Jump to Content
Radu Soricut

Radu Soricut

I am a Principal Scientist & Research Director at Google Research, leading teams that conduct natural language modeling and machine learning research, with projects that focus on multimodal (language & vision) understanding and generation. I completed my PhD at USC, where I worked with Daniel Marcu and Kevin Knight.
See also my personal webpage for more information.
Authored Publications
Google Publications
Other Publications
Sort By
  • Title
  • Title, desc
  • Year
  • Year, desc
    Preview abstract Effective scaling and a flexible task interface enable large-capacity language models to excel at many tasks. PaLI (Pathways Language and Image model) extends these ideas to the joint modeling of language and vision. PaLI is a model that generates text based on visual and textual inputs. Using this API, PaLI is able to perform many vision, language, and multimodal tasks, across many languages. We train PaLI with two main principles: reuse of pretrained unimodal components, and joint scaling of modalities. Using large-capacity pretrained language models and vision models allows us to capitalize on their existing capabilities, while leveraging the substantial cost of training them. We scale PaLI models across three axes:the language component, the vision component, and the training data that fuses them. For the vision component, we train the largest and best-performing VisionTransformer (ViT) to date. For the data, we build an image-text training set over10B images and covering over 100 languages. PaLI inherits and enhances language-understanding capabilities, and achieves state-of-the-art in multiple vision and language tasks (image classification, image captioning, visual question-answering, scene-text understanding, etc.), based on a simple, modular, and reuse-friendly platform for modeling and scaling. View details
    Preview abstract Visual Question Answering (VQA) has been primarily studied through the lens of the English language. Yet, tackling VQA in other languages in the same manner would require a considerable amount of resources. In this paper, we propose scalable solutions to multilingual visual question answering (mVQA), on both data and modeling fronts. We first propose a translation-based framework to mVQA data generation that requires much less human annotation efforts than the conventional approach of directly collection questions and answers. Then, we apply our framework to the multilingual captions in the Crossmodal-3600 dataset and develop an efficient annotation protocol to create MaXM, a test-only VQA benchmark in 7 diverse languages. Finally, we develop a simple, lightweight, and effective approach as well as benchmark state-of-the-art English and multilingual VQA models. We hope that our benchmark encourages further research on mVQA. View details
    Preview abstract Text-guided image editing can have a transformative impact in supporting creative applications. A key challenge is to generate edits that are faithful to the input text prompt, while consistent with the input image. We present Imagen Editor, a cascaded diffusion model, built by fine-tuning Imagen on text-guided image inpainting. Imagen Editor's edits are faithful to the text prompts, which is accomplished by incorporating object detectors for proposing inpainting masks during training. In addition, text-guided image inpainting captures fine details in the input image by conditioning the cascaded pipeline on the original high resolution image. To improve qualitative and quantitative evaluation, we introduce EditBench, a systematic benchmark for text-guided image inpainting. EditBench evaluates inpainting edits on natural and generated images exploring objects, attributes, and scenes. Through extensive human evaluation on EditBench, we find that object-masking during training leads to across-the-board improvements in text-image alignment -- such that Imagen Editor is preferred over DALL-E 2 and Stable Diffusion -- and, as a cohort, these models are better at object-rendering than text-rendering, and handle material/color/size attributes better than count/shape attributes. View details
    Connecting Vision and Language with Video Localized Narratives
    Vittorio Ferrari
    IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2023 (to appear)
    Preview abstract We propose Video Localized Narratives, a new form of multimodal video annotations connecting vision and language. In the original Localized Narratives, annotators speak and move their mouse simultaneously on an image, thus grounding each word with a mouse trace segment. However, this is challenging on a video. Our new protocol empowers annotators to tell the story of a video with Localized Narratives, capturing even complex events involving multiple actors interacting with each other and with several passive objects. We annotated 20k videos of the OVIS, UVO, and Oops datasets, totalling 1.7M words. Based on this data, we also construct new benchmarks for the video narrative grounding and video question-answering tasks, and provide reference results from strong baseline models. Our annotations are available at https://google.github.io/video-localized-narratives/. View details
    Preview abstract The ability to recognize and reason about text embedded in visual inputs is often lacking in vision-and-language (V&L) models, perhaps because V&L pre-training methods have often failed to include such an ability in their training objective. In this paper, we propose PreSTU, a novel pre-training recipe dedicated to scene-text understanding (STU). PreSTU introduces OCR-aware pre-training objectives that encourage the model to recognize text from an image and connect it to the rest of the image content. We implement PreSTU using a simple transformer-based encoder-decoder architecture, combined with large-scale image-text datasets with scene text obtained from an off-the-shelf OCR system. We empirically demonstrate the effectiveness of this pre-training approach on eight visual question answering and four image captioning benchmarks. View details
    Preview abstract Visual Question Answering (VQA) has benefited from increasingly sophisticated models, but has not enjoyed the same level of engagement in terms of data creation. In this paper, we propose a method that automatically derives VQA examples at volume, by leveraging the abundance of existing image-caption annotations combined with neural models for textual question generation. We show that the resulting data is of high-quality. VQA models trained on our data improve state-of-the-art zero-shot accuracy by double digits and achieve a level of robustness that lacks in the same model trained on human-annotated VQA data. View details
    Preview abstract Research in massively multilingual image captioning has been severely hampered by a lack of high-quality evaluation datasets. In this paper we present the Crossmodal-3600 dataset (XM3600 in short), a geographically-diverse set of 3600 images annotated with human-generated reference captions in 36 languages. The images were selected from across the world, covering regions where the 36 languages are spoken, and annotated with captions that achieve consistency in terms of style across all languages, while avoiding annotation artifacts due to direct translation. We apply this benchmark to model selection for massively multilingual image captioning models, and show strong correlation results with human evaluations when using XM3600 as golden references for automatic metrics. View details
    Preview abstract One challenge in evaluating visual question answering (VQA) models in the cross-dataset adaptation setting is that the distribution shifts are multi-modal, making it difficult to identify if it is the shifts in visual or language features that play a key role. In this paper, we propose a semi-automatic framework for generating disentangled shifts by introducing a controllable visual question-answer generation (VQAG) module that is capable of generating highly-relevant and diverse question-answer pairs with the desired dataset style. We use it to create CrossVQA, a collection of test splits for assessing VQA generalization based on the VQA2, VizWiz, and Open Images datasets. We provide an analysis of our generated datasets and demonstrate its utility by using them to evaluate several state-of-the-art VQA systems. One important finding is that the visual shifts in cross-dataset VQA matter more than the language shifts. More broadly, we present a scalable framework for systematically evaluating the machine with little human intervention. View details
    Preview abstract Most existing image retrieval systems use text queries as a way for the user to express what they are looking for. However, fine-grained image retrieval often requires the ability to also express the where in the image the content they are looking for is. The text modality can only cumbersomely express such localization preferences, whereas pointing is a more natural fit. In this paper, we propose an image retrieval setup with a new form of multimodal queries, where the user simultaneously uses both spoken natural language (the what) and mouse traces over an empty canvas (the where) to express the characteristics of the desired target image. We then describe simple modifications to an existing image retrieval model, enabling it to operate in this setup. Qualitative and quantitative experiments show that our model effectively takes this spatial guidance into account, and provides significantly more accurate retrieval results compared to text-only equivalent systems. View details
    Preview abstract Despite recent advances in its theoretical understanding, there still remains a significant gap in the ability of existing meta-learning theorems to explain the performance improvements in the few-shot learning setting, where the number of samples in the target tasks is severely limited. This gap originates from an assumption in the existing theories which supposes that the number of samples in the observed tasks and the number of samples in the target tasks follow the same distribution, an assumption that rarely holds in practice. By relaxing this assumption we develop two PAC-Bayesian bounds tailored for the few-shot learning setting and show that two existing meta-learning algorithms (MAML and Reptile) can be derived from our bounds, thereby bridging the gap between practice and PAC-Bayesian theorems. Furthermore, we derive a new computationally efficient PAC-Bayesian algorithm, and show it outperforms existing meta-learning algorithms on several few-shot benchmark datasets. View details
    Preview abstract The availability of large-scale image captioning and visual question answering datasets has contributed significantly to recent successes in vision-and-language pre-training. However, these datasets are often collected with overrestrictive requirements, inherited from their original target tasks (e.g., image caption generation), which limit the resulting dataset scale and diversity. We take a step further in pushing the limits of vision-and-language pre-training data by relaxing the data collection pipeline used in Conceptual Captions 3M (CC3M) [Sharma et al. 2018] and introduce the Conceptual 12M (CC12M), a dataset with 12 million image-text pairs specifically meant to be used for vision-and-language pre-training. We perform an analysis of this dataset, as well as benchmark its effectiveness against CC3M on multiple downstream tasks with an emphasis on long-tail visual recognition. The quantitative and qualitative results clearly illustrate the benefit of scaling up pre-training data for vision-and-language tasks, as indicated by the new state-of-the-art results on both the nocaps and Conceptual Captions benchmarks. View details
    Preview abstract This paper introduces TeaForN, an extension of the teacher-forcing method to N-grams. Sequence generation models trained with teacher-forcing suffer from problems such as exposure bias and lack of differentiability across timesteps. TeaForN addresses both these problems directly, through the use of a stack of N decoders trained to decode along a secondary time axis that allows model-parameter updates based on N prediction steps. Unlike other approaches, TeaForN can be used with a wide class of decoder architectures and requires minimal modifications from a standard teacher-forcing setup. Empirically, we show that TeaForN boosts model quality and beam-efficiency against several sequence generation benchmarks. View details
    Preview abstract We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning. View details
    Preview abstract Human ratings are currently the most accurate way to assess the quality of an image captioning model, yet most often the only used outcome of an expensive human rating evaluation is a few overall statistics over the evaluation dataset. In this paper, we show that the signal from instance-level human caption ratings can be leveraged to achieve improved captioning models, even when the amount of caption ratings is several orders of magnitude less than the caption training data. We employ a policy gradient method to maximize the human ratings as rewards in an off-policy reinforcement learning setting, using a technique that makes use of a sampling distribution that focuses on the captions that are present in a caption-ratings dataset. We present empirical evidence that indicates that our models learn to generalize the human raters’judgments in the caption-ratings training data to a previously unseen set of images, as judged by a different set of human judges and additionally on a different, multi-dimensional side-by-side human evaluation procedure. View details
    Preview abstract An image caption should fluently present the essential information in a given image, including informative, fine-grained entity mentions and the manner in which these entities interact. However, current captioning models are usually trained to generate captions that only contain common object names, thus falling short on an important “informativeness” dimension. We present a mechanism for integrating image information together with fine-grained labels (assumed to be generated by some upstream models) into a caption that describes the image in a fluent and informative manner. We introduce a multimodal, multi-encoder model based on Transformer that ingests both image features and multiple sources of entity labels. We demonstrate that we can learn to control the appearance of these entity labels in the output, resulting in captions that are both fluent and informative. View details
    Preview abstract Object detection plays an important role in current solutions to vision and language tasks like image captioning and visual question answering. However, popular models like Faster R-CNN rely on a costly process of annotating ground-truths for both the bounding boxes and their corresponding semantic labels, making it less amenable as a primitive task for transfer learning. In this paper, we examine the effect of decoupling box proposal and featurization for down-stream tasks. The key insight is that this allows us to leverage a large amount of labeled annotations that were previously unavailable for standard object detection benchmarks. Empirically, we demonstrate that this leads to effective transfer learning and improved image captioning and visual question answering models, as measured on publicly-available benchmarks. View details
    Preview abstract We present a new dataset of image caption annotations, CHIA, which contains an order of magnitude more images than the MS-COCO dataset and represents a wider variety of both image and image caption styles. We achieve this by extracting and filtering image caption annotations from billions of Internet webpages. We also present quantitative evaluations of a number of image captioning models and show that a model architecture based on Inception-ResNet-v2 CNN for image-feature extraction and Transformer for sequence modeling achieves best performance when trained on the CHIA dataset. We present a new dataset of image caption annotations, Conceptual Captions, which contains an order of magnitude more images than the MS-COCO dataset and represents a wider variety of both images and image caption styles. We achieve this by extracting and filtering image caption annotations from billions of webpages. We also present quantitative evaluations of a number of image captioning models and show that a model architecture based on Inception-ResNet-v2 for image-feature extraction and Transformer for sequence modeling achieves the best performance when trained on the Conceptual Captions dataset. View details
    Preview abstract Supervised training of abstractive language generation models results in learning conditional probabilities over language sequences based on the supervised training signal. When the training signal contains a variety of writing styles, such models may end up learning an 'average' style that is directly influenced by the training data make-up and cannot be controlled by the needs of an application. We describe a family of model architectures capable of capturing both generic language characteristics via shared model parameters, as well as particular style characteristics via private model parameters. Such models are able to generate language according to a specific learned style, while still taking advantage of their power to model generic language phenomena. Furthermore, we describe an extension that uses a mixture of output distributions from all learned styles to perform on-the-fly style adaptation based on the textual input alone. Experimentally, we find that the proposed models consistently outperform models that encapsulate single-style or average-style language generation capabilities. View details
    Points, Paths, and Playscapes: Large-scale Spatial Language Understanding Tasks Set in the Real World
    Daphne Luong
    Proceedings of the First International Workshop on Spatial Language Understanding, Association for Computational Linguistics, New Orleans, Louisiana, USA (2018), pp. 46-52
    Preview abstract Spatial language understanding is important for practical applications and as a building block for better abstract language understanding. Much progress has been made through work on understanding spatial relations and values in images and texts as well as on giving and following navigation instructions in restricted domains. We argue that the next big advances in spatial language understanding can be best supported by creating large-scale datasets that focus on points and paths based in the real world, and then extending these to create online, persistent playscapes that mix human and bot players. The bot players can begin play having undergone a prior training regime, but then must learn, evolve, and survive according to their depth of understanding of scenes, navigation, and interactions. View details
    Preview abstract Policy-gradient approaches to reinforcement learning have two common and undesirable overhead procedures, namely warm-start training and sample variance reduction. In this paper, we describe a reinforcement learning method based on a softmax policy that requires neither of these procedures. Our method combines the advantages of policy-gradient methods with the efficiency and simplicity of maximum-likelihood approaches. We apply this new cold-start reinforcement learning method in training sequence generation models for structured output prediction problems. Empirical evidence validates this method on automatic summarization and image captioning tasks. View details
    Preview abstract We describe a new multi-modal task for computer systems, posed as a combined vision-language comprehension challenge: identify the most suitable \emph{text} describing a scene, given several similar options. Accomplishing the task entails demonstrating comprehension beyond just recognizing ``keywords'' (or key-phrases) and their corresponding visual concepts, and instead requires an alignment between the representations of the two modalities that achieves a visually-grounded ``understanding'' of various linguistic elements and their dependencies. This new task also admits an easy-to-compute and well-understood metric: the accuracy in detecting the true target among the decoys. The paper makes several contributions: a generic mechanism for generating decoys from (human-created) image captions; an instance of applying this mechanism, yielding a large-scale machine comprehension dataset (based on the COCO images and captions) that we make publicly available; results on a human evaluation on this dataset, thus providing a performance ceiling; and several baseline and competitive learning approaches that illustrate the utility of the proposed framework in advancing both image and language machine comprehension. In particular, there is a large gap between human performance and state-of-the-art learning methods, suggesting a fruitful direction for future research. View details
    Preview abstract We present a dual contribution to the task of machine reading-comprehension: a technique for creating large-sized machine-comprehension (MC) datasets using paragraph-vector models; and a novel, hybrid neural-network architecture that combines the representation power of recursive neural networks with the discriminative power of fully-connected multi-layered networks. We use the MC-dataset generation technique to build a dataset of around 2 million examples, for which we empirically determine the high-ceiling of human performance (around 91\% accuracy), as well as the performance of a variety of computer models. Among all the models we have experimented with, our hybrid neural-network architecture achieves the highest performance (83.2\% accuracy). The remaining gap to the human-performance ceiling provides enough room for future model improvements. View details
    Preview abstract We present a family of neural-network–inspired models for computing continuous word representation, specifically designed to exploit monolingual and multilingual text, without and with annotations (syntactic dependencies, word alignments, etc.). We find that this framework allows us to train embeddings with significantly higher accuracy on syntactic and semantic compositionality, as well as multilingual semantic similarity, compared to previous models. We also show that some of these embeddings can be used to improve the performance of a state-of-the-art machine translation system for words outside the vocabulary of the parallel training data. View details
    Preview abstract We present a language agnostic, unsupervised method for inducing morphological transformations between words. The method relies heavily on certain regularities that manifest in high-dimensional vector spaces. We show that this method is capable of discovering a wide-range of morphological rules, which can be successfully used towards improved natural language processing. We evaluate this method across six different languages and nine datasets, and show significant improvements across all languages. View details
    No Results Found