Recent advances in visual generation have made significant strides in producing content of exceptional quality. However, most methods suffer from a fundamental problem - a bottleneck of inference computational efficiency. Most of these algorithms involve multiple passes over a transformer model to generate tokens or denoise inputs. However, the model size is kept consistent throughout all iterations, which makes it computationally expensive. In this work, we aim to address this issue primarily through two key ideas - (a) not all parts of the generation process need equal compute, and we design a decode time model scaling schedule to utilize compute effectively, and (b) we can cache and reuse some of the computation. Combining these two ideas leads to using smaller models to process more tokens while large models process fewer tokens. These different-sized models do not increase the parameter size, as they share parameters. We rigorously experiment with ImageNet256$\times$256 , UCF101, and Kinetics600 to showcase the efficacy of the proposed method for image/video generation and frame prediction. Our experiments show that with almost $3\times$ less compute than baseline, our model obtains competitive performance.

Vision Transformers (ViT) have emerged as the de-facto choice for numerous industry grade vision solutions. But their inference cost can be prohibitive for many settings, as they compute self-attention in each layer which suffers from quadratic computational complexity in the number of tokens. On the other hand, spatial information in images and spatio-temporal information in videos is usually sparse and redundant. In this work, we introduce LookupViT, that aims to exploit this information sparsity to reduce ViT inference cost. LookupViT provides a novel general purpose vision transformer block that operates by compressing information from higher resolution tokens to a fixed number of tokens. These few compressed tokens undergo meticulous processing, while the higher-resolution tokens are passed through computationally cheaper layers. Information sharing between these two token sets is enabled through a bidirectional cross-attention mechanism. The approach offers multiple advantages - (a) easy to implement on standard ML accelerators (GPUs/TPUs) via standard high-level operators, (b) applicable to standard ViT and its variants, thus generalizes to various tasks, (c) can handle different tokenization and attention approaches. LookupViT also offers flexibility for the compressed tokens, enabling performance-computation trade-offs in a single trained model. We show LookupViT's effectiveness on multiple domains - (a) for image-classification (ImageNet-1K and ImageNet-21K), (b) video classification (Kinetics400 and Something-Something V2), (c) image captioning (COCO-Captions) with a frozen encoder. LookupViT provides $2\times$ reduction in FLOPs while upholding or improving accuracy across these domains. In addition, LookupViT also demonstrates out-of-the-box robustness and generalization on image classification (ImageNet-C,R,A,O), improving by up to $4\%$ over ViT.

Recent document question answering models consist of two key components: the vision encoder, which captures layout and visual elements in images, and a Large Language Model (LLM) that helps contextualize questions to the image and supplements them with external world knowledge to generate accurate answers. However, the relative contributions of the vision encoder and the language model in these tasks remain unclear. This is especially interesting given the effectiveness of instruction-tuned LLMs, which exhibit remarkable adaptability to new tasks. To this end, we explore the following aspects in this work: (1) The efficacy of an LLM-only approach on document question answering tasks (2) strategies for serializing textual information within document images and feeding it directly to an instruction-tuned LLM, thus bypassing the need for an explicit vision encoder (3) thorough quantitative analysis on the feasibility of such an approach. Our comprehensive analysis encompasses six diverse benchmark datasets, utilizing LLMs of varying scales. Our findings reveal that a strategy exclusively reliant on the LLM yields results that are on par with or closely approach state-of-the-art performance across a range of datasets. We posit that this evaluation framework will serve as a guiding resource for selecting appropriate datasets for future research endeavors that emphasize the fundamental importance of layout and image content information.

Current visual detectors, though impressive within their training distribution, often fail to parse out-of-distribution scenes into their constituent entities. Recent test-time adaptation methods use auxiliary self-supervised losses to adapt the network parameters to each test example independently and have shown promising results towards generalization outside the training distribution for the task of image classification. In our work, we find evidence that these losses are insufficient for the task of scene decomposition, without also considering architectural inductive biases. Recent slot-centric generative models attempt to decompose scenes into entities in a self-supervised manner by reconstructing pixels. Drawing upon these two lines of work, we propose Slot-TTA, a semi-supervised slot-centric scene decomposition model that at test time is adapted per scene through gradient descent on reconstruction or cross-view synthesis objectives. We evaluate Slot-TTA across multiple input modalities, images or 3D point clouds, and show substantial out-of-distribution performance improvements against state-of-the-art supervised feed-forward detectors, and alternative test-time adaptation methods.

State-of-the-art information extraction methods are limited by OCR errors. They work well for printed text in form-like documents, but unstructured, handwritten documents still remain a challenge. Adapting existing models to domain-specific training data is quite expensive, because of two factors, 1) limited availability of the domain-specific documents (such as handwritten prescriptions, lab notes, etc.), and 2) annotations become even more challenging as one needs domain-specific knowledge to decode inscrutable handwritten document images. In this work, we focus on the complex problem of extracting medicine names from handwritten prescriptions using only weakly labeled data. The data consists of images along with the list of medicine names in it, but not their location in the image. We solve the problem by first identifying the regions of interest, i.e., medicine lines from just weak labels and then injecting a domain-specific medicine language model learned using only synthetically generated data. Compared to off-the-shelf state-of-the-art methods, our approach performs > 2.5 better in medicine names extraction from prescriptions.

Imitation learning seeks to circumvent the difficulty in designing proper reward functions for training agents by utilizing expert behavior. With environments modeled as Markov Decision Processes (MDP), most of the existing imitation algorithms are contingent on the availability of expert demonstrations in the same MDP as the one in which a new imitation policy is to be learned. In this paper, we study the problem of how to imitate tasks when there exist discrepancies between the expert and agent MDP. These discrepancies across domains could include differing dynamics, viewpoint, or morphology; we present a novel framework to learn correspondences across such domains. Importantly, in contrast to prior works, we use unpaired and unaligned trajectories containing only states in the expert domain, to learn this correspondence. We utilize a cycle-consistency constraint on both the state space and a domain agnostic latent space to do this. In addition, we enforce consistency on the temporal position of states via a normalized position estimator function, to align the trajectories across the two domains. Once this correspondence is found, we can directly transfer the demonstrations on one domain to the other and use it for imitation. Experiments across a wide variety of challenging domains demonstrate the efficacy of our approach.

Deep Reinforcement Learning has shown tremendous success in solving several games and tasks in robotics. However, unlike humans, it generally requires a lot of training instances. Trajectories imitating to solve the task at hand can help to increase sample-efficiency of deep RL methods. In this paper, we present a simple approach to use such trajectories, applied to the challenging Ball-in-Maze Games, recently introduced in the literature. We show that in spite of not using human-generated trajectories and just using the simulator as a model to generate a limited number of trajectories, we can get a speed-up of about 2-3x in the learning process. We also discuss some challenges we observed while using trajectory-based learning for very sparse reward functions.

Recent research has demonstrated the brittleness of machine learning systems to adversarial perturbations. However, the studies have been mostly limited to perturbations on images and more generally, classification that does not deal with temporally varying inputs. In this paper we ask "Are adversarial perturbations possible in real-time video classification systems and if so, what properties must they satisfy?" Such systems find application in surveillance applications, smart vehicles, and smart elderly care and thus, misclassification could be particularly harmful (e.g., a mishap at an elderly care facility may be missed). We show that accounting for temporal structure is key to generating adversarial examples in such systems. We exploit recent advances in generative adversarial network (GAN) architectures to account for temporal correlations and generate adversarial samples that can cause misclassification rates of over 80% for targeted activities. More importantly, the samples also leave other activities largely unaffected making them extremely stealthy. Finally, we also surprisingly find that in many scenarios, the same perturbation can be applied to every frame in a video clip that makes the adversary's ability to achieve misclassification relatively easy.