We present Visual AutoRegressive modeling (VAR), a new generation paradigm that redefines the autoregressive learning on images as coarse-to-fine "next-scale prediction" or "next-resolution prediction", diverging from the standard raster-scan "next-token prediction". This simple, intuitive methodology allows autoregressive (AR) transformers to learn visual distributions fast and generalize well: VAR, for the first time, makes AR models surpass diffusion transformers in image generation. On ImageNet 256x256 benchmark, VAR significantly improve AR baseline by improving Frechet inception distance (FID) from 18.65 to 1.80, inception score (IS) from 80.4 to 356.4, with around 20x faster inference speed. It is also empirically verified that VAR outperforms the Diffusion Transformer (DiT) in multiple dimensions including image quality, inference speed, data efficiency, and scalability. Scaling up VAR models exhibits clear power-law scaling laws similar to those observed in LLMs, with linear correlation coefficients near -0.998 as solid evidence. VAR further showcases zero-shot generalization ability in downstream tasks including image in-painting, out-painting, and editing. These results suggest VAR has initially emulated the two important properties of LLMs: Scaling Laws and zero-shot task generalization. We have released all models and codes to promote the exploration of AR/VAR models for visual generation and unified learning.
Visual representation learning has been a cornerstone in computer vision, evolving from supervised learning with human-annotated labels to aligning image-text pairs from the Internet. Despite recent advancements in multi-modal large language models (MLLMs), the visual representations they rely on, such as CLIP embeddings, often lack access to external world knowledge critical for real-world visual reasoning. In this work, we propose Visual Table, a novel visual representation tailored for MLLMs. It provides hierarchical text descriptions of holistic visual scenes, consisting of a scene description and multiple object-centric descriptions that encompass categories, attributes, and knowledge at instance level. We further develop a scalable generator for visual table generation and train it on small-scale annotations from GPT4V. Extensive evaluations demonstrate that, with generated visual tables as additional visual representations, our model can consistently outperform the state-of-the-art (SOTA) MLLMs across diverse benchmarks. When visual tables serve as standalone visual representations, our model can closely match or even beat the SOTA MLLMs that are built on CLIP visual embeddings. Our code is available at https://github.com/LaVi-Lab/Visual-Table.
This paper proposes a simple, yet effective framework, called GiT, simultaneously applicable for various vision tasks only with a vanilla ViT. Motivated by the universality of the Multi-layer Transformer architecture (e.g, GPT) widely used in large language models (LLMs), we seek to broaden its scope to serve as a powerful vision foundation model (VFM). However, unlike language modeling, visual tasks typically require specific modules, such as bounding box heads for detection and pixel decoders for segmentation, greatly hindering the application of powerful multi-layer transformers in the vision domain. To solve this, we design a universal language interface that empowers the successful auto-regressive decoding to adeptly unify various visual tasks, from image-level understanding (e.g., captioning), over sparse perception (e.g., detection), to dense prediction (e.g., segmentation). Based on the above designs, the entire model is composed solely of a ViT, without any specific additions, offering a remarkable architectural simplification. GiT is a multi-task visual model, jointly trained across five representative benchmarks without task-specific fine-tuning. Interestingly, our GiT builds a new benchmark in generalist performance, and fosters mutual enhancement across tasks, leading to significant improvements compared to isolated training. This reflects a similar impact observed in LLMs. Further enriching training with 27 datasets, GiT achieves strong zero-shot results over various tasks. Due to its simple design, this paradigm holds promise for narrowing the architectural gap between vision and language. Code and models will be available at \url{https://github.com/Haiyang-W/GiT}.
Cyclical MCMC is a novel MCMC framework recently proposed by Zhang et al. (2019) to address the challenge posed by high-dimensional multimodal posterior distributions like those arising in deep learning. The algorithm works by generating a nonhomogeneous Markov chain that tracks -- cyclically in time -- tempered versions of the target distribution. We show in this work that cyclical MCMC converges to the desired probability distribution in settings where the Markov kernels used are fast mixing, and sufficiently long cycles are employed. However in the far more common settings of slow mixing kernels, the algorithm may fail to produce samples from the desired distribution. In particular, in a simple mixture example with unequal variance, we show by simulation that cyclical MCMC fails to converge to the desired limit. Finally, we show that cyclical MCMC typically estimates well the local shape of the target distribution around each mode, even when we do not have convergence to the target.
As transformer-based language models are trained on increasingly large datasets and with vast numbers of parameters, finding more efficient alternatives to the standard Transformer has become very valuable. While many efficient Transformers and Transformer alternatives have been proposed, none provide theoretical guarantees that they are a suitable replacement for the standard Transformer. This makes it challenging to identify when to use a specific model and what directions to prioritize for further investigation. In this paper, we aim to understand the capabilities and limitations of efficient Transformers, specifically the Sparse Transformer and the Linear Transformer. We focus on their reasoning capability as exhibited by Chain-of-Thought (CoT) prompts and follow previous works to model them as Dynamic Programming (DP) problems. Our results show that while these models are expressive enough to solve general DP tasks, contrary to expectations, they require a model size that scales with the problem size. Nonetheless, we identify a class of DP problems for which these models can be more efficient than the standard Transformer. We confirm our theoretical results through experiments on representative DP tasks, adding to the understanding of efficient Transformers' practical strengths and weaknesses.
Solving partial differential equations (PDEs) efficiently is essential for analyzing complex physical systems. Recent advancements in leveraging deep learning for solving PDE have shown significant promise. However, machine learning methods, such as Physics-Informed Neural Networks (PINN), face challenges in handling high-order derivatives of neural network-parameterized functions. Inspired by Forward Laplacian, a recent method of accelerating Laplacian computation, we propose an efficient computational framework, Differential Operator with Forward-propagation (DOF), for calculating general second-order differential operators without losing any precision. We provide rigorous proof of the advantages of our method over existing methods, demonstrating two times improvement in efficiency and reduced memory consumption on any architectures. Empirical results illustrate that our method surpasses traditional automatic differentiation (AutoDiff) techniques, achieving 2x improvement on the MLP structure and nearly 20x improvement on the MLP with Jacobian sparsity.
In this work, we leverage the intrinsic segmentation of language sequences and design a new positional encoding method called Bilevel Positional Encoding (BiPE). For each position, our BiPE blends an intra-segment encoding and an inter-segment encoding. The intra-segment encoding identifies the locations within a segment and helps the model capture the semantic information therein via absolute positional encoding. The inter-segment encoding specifies the segment index, models the relationships between segments, and aims to improve extrapolation capabilities via relative positional encoding. Theoretical analysis shows this disentanglement of positional information makes learning more effective. The empirical results also show that our BiPE has superior length extrapolation capabilities across a wide range of tasks in diverse text modalities.
Long-term fetal heart rate (FHR) monitoring during the antepartum period, increasingly popularized by electronic FHR monitoring, represents a growing approach in FHR monitoring. This kind of continuous monitoring, in contrast to the short-term one, collects an extended period of fetal heart data. This offers a more comprehensive understanding of fetus's conditions. However, the interpretation of long-term antenatal fetal heart monitoring is still in its early stages, lacking corresponding clinical standards. Furthermore, the substantial amount of data generated by continuous monitoring imposes a significant burden on clinical work when analyzed manually. To address above challenges, this study develops an automatic analysis system named LARA (Long-term Antepartum Risk Analysis system) for continuous FHR monitoring, combining deep learning and information fusion methods. LARA's core is a well-established convolutional neural network (CNN) model. It processes long-term FHR data as input and generates a Risk Distribution Map (RDM) and Risk Index (RI) as the analysis results. We evaluate LARA on inner test dataset, the performance metrics are as follows: AUC 0.872, accuracy 0.816, specificity 0.811, sensitivity 0.806, precision 0.271, and F1 score 0.415. In our study, we observe that long-term FHR monitoring data with higher RI is more likely to result in adverse outcomes (p=0.0021). In conclusion, this study introduces LARA, the first automated analysis system for long-term FHR monitoring, initiating the further explorations into its clinical value in the future.
Designing expressive Graph Neural Networks (GNNs) is a fundamental topic in the graph learning community. So far, GNN expressiveness has been primarily assessed via the Weisfeiler-Lehman (WL) hierarchy. However, such an expressivity measure has notable limitations: it is inherently coarse, qualitative, and may not well reflect practical requirements (e.g., the ability to encode substructures). In this paper, we introduce a unified framework for quantitatively studying the expressiveness of GNN architectures, addressing all the above limitations. Specifically, we identify a fundamental expressivity measure termed homomorphism expressivity, which quantifies the ability of GNN models to count graphs under homomorphism. Homomorphism expressivity offers a complete and practical assessment tool: the completeness enables direct expressivity comparisons between GNN models, while the practicality allows for understanding concrete GNN abilities such as subgraph counting. By examining four classes of prominent GNNs as case studies, we derive simple, unified, and elegant descriptions of their homomorphism expressivity for both invariant and equivariant settings. Our results provide novel insights into a series of previous work, unify the landscape of different subareas in the community, and settle several open questions. Empirically, extensive experiments on both synthetic and real-world tasks verify our theory, showing that the practical performance of GNN models aligns well with the proposed metric.
Powder X-ray diffraction (PXRD) is a crucial means for crystal structure determination. Such determination often involves external database matching to find a structural analogue and Rietveld refinement to obtain finer structure. However, databases may be incomplete and Rietveld refinement often requires intensive trial-and-error efforts from trained experimentalists, which remains ineffective in practice. To settle these issues, we propose XtalNet, the first end-to-end deep learning-based framework capable of ab initio generation of crystal structures that accurately match given PXRD patterns. The model employs contrastive learning and Diffusion-based conditional generation to enable the simultaneous execution of two tasks: crystal structure retrieval based on PXRD patterns and conditional structure generations. To validate the effectiveness of XtalNet, we curate a much more challenging and practical dataset hMOF-100, XtalNet performs well on this dataset, reaching 96.3\% top-10 hit ratio on the database retrieval task and 95.0\% top-10 match rate on the ranked structure generation task.