Autoregressive models, built based on the Next Token Prediction (NTP) paradigm, show great potential in developing a unified framework that integrates both language and vision tasks. Pioneering works introduce NTP to autoregressive visual generation tasks. In this work, we rethink the NTP for autoregressive image generation and extend it to a novel Next Patch Prediction (NPP) paradigm. Our key idea is to group and aggregate image tokens into patch tokens with higher information density. By using patch tokens as a more compact input sequence, the autoregressive model is trained to predict the next patch, significantly reducing computational costs. To further exploit the natural hierarchical structure of image data, we propose a multi-scale coarse-to-fine patch grouping strategy. With this strategy, the training process begins with a large patch size and ends with vanilla NTP where the patch size is 1x1, thus maintaining the original inference process without modifications. Extensive experiments across a diverse range of model sizes demonstrate that NPP could reduce the training cost to around 0.6 times while improving image generation quality by up to 1.0 FID score on the ImageNet 256x256 generation benchmark. Notably, our method retains the original autoregressive model architecture without introducing additional trainable parameters or specifically designing a custom image tokenizer, offering a flexible and plug-and-play solution for enhancing autoregressive visual generation.

Recent advances insingle-view3D scene reconstruction have highlighted the challenges in capturing fine geometric details and ensuring structural consistency, particularly in high-fidelity outdoor scene modeling. This paper presents Niagara, a new single-view 3D scene reconstruction framework that can faithfully reconstruct challenging outdoor scenes from a single input image for the first time. Our approach integrates monocular depth and normal estimation as input, which substantially improves its ability to capture fine details, mitigating common issues like geometric detail loss and deformation. Additionally, we introduce a geometric affine field (GAF) and 3D self-attention as geometry-constraint, which combines the structural properties of explicit geometry with the adaptability of implicit feature fields, striking a balance between efficient rendering and high-fidelity reconstruction. Our framework finally proposes a specialized encoder-decoder architecture, where a depth-based 3D Gaussian decoder is proposed to predict 3D Gaussian parameters, which can be used for novel view synthesis. Extensive results and analyses suggest that our Niagara surpasses prior SoTA approaches such as Flash3D in both single-view and dual-view settings, significantly enhancing the geometric accuracy and visual fidelity, especially in outdoor scenes. Webpage: https://ai-kunkun.github.io/Niagara page/.

Multimodal Large Language Models (MLLMs) have displayed remarkable performance in multi-modal tasks, particularly in visual comprehension. However, we reveal that MLLMs often generate incorrect answers even when they understand the visual content. To this end, we manually construct a benchmark with 12 categories and design evaluation metrics that assess the degree of error in MLLM responses even when the visual content is seemingly understood. Based on this benchmark, we test 15 leading MLLMs and analyze the distribution of attention maps and logits of some MLLMs. Our investigation identifies two primary issues: 1) most instruction tuning datasets predominantly feature questions that "directly" relate to the visual content, leading to a bias in MLLMs’ responses to other indirect questions, and 2) MLLMs’ attention to visual tokens is notably lower than to system and question tokens. We further observe that attention scores between questions and visual tokens as well as the model’s confidence in the answers are lower in response to misleading questions than to straightforward ones. To address the first challenge, we introduce a paired positive and negative data construction pipeline to diversify the dataset. For the second challenge, we propose to enhance the model’s focus on visual content during decoding by refining the text and visual prompt. For the text prompt, we propose a content guided refinement strategy that performs preliminary visual content analysis to generate structured information before answering the question. Additionally, we employ a visual attention refinement strategy that highlights question-relevant visual tokens to increase the model’s attention to visual content that aligns with the question. Extensive experiments demonstrate that these challenges can be significantly mitigated with our proposed dataset and techniques.

In this work, we present DreamDance, a novel method for animating human images using only skeleton pose sequences as conditional inputs. Existing approaches struggle with generating coherent, high-quality content in an efficient and user-friendly manner. Concretely, baseline methods relying on only 2D pose guidance lack the cues of 3D information, leading to suboptimal results, while methods using 3D representation as guidance achieve higher quality but involve a cumbersome and time-intensive process. To address these limitations, DreamDance enriches 3D geometry cues from 2D poses by introducing an efficient diffusion model, enabling high-quality human image animation with various guidance. Our key insight is that human images naturally exhibit multiple levels of correlation, progressing from coarse skeleton poses to fine-grained geometry cues, and further from these geometry cues to explicit appearance details. Capturing such correlations could enrich the guidance signals, facilitating intra-frame coherency and inter-frame consistency. Specifically, we construct the TikTok-Dance5K dataset, comprising 5K high-quality dance videos with detailed frame annotations, including human pose, depth, and normal maps. Next, we introduce a Mutually Aligned Geometry Diffusion Model to generate fine-grained depth and normal maps for enriched guidance. Finally, a Cross-domain Controller incorporates multi-level guidance to animate human images effectively with a video diffusion model. Extensive experiments demonstrate that our method achieves state-of-the-art performance in animating human images.

Multimodal large language models (MLLMs) offer a powerful mechanism for interpreting visual information. However, they often suffer from hallucinations, which impede the real-world usage of these models. Existing methods attempt to alleviate this issue by designing special decoding strategies that penalize the summary tokens. However, these methods lack analysis of the relationship between hallucination and summarization mechanism of LLMs. Interestingly, we find that penalizing summary tokens is not necessary: merely intervening the query-key parameters variance, without costing extra inference time, still alleviates hallucinations. Specifically, we explore the causes of hallucinations by analyzing localized self-attention patterns called ``anchor" tokens and define the attention localization degree of the model as token propagation probabilities. Our analysis reveals that over-propagation of anchor tokens occurs when the distribution of eigenvalues of the query and key matrices has a non-zero mean and a polarized variance, leading to excessive dependence on anchor tokens while neglecting vision information and describes the image content with hallucination. Based on the observation, we propose a versatile plug-and-play decoding strategy, Dynamic Token Propagation Mechanism (TAME), to alleviate excessive propagation by dynamically intervening the eigenspectrum variance of the attention weight, thereby alleviating hallucinations without relying on complex decoding strategies. Extensive experiments reveal a correlation between the eigenspectrum and hallucinations across various MLLMs, and show that TAME reduces the percentage of hallucinated objects.

Recent advancements in video generation have enabled the creation of high-quality, visually compelling videos. However, generating videos that adhere to the laws of physics remains a critical challenge for applications requiring realism and accuracy. In this work, we propose **PhysHPO**, a novel framework for Hierarchical Cross-Modal Direct Preference Optimization, to tackle this challenge by enabling fine-grained preference alignment for physically plausible video generation. PhysHPO optimizes video alignment across four hierarchical granularities: a) ***Instance Level***, aligning the overall video content with the input prompt; b) ***State Level***, ensuring temporal consistency using boundary frames as anchors; c) ***Motion Level***, modeling motion trajectories for realistic dynamics; and d) ***Semantic Level***, maintaining logical consistency between narrative and visuals. Recognizing that real-world videos are the best reflections of physical phenomena, we further introduce an automated data selection pipeline to efficiently identify and utilize *"good data"* from existing large-scale text-video datasets, thereby eliminating the need for costly and time-intensive dataset construction. Extensive experiments on both physics-focused and general capability benchmarks demonstrate that PhysHPO significantly improves physical plausibility and overall video generation quality of advanced models. To the best of our knowledge, this is the first work to explore fine-grained preference alignment and data selection for video generation, paving the way for more realistic and human-preferred video generation paradigms.

Large Multimodal Models (LMMs) have achieved impressive progress in visual perception and reasoning. However, when confronted with visually ambiguous or non-semantic scene text, they often struggle to accurately spot and understand the content, frequently generating semantically plausible yet visually incorrect answers, which we refer to as semantic hallucination. In this work, we investigate the underlying causes of semantic hallucination and identify a key finding: Transformer layers in LLM with stronger attention focus on scene text regions are less prone to producing semantic hallucinations. Thus, we propose a training-free semantic hallucination mitigation framework comprising two key components: (1) ZoomText, a coarse-to-fine strategy that identifies potential text regions without external detectors; and (2) Grounded Layer Correction, which adaptively leverages the internal representations from layers less prone to hallucination to guide decoding, correcting hallucinated outputs for non-semantic samples while preserving the semantics of meaningful ones. To enable rigorous evaluation, we introduce TextHalu-Bench, a benchmark of 1,740 samples spanning both semantic and non-semantic cases, with manually curated question–answer pairs designed to probe model hallucinations. Extensive experiments demonstrate that our method not only effectively mitigates semantic hallucination but also achieves strong performance on public benchmarks for scene text spotting and understanding.

We show that the effectiveness of the well celebrated Mixup can be further improved if instead of using it as the sole learning objective, it is utilized as an additional regularizer to the standard cross-entropy loss. This simple change not only improves accuracy but also significantly improves the quality of the predictive uncertainty estimation of Mixup in most cases under various forms of covariate shifts and out-of-distribution detection experiments. In fact, we observe that Mixup otherwise yields much degraded performance on detecting out-of-distribution samples possibly, as we show empirically, due to its tendency to learn models exhibiting high-entropy throughout; making it difficult to differentiate in-distribution samples from out-of-distribution ones. To show the efficacy of our approach (RegMixup), we provide thorough analyses and experiments on vision datasets (ImageNet & CIFAR-10/100) and compare it with a suite of recent approaches for reliable uncertainty estimation.

While deep neural networks have achieved remarkable success in various computer vision tasks, they often fail to generalize to new domains and subtle variations of input images. Several defenses have been proposed to improve the robustness against these variations. However, current defenses can only withstand the specific attack used in training, and the models often remain vulnerable to other input variations. Moreover, these methods often degrade performance of the model on clean images and do not generalize to out-of-domain samples. In this paper we present Generative Adversarial Training, an approach to simultaneously improve the model’s generalization to the test set and out-of-domain samples as well as its robustness to unseen adversarial attacks. Instead of altering a low-level pre-defined aspect of images, we generate a spectrum of low-level, mid-level and high-level changes using generative models with a disentangled latent space. Adversarial training with these examples enable the model to withstand a wide range of attacks by observing a variety of input alterations during training. We show that our approach not only improves performance of the model on clean images and out-of-domain samples but also makes it robust against unforeseen attacks and outperforms prior work. We validate effectiveness of our method by demonstrating results on various tasks such as classification, segmentation and object detection.

In this paper, we study the problem of recognizing man-made objects and present a novel affine-invariant feature, Low-rank SIFT, which exploits the regular appearance property in man-made objects. The proposed feature achieves full affine invariance without needing to simulate over affine parameter space. We rectify local patches by converting them to their low-rank forms to achieve skew invariance, and perform the way similar to conventional SIFT to resolve rotation, translation and scaling ambiguity. The main contributions lie in two-fold: our method seeks to leverage low-rank prior to estimate affine parameters for local patches directly and we propose a fast algorithm to compute such parameters by introducing the Low-rank Integral Map. Besides, we describe a pipeline of constructing a geotagged building database from the ground up. We demonstrate the effectiveness of our approach in the application to place recognition.