Jia Wang

Yifan Sui, Xin Huang, Hongbing Li et al.

The rapid development of GUI foundation models and mobile GUI agents has spurred numerous evaluation benchmarks, yet most rely on simulated environments or open-source applications, leaving real-world closed-source applications largely unevaluated. The core difficulty is that closed-source applications do not expose internal states, making traditional automatic verification inapplicable. To bridge this gap, we introduce AndroidDaily, a large-scale benchmark comprising 350 realistic daily-use tasks across 94 high-frequency Android applications spanning transportation, shopping, local services, entertainment, content creation, social media, and everyday utilities. To enable automatic and verifiable assessment in these opaque environments, we propose Guideline-grounded Reviewer for Automatic Diagnostic Evaluation (GRADE), a process-aware evaluator built on a three-tiered system of observable external guidelines: operational obligations, output quality, and negative constraints. GRADE tracks the agent's visual trajectory against these criteria and produces step-level diagnostic judgments, turning long-horizon, open-ended mobile interactions into verifiable evaluation without relying on hidden internal states. Experiments show that GRADE achieves 87.37\% agreement with human evaluators. The strongest model reaches a 62.0\% success rate on AndroidDaily, highlighting a substantial gap between current reasoning capabilities and practical execution in realistic mobile workflows.

Chong Zhang, Xiang Li, Jia Wang et al.

The robustness of deep neural networks is crucial for safety-critical deployments, yet existing evaluation methods are often attack-dependent and lack interpretability. We propose a principled, attack-agnostic robustness metric based on the spectral norm of the Fisher Information Matrix (FIM), which quantifies the worst-case sensitivity of the model's output distribution to input perturbations. Theoretically, we establish that the FIM equals the variance of the input Jacobian and derive closed-form spectral bounds for common architectures, including VGG, ResNet, DenseNet, and Transformer, providing the first theoretical robustness ranking. To enable scalable evaluation, we develop efficient algorithms, including power iteration and Hutchinson-based estimation, that support both white-box and black-box settings. Extensive experiments across multiple datasets, including CIFAR, ImageNet, and medical images, and across multiple architectures show a strong correlation between our metric and adversarial vulnerability. Our framework serves as an interpretable diagnostic tool that complements attack-based evaluations, offering insights into architectural sensitivity and guiding the design of more robust models. Code is available at: https://github.com/franz-chang/SRP/.

Haoran Wei, Chenglong Liu, Jinyue Chen et al. · tsinghua

Traditional OCR systems (OCR-1.0) are increasingly unable to meet people's usage due to the growing demand for intelligent processing of man-made optical characters. In this paper, we collectively refer to all artificial optical signals (e.g., plain texts, math/molecular formulas, tables, charts, sheet music, and even geometric shapes) as "characters" and propose the General OCR Theory along with an excellent model, namely GOT, to promote the arrival of OCR-2.0. The GOT, with 580M parameters, is a unified, elegant, and end-to-end model, consisting of a high-compression encoder and a long-contexts decoder. As an OCR-2.0 model, GOT can handle all the above "characters" under various OCR tasks. On the input side, the model supports commonly used scene- and document-style images in slice and whole-page styles. On the output side, GOT can generate plain or formatted results (markdown/tikz/smiles/kern) via an easy prompt. Besides, the model enjoys interactive OCR features, i.e., region-level recognition guided by coordinates or colors. Furthermore, we also adapt dynamic resolution and multi-page OCR technologies to GOT for better practicality. In experiments, we provide sufficient results to prove the superiority of our model.

Ailin Huang, Chengyuan Yao, Chunrui Han et al.

We present STEP3-VL-10B, a lightweight open-source foundation model designed to redefine the trade-off between compact efficiency and frontier-level multimodal intelligence. STEP3-VL-10B is realized through two strategic shifts: first, a unified, fully unfrozen pre-training strategy on 1.2T multimodal tokens that integrates a language-aligned Perception Encoder with a Qwen3-8B decoder to establish intrinsic vision-language synergy; and second, a scaled post-training pipeline featuring over 1k iterations of reinforcement learning. Crucially, we implement Parallel Coordinated Reasoning (PaCoRe) to scale test-time compute, allocating resources to scalable perceptual reasoning that explores and synthesizes diverse visual hypotheses. Consequently, despite its compact 10B footprint, STEP3-VL-10B rivals or surpasses models 10$\times$-20$\times$ larger (e.g., GLM-4.6V-106B, Qwen3-VL-235B) and top-tier proprietary flagships like Gemini 2.5 Pro and Seed-1.5-VL. Delivering best-in-class performance, it records 92.2% on MMBench and 80.11% on MMMU, while excelling in complex reasoning with 94.43% on AIME2025 and 75.95% on MathVision. We release the full model suite to provide the community with a powerful, efficient, and reproducible baseline.

Yichen Luo, Peiyu Zhu, Dongxiao Hu et al.

While Physics-Informed Neural Networks (PINNs) are powerful for solving Partial Differential Equations (PDEs), their training is often paralyzed by gradient pathology. The gradients from the PDE residuals and boundary constraints oppose each other, trapping the model in local minima. Current solutions, such as adaptive weighting or hard constraints, either fail to fundamentally resolve this ill-conditioning or are limited to simple geometries. In this study, we systematically analyze the possible causes of this gradient pathology from the perspectives of loss landscapes and optimization dynamics. Based on the obtained conclusion, we propose Constraint-Aligned loss with Manifold Lifting (CAML). By reformulating all zeroth-order terms into aligned constraints, our method effectively mitigates gradient conflicts. In addition, we introduce a delay factor to help the optimizer skip the high-curvature area. Experiments demonstrate that our CAML significantly enhances numerical stability and efficiency in highly complex PINN problems. Our code is open-sourced on https://github.com/YichenLuo-0/CAML.

Ziheng Chen, Fabrizio Silvestri, Jia Wang et al.

Recently, graph neural networks (GNNs) have been widely used to develop successful recommender systems. Although powerful, it is very difficult for a GNN-based recommender system to attach tangible explanations of why a specific item ends up in the list of suggestions for a given user. Indeed, explaining GNN-based recommendations is unique, and existing GNN explanation methods are inappropriate for two reasons. First, traditional GNN explanation methods are designed for node, edge, or graph classification tasks rather than ranking, as in recommender systems. Second, standard machine learning explanations are usually intended to support skilled decision-makers. Instead, recommendations are designed for any end-user, and thus their explanations should be provided in user-understandable ways. In this work, we propose GREASE, a novel method for explaining the suggestions provided by any black-box GNN-based recommender system. Specifically, GREASE first trains a surrogate model on a target user-item pair and its $l$-hop neighborhood. Then, it generates both factual and counterfactual explanations by finding optimal adjacency matrix perturbations to capture the sufficient and necessary conditions for an item to be recommended, respectively. Experimental results conducted on real-world datasets demonstrate that GREASE can generate concise and effective explanations for popular GNN-based recommender models.

Naifu Zhang, Meixia Tao, Jia Wang et al.

One of the main focuses in distributed learning is communication efficiency, since model aggregation at each round of training can consist of millions to billions of parameters. Several model compression methods, such as gradient quantization and sparsification, have been proposed to improve the communication efficiency of model aggregation. However, the information-theoretic minimum communication cost for a given distortion of gradient estimators is still unknown. In this paper, we study the fundamental limit of communication cost of model aggregation in distributed learning from a rate-distortion perspective. By formulating the model aggregation as a vector Gaussian CEO problem, we derive the rate region bound and sum-rate-distortion function for the model aggregation problem, which reveals the minimum communication rate at a particular gradient distortion upper bound. We also analyze the communication cost at each iteration and total communication cost based on the sum-rate-distortion function with the gradient statistics of real-world datasets. It is found that the communication gain by exploiting the correlation between worker nodes is significant for SignSGD, and a high distortion of gradient estimator can achieve low total communication cost in gradient compression.

Huiyu Duan, Wei Shen, Xiongkuo Min et al.

Transformers have shown significant effectiveness for various vision tasks including both high-level vision and low-level vision. Recently, masked autoencoders (MAE) for feature pre-training have further unleashed the potential of Transformers, leading to state-of-the-art performances on various high-level vision tasks. However, the significance of MAE pre-training on low-level vision tasks has not been sufficiently explored. In this paper, we show that masked autoencoders are also scalable self-supervised learners for image processing tasks. We first present an efficient Transformer model considering both channel attention and shifted-window-based self-attention termed CSformer. Then we develop an effective MAE architecture for image processing (MAEIP) tasks. Extensive experimental results show that with the help of MAEIP pre-training, our proposed CSformer achieves state-of-the-art performance on various image processing tasks, including Gaussian denoising, real image denoising, single-image motion deblurring, defocus deblurring, and image deraining.

Yanjie Dong, Luya Wang, Yuanfang Chi et al.

A wireless federated learning system is investigated by allowing a server and workers to exchange uncoded information via orthogonal wireless channels. Since the workers frequently upload local gradients to the server via bandwidth-limited channels, the uplink transmission from the workers to the server becomes a communication bottleneck. Therefore, a one-shot distributed principle component analysis (PCA) is leveraged to reduce the dimension of uploaded gradients such that the communication bottleneck is relieved. A PCA-based wireless federated learning (PCA-WFL) algorithm and its accelerated version (i.e., PCA-AWFL) are proposed based on the low-dimensional gradients and the Nesterov's momentum. For the non-convex loss functions, a finite-time analysis is performed to quantify the impacts of system hyper-parameters on the convergence of the PCA-WFL and PCA-AWFL algorithms. The PCA-AWFL algorithm is theoretically certified to converge faster than the PCA-WFL algorithm. Besides, the convergence rates of PCA-WFL and PCA-AWFL algorithms quantitatively reveal the linear speedup with respect to the number of workers over the vanilla gradient descent algorithm. Numerical results are used to demonstrate the improved convergence rates of the proposed PCA-WFL and PCA-AWFL algorithms over the benchmarks.

Haodong Li, Jingwei Wu, Quan Sun et al.

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.

Mohsen Karami, Khadijeh, Jahanian et al.

In recent years, Reinforcement Learning (RL) has emerged as a powerful tool for solving a wide range of problems, including decision-making and genomics. The exponential growth of raw genomic data over the past two decades has exceeded the capacity of manual analysis, leading to a growing interest in automatic data analysis and processing. RL algorithms are capable of learning from experience with minimal human supervision, making them well-suited for genomic data analysis and interpretation. One of the key benefits of using RL is the reduced cost associated with collecting labeled training data, which is required for supervised learning. While there have been numerous studies examining the applications of Machine Learning (ML) in genomics, this survey focuses exclusively on the use of RL in various genomics research fields, including gene regulatory networks (GRNs), genome assembly, and sequence alignment. We present a comprehensive technical overview of existing studies on the application of RL in genomics, highlighting the strengths and limitations of these approaches. We then discuss potential research directions that are worthy of future exploration, including the development of more sophisticated reward functions as RL heavily depends on the accuracy of the reward function, the integration of RL with other machine learning techniques, and the application of RL to new and emerging areas in genomics research. Finally, we present our findings and conclude by summarizing the current state of the field and the future outlook for RL in genomics.

Junwei Yin, Min Gao, Kai Shu et al.

The wide dissemination of fake news has affected our lives in many aspects, making fake news detection important and attracting increasing attention. Existing approaches make substantial contributions in this field by modeling news from a single-modal or multi-modal perspective. However, these modal-based methods can result in sub-optimal outcomes as they ignore reader behaviors in news consumption and authenticity verification. For instance, they haven't taken into consideration the component-by-component reading process: from the headline, images, comments, to the body, which is essential for modeling news with more granularity. To this end, we propose an approach of Emulating the behaviors of readers (Ember) for fake news detection on social media, incorporating readers' reading and verificating process to model news from the component perspective thoroughly. Specifically, we first construct intra-component feature extractors to emulate the behaviors of semantic analyzing on each component. Then, we design a module that comprises inter-component feature extractors and a sequence-based aggregator. This module mimics the process of verifying the correlation between components and the overall reading and verification sequence. Thus, Ember can handle the news with various components by emulating corresponding sequences. We conduct extensive experiments on nine real-world datasets, and the results demonstrate the superiority of Ember.

Ailin Huang, Ang Li, Aobo Kong et al.

We introduce Step 3.5 Flash, a sparse Mixture-of-Experts (MoE) model that bridges frontier-level agentic intelligence and computational efficiency. We focus on what matters most when building agents: sharp reasoning and fast, reliable execution. Step 3.5 Flash pairs a 196B-parameter foundation with 11B active parameters for efficient inference. It is optimized with interleaved 3:1 sliding-window/full attention and Multi-Token Prediction (MTP-3) to reduce the latency and cost of multi-round agentic interactions. To reach frontier-level intelligence, we design a scalable reinforcement learning framework that combines verifiable signals with preference feedback, while remaining stable under large-scale off-policy training, enabling consistent self-improvement across mathematics, code, and tool use. Step 3.5 Flash demonstrates strong performance across agent, coding, and math tasks, achieving 85.4% on IMO-AnswerBench, 86.4% on LiveCodeBench-v6 (2024.08-2025.05), 88.2% on tau2-Bench, 69.0% on BrowseComp (with context management), and 51.0% on Terminal-Bench 2.0, comparable to frontier models such as GPT-5.2 xHigh and Gemini 3.0 Pro. By redefining the efficiency frontier, Step 3.5 Flash provides a high-density foundation for deploying sophisticated agents in real-world industrial environments.

Zehao Zhu, Wei Sun, Jun Jia et al.

In recent years, live video streaming has gained widespread popularity across various social media platforms. Quality of experience (QoE), which reflects end-users' satisfaction and overall experience, plays a critical role for media service providers to optimize large-scale live compression and transmission strategies to achieve perceptually optimal rate-distortion trade-off. Although many QoE metrics for video-on-demand (VoD) have been proposed, there remain significant challenges in developing QoE metrics for live video streaming. To bridge this gap, we conduct a comprehensive study of subjective and objective QoE evaluations for live video streaming. For the subjective QoE study, we introduce the first live video streaming QoE dataset, TaoLive QoE, which consists of $42$ source videos collected from real live broadcasts and $1,155$ corresponding distorted ones degraded due to a variety of streaming distortions, including conventional streaming distortions such as compression, stalling, as well as live streaming-specific distortions like frame skipping, variable frame rate, etc. Subsequently, a human study was conducted to derive subjective QoE scores of videos in the TaoLive QoE dataset. For the objective QoE study, we benchmark existing QoE models on the TaoLive QoE dataset as well as publicly available QoE datasets for VoD scenarios, highlighting that current models struggle to accurately assess video QoE, particularly for live content. Hence, we propose an end-to-end QoE evaluation model, Tao-QoE, which integrates multi-scale semantic features and optical flow-based motion features to predicting a retrospective QoE score, eliminating reliance on statistical quality of service (QoS) features.

Jia Wang, Junping Du, Yingxia Shao et al.

With the development of online travel services, it has great application prospects to timely mine users' evaluation emotions for travel services and use them as indicators to guide the improvement of online travel service quality. In this paper, we study the text sentiment classification of online travel reviews based on social media online comments and propose the SCCL model based on capsule network and sentiment lexicon. SCCL model aims at the lack of consideration of local features and emotional semantic features of the text in the language model that can efficiently extract text context features like BERT and GRU. Then make the following improvements to their shortcomings. On the one hand, based on BERT-BiGRU, the capsule network is introduced to extract local features while retaining good context features. On the other hand, the sentiment lexicon is introduced to extract the emotional sequence of the text to provide richer emotional semantic features for the model. To enhance the universality of the sentiment lexicon, the improved SO-PMI algorithm based on TF-IDF is used to expand the lexicon, so that the lexicon can also perform well in the field of online travel reviews.

Shiqi Gao, Kang Fu, Zitong Xu et al.

Current no-reference image quality assessment (NR-IQA) models for enhanced images often struggle to generalize, as they tend to overfit to the distinct patterns of specific enhancement algorithms rather than evaluating genuine perceptual quality. To address this issue, we propose a preference-guided debiasing framework for no-reference enhancement image quality assessment (EIQA). Specifically, we first learn a continuous enhancement-preference embedding space using supervised contrastive learning, where images generated by similar enhancement styles are encouraged to have closer representations. Based on this, we further estimate the enhancement-induced nuisance component contained in the raw quality representation and remove it before quality regression. In this way, the model is guided to focus on algorithm-invariant perceptual quality cues instead of enhancement-specific visual fingerprints. To facilitate stable optimization, we adopt a two-stage training strategy that first learns the enhancement-preference space and then performs debiased quality prediction. Extensive experiments on public EIQA benchmarks demonstrate that the proposed method effectively mitigates algorithm-induced representation bias and achieves superior robustness and cross-algorithm generalization compared with existing approaches.

Shiqi Gao, Zitong Xu, Kang Fu et al.

Evaluating text-guided image editing (TIE) methods remains a challenging problem, as reliable assessment should simultaneously consider perceptual quality, alignment with textual instructions, and preservation of original image content. Despite rapid progress in TIE models, existing evaluation benchmarks remain limited in scale and often show weak correlation with human perceptual judgments. In this work, we introduce TIEdit, a benchmark for systematic evaluation of text-guided image editing methods. TIEdit consists of 512 source images paired with editing prompts across eight representative editing tasks, producing 5,120 edited images generated by ten state-of-the-art TIE models. To obtain reliable subjective ratings, 20 experts are recruited to produce 307,200 raw subjective ratings, which accumulates into 15,360 mean opinion scores (MOSs) across three evaluation dimensions: perceptual quality, editing alignment, and content preservation. Beyond the benchmark itself, we further propose EditProbe, an LLM-based evaluator that estimates editing quality via intermediate-layer probing of hidden representations. Instead of relying solely on final model outputs, EditProbe extracts informative representations from intermediate layers of multimodal large language models to better capture semantic and perceptual relationships between source images, editing instructions, and edited results. Experimental results demonstrate that widely used automatic evaluation metrics show limited correlation with human judgments on editing tasks, while EditProbe achieves substantially stronger alignment with human perception. Together, TIEdit and EditProbe provide a foundation for more reliable and perceptually aligned evaluation of text-guided image editing methods.

Kang Fu, Huiyu Duan, Zicheng Zhang et al.

Large Multimodal Models (LMMs) have recently shown remarkable promise in low-level visual perception tasks, particularly in Image Quality Assessment (IQA), demonstrating strong zero-shot capability. However, achieving state-of-the-art performance often requires computationally expensive fine-tuning methods, which aim to align the distribution of quality-related token in output with image quality levels. Inspired by recent training-free works for LMM, we introduce IQARAG, a novel, training-free framework that enhances LMMs' IQA ability. IQARAG leverages Retrieval-Augmented Generation (RAG) to retrieve some semantically similar but quality-variant reference images with corresponding Mean Opinion Scores (MOSs) for input image. These retrieved images and input image are integrated into a specific prompt. Retrieved images provide the LMM with a visual perception anchor for IQA task. IQARAG contains three key phases: Retrieval Feature Extraction, Image Retrieval, and Integration & Quality Score Generation. Extensive experiments across multiple diverse IQA datasets, including KADID, KonIQ, LIVE Challenge, and SPAQ, demonstrate that the proposed IQARAG effectively boosts the IQA performance of LMMs, offering a resource-efficient alternative to fine-tuning for quality assessment.

Yanbing Zeng, Jia Wang, Hanghang Ma et al.

Integrating image generation and understanding into a single framework has become a pivotal goal in the multimodal domain. However, how understanding can effectively assist generation has not been fully explored. Unlike previous works that focus on leveraging reasoning abilities and world knowledge from understanding models, this paper introduces a novel perspective: leveraging understanding to enhance the fidelity and detail richness of generated images. To this end, we propose Forge-and-Quench, a new unified framework that puts this principle into practice. In the generation process of our framework, an MLLM first reasons over the entire conversational context, including text instructions, to produce an enhanced text instruction. This refined instruction is then mapped to a virtual visual representation, termed the Bridge Feature, via a novel Bridge Adapter. This feature acts as a crucial link, forging insights from the understanding model to quench and refine the generation process. It is subsequently injected into the T2I backbone as a visual guidance signal, alongside the enhanced text instruction that replaces the original input. To validate this paradigm, we conduct comprehensive studies on the design of the Bridge Feature and Bridge Adapter. Our framework demonstrates exceptional extensibility and flexibility, enabling efficient migration across different MLLM and T2I models with significant savings in training overhead, all without compromising the MLLM's inherent multimodal understanding capabilities. Experiments show that Forge-and-Quench significantly improves image fidelity and detail across multiple models, while also maintaining instruction-following accuracy and enhancing world knowledge application. Models and codes are available at https://github.com/YanbingZeng/Forge-and-Quench.

Chenyi Liu, Fei Chen, Lu Deng et al.

The point pair feature (PPF) is widely used for 6D pose estimation. In this paper, we propose an efficient 6D pose estimation method based on the PPF framework. We introduce a well-targeted down-sampling strategy that focuses more on edge area for efficient feature extraction of complex geometry. A pose hypothesis validation approach is proposed to resolve the symmetric ambiguity by calculating edge matching degree. We perform evaluations on two challenging datasets and one real-world collected dataset, demonstrating the superiority of our method on pose estimation of geometrically complex, occluded, symmetrical objects. We further validate our method by applying it to simulated punctures.

Dinging Li, Yingxiu Zhao, Xinrui Cheng et al.

Spatial reasoning over three-dimensional scenes is a core capability for embodied intelligence, yet continuous model improvement remains bottlenecked by the cost of geometric annotation. The self-evolving paradigm offers a promising path, but its reliance on model consensus to construct pseudo-labels causes training to reinforce rather than correct the model's own geometric errors. We identify a property unique to 3D spatial reasoning that circumvents this limitation: ground truth is a deterministic consequence of the underlying geometry, computable exactly from point clouds and camera poses without any model involvement. Building on this insight, we present SpatialEvo, a self-evolving framework for 3D spatial reasoning, centered on the Deterministic Geometric Environment (DGE). The DGE formalizes 16 spatial reasoning task categories under explicit geometric validation rules and converts unannotated 3D scenes into zero-noise interactive oracles, replacing model consensus with objective physical feedback. A single shared-parameter policy co-evolves across questioner and solver roles under DGE constraints: the questioner generates physically valid spatial questions grounded in scene observations, while the solver derives precise answers against DGE-verified ground truth. A task-adaptive scheduler endogenously concentrates training on the model's weakest categories, producing a dynamic curriculum without manual design. Experiments across nine benchmarks demonstrate that SpatialEvo achieves the highest average score at both 3B and 7B scales, with consistent gains on spatial reasoning benchmarks and no degradation on general visual understanding.

Haolong Yan, Yeqing Shen, Xin Huang et al.

With the rapid development of Large Vision Language Models, the focus of Graphical User Interface (GUI) agent tasks shifts from single-screen tasks to complex screen navigation challenges. However, real-world GUI environments, such as PC software and mobile Apps, are often complex and proprietary, making it difficult to obtain the comprehensive environment information needed for agent training and evaluation. This limitation hinders systematic investigation and benchmarking of agent navigation capabilities. To address this limitation, we introduce GUI Exploration Lab, a simulation environment engine for GUI agent navigation research that enables flexible definition and composition of screens, icons, and navigation graphs, while providing full access to environment information for comprehensive agent training and evaluation. Through extensive experiments, we find that supervised fine-tuning enables effective memorization of fundamental knowledge, serving as a crucial foundation for subsequent training. Building on this, single-turn reinforcement learning further enhances generalization to unseen scenarios. Finally, multi-turn reinforcement learning encourages the development of exploration strategies through interactive trial and error, leading to further improvements in screen navigation performance. We validate our methods on both static and interactive benchmarks, demonstrating that our findings generalize effectively to real-world scenarios. These findings demonstrate the advantages of reinforcement learning approaches in GUI navigation and offer practical guidance for building more capable and generalizable GUI agents.

NextStep Team, Chunrui Han, Guopeng Li et al. · tsinghua

Prevailing autoregressive (AR) models for text-to-image generation either rely on heavy, computationally-intensive diffusion models to process continuous image tokens, or employ vector quantization (VQ) to obtain discrete tokens with quantization loss. In this paper, we push the autoregressive paradigm forward with NextStep-1, a 14B autoregressive model paired with a 157M flow matching head, training on discrete text tokens and continuous image tokens with next-token prediction objectives. NextStep-1 achieves state-of-the-art performance for autoregressive models in text-to-image generation tasks, exhibiting strong capabilities in high-fidelity image synthesis. Furthermore, our method shows strong performance in image editing, highlighting the power and versatility of our unified approach. To facilitate open research, we will release our code and models to the community.

Yana Wei, Liang Zhao, Jianjian Sun et al. · tsinghua

The remarkable reasoning capability of large language models (LLMs) stems from cognitive behaviors that emerge through reinforcement with verifiable rewards. This work investigates how to transfer this principle to Multimodal LLMs (MLLMs) to unlock advanced visual reasoning. We introduce a two-stage paradigm built on Qwen2.5-VL-7B: a massive linguistic cold-start fine-tuning, followed by multimodal reinforcement learning (RL) spanning nearly 1,000 steps, surpassing all previous open-source efforts in scale. This pioneering work reveals three fundamental insights: 1) Behavior transfer emerges surprisingly early in cold start due to linguistic mental imagery. 2) Cold start broadly memorizes visual behaviors, while RL critically discerns and scales up effective patterns. 3) Transfer strategically favors high-utility behaviors such as visual reflection. Our resulting model, Open-Vision-Reasoner (OVR), achieves state-of-the-art performance on a suite of reasoning benchmarks, including 95.3% on MATH500, 51.8% on MathVision and 54.6% on MathVerse. We release our model, data, and training dynamics to catalyze the development of more capable, behavior-aligned multimodal reasoners.

Jiayang Zou, Luyao Fan, Jiayang Gao et al.

We construct a smooth, strictly positive, Gaussian-decaying density on $\R^2$ for which Fisher information along the heat flow is not log-convex. This disproves the Cheng--Geng log-convexity conjecture in dimension two and, by tensorization, in every dimension $d\ge2$. Consequently, the multidimensional forms of the Gaussian completely monotone conjecture, McKean's conjecture, and Toscani's entropy power conjecture also fail. This complements the recent one-dimensional counterexample of Gu and Sellke. The counterexample is a small hexagonal perturbation on the triangular torus, transferred to $\R^2$ by a Gaussian envelope, and supported by explicit two-dimensional numerics. Finally, we initiate the study of the sharp constants $θ_d^*$ by proving $θ_1^*=1$, establishing monotonicity in the dimension, and recording an intrinsic simplex resonance family in every fixed dimension. The explicit two-dimensional counterexample was found by GPT-5.5 Pro.

Chunyi Li, Xiaozhe Li, Zicheng Zhang et al.

With the emergence of Multimodal Large Language Models (MLLMs), hundreds of benchmarks have been developed to ensure the reliability of MLLMs in downstream tasks. However, the evaluation mechanism itself may not be reliable. For developers of MLLMs, questions remain about which benchmark to use and whether the test results meet their requirements. Therefore, we propose a critical principle of Information Density, which examines how much insight a benchmark can provide for the development of MLLMs. We characterize it from four key dimensions: (1) Fallacy, (2) Difficulty, (3) Redundancy, (4) Diversity. Through a comprehensive analysis of more than 10,000 samples, we measured the information density of 19 MLLM benchmarks. Experiments show that using the latest benchmarks in testing can provide more insight compared to previous ones, but there is still room for improvement in their information density. We hope this principle can promote the development and application of future MLLM benchmarks. Project page: https://github.com/lcysyzxdxc/bench4bench

Zitong Xu, Huiyu Duan, Bingnan Liu et al.

The rapid advancement of Text-guided Image Editing (TIE) enables image modifications through text prompts. However, current TIE models still struggle to balance image quality, editing alignment, and consistency with the original image, limiting their practical applications. Existing TIE evaluation benchmarks and metrics have limitations on scale or alignment with human perception. To this end, we introduce EBench-18K, the first large-scale image Editing Benchmark including 18K edited images with fine-grained human preference annotations for evaluating TIE. Specifically, EBench-18K includes 1,080 source images with corresponding editing prompts across 21 tasks, 18K+ edited images produced by 17 state-of-the-art TIE models, 55K+ mean opinion scores (MOSs) assessed from three evaluation dimensions, and 18K+ question-answering (QA) pairs. Based on EBench-18K, we employ outstanding LMMs to assess edited images, while the evaluation results, in turn, provide insights into assessing the alignment between the LMMs' understanding ability and human preferences. Then, we propose LMM4Edit, a LMM-based metric for evaluating image Editing models from perceptual quality, editing alignment, attribute preservation, and task-specific QA accuracy in an all-in-one manner. Extensive experiments show that LMM4Edit achieves outstanding performance and aligns well with human preference. Zero-shot validation on the other datasets also shows the generalization ability of our model. The dataset and code are available at https://github.com/IntMeGroup/LMM4Edit.

Jiayang Zou, Luyao Fan, Jiayang Gao et al.

In this paper, we study rate-distortion theory for general sources with an emphasis on the existence of optimal reconstruction distributions on noncompact alphabets. Classical attainability results typically rely on compactness of the reproduction alphabet together with continuity of the distortion function, which may fail in many noncompact settings. We identify two complementary existence mechanisms under lower semi-continuity on locally compact Polish alphabets. For bounded distortions, we prove that the rate-distortion infimum is attained via the one-point compactification argument. For unbounded coercive distortions, we establish existence via concentration-compactness. We also give several counterexamples showing that our attainability results are close to sharp. Our results provide a unified and transparent existence theorem for rate-distortion problems with lower semi-continuous distortions.

Weikang Chen, Junping Du, Yingxia Shao et al.

Federated learning enables a collaborative training and optimization of global models among a group of devices without sharing local data samples. However, the heterogeneity of data in federated learning can lead to unfair representation of the global model across different devices. To address the fairness issue in federated learning, we propose a dynamic q fairness federated learning algorithm with reinforcement learning, called DQFFL. DQFFL aims to mitigate the discrepancies in device aggregation and enhance the fairness of treatment for all groups involved in federated learning. To quantify fairness, DQFFL leverages the performance of the global federated model on each device and incorporates α-fairness to transform the preservation of fairness during federated aggregation into the distribution of client weights in the aggregation process. Considering the sensitivity of parameters in measuring fairness, we propose to utilize reinforcement learning for dynamic parameters during aggregation. Experimental results demonstrate that our DQFFL outperforms the state-of-the-art methods in terms of overall performance, fairness and convergence speed.

Mingyu Zheng, Zhifan Feng, Jia Wang et al.

Despite the commendable progress of recent LLM-based data synthesis methods, they face two limitations in generating table instruction tuning data. First, they can not thoroughly explore the vast input space of table understanding tasks, leading to limited data diversity. Second, they ignore the weaknesses in table understanding ability of the target LLM and blindly pursue the increase of data quantity, resulting in suboptimal data efficiency. In this paper, we introduce a progressive and weakness-guided data synthesis framework tailored for table instruction tuning, named TableDreamer, to mitigate the above issues. Specifically, we first synthesize diverse tables and related instructions as seed data, and then perform an iterative exploration of the input space under the guidance of the newly identified weakness data, which eventually serve as the final training data for fine-tuning the target LLM. Extensive experiments on 10 tabular benchmarks demonstrate the effectiveness of the proposed framework, which boosts the average accuracy of Llama3.1-8B-instruct by 11.62% (49.07% to 60.69%) with 27K GPT-4o synthetic data and outperforms state-of-the-art data synthesis baselines which use more training data. The code and data is available at https://github.com/SpursGoZmy/TableDreamer

Xiaorong Zhu, Ziheng Jia, Jiarui Wang et al.

The rapid evolution of Multi-modality Large Language Models (MLLMs) is driving significant advancements in visual understanding and generation. Nevertheless, a comprehensive assessment of their capabilities, concerning the fine-grained physical principles especially in geometric optics, remains underexplored. To address this gap, we introduce GOBench, the first benchmark to systematically evaluate MLLMs' ability across two tasks: 1) Generating Optically Authentic Imagery and 2) Understanding Underlying Optical Phenomena. We curates high-quality prompts of geometric optical scenarios and use MLLMs to construct GOBench-Gen-1k dataset.We then organize subjective experiments to assess the generated imagery based on Optical Authenticity, Aesthetic Quality, and Instruction Fidelity, revealing MLLMs' generation flaws that violate optical principles. For the understanding task, we apply crafted evaluation instructions to test optical understanding ability of eleven prominent MLLMs. The experimental results demonstrate that current models face significant challenges in both optical generation and understanding. The top-performing generative model, GPT-4o-Image, cannot perfectly complete all generation tasks, and the best-performing MLLM model, Gemini-2.5Pro, attains a mere 37.35\% accuracy in optical understanding. Database and codes are publicly available at https://github.com/aiben-ch/GOBench.

Jia Wang, Chuanyu Qin, Mingyu Zheng et al.

Despite the success of Large Language Models (LLMs) in table understanding, their internal mechanisms remain unclear. In this paper, we conduct an empirical study on 16 LLMs, covering general LLMs, specialist tabular LLMs, and Mixture-of-Experts (MoE) models, to explore how LLMs understand tabular data and perform downstream tasks. Our analysis focus on 4 dimensions including the attention dynamics, the effective layer depth, the expert activation, and the impacts of input designs. Key findings include: (1) LLMs follow a three-phase attention pattern -- early layers scan the table broadly, middle layers localize relevant cells, and late layers amplify their contributions; (2) tabular tasks require deeper layers than math reasoning to reach stable predictions; (3) MoE models activate table-specific experts in middle layers, with early and late layers sharing general-purpose experts; (4) Chain-of-Thought prompting increases table attention, further enhanced by table-tuning. We hope these findings and insights can facilitate interpretability and future research on table-related tasks.

Jie Zhu, Hanghang Ma, Jia Wang et al.

In this work, we introduce Wallaroo, a simple autoregressive baseline that leverages next-token prediction to unify multi-modal understanding, image generation, and editing at the same time. Moreover, Wallaroo supports multi-resolution image input and output, as well as bilingual support for both Chinese and English. We decouple the visual encoding into separate pathways and apply a four-stage training strategy to reshape the model's capabilities. Experiments are conducted on various benchmarks where Wallaroo produces competitive performance or exceeds other unified models, suggesting the great potential of autoregressive models in unifying multi-modality understanding and generation. Our code is available at https://github.com/JiePKU/Wallaroo.

Xiao Wang, Jia Wang, Yijie Wang et al.

Large language models (LLMs) have achieved impressive results in natural language understanding, yet their reasoning capabilities remain limited when operating as single agents. Multi-Agent Debate (MAD) has been proposed to address this limitation by enabling collaborative reasoning among multiple models in a round-table debate manner. While effective, MAD introduces substantial computational overhead due to the number of agents involved and the frequent communication required. In this paper, we propose MARS (Multi-Agent Review System), a role-based collaboration framework inspired by the review process. In MARS, an author agent generates an initial solution, reviewer agents provide decisions and comments independently, and a meta-reviewer integrates the feedback to make the final decision and guide further revision. This design enhances reasoning quality while avoiding costly reviewer-to-reviewer interactions, thereby controlling token consumption and inference time. We compared MARS with both MAD and other state-of-the-art reasoning strategies across multiple benchmarks. Extensive experiments with different LLMs show that MARS matches the accuracy of MAD while reducing both token usage and inference time by approximately 50\%. Code is available at https://github.com/xwang97/MARS.

Haichao Zhang, Chong Zhang, Peiyu Hu et al.

Modern recommender systems face a critical challenge in complying with privacy regulations like the 'right to be forgotten': removing a user's data without disrupting recommendations for others. Traditional unlearning methods address this by partial model updates, but introduce propagation bias--where unlearning one user's data distorts recommendations for behaviorally similar users, degrading system accuracy. While retraining eliminates bias, it is computationally prohibitive for large-scale systems. To address this challenge, we propose CRAGRU, a novel framework leveraging Retrieval-Augmented Generation (RAG) for efficient, user-specific unlearning that mitigates bias while preserving recommendation quality. CRAGRU decouples unlearning into distinct retrieval and generation stages. In retrieval, we employ three tailored strategies designed to precisely isolate the target user's data influence, minimizing collateral impact on unrelated users and enhancing unlearning efficiency. Subsequently, the generation stage utilizes an LLM, augmented with user profiles integrated into prompts, to reconstruct accurate and personalized recommendations without needing to retrain the entire base model. Experiments on three public datasets demonstrate that CRAGRU effectively unlearns targeted user data, significantly mitigating unlearning bias by preventing adverse impacts on non-target users, while maintaining recommendation performance comparable to fully trained original models. Our work highlights the promise of RAG-based architectures for building robust and privacy-preserving recommender systems. The source code is available at: https://github.com/zhanghaichao520/LLM_rec_unlearning.

Yichen Luo, Jia Wang, Dapeng Lan et al.

Partial Differential Equations (PDEs) are fundamental for modeling physical systems, yet solving them in a generic and efficient manner using machine learning-based approaches remains challenging due to limited multi-input and multi-scale generalization capabilities, as well as high computational costs. This paper proposes the Multi-input and Multi-scale Efficient Transformer (MMET), a novel framework designed to address the above challenges. MMET decouples mesh and query points as two sequences and feeds them into the encoder and decoder, respectively, and uses a Gated Condition Embedding (GCE) layer to embed input variables or functions with varying dimensions, enabling effective solutions for multi-scale and multi-input problems. Additionally, a Hilbert curve-based reserialization and patch embedding mechanism decrease the input length. This significantly reduces the computational cost when dealing with large-scale geometric models. These innovations enable efficient representations and support multi-scale resolution queries for large-scale and multi-input PDE problems. Experimental evaluations on diverse benchmarks spanning different physical fields demonstrate that MMET outperforms SOTA methods in both accuracy and computational efficiency. This work highlights the potential of MMET as a robust and scalable solution for real-time PDE solving in engineering and physics-based applications, paving the way for future explorations into pre-trained large-scale models in specific domains. This work is open-sourced at https://github.com/YichenLuo-0/MMET.

Jihao Liu, Xin Huang, Jinliang Zheng et al.

This paper introduces MM-Instruct, a large-scale dataset of diverse and high-quality visual instruction data designed to enhance the instruction-following capabilities of large multimodal models (LMMs). While existing visual instruction datasets often focus on question-answering, they struggle to generalize to broader application scenarios such as creative writing, summarization, or image analysis. To address these limitations, we propose a novel approach to constructing MM-Instruct that leverages the strong instruction-following capabilities of existing LLMs to generate novel visual instruction data from large-scale but conventional image captioning datasets. MM-Instruct first leverages ChatGPT to automatically generate diverse instructions from a small set of seed instructions through augmenting and summarization. It then matches these instructions with images and uses an open-sourced large language model (LLM) to generate coherent answers to the instruction-image pairs. The LLM is grounded by the detailed text descriptions of images in the whole answer generation process to guarantee the alignment of the instruction data. Moreover, we introduce a benchmark based on the generated instruction data to evaluate the instruction-following capabilities of existing LMMs. We demonstrate the effectiveness of MM-Instruct by training a LLaVA-1.5 model on the generated data, denoted as LLaVA-Instruct, which exhibits significant improvements in instruction-following capabilities compared to LLaVA-1.5 models. The MM-Instruct dataset, benchmark, and pre-trained models are available at https://github.com/jihaonew/MM-Instruct.

Shiqi Gao, Huiyu Duan, Xinyue Li et al.

In recent years, learning-based color and tone enhancement methods for photos have become increasingly popular. However, most learning-based image enhancement methods just learn a mapping from one distribution to another based on one dataset, lacking the ability to adjust images continuously and controllably. It is important to enable the learning-based enhancement models to adjust an image continuously, since in many cases we may want to get a slighter or stronger enhancement effect rather than one fixed adjusted result. In this paper, we propose a quality-guided image enhancement paradigm that enables image enhancement models to learn the distribution of images with various quality ratings. By learning this distribution, image enhancement models can associate image features with their corresponding perceptual qualities, which can be used to adjust images continuously according to different quality scores. To validate the effectiveness of our proposed method, a subjective quality assessment experiment is first conducted, focusing on skin tone adjustment in portrait photography. Guided by the subjective quality ratings obtained from this experiment, our method can adjust the skin tone corresponding to different quality requirements. Furthermore, an experiment conducted on 10 natural raw images corroborates the effectiveness of our model in situations with fewer subjects and fewer shots, and also demonstrates its general applicability to natural images. Our project page is https://github.com/IntMeGroup/quality-guided-enhancement .

Jia Wang, Ping Wang, Biao Li et al.

Remote sensing image registration is valuable for image-based navigation system despite posing many challenges. As the search space of registration is usually non-convex, the optimization algorithm, which aims to search the best transformation parameters, is a challenging step. Conventional optimization algorithms can hardly reconcile the contradiction of simultaneous rapid convergence and the global optimization. In this paper, a novel learning-based optimization algorithm named Image Registration Optimizer Network (IRON) is proposed, which can predict the global optimum after single iteration. The IRON is trained by a 3D tensor (9x9x9), which consists of similar metric values. The elements of the 3D tensor correspond to the 9x9x9 neighbors of the initial parameters in the search space. Then, the tensor's label is a vector that points to the global optimal parameters from the initial parameters. Because of the special architecture, the IRON could predict the global optimum directly for any initialization. The experimental results demonstrate that the proposed algorithm performs better than other classical optimization algorithms as it has higher accuracy, lower root of mean square error (RMSE), and more efficiency. Our IRON codes are available for further study.https://www.github.com/jaxwangkd04/IRON

Jia Wang, Vincent W. Zheng, Zemin Liu et al.

In this paper, we study the problem of using representation learning to assist information diffusion prediction on graphs. In particular, we aim at estimating the probability of an inactive node to be activated next in a cascade. Despite the success of recent deep learning methods for diffusion, we find that they often underexplore the cascade structure. We consider a cascade as not merely a sequence of nodes ordered by their activation time stamps; instead, it has a richer structure indicating the diffusion process over the data graph. As a result, we introduce a new data model, namely diffusion topologies, to fully describe the cascade structure. We find it challenging to model diffusion topologies, which are dynamic directed acyclic graphs (DAGs), with the existing neural networks. Therefore, we propose a novel topological recurrent neural network, namely Topo-LSTM, for modeling dynamic DAGs. We customize Topo-LSTM for the diffusion prediction task, and show it improves the state-of-the-art baselines, by 20.1%--56.6% (MAP) relatively, across multiple real-world data sets. Our code and data sets are available online at https://github.com/vwz/topolstm.

Jia Wang, Qian Xu, Xuanwen Ding et al.

In this paper, we introduce ProcureGym, an data-driven multi-agent simulation platform that models China's National Volume-Based drug Procurement (NVBP) as a Markov Game. Based on real-world data from 7 rounds of NVBP (covering 325 drugs and 2,267 firms), the platform establishes a high-fidelity simulation environment. Within this framework, we evaluate diverse agent models, including Reinforcement Learning (RL), Large Language Model (LLM), and Rule-based algorithms. Experimental results demonstrate that RL agents achieve superior winner alignment and profits. Further analyses show that maximum valid bidding price and procurement volume dominate strategic outcomes. ProcureGym thus serves as a rigorous instrument for assessing policy impacts and formulating future procurement strategies.

Ben Wan, Yan Feng, Zihan Tang et al.

DeepSeek-OCR leverages visual-text compression to reduce long-text processing costs and accelerate inference, yet visual tokens remain prone to redundant textual and structural information. Moreover, current token pruning methods for conventional vision-language models (VLMs) fail to preserve textual fidelity due to improper compression mechanisms. By analyzing the decoding process of DeepSeek-OCR, we find that a distinct two-stage reading trajectory: the model initially prioritizes the majority of high-norm tokens, then subsequently redistributes its attention to the remaining ones. Motivated by this insight, we propose RTPrune, a two-stage token pruning method tailored for DeepSeek-OCR. In the first stage, we prioritize high-norm visual tokens that capture salient textual and structural information. In the second stage, the remaining tokens are paired and merged based on optimal transport theory to achieve efficient feature aggregation. We further introduce a dynamic pruning ratio that adapts to token similarity and textual density for OCR tasks, enabling a better efficiency-accuracy trade-off. Extensive experiments demonstrate state-of-the-art performance, as evidenced by 99.47% accuracy and 1.23$\times$ faster prefill on OmniDocBench, achieved with 84.25% token retention when applied to DeepSeek-OCR-Large.

Bo Wu, Peiye Liu, Wen-Huang Cheng et al.

Social Media Popularity Prediction (SMPP) is a crucial task that involves automatically predicting future popularity values of online posts, leveraging vast amounts of multimodal data available on social media platforms. Studying and investigating social media popularity becomes central to various online applications and requires novel methods of comprehensive analysis, multimodal comprehension, and accurate prediction. SMP Challenge is an annual research activity that has spurred academic exploration in this area. This paper summarizes the challenging task, data, and research progress. As a critical resource for evaluating and benchmarking predictive models, we have released a large-scale SMPD benchmark encompassing approximately half a million posts authored by around 70K users. The research progress analysis provides an overall analysis of the solutions and trends in recent years. The SMP Challenge website (www.smp-challenge.com) provides the latest information and news.

Haoran Wei, Youyang Yin, Yumeng Li et al.

Recently, "visual o1" began to enter people's vision, with expectations that this slow-thinking design can solve visual reasoning tasks, especially geometric math problems. However, the reality is that current LVLMs (Large Vision Language Models) can hardly even accurately copy a geometric figure, let alone truly understand the complex inherent logic and spatial relationships within geometric shapes. We believe accurate copying (strong perception) is the first step to visual o1. Accordingly, we introduce the concept of "slow perception" (SP), which guides the model to gradually perceive basic point-line combinations, as our humans, reconstruct complex geometric structures progressively. There are two-fold stages in SP: a) perception decomposition. Perception is not instantaneous. In this stage, complex geometric figures are broken down into basic simple units to unify geometry representation. b) perception flow, which acknowledges that accurately tracing a line is not an easy task. This stage aims to avoid "long visual jumps" in regressing line segments by using a proposed "perceptual ruler" to trace each line stroke-by-stroke. Surprisingly, such a human-like perception manner enjoys an inference time scaling law -- the slower, the better. Researchers strive to speed up the model's perception in the past, but we slow it down again, allowing the model to read the image step-by-step and carefully.

StepFun, Bin Wang, Bojun Wang et al.

Large language models (LLMs) face low hardware efficiency during decoding, especially for long-context reasoning tasks. This paper introduces Step-3, a 321B-parameter VLM with hardware-aware model-system co-design optimized for minimizing decoding costs. Step-3 innovates in two key dimensions: (1) A novel Multi-Matrix Factorization Attention (MFA) mechanism that significantly reduces both KV cache size and computation while maintaining high attention expressiveness, and (2) Attention-FFN Disaggregation (AFD), a distributed inference system that decouples attention and Feed-Forward Network (FFN) layers into specialized subsystems. This co-design achieves unprecedented cost efficiency: Step-3 significantly reduces theoretical decoding costs compared with models like DeepSeek-V3 and Qwen3 MoE 235B, with the gains widening at longer context. Step-3 achieves low cost while activating 38B parameters per token (more than DeepSeek-V3 and Qwen3 MoE 235B), demonstrating that hardware-aligned attention arithmetic intensity, MoE sparsity, and AFD are critical to cost-effectiveness. We perform a head-to-head comparison with DeepSeek-V3 in its favorable scenarios. Our implementation on Hopper GPUs achieves a decoding throughput of up to 4,039 tokens per second per GPU under 50ms TPOT SLA (4K context, FP8, no MTP). It is higher than DeepSeek-V3's 2,324 in the same setup and sets a new Pareto frontier for LLM decoding.

Kang Fu, Yicong Peng, Zicheng Zhang et al.

Recently, many algorithms have employed image-adaptive lookup tables (LUTs) to achieve real-time image enhancement. Nonetheless, a prevailing trend among existing methods has been the employment of linear combinations of basic LUTs to formulate image-adaptive LUTs, which limits the generalization ability of these methods. To address this limitation, we propose a novel framework named AttentionLut for real-time image enhancement, which utilizes the attention mechanism to generate image-adaptive LUTs. Our proposed framework consists of three lightweight modules. We begin by employing the global image context feature module to extract image-adaptive features. Subsequently, the attention fusion module integrates the image feature with the priori attention feature obtained during training to generate image-adaptive canonical polyadic tensors. Finally, the canonical polyadic reconstruction module is deployed to reconstruct image-adaptive residual 3DLUT, which is subsequently utilized for enhancing input images. Experiments on the benchmark MIT-Adobe FiveK dataset demonstrate that the proposed method achieves better enhancement performance quantitatively and qualitatively than the state-of-the-art methods.

Ziheng Chen, Jia Wang, Jun Zhuang et al.

The right to be forgotten (RTBF) seeks to safeguard individuals from the enduring effects of their historical actions by implementing machine-learning techniques. These techniques facilitate the deletion of previously acquired knowledge without requiring extensive model retraining. However, they often overlook a critical issue: unlearning processes bias. This bias emerges from two main sources: (1) data-level bias, characterized by uneven data removal, and (2) algorithm-level bias, which leads to the contamination of the remaining dataset, thereby degrading model accuracy. In this work, we analyze the causal factors behind the unlearning process and mitigate biases at both data and algorithmic levels. Typically, we introduce an intervention-based approach, where knowledge to forget is erased with a debiased dataset. Besides, we guide the forgetting procedure by leveraging counterfactual examples, as they maintain semantic data consistency without hurting performance on the remaining dataset. Experimental results demonstrate that our method outperforms existing machine unlearning baselines on evaluation metrics.

Kang Fu, Huiyu Duan, Zicheng Zhang et al.

Recent advancements in text-to-image (T2I) generation have spurred the development of text-to-3D asset (T23DA) generation, leveraging pretrained 2D text-to-image diffusion models for text-to-3D asset synthesis. Despite the growing popularity of text-to-3D asset generation, its evaluation has not been well considered and studied. However, given the significant quality discrepancies among various text-to-3D assets, there is a pressing need for quality assessment models aligned with human subjective judgments. To tackle this challenge, we conduct a comprehensive study to explore the T23DA quality assessment (T23DAQA) problem in this work from both subjective and objective perspectives. Given the absence of corresponding databases, we first establish the largest text-to-3D asset quality assessment database to date, termed the AIGC-T23DAQA database. This database encompasses 969 validated 3D assets generated from 170 prompts via 6 popular text-to-3D asset generation models, and corresponding subjective quality ratings for these assets from the perspectives of quality, authenticity, and text-asset correspondence, respectively. Subsequently, we establish a comprehensive benchmark based on the AIGC-T23DAQA database, and devise an effective T23DAQA model to evaluate the generated 3D assets from the aforementioned three perspectives, respectively.

Jiaxin Liu, Jia Wang, Saihui Hou et al.

In recent years, the explosive advancement of deepfake technology has posed a critical and escalating threat to public security: diffusion-based digital human generation. Unlike traditional face manipulation methods, such models can generate highly realistic videos with consistency via multimodal control signals. Their flexibility and covertness pose severe challenges to existing detection strategies. To bridge this gap, we introduce DigiFakeAV, the new large-scale multimodal digital human forgery dataset based on diffusion models. Leveraging five of the latest digital human generation methods and a voice cloning method, we systematically construct a dataset comprising 60,000 videos (8.4 million frames), covering multiple nationalities, skin tones, genders, and real-world scenarios, significantly enhancing data diversity and realism. User studies demonstrate that the misrecognition rate by participants for DigiFakeAV reaches as high as 68%. Moreover, the substantial performance degradation of existing detection models on our dataset further highlights its challenges. To address this problem, we propose DigiShield, an effective detection baseline based on spatiotemporal and cross-modal fusion. By jointly modeling the 3D spatiotemporal features of videos and the semantic-acoustic features of audio, DigiShield achieves state-of-the-art (SOTA) performance on the DigiFakeAV and shows strong generalization on other datasets.

Jia Wang, Xinfeng Zhang, Gai Zhang et al.

Implicit Neural Representations (INRs) have demonstrated significant potential in video compression by representing videos as neural networks. However, as the number of frames increases, the memory consumption for training and inference increases substantially, posing challenges in resource-constrained scenarios. Inspired by the success of traditional video compression frameworks, which process video frame by frame and can efficiently compress long videos, we adopt this modeling strategy for INRs to decrease memory consumption, while aiming to unify the frameworks from the perspective of timeline-based autoregressive modeling. In this work, we present a novel understanding of INR models from an autoregressive (AR) perspective and introduce a Unified AutoRegressive Framework for memory-efficient Neural Video Compression (UAR-NVC). UAR-NVC integrates timeline-based and INR-based neural video compression under a unified autoregressive paradigm. It partitions videos into several clips and processes each clip using a different INR model instance, leveraging the advantages of both compression frameworks while allowing seamless adaptation to either in form. To further reduce temporal redundancy between clips, we design two modules to optimize the initialization, training, and compression of these model parameters. UAR-NVC supports adjustable latencies by varying the clip length. Extensive experimental results demonstrate that UAR-NVC, with its flexible video clip setting, can adapt to resource-constrained environments and significantly improve performance compared to different baseline models. The project page: "https://wj-inf.github.io/UAR-NVC-page/".