arXiv subjects

Zihan Zhou

Publications and source records attributed to Zihan Zhou.

At least 37 records · Page 2Linked to original sources

arXiv2026

Modeling Relativistic Tidal Disruptions of MESA Stars

Tidal disruption events (TDEs) occur when a star passes so close to a black hole that its self-gravity is overcome by the external tidal field. As the star passes, it initially deforms, then is ripped apart, and some of its material eventually falls back on bound orbits, forming an accretion disk around the black hole. A Newtonian model of TDEs, based on stellar perturbation theory of MESA stars, was recently introduced as an alternative to computationally intensive hydrodynamical simulations. In this work, we add relativistic corrections to the model, incorporating equatorial Kerr geodesics, relativistic tidal fields, and relativistic fallback times. Compared to the Newtonian case, we find that stars are disrupted earlier in their orbit, which gives them less time to accumulate physical deformations. Additionally, we find that the increased distance from the black hole at the time of disruption makes the fallback time longer. However, the black hole spin has a negligible impact on fallback time, except for orbits with exceptionally close pericenter. Our results allow for a more accurate calculation of fallback rates than the Newtonian model, while also remaining computationally cheap. The code is available on GitHub.

astro-ph.HE↗

arXiv2026

Order Matters in Retrosynthesis: Structure-aware Generation via Reaction-Center-Guided Discrete Flow Matching

Template-free retrosynthesis methods treat the task as black-box sequence generation, limiting learning efficiency, while semi-template approaches rely on rigid reaction libraries that constrain generalization. We address this gap with a key insight: atom ordering in neural representations matters. Building on this insight, we propose a structure-aware template-free framework that encodes the two-stage nature of chemical reactions as a positional inductive bias. By placing reaction center atoms at the sequence head, our method transforms implicit chemical knowledge into explicit positional patterns that the model can readily capture. The proposed RetroDiT backbone, a graph transformer with rotary position embeddings, exploits this ordering to prioritize chemically critical regions. Combined with discrete flow matching, our approach decouples training from sampling and enables generation in 20--50 steps versus 500 for prior diffusion methods. Our method achieves state-of-the-art performance on both USPTO-50k (61.2% top-1) and the large-scale USPTO-Full (51.3% top-1) with predicted reaction centers. With oracle centers, performance reaches 71.1% and 63.4% respectively, surpassing foundation models trained on 10 billion reactions while using orders of magnitude less data. Ablation studies further reveal that structural priors outperform brute-force scaling: a 280K-parameter model with proper ordering matches a 65M-parameter model without it.

cs.LG↗

arXiv2026

Gradient Under Microscope: Benchmarking Resource Utilization of Memory-Efficient Gradient Computation Methods

AI training's rising resource intensity is straining electricity supplies and carbon budgets, motivating systematic study of memory-efficient training on constrained hardware. We benchmark five gradient optimizers (SGD, Adam, Adagrad, Adadelta, and Conjugate Gradient Descent) under three memory strategies (standard training, gradient checkpointing, and gradient accumulation) across four transformer architectures (ViT, ModernBERT, Llama 3.1 1B, and NanoVLM), measuring training loss, GPU utilization, training time, and memory usage. Gradient accumulation emerges as the most reliable strategy, cutting training loss by roughly an order of magnitude on the vision-language model and about four-fold on the language model without additional GPU memory. Contrary to common practice, Adam is not universally superior: Adadelta and SGD outperform it on the encoder and autoregressive architectures. Gradient checkpointing's effect is strongly architecture-dependent, improving vision transformer loss while severely degrading the encoder model, and it increases training time by up to 60% on memory-bound models. GPU utilization is governed primarily by architecture, ranging from 8-15% for the memory-bound language model to 96-99% for compute-bound vision models. These findings provide practical guidelines for optimizer and gradient-strategy selection in resource-efficient model training and deployment.

cs.LG↗

arXiv2026

SR-JEPA: Learning Predictive Latent State in 3D Scenes

Joint-embedding predictive architectures learn by predicting latent representations of missing observations, yet many masked JEPAs are evaluated primarily through the encoders they produce. We ask what a trained predictive pathway itself infers when an entire entity is absent from a native 3D scene. We introduce SR-JEPA, a point-native JEPA for scene-scale point clouds whose original frozen predictive pathway can be queried at a supplied location. At evaluation, every point of one object is removed before encoding and replaced by the same shape-free 32-point query at its centroid. Training uses only self-contained 3D EMA targets: no reconstruction, semantic labels, language, or lifted 2D features. On 5,953 held-out ARKitScenes objects, the imputed latent reaches 43.13% semantic-identity macro accuracy, 22.18 points above the strongest floor. Randomizing the prediction path removes 9.78 points, while substituting matched donor context removes 21.98 points. On 8,570 Sr3D support pairs, the full latent reaches 41.15 AP; identity decoded from the missing-object latent, combined with anchor identity and geometry, reaches 39.37 AP, leaving an unresolved 1.78-point residual. These results reveal a queryable, compositional 3D predictive state: the model completes context-dependent entity content, which downstream computation combines with metric geometry.

cs.CV↗

arXiv2026

GeoWorld: Providing Full-frame Geometry Features to Facilitate 3D Scene Generation

Previous works that leverage video models for image-to-3D scene generation often suffer from geometric distortions and blurry content. Using video generation models to implicitly maintain geometric consistency according to a single-frame input is ineffective. In this paper, we present a two-stage method, named $\textbf{GeoWorld}$, that renovates the image-to-3D scene generation pipeline by providing full-frame geometry features. The first-stage video generation model, followed by a multi-view geometry model, produces $\textbf{full-frame}$ geometry features, which are then used as a mental draft of geometric conditions to aid the second-stage video-generation model. A geometric loss is proposed to impose real-world geometric constraints, and a geometry adaptation module is introduced to ensure the effective utilization of geometry features. Thanks to full-frame geometric modeling, the two smaller video models in our two-stage method can generate higher-fidelity 3D scenes than SOTA methods, while being even faster, e.g. 7.5$\times$ faster than Hunyuan-Voyager. Project page: https://peaes.github.io/GeoWorld.

cs.CV↗

arXiv2026

ReinforceGen: Hybrid Skill Policies with Automated Data Generation and Reinforcement Learning

Long-horizon manipulation has been a long-standing challenge in the robotics community. We propose ReinforceGen, a system that combines task decomposition, data generation, imitation learning, and motion planning to form an initial solution, and improves each component through reinforcement-learning-based fine-tuning. ReinforceGen first segments the task into multiple localized skills, which are connected through motion planning. The skills and motion planning targets are trained with imitation learning on a dataset generated from 10 human demonstrations, and then fine-tuned through online adaptation and reinforcement learning. When benchmarked on the Robosuite dataset, ReinforceGen reaches 80% success rate on all tasks with visuomotor controls in the highest reset range setting. Additional ablation studies show that our fine-tuning approaches contribute to an 89% average performance increase. Finally, ReinforceGen demonstrates significant improvement through fine-tuning in our real-world evaluations. More results and videos are available at https://reinforcegen.github.io.

cs.RO↗

arXiv2026

Bricker to BRACE: A Bracket Exposure RAW Dataset and Restoration Model for Flicker-Banding

Flicker-banding (FB), arises from temporal aliasing between a camera's rolling shutter and a display's brightness modulation, degrading screen-captured image readability with color shifts and jagged patterns. Existing single-frame methods with simplified parametric stripe models cannot reliably distinguish these artifacts from genuine texture. To address this, we conduct a systematic analysis of complex FB morphologies and reveal their significant variation across exposure settings, motivating a multi-frame bracketed RAW restoration paradigm. We construct Bricker, a synthetic-real bracketed RAW dataset built via ray-tracing-based physical simulation and automated multi-exposure capture tool. We further propose BRACE: Bracketed RAW Flicker-Banding Removal, a multi-frame restoration model that utilizes frequency-aware banding prior and a multi-scale spatial cross-attention modulator (MSCAM) for cross-exposure spatial fusion. We also introduce the Stripe Frequency Consistency (SFC) metric to evaluate banding removal. Experiments demonstrate state-of-the-art performance on both synthetic and real benchmarks. Our dataset and code are available at: https://github.com/ZZH-qwq/BRACE.

cs.CV↗

arXiv2026

Freqformer: Image-Demoiréing Transformer via Effective Frequency Decomposition

Image demoiréing remains a challenging task due to the complex interplay between texture corruption and color distortions caused by moiré patterns. Existing methods, especially those relying on direct image-to-image restoration, often fail to disentangle these intertwined artifacts effectively. While frequency-aware approaches offer a promising direction, their potential is hindered by the discrete transform (e.g., Haar wavelet or block-based DCT), which may suffer from spatial discontinuity, channel redundancy, and further cause error accumulation during their fixed inverse processes. In this paper, we present Freqformer, a Transformer-based framework specifically designed for image demoiréing through targeted frequency separation. Our method performs an effective frequency decomposition that splits moiré patterns into high-frequency spatially-localized textures and low-frequency scale-robust color distortions, which are then handled by a dual-branch architecture and an asymmetric training scheme tailored to their distinct characteristics. We further propose a learnable Frequency Composition Transform (FCT) module to adaptively fuse the frequency-specific outputs, enabling consistent and high-fidelity reconstruction. To better aggregate the spatial dependencies and the inter-channel complementary information, we introduce a Spatial-Aware Channel Attention (SA-CA) module that refines moiré-sensitive regions without incurring high computational cost. Extensive experiments on various demoiréing benchmarks demonstrate that Freqformer achieves state-of-the-art performance with a compact model size. The code will be made publicly available at https://github.com/xyLiu339/Freqformer.

cs.CV↗

arXiv2026

Time-Invariant Neural Operators with Applications in Solving Time-Dependent PDEs

The deep operator network (DeepONet) is one of the basic architectures for learning nonlinear operators with neural networks. However, for operators that describe the dynamic response of physical systems, DeepONet does not naturally respect fundamental time properties, including time causality and time invariance. We propose time-invariant neural operator (TINO) to overcome these limitations, by creating the connection between time-delay neural network (TDNN) and DeepONet. We then extend the proposal to solving time-dependent PDEs with initial values, which can be treated as truncated time-invariant systems, with spatial proper orthogonal decomposition (POD) implementation in the output function space. Various numerical tests comparing TINO, neural operators with time causality, and those without time properties justify the enhanced precision of the proposed neural operator frameworks.

math.NA↗

arXiv2026

Surface Water Wave Scattering and the Hydrotope

We study the classical tree-level scattering amplitudes of deep-water surface gravity waves using the methods of high-energy physics. For scattering in one horizontal dimension and in the two-negative-wavenumber sector we obtain a closed formula for $n$-wave scattering. Up to a kinematic prefactor, the amplitude is the volume of a classic polytope -- a box sliced by a hyperplane, which we dub the hydrotope, whose purpose in life is simply to organize the sign patterns of the "chambers" characterizing all the different regions of the two-minus kinematic space. The general formula was discovered by Claude Opus 4.6 working under our guidance, beginning with our earlier discovery of a one-term expression valid in the "simplest" kinematic chamber. Our results resolve the puzzle raised by Y.V. Lvov's 1997 computation of the five-wave amplitudes, unifying and extending it to all multiplicities.

hep-th↗

arXiv2026

Rethinking the Role of Efficient Attention in Hybrid Architectures

Modern language models increasingly adopt hybrid architectures that combine full attention with efficient attention modules, such as sliding-window attention (SWA) and recurrent sequence mixers. However, how these efficient modules shape model capabilities remains poorly understood. To address this gap, we conduct a systematic analysis across hybrid architectures from three perspectives: scaling behavior, mechanism analysis, and architecture design. First, from a scaling perspective, we find that efficient-attention design primarily affects how fast long-context capability emerges, while different hybrids eventually converge to comparable long-context performance under sufficient training. Second, mechanistically, we show that long-range retrieval is mainly carried by full attention, whereas efficient attention shapes its optimization trajectory. This explains a counter-intuitive phenomenon we call Large-Window Laziness: larger SWA windows can delay the formation of retrieval heads in full-attention layers. Third, guided by this mechanism, we show that applying NoPE to only the full-attention layers of a small-window SWA hybrid substantially improves long-context performance with negligible impact on short-context performance.

cs.CL↗

arXiv2026

Translationally Covariant Modulated Symmetries: Classification and Goldstone

Modulated symmetries are global symmetries with a spatially dependent unit of charge, such as the dipole symmetry and the exponential symmetry. We give the generic condition for a modulated symmetry to be compatible with translationally symmetric Hamiltonians, which we define as a translationally covariant modulated symmetry (TCMS). For Abelian TCMSs, we prove that their units of charge can only contain multipole, exponential and harmonic components. Particularly, we classify all the one-dimensional TCMSs by real Jordan normal form blocks. We further derive the generic Goldstone action for SSB phases of continuous TCMSs, by which we show that a broken multipole symmetry gives higher-order gapless Goldstone modes, a broken harmonic symmetry gives gapless Goldstone modes at finite momenta, and a broken exponential symmetry gives no gapless Goldstone modes, modifying the conventional Goldstone theorem.

cond-mat.str-el↗

arXiv2026

CollabBench: Benchmarking and Unleashing Collaborative Ability of LLMs with Diverse Players via Proactive Engagement

While LLM-based agents excel at individual tasks, effective collaboration with realistic human partners remains challenging. Most of the existing conversation-level collaborative studies lack grounded interaction and behavioral execution, motivating the need for cooperative game environments that enable contextualized and immersive collaboration. To this end, this paper proposes CollabBench, a benchmark for evaluating and training collaborative agents in cooperative games. CollabBench features a Diverse Player Profile Simulation pipeline to model varied players behaviors, and a Collaborative Agentic Training paradigm that unifies reasoning, communication, and action via agentic rollouts, optimized with a hybrid reward balancing task efficiency and affective adaptation. We further extend classic environments to CWAH-MultiPlayer and Cook-MultiPlayer for systematic evaluation under diverse personalities. Experiments with efficiency and affective metrics show that our trained models outperform base models, achieving 19.5% higher efficiency and 24.4% improved affective performance. Further analysis reveals key collaborative limitations of existing models and offers insights for future collaborative training.

cs.CL↗

arXiv2026

AMix-2: Establishing Protein as a Native Modality in Large Language Models

We present AMix-2, a protein-text foundation model that establishes protein as a native modality in large language models (LLMs), unifying protein understanding and sequence design within a single foundation model. AMix-2 is built upon two key ideas: (1) a unified protein-text formulation that embeds natural language and protein sequence in a shared token space, enabling one model to perform biological reasoning and conditional design instead of separate downstream task-specialized models; and (2) a block-wise diffusion language modeling backbone that combines causal generation across blocks with bidirectional context and iterative refinement within blocks. This scheme better matches the intrinsic nature of proteins than a strict left-to-right factorization. To evaluate protein foundation models under realistic generalization settings, we further introduce ProteinArena, a comprehensive benchmark with time-aware and homology-aware protocols across various understanding and design tasks, and with baselines covering classical bioinformatics tools, protein-specialized models and LLMs. On ProteinArena, AMix-2 outperforms frontier LLMs and demonstrates competitive performance to task-specific protein models. Controlled experiments further show that the diffusion-based paradigm generally surpasses its autoregressive counterpart, highlighting the advantage of flexible generation order for protein sequences. We release both AMix-2 and ProteinArena to facilitate open research in protein foundation models.

q-bio.BM↗

arXiv2026

VDFP: Video Deflickering with Flicker-banding Priors

Capturing digital screens with smartphones frequently induces severe banding due to hardware synchronization mismatches. Existing video restoration methods struggle with these structured, periodic luminance fluctuations, often resulting in residual artifacts or over-smoothed textures. We firstly construct DeViD, a real-world dataset in various scenes to deal with the lack of available datasets. Then we propose VDFP (Video Deflickering with Flicker-banding Priors), a novel perception-guided generation framework. First, we introduce a Degradation Field Modeling Based on Rolling Shutter Mechanism (DFM) capable of synthesizing complex multi-banding scenarios. Second, we present a spatial-temporal continuous prior perception (CPP). Unlike traditional binary segmentation, this module is optimized via a Flicker-Aware Mean Squared Error (FA-MSE) to capture the luminance transitions. By zero-initializing an augmented input layer, our model preserves pre-trained generative priors as well as spatial-temporal prior perception. Extensive experiments demonstrate that VDFP significantly outperforms other methods, eliminating complex banding with high-fidelity spatial details and temporal consistency. Our dataset and code will be released at https://github.com/ZhiyiZZhou/VDFP.

cs.CV↗

arXiv2026

Core-Halo Decomposition: Decentralizing Large-Scale Fixed-Point Problems

We study solving large-scale fixed-point equation $x^\star=\bar F(x^\star)$ with decomposition. Standard strict decomposition assigns each agent a disjoint block and evaluates updates using only owned coordinates. For most operators, however, a block update may depend on variables outside the block. Truncating these dependencies by strict decomposition changes the mean operator and creates structural bias that cannot be removed by more samples, smaller stepsizes, or additional consensus. We therefore propose Core-Halo decomposition, which separates write ownership from read-only evaluation context: each agent updates its own core and reads from an overlapping halo. By aligning the Core-Halo decomposition with the block-dependence structure of $\bar F$, the original fixed-point problem can be implemented faithfully in a decentralized multi-agent system. We further characterize the fundamental obstruction faced by strict decomposition through a Bellman closure condition and a blockwise bias lower bound, showing that local-only updates can alter the original fixed-point operator. Finally, we conduct extensive experiments across a range of application settings, and demonstrate that Core-Halo achieves near-centralized performance while retaining the parallelism benefits of decentralization.

stat.ML↗

arXiv2026

Towards Reliable LLM Evaluation: Correcting the Winner's Curse in Adaptive Benchmarking

Adaptive prompt and program search makes LLM evaluation selection-sensitive. Once benchmark items are reused inside tuning, the observed winner's score need not estimate the fresh-data performance of the full tune-then-deploy procedure. We study inference for this procedure-level target under explicit tuning budgets. We propose SIREN, a selection-aware repeated-split reporting protocol that freezes the post-search shortlist, separates splitwise selection from held-out evaluation, and uses an item-level Gaussian multiplier bootstrap for uncertainty quantification. In a fixed-shortlist regime with smooth stabilized selection, the estimator admits a first-order item-level representation, and the bootstrap yields valid simultaneous inference on a finite budget grid. This supports confidence intervals for procedure-performance curves and pre-specified equal-budget and cross-budget comparisons. Controlled simulations and MMLU-Pro tuning experiments show that winner-based reporting can be optimistic and can change deployment conclusions, while SIREN remains close to the finite-sample reporting target.

stat.ML↗

arXiv2026

Incomplete Data, Complete Dynamics: A Diffusion Approach

Learning physical dynamics from data is a fundamental challenge in machine learning and scientific modeling. Real-world observational data are inherently incomplete and irregularly sampled, posing significant challenges for existing data-driven approaches. In this work, we propose a principled diffusion-based framework for learning physical systems from incomplete training samples. To this end, our method strategically partitions each such sample into observed context and unobserved query components through a carefully designed splitting strategy, then trains a conditional diffusion model to reconstruct the missing query portions given available contexts. This formulation enables accurate imputation across arbitrary observation patterns without requiring complete data supervision. Specifically, we provide theoretical analysis demonstrating that our diffusion training paradigm on incomplete data achieves asymptotic convergence to the true complete generative process under mild regularity conditions. Empirically, we show that our method significantly outperforms existing baselines on synthetic and real-world physical dynamics benchmarks, including fluid flows and weather systems, with particularly strong performance in limited and irregular observation regimes. These results demonstrate the effectiveness of our theoretically principled approach for learning and imputing partially observed dynamics.

cs.LG↗