arXiv subjects

Huanyu Zhang

Publications and source records attributed to Huanyu Zhang.

At least 19 recordsLinked to original sources

arXiv2026

OpenGPT-4o-Image: A Comprehensive Dataset for Advanced Image Generation and Editing

The performance of unified multimodal models for image generation and editing is fundamentally constrained by the quality and comprehensiveness of their training data. While existing datasets have covered basic tasks like style transfer and simple object manipulation, they often lack the systematic structure and challenging scenarios required for real-world applications. To address this bottleneck, we introduce OpenGPT-4o-Image, a large-scale dataset constructed using a novel methodology that combines hierarchical task taxonomy with automated data generation. Our taxonomy not only includes fundamental capabilities such as text rendering and style control but also introduces highly practical yet challenging categories like scientific imagery for chemistry illustrations and complex instruction editing requiring simultaneous execution of multiple operations. Through an automated pipeline leveraging structured resource pools and GPT-4o, we generate 80k high-quality instruction-image pairs with controlled diversity, covering 11 major domains and 51 subtasks. Extensive experiments show that fine-tuning leading models on our dataset achieves significant performance gains across multiple benchmarks, with improvements of up to 18\% on editing tasks (UniWorld-V1 on ImgEdit-Bench) and 13% on generation tasks (Harmon on GenEval). Our work demonstrates that systematic data construction is key to advancing multimodal AI capabilities.

cs.CV↗

arXiv2026

CoCo: Code as CoT for Text-to-Image Preview and Rare Concept Generation

Recent advancements in Unified Multimodal Models (UMMs) have significantly advanced text-to-image (T2I) generation, particularly through the integration of Chain-of-Thought (CoT) reasoning. However, existing CoT-based T2I methods largely rely on abstract natural-language planning, which lacks the precision required for complex spatial layouts, structured visual elements, and dense textual content. In this work, we propose CoCo (Code-as-CoT), a code-driven reasoning framework that represents the reasoning process as executable code, enabling explicit and verifiable intermediate planning for image generation. Given a text prompt, CoCo first generates executable code that specifies the structural layout of the scene. The code is then executed in a sandboxed environment to render a deterministic draft image. Subsequently, the model refines this draft through fine-grained image editing to produce the final high-fidelity result. To support this training paradigm, we construct CoCo-10K, a curated dataset containing structured draft-final image pairs designed to teach both structured draft construction and corrective visual refinement. Empirical evaluations on StructT2IBench, OneIG-Bench, and LongText-Bench show that CoCo achieves improvements of 68.83%, 54.8%, and 41.23%, respectively, over direct generation, while also outperforming other CoT-enhanced generation methods. These results demonstrate that executable code is an effective and reliable reasoning paradigm for precise, controllable, and structured text-to-image generation.

cs.AI↗

arXiv2026

Minimizing Quantized Semantic Age of Information (QSAoI) in Foundation Model-Based Semantic Communications

The emerging techniques of semantic communications and edge computing in 6G networks necessitate a paradigm shift toward co-designed semantic-aware and adaptive resource allocation for short-packet transmissions. However, there is a fundamental gap between the semantic layer and the physical layer under low-latency finite blocklength (FBL) effects. To bridge this gap, we introduce the Quantized Semantic Age of Information (QSAoI), a novel metric that rigorously captures the trade-offs among freshness and semantic efficiency of high-level features in real-time communication in the FBL regime. Guided by this metric, we propose a novel foundation model-based efficient co-designed framework to minimize the expected QSAoI over wireless fading channels in latency-constrained semantic communication. Specifically, we formulate a non-linear joint optimization problem to dynamically optimize the block-wise mixed-precision quantization (MPQ) strategy and the physical blocklength. To efficiently resolve this complex problem, we develop a high-efficiency low-complexity algorithm based on fixpoint inspection and bisection search. Extensive simulations validate that our proposed algorithm dynamically adapts the semantic quantization precision to varying channel conditions, effectively minimizing the expected QSAoI compared to baselines.

eess.SP↗

arXiv2026

Multiferroicity in the two-dimensional limit in hexagonal LuFeO3 films

Multiferroic oxides, which combine coupled ferroelectric and magnetic orders, are central to understanding correlated quantum phenomena. Yet, as thickness approaches the two-dimensional (2D) limit, both ferroelectricity and magnetism are conventionally expected to vanish due to depolarization fields and finite-size effects, respectively. Here, we demonstrate that hexagonal LuFeO3 (h-LuFeO3) retains coupled ferroelectricity and magnetism at the 2D limit, with a thickness of just one and a half unit cells. Remarkably, the ferroelectric polarization remains comparable to bulk values at room temperature, while long-range magnetism and magnetoelectric coupling persist at low temperatures. We further show that the K3 phonon mode, which mediates the polarization-magnetism coupling, is stable down to the 2D limit. Our results establish h-LuFeO3 as the first oxide system to exhibit genuine 2D-limit multiferroicity, providing a fundamental breakthrough in the long-standing quest to understand and control coupled ferroic orders at the atomic scale.

cond-mat.mtrl-sci↗

arXiv2026

RIFTES: An RTM- and iteration-free temperature-emissivity separation framework for accurate and efficient clear-sky land surface temperature retrieval

This study proposes an RTM- and iteration-free TES (RIFTES) framework to improve both computational efficiency and retrieval accuracy of the temperature-emissivity separation (TES) algorithm for clear-sky land surface temperature (LST) retrieval. Based on physical derivations, a non-iterative TES algorithm was first developed by reformulating the original iterative procedure into a mathematically equivalent closed-form solution, thereby eliminating the need for cumbersome iterations. To further reduce error propagation risks and computational burdens, a deep residual neural network that integrates atmospheric radiative transfer physics was adopted to conduct atmospheric correction using easily accessible parameters, with a masking mechanism introduced to flexibly incorporate atmospheric constraints when available. Comprehensive validations demonstrate the effectiveness of the proposed algorithm. Simulation results show that RIFTES remains robust to input uncertainties and achieves the lowest root mean squared error (RMSE) of 1.06 K among representative existing algorithms, including split-window (SW), TES, and SW-TES hybrid methods. In-situ measurements from globally distributed sites were then used to evaluate the practical performance of RIFTES when applied to both the ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) and the Advanced Baseline Imager (ABI). The new algorithm achieves RMSE values of 1.51 K and 1.97 K for ECOSTRESS and ABI, respectively, reducing retrieval uncertainties by up to 24% and 32% compared with existing methods. Furthermore, by simplifying both the iterative procedure and atmospheric correction, RIFTES reduces the overall computational time by 74.0% and 62.5% compared with the TES and hybrid algorithms, respectively.

physics.geo-ph↗

arXiv2026

Efficient DP-SGD for LLMs with Randomized Clipping

Large language models (LLMs) are trained on vast datasets that may contain sensitive information. Differential privacy (DP), the de facto standard for formal privacy guarantees, provides a principled framework for training LLMs with provable privacy protection. However, state-of-the-art DP training implementations rely on fast gradient clipping techniques with memory overhead $O(B \min\{T^2, d^2\})$, where $B$ is the batch size, $T$ is the sequence length, and $d$ is the model width. This becomes prohibitive as both model size and context length grow. We propose DP-SGD-RC, a novel variant of DP-SGD with randomized clipping that reduces memory and compute complexity. DP-SGD-RC leverages stochastic trace estimation methods, specifically Hutchinson's estimator[Hutchinson, 1989] and its improved variant, Hutch++[Meyer et al., 2021], to reduce the memory footprint of per-sample gradient norm estimation. We provide a tight privacy analysis showing that DP-SGD-RC achieves noise multipliers competitive with deterministic clipping. Experiments fine-tuning Llama~3.2-1B on long-context benchmarks spanning classification, question answering, and summarization tasks demonstrate that DP-SGD-RC matches baseline utility while significantly reducing memory and compute requirements.

cs.LG↗

arXiv2026

How Well Do Models Follow Visual Instructions? VIBE: A Systematic Benchmark for Visual Instruction-Driven Image Editing

Recent generative models have achieved remarkable progress in image editing. However, existing systems and benchmarks remain largely text-guided. In contrast, human communication is inherently multimodal, where visual instructions such as sketches efficiently convey spatial and structural intent. To address this gap, we introduce VIBE, the Visual Instruction Benchmark for Image Editing with a three-level interaction hierarchy that captures deictic grounding, morphological manipulation, and causal reasoning. Across these levels, we curate high-quality and diverse test cases that reflect progressively increasing complexity in visual instruction following. We further propose a robust LMM-as-a-judge evaluation framework with task-specific metrics to enable scalable and fine-grained assessment. Through a comprehensive evaluation of 17 representative open-source and proprietary image editing models, we find that proprietary models exhibit early-stage visual instruction-following capabilities and consistently outperform open-source models. However, performance degrades markedly with increasing task difficulty even for the strongest systems, highlighting promising directions for future research.

cs.CV↗

arXiv2026

Memory-Efficient Differentially Private Training with Gradient Random Projection

Differential privacy (DP) protects sensitive data during neural network training, but standard methods like DP-Adam suffer from high memory overhead due to per-sample gradient clipping, limiting scalability. We introduce DP-GRAPE (Gradient RAndom ProjEction), a DP training method that significantly reduces memory usage while maintaining utility on par with first-order DP approaches. DP-GRAPE is motivated by our finding that privatization flattens the gradient singular value spectrum, making SVD-based projections (as in GaLore (Zhao et al., 2024)) unnecessary. Consequently, DP-GRAPE employs three key components: (1) random Gaussian matrices replace SVD-based subspaces, (2) gradients are privatized after projection, and (3) projection is applied during backpropagation. These contributions eliminate the need for costly SVD computations, enable substantial memory savings, and lead to improved utility. Despite operating in lower-dimensional subspaces, our theoretical analysis shows that DP-GRAPE achieves a privacy-utility tradeoff comparable to DP-SGD. Our extensive empirical experiments show that DP-GRAPE can significantly reduce the memory footprint of DP training without sacrificing accuracy or training time. In particular, DP-GRAPE reduces memory usage by over 63% when pre-training Vision Transformers and over 70% when fine-tuning RoBERTa-Large as compared to DP-Adam, while achieving similar performance. We further demonstrate that DP-GRAPE scales to fine-tuning large models such as OPT with up to 6.7 billion parameters, a scale at which DP-Adam fails due to memory constraints. Our code is available at https://github.com/alexmul1114/DP_GRAPE.

cs.LG↗

arXiv2026

Edgeworth Accountant: An Analytical Approach to Differential Privacy Composition

In privacy-preserving data analysis, many procedures and algorithms are structured as compositions of multiple private building blocks. As such, an important question is how to efficiently compute the overall privacy loss under composition. This paper introduces the Edgeworth Accountant, an analytical approach to composing differential privacy guarantees for private algorithms. Leveraging the $f$-differential privacy framework, the Edgeworth Accountant accurately tracks privacy loss under composition, enabling a closed-form expression of privacy guarantees through privacy-loss log-likelihood ratios (PLLRs). As implied by its name, this method applies the Edgeworth expansion to estimate and define the probability distribution of the sum of the PLLRs. Furthermore, by using a technique that simplifies complex distributions into simpler ones, we demonstrate the Edgeworth Accountant's applicability to any noise-addition mechanism. Its main advantage is providing $(ε, δ)$-differential privacy bounds that are non-asymptotic and do not significantly increase computational cost. This feature sets it apart from previous approaches, in which the running time increases with the number of mechanisms under composition. We conclude by showing how our Edgeworth Accountant offers accurate estimates and tight upper and lower bounds on $(ε, δ)$-differential privacy guarantees, especially tailored for training private models in deep learning and federated analytics.

cs.CR↗

arXiv2026

PEARL: Personalized Streaming Video Understanding Model

Human cognition of new concepts is inherently a streaming process: we continuously recognize new objects or identities and update our memories over time. However, current multimodal personalization methods are largely limited to static images or offline videos. This disconnects continuous visual input from instant real-world feedback, limiting their ability to provide the real-time, interactive personalized responses essential for future AI assistants. To bridge this gap, we first propose and formally define the novel task of Personalized Streaming Video Understanding (PSVU). To facilitate research in this new direction, we introduce PEARL-Bench, the first comprehensive benchmark designed specifically to evaluate this challenging setting. It evaluates a model's ability to respond to personalized concepts at exact timestamps under two modes: (1) Frame-level, focusing on a specific person or object in discrete frames, and (2) a novel Video-level, focusing on personalized actions unfolding across continuous frames. PEARL-Bench comprises 132 unique videos and 2,173 fine-grained annotations with precise timestamps. Concept diversity and annotation quality are strictly ensured through a combined pipeline of automated generation and human verification. To tackle this challenging new setting, we further propose PEARL, a plug-and-play, training-free strategy that serves as a strong baseline. Extensive evaluations across 8 offline and online models demonstrate that PEARL achieves state-of-the-art performance. Notably, it brings consistent PSVU improvements when applied to 3 distinct architectures, proving to be a highly effective and robust strategy. We hope this work advances vision-language model (VLM) personalization and inspires further research into streaming personalized AI assistants. Code is available at https://github.com/Yuanhong-Zheng/PEARL.

cs.CV↗

arXiv2026

GEBench: Benchmarking Image Generation Models as GUI Environments

Recent advancements in image generation models have enabled the prediction of future Graphical User Interface (GUI) states based on user instructions. However, existing benchmarks primarily focus on general domain visual fidelity, leaving the evaluation of state transitions and temporal coherence in GUI-specific contexts underexplored. To address this gap, we introduce GEBench, a comprehensive benchmark for evaluating dynamic interaction and temporal coherence in GUI generation. GEBench comprises 700 carefully curated samples spanning five task categories, covering both single-step interactions and multi-step trajectories across real-world and fictional scenarios, as well as grounding point localization. To support systematic evaluation, we propose GE-Score, a novel five-dimensional metric that assesses Goal Achievement, Interaction Logic, Content Consistency, UI Plausibility, and Visual Quality. Extensive evaluations on current models indicate that while they perform well on single-step transitions, they struggle significantly with maintaining temporal coherence and spatial grounding over longer interaction sequences. Our findings identify icon interpretation, text rendering, and localization precision as critical bottlenecks. This work provides a foundation for systematic assessment and suggests promising directions for future research toward building high-fidelity generative GUI environments. The code is available at: https://github.com/stepfun-ai/GEBench.

cs.AI↗

arXiv2026

Thinking in Frames: How Visual Context and Test-Time Scaling Empower Video Reasoning

Vision-Language Models have excelled at textual reasoning, but they often struggle with fine-grained spatial understanding and continuous action planning, failing to simulate the dynamics required for complex visual reasoning. In this work, we formulate visual reasoning by means of video generation models, positing that generated frames can act as intermediate reasoning steps between initial states and solutions. We evaluate their capacity in two distinct regimes: Maze Navigation for sequential discrete planning with low visual change and Tangram Puzzle for continuous manipulation with high visual change. Our experiments reveal three critical insights: (1) Robust Zero-Shot Generalization: In both tasks, the model demonstrates strong performance on unseen data distributions without specific finetuning. (2) Visual Context: The model effectively uses visual context as explicit control, such as agent icons and tangram shapes, enabling it to maintain high visual consistency and adapt its planning capability robustly to unseen patterns. (3) Visual Test-Time Scaling: We observe a test-time scaling law in sequential planning; increasing the generated video length (visual inference budget) empowers better zero-shot generalization to spatially and temporally complex paths. These findings suggest that video generation is not merely a media tool, but a scalable, generalizable paradigm for visual reasoning.

cs.LG↗

arXiv2025

PGDM: Physically guided diffusion model for land surface temperature downscaling

Land surface temperature (LST) is a fundamental parameter in thermal infrared remote sensing, while current LST products are often constrained by the trade-off between spatial and temporal resolutions. To mitigate this limitation, numerous studies have been conducted to enhance the resolutions of LST data, with a particular emphasis on the spatial dimension (commonly known as LST downscaling). Nevertheless, a comprehensive benchmark dataset tailored for this task remains scarce. In addition, existing downscaling models face challenges related to accuracy, practical usability, and the capability to self-evaluate their uncertainties. To overcome these challenges, this study first compiled three representative datasets, including one dataset over mainland China containing 22,909 image patches for model training and evaluation, as well as two datasets covering 40 heterogeneous regions worldwide for external evaluation. Subsequently, grounded in the surface energy balance (SEB)-based geophysical reasoning, we proposed the physically guided diffusion model (PGDM) for LST downscaling. In this framework, the downscaling task was formulated as an inference problem, aiming to sample from the posterior distribution of high-spatial-resolution (HR) LST conditioned on low-spatial-resolution (LR) LST observations and a suite of HR geophysical priors. Comprehensive evaluations demonstrate the effectiveness of PGDM, which generates high-quality downscaling results and outperforms existing representative interpolation, kernel-driven, hybrid, and deep learning approaches. Finally, by exploiting the inherent stochasticity of PGDM, the scene-level standard deviation of multiple generations was computed, revealing a strong positive linear correlation with the actual downscaling error...

physics.geo-ph↗

arXiv2025

Latent Sketchpad: Sketching Visual Thoughts to Elicit Multimodal Reasoning in MLLMs

While Multimodal Large Language Models (MLLMs) excel at visual understanding, they often struggle in complex scenarios that require visual planning and imagination. Inspired by how humans use sketching as a form of visual thinking to develop and communicate ideas, we introduce Latent Sketchpad, a framework that equips MLLMs with an internal visual scratchpad. The internal visual representations of MLLMs have traditionally been confined to perceptual understanding. We repurpose them to support generative visual thought without compromising reasoning ability. Building on frontier MLLMs, our approach integrates visual generation directly into their native autoregressive reasoning process. It allows the model to interleave textual reasoning with the generation of visual latents. These latents guide the internal thought process and can be translated into sketch images for interpretability. To realize this, we introduce two components: a Context-Aware Vision Head autoregressively produces visual representations, and a pretrained Sketch Decoder renders these into human-interpretable images. We evaluate the framework on our new dataset MazePlanning. Experiments across various MLLMs show that Latent Sketchpad delivers comparable or even superior reasoning performance to their backbone. It further generalizes across distinct frontier MLLMs, including Gemma3 and Qwen2.5-VL. By extending model's textual reasoning to visual thinking, our framework opens new opportunities for richer human-computer interaction and broader applications. More details and resources are available on our project page: https://latent-sketchpad.github.io/.

cs.CV↗

arXiv2025

Items Proxy Bridging: Enabling Frictionless Critiquing in Knowledge Graph Recommendations

Modern recommender systems place great inclination towards facilitating user experience, as more applications enabling users to critique and then refine recommendations immediately. Considering the real-time requirements, critique-able recommender systems typically straight modify the model parameters and update the recommend list through analyzing the user critiquing keyphrases in the inference phase. Current critiquing methods require first constructing a specially designated model which establish direct correlations between users and keyphrases during the training phase allowing for innovative recommendations upon the critiquing,restricting the applicable scenarios. Additionally, all these approaches ignore the catastrophic forgetting problem, where the cumulative changes in parameters during continuous multi-step critiquing may lead to a collapse in model performance. Thus, We conceptualize a proxy bridging users and keyphrases, proposing a streamlined yet potent Items Proxy Generic Critiquing Framework (IPGC) framework, which can serve as a universal plugin for most knowledge graph recommender models based on collaborative filtering (CF) strategies. IPGC provides a new paradigm for frictionless integration of critique mechanisms to enable iterative recommendation refinement in mainstream recommendation scenarios. IPGC describes the items proxy mechanism for transforming the critiquing optimization objective of user-keyphrase pairs into user-item pairs, adapting it for general CF recommender models without the necessity of specifically designed user-keyphrase correlation module. Furthermore, an anti-forgetting regularizer is introduced in order to efficiently mitigate the catastrophic forgetting problem of the model as a prior for critiquing optimization.

cs.IR↗

arXiv2025

BaseReward: A Strong Baseline for Multimodal Reward Model

The rapid advancement of Multimodal Large Language Models (MLLMs) has made aligning them with human preferences a critical challenge. Reward Models (RMs) are a core technology for achieving this goal, but a systematic guide for building state-of-the-art Multimodal Reward Models (MRMs) is currently lacking in both academia and industry. Through exhaustive experimental analysis, this paper aims to provide a clear ``recipe'' for constructing high-performance MRMs. We systematically investigate every crucial component in the MRM development pipeline, including \textit{reward modeling paradigms} (e.g., Naive-RM, Critic-based RM, and Generative RM), \textit{reward head architecture}, \textit{training strategies}, \textit{data curation} (covering over ten multimodal and text-only preference datasets), \textit{backbone model} and \textit{model scale}, and \textit{ensemble methods}. Based on these experimental insights, we introduce \textbf{BaseReward}, a powerful and efficient baseline for multimodal reward modeling. BaseReward adopts a simple yet effective architecture, built upon a {Qwen2.5-VL} backbone, featuring an optimized two-layer reward head, and is trained on a carefully curated mixture of high-quality multimodal and text-only preference data. Our results show that BaseReward establishes a new SOTA on major benchmarks such as MM-RLHF-Reward Bench, VL-Reward Bench, and Multimodal Reward Bench, outperforming previous models. Furthermore, to validate its practical utility beyond static benchmarks, we integrate BaseReward into a real-world reinforcement learning pipeline, successfully enhancing an MLLM's performance across various perception, reasoning, and conversational tasks. This work not only delivers a top-tier MRM but, more importantly, provides the community with a clear, empirically-backed guide for developing robust reward models for the next generation of MLLMs.

cs.CV↗

arXiv2025

11Plus-Bench: Demystifying Multimodal LLM Spatial Reasoning with Cognitive-Inspired Analysis

For human cognitive process, spatial reasoning and perception are closely entangled, yet the nature of this interplay remains underexplored in the evaluation of multimodal large language models (MLLMs). While recent MLLM advancements show impressive performance on reasoning, their capacity for human-like spatial cognition remains an open question. In this work, we introduce a systematic evaluation framework to assess the spatial reasoning abilities of state-of-the-art MLLMs relative to human performance. Central to our work is 11Plus-Bench, a high-quality benchmark derived from realistic standardized spatial aptitude tests. 11Plus-Bench also features fine-grained expert annotations of both perceptual complexity and reasoning process, enabling detailed instance-level analysis of model behavior. Through extensive experiments across 14 MLLMs and human evaluation, we find that current MLLMs exhibit early signs of spatial cognition. Despite a large performance gap compared to humans, MLLMs' cognitive profiles resemble those of humans in that cognitive effort correlates strongly with reasoning-related complexity. However, instance-level performance in MLLMs remains largely random, whereas human correctness is highly predictable and shaped by abstract pattern complexity. These findings highlight both emerging capabilities and limitations in current MLLMs' spatial reasoning capabilities and provide actionable insights for advancing model design.

cs.CL↗

arXiv2025

Scaling and Beyond: Advancing Spatial Reasoning in MLLMs Requires New Recipes

Multimodal Large Language Models (MLLMs) have demonstrated impressive performance in general vision-language tasks. However, recent studies have exposed critical limitations in their spatial reasoning capabilities. This deficiency in spatial reasoning significantly constrains MLLMs' ability to interact effectively with the physical world, thereby limiting their broader applications. We argue that spatial reasoning capabilities will not naturally emerge from merely scaling existing architectures and training methodologies. Instead, this challenge demands dedicated attention to fundamental modifications in the current MLLM development approach. In this position paper, we first establish a comprehensive framework for spatial reasoning within the context of MLLMs. We then elaborate on its pivotal role in real-world applications. Through systematic analysis, we examine how individual components of the current methodology, from training data to reasoning mechanisms, influence spatial reasoning capabilities. This examination reveals critical limitations while simultaneously identifying promising avenues for advancement. Our work aims to direct the AI research community's attention toward these crucial yet underexplored aspects. By highlighting these challenges and opportunities, we seek to catalyze progress toward achieving human-like spatial reasoning capabilities in MLLMs.

cs.LG↗