We introduce VGAU-Diag, a fine-grained evaluation framework for vision generation-assisted understanding. It stratifies samples by difficulty, enables unified evaluation of multiple reasoning paradigms, and uses Oracle-Assisted Reference Protocols. Our analysis shows that generated visual aids help on easier instances but become unreliable as reasoning complexity increases. Oracle-assisted diagnosis further reveals that the main bottleneck often lies on the visual-understanding side rather than the visual-generation side, as current UMMs struggle to leverage even faithful visual aids. We also show that effective visual generation should target visual-understanding bottlenecks rather than add more reasoning steps, and identify a three-stage transition from task-irrelevant noise, to misleading plausible guidance, and finally to useful assistance. These findings would be useful to guide the development of better UMMs.

cs.CV↗

Distinguishing the nature of dark matter by mapping cosmic filaments from Lyman-alpha emission

The standard $Λ$CDM cosmological model predicts that cosmic filaments are highly clumpy, whereas warm dark matter -- invoked to address small-scale challenges in $Λ$CDM -- produces filaments that are noticeably smoother and less structured. In this work, we investigate the potential of Lyman $α$ (Ly$α$) emission to trace cosmic filaments at redshifts $z=2.5$ and $z=4$, and assess their potential for constraining the nature of dark matter. Our analysis shows that Ly$α$ filaments provide a promising observational probe of dark matter: at $z=4$, differences in filament smoothness and surface brightness serve as distinctive signatures between models. Looking ahead, the upcoming generation of 30-meter class telescopes will be critical for enabling these measurements, offering a compelling opportunity to distinguish the nature of dark matter by mapping the structure of cosmic filaments.

astro-ph.CO↗

Open-AoE: An Open Egocentric Manipulation Dataset and Toolchain for Embodied Learning

Egocentric videos of human manipulation provide scalable supervision for embodied intelligence, yet existing resources rarely combine low-cost continuous capture, manipulation-level structured annotations, and reusable tools for robot learning. We present Open-AoE, an open, community-oriented egocentric manipulation dataset and toolchain spanning the full pipeline from smartphone capture to model training. Its first release contains approximately 2,000 hours of manipulation video collected in natural environments by 500+ contributors using 400+ smartphones. The dataset provides text annotations, MANO-based hand poses, camera trajectories, and temporally localized atomic actions. Open-AoE further includes a data processing pipeline that transforms raw recordings into structured samples through temporal action segmentation, semantic annotation, hand reconstruction, and camera trajectory reconstruction. Meanwhile, we provide a separate downstream toolchain supports visualization, cross-embodiment retargeting, model-specific data conversion, and training recipes for VLA policies, WAMs, and World Models. By integrating scalable capture, structured processing, and downstream adaptation, Open-AoE reduces the barriers to both data contribution and reuse, providing practical open infrastructure for embodied model training, human-to-robot transfer, and world modeling.

cs.RO↗

Video = World + Event Stream

We present Wan-Streamer v0.3, which reframes our native-streaming interaction model under a single organizing view: a video is a world plus an event stream. The world is the persistent context in which a video unfolds, including the environment, scene, subjects, ambient acoustic conditions, voice characteristics, and other relatively stable conditions. The event stream is everything that changes over time within that world, including scene or environmental changes, subject behavior, speech, and other sounds. This yields a general-purpose pretraining task over large amounts of real video: given a world and incoming input, predict how the world moves, changes, and responds in real time. The resulting competence can be specialized to a broad family of real-time downstream tasks. We instantiate it on real-time full-duplex audio-visual interaction, where the event stream is the agent's speech together with free-form behavior. Functionally, the model's multimodal understanding process is vision-language-action-like: it maps multimodal user input to language-form speech and behavior actions. Wan-Streamer v0.3 preserves the v0.2 operating point: 640x368 video at 25 FPS, a 160 ms streaming unit, approximately 200 ms model-side response latency, and approximately 550 ms total interaction latency under a 350 ms bidirectional network budget.

cs.CV↗