Radu Soricut

Anthony Brohan, Noah Brown, Justice Carbajal et al. · stanford

We study how vision-language models trained on Internet-scale data can be incorporated directly into end-to-end robotic control to boost generalization and enable emergent semantic reasoning. Our goal is to enable a single end-to-end trained model to both learn to map robot observations to actions and enjoy the benefits of large-scale pretraining on language and vision-language data from the web. To this end, we propose to co-fine-tune state-of-the-art vision-language models on both robotic trajectory data and Internet-scale vision-language tasks, such as visual question answering. In contrast to other approaches, we propose a simple, general recipe to achieve this goal: in order to fit both natural language responses and robotic actions into the same format, we express the actions as text tokens and incorporate them directly into the training set of the model in the same way as natural language tokens. We refer to such category of models as vision-language-action models (VLA) and instantiate an example of such a model, which we call RT-2. Our extensive evaluation (6k evaluation trials) shows that our approach leads to performant robotic policies and enables RT-2 to obtain a range of emergent capabilities from Internet-scale training. This includes significantly improved generalization to novel objects, the ability to interpret commands not present in the robot training data (such as placing an object onto a particular number or icon), and the ability to perform rudimentary reasoning in response to user commands (such as picking up the smallest or largest object, or the one closest to another object). We further show that incorporating chain of thought reasoning allows RT-2 to perform multi-stage semantic reasoning, for example figuring out which object to pick up for use as an improvised hammer (a rock), or which type of drink is best suited for someone who is tired (an energy drink).

Xi Chen, Xiao Wang, Soravit Changpinyo et al. · deepmind

Effective scaling and a flexible task interface enable large language models to excel at many tasks. We present PaLI (Pathways Language and Image model), a model that extends this approach to the joint modeling of language and vision. PaLI generates text based on visual and textual inputs, and with this interface performs many vision, language, and multimodal tasks, in many languages. To train PaLI, we make use of large pre-trained encoder-decoder language models and Vision Transformers (ViTs). This allows us to capitalize on their existing capabilities and leverage the substantial cost of training them. We find that joint scaling of the vision and language components is important. Since existing Transformers for language are much larger than their vision counterparts, we train a large, 4-billion parameter ViT (ViT-e) to quantify the benefits from even larger-capacity vision models. To train PaLI, we create a large multilingual mix of pretraining tasks, based on a new image-text training set containing 10B images and texts in over 100 languages. PaLI achieves state-of-the-art in multiple vision and language tasks (such as captioning, visual question-answering, scene-text understanding), while retaining a simple, modular, and scalable design.

Lucas Beyer, Andreas Steiner, André Susano Pinto et al. · deepmind, oxford

PaliGemma is an open Vision-Language Model (VLM) that is based on the SigLIP-So400m vision encoder and the Gemma-2B language model. It is trained to be a versatile and broadly knowledgeable base model that is effective to transfer. It achieves strong performance on a wide variety of open-world tasks. We evaluate PaliGemma on almost 40 diverse tasks including standard VLM benchmarks, but also more specialized tasks such as remote-sensing and segmentation.

Xi Chen, Xiao Wang, Lucas Beyer et al. · deepmind

This paper presents PaLI-3, a smaller, faster, and stronger vision language model (VLM) that compares favorably to similar models that are 10x larger. As part of arriving at this strong performance, we compare Vision Transformer (ViT) models pretrained using classification objectives to contrastively (SigLIP) pretrained ones. We find that, while slightly underperforming on standard image classification benchmarks, SigLIP-based PaLI shows superior performance across various multimodal benchmarks, especially on localization and visually-situated text understanding. We scale the SigLIP image encoder up to 2 billion parameters, and achieves a new state-of-the-art on multilingual cross-modal retrieval. We hope that PaLI-3, at only 5B parameters, rekindles research on fundamental pieces of complex VLMs, and could fuel a new generation of scaled-up models.

Paul Voigtlaender, Soravit Changpinyo, Jordi Pont-Tuset et al. · deepmind

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/.

Ashish V. Thapliyal, Jordi Pont-Tuset, Xi Chen et al. · deepmind

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 superior correlation results with human evaluations when using XM3600 as golden references for automatic metrics.

Soravit Changpinyo, Doron Kukliansky, Idan Szpektor et al. · deepmind

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.

Jihyung Kil, Soravit Changpinyo, Xi Chen et al. · deepmind

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.

Nan Ding, Xi Chen, Tomer Levinboim et al. · deepmind

With the increasing abundance of pretrained models in recent years, the problem of selecting the best pretrained checkpoint for a particular downstream classification task has been gaining increased attention. Although several methods have recently been proposed to tackle the selection problem (e.g. LEEP, H-score), these methods resort to applying heuristics that are not well motivated by learning theory. In this paper we present PACTran, a theoretically grounded family of metrics for pretrained model selection and transferability measurement. We first show how to derive PACTran metrics from the optimal PAC-Bayesian bound under the transfer learning setting. We then empirically evaluate three metric instantiations of PACTran on a number of vision tasks (VTAB) as well as a language-and-vision (OKVQA) task. An analysis of the results shows PACTran is a more consistent and effective transferability measure compared to existing selection methods.

Su Wang, Chitwan Saharia, Ceslee Montgomery et al.

Text-guided image editing can have a transformative impact in supporting creative applications. A key challenge is to generate edits that are faithful to input text prompts, while consistent with input images. 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 using object detectors to propose inpainting masks during training. In addition, Imagen Editor 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.

Soravit Changpinyo, Linting Xue, Michal Yarom et al. · deepmind

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.

Zifan Wang, Nan Ding, Tomer Levinboim et al. · deepmind

Recent research in robust optimization has shown an overfitting-like phenomenon in which models trained against adversarial attacks exhibit higher robustness on the training set compared to the test set. Although previous work provided theoretical explanations for this phenomenon using a robust PAC-Bayesian bound over the adversarial test error, related algorithmic derivations are at best only loosely connected to this bound, which implies that there is still a gap between their empirical success and our understanding of adversarial robustness theory. To close this gap, in this paper we consider a different form of the robust PAC-Bayesian bound and directly minimize it with respect to the model posterior. The derivation of the optimal solution connects PAC-Bayesian learning to the geometry of the robust loss surface through a Trace of Hessian (TrH) regularizer that measures the surface flatness. In practice, we restrict the TrH regularizer to the top layer only, which results in an analytical solution to the bound whose computational cost does not depend on the network depth. Finally, we evaluate our TrH regularization approach over CIFAR-10/100 and ImageNet using Vision Transformers (ViT) and compare against baseline adversarial robustness algorithms. Experimental results show that TrH regularization leads to improved ViT robustness that either matches or surpasses previous state-of-the-art approaches while at the same time requires less memory and computational cost.

Wanrong Zhu, Bo Pang, Ashish V. Thapliyal et al.

Dense video captioning aims to identify the events of interest in an input video, and generate descriptive captions for each event. Previous approaches usually follow a two-stage generative process, which first proposes a segment for each event, then renders a caption for each identified segment. Recent advances in large-scale sequence generation pretraining have seen great success in unifying task formulation for a great variety of tasks, but so far, more complex tasks such as dense video captioning are not able to fully utilize this powerful paradigm. In this work, we show how to model the two subtasks of dense video captioning jointly as one sequence generation task, and simultaneously predict the events and the corresponding descriptions. Experiments on YouCook2 and ViTT show encouraging results and indicate the feasibility of training complex tasks such as end-to-end dense video captioning integrated into large-scale pretrained models.

Valentin Gabeur, Shangbang Long, Songyou Peng et al.

Recent works show that image and video generators exhibit zero-shot visual understanding behaviors, in a way reminiscent of how LLMs develop emergent capabilities of language understanding and reasoning from generative pretraining. While it has long been conjectured that the ability to create visual content implies an ability to understand it, there has been limited evidence that generative vision models have developed strong understanding capabilities. In this work, we demonstrate that image generation training serves a role similar to LLM pretraining, and lets models learn powerful and general visual representations that enable SOTA performance on various vision tasks. We introduce Vision Banana, a generalist model built by instruction-tuning Nano Banana Pro (NBP) on a mixture of its original training data alongside a small amount of vision task data. By parameterizing the output space of vision tasks as RGB images, we seamlessly reframe perception as image generation. Our generalist model, Vision Banana, achieves SOTA results on a variety of vision tasks involving both 2D and 3D understanding, beating or rivaling zero-shot domain-specialists, including Segment Anything Model 3 on segmentation tasks, and the Depth Anything series on metric depth estimation. We show that these results can be achieved with lightweight instruction-tuning without sacrificing the base model's image generation capabilities. The superior results suggest that image generation pretraining is a generalist vision learner. It also shows that image generation serves as a unified and universal interface for vision tasks, similar to text generation's role in language understanding and reasoning. We could be witnessing a major paradigm shift for computer vision, where generative vision pretraining takes a central role in building Foundational Vision Models for both generation and understanding.

Nan Ding, Tomer Levinboim, Jialin Wu et al.

Recent empirical evidence indicates that transformer based in-context learning performs better when using a prefix language model (prefixLM), in which in-context samples can all attend to each other, compared to causal language models (causalLM), which use auto-regressive attention that prohibits in-context samples to attend to future samples. While this result is intuitive, it is not understood from a theoretical perspective. In this paper we take a theoretical approach and analyze the convergence behavior of prefixLM and causalLM under a certain parameter construction. Our analysis shows that both LM types converge to their stationary points at a linear rate, but that while prefixLM converges to the optimal solution of linear regression, causalLM convergence dynamics follows that of an online gradient descent algorithm, which is not guaranteed to be optimal even as the number of samples grows infinitely. We supplement our theoretical claims with empirical experiments over synthetic and real tasks and using various types of transformers. Our experiments verify that causalLM consistently underperforms prefixLM in all settings.

Yaqing Wang, Jialin Wu, Tanmaya Dabral et al.

Large language models (LLMs) and vision language models (VLMs) demonstrate excellent performance on a wide range of tasks by scaling up parameter counts from O(10^9) to O(10^{12}) levels and further beyond. These large scales make it impossible to adapt and deploy fully specialized models given a task of interest. Parameter-efficient fine-tuning (PEFT) emerges as a promising direction to tackle the adaptation and serving challenges for such large models. We categorize PEFT techniques into two types: intrusive and non-intrusive. Intrusive PEFT techniques directly change a model's internal architecture. Though more flexible, they introduce significant complexities for training and serving. Non-intrusive PEFT techniques leave the internal architecture unchanged and only adapt model-external parameters, such as embeddings for input. In this work, we describe AdaLink as a non-intrusive PEFT technique that achieves competitive performance compared to SoTA intrusive PEFT (LoRA) and full model fine-tuning (FT) on various tasks. We evaluate using both text-only and multimodal tasks, with experiments that account for both parameter-count scaling and training regime (with and without instruction tuning).

Gemini Team, Petko Georgiev, Ving Ian Lei et al. · deepmind, mila

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.

Gheorghe Comanici, Eric Bieber, Mike Schaekermann et al. · amazon-science, baidu

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Yu-Chuan Su, Soravit Changpinyo, Xiangning Chen et al.

Visual 2.5D perception involves understanding the semantics and geometry of a scene through reasoning about object relationships with respect to the viewer in an environment. However, existing works in visual recognition primarily focus on the semantics. To bridge this gap, we study 2.5D visual relationship detection (2.5VRD), in which the goal is to jointly detect objects and predict their relative depth and occlusion relationships. Unlike general VRD, 2.5VRD is egocentric, using the camera's viewpoint as a common reference for all 2.5D relationships. Unlike depth estimation, 2.5VRD is object-centric and not only focuses on depth. To enable progress on this task, we create a new dataset consisting of 220k human-annotated 2.5D relationships among 512K objects from 11K images. We analyze this dataset and conduct extensive experiments including benchmarking multiple state-of-the-art VRD models on this task. Our results show that existing models largely rely on semantic cues and simple heuristics to solve 2.5VRD, motivating further research on models for 2.5D perception. The new dataset is available at https://github.com/google-research-datasets/2.5vrd.

Zhenzhong Lan, Mingda Chen, Sebastian Goodman et al.

Increasing model size when pretraining natural language representations often results in improved performance on downstream tasks. However, at some point further model increases become harder due to GPU/TPU memory limitations and longer training times. To address these problems, we present two parameter-reduction techniques to lower memory consumption and increase the training speed of BERT. Comprehensive empirical evidence shows that our proposed methods lead to models that scale much better compared to the original BERT. We also use a self-supervised loss that focuses on modeling inter-sentence coherence, and show it consistently helps downstream tasks with multi-sentence inputs. As a result, our best model establishes new state-of-the-art results on the GLUE, RACE, and \squad benchmarks while having fewer parameters compared to BERT-large. The code and the pretrained models are available at https://github.com/google-research/ALBERT.

Mehran Kazemi, Hamidreza Alvari, Ankit Anand et al.

Large language models have shown impressive results for multi-hop mathematical reasoning when the input question is only textual. Many mathematical reasoning problems, however, contain both text and image. With the ever-increasing adoption of vision language models (VLMs), understanding their reasoning abilities for such problems is crucial. In this paper, we evaluate the reasoning capabilities of VLMs along various axes through the lens of geometry problems. We procedurally create a synthetic dataset of geometry questions with controllable difficulty levels along multiple axes, thus enabling a systematic evaluation. The empirical results obtained using our benchmark for state-of-the-art VLMs indicate that these models are not as capable in subjects like geometry (and, by generalization, other topics requiring similar reasoning) as suggested by previous benchmarks. This is made especially clear by the construction of our benchmark at various depth levels, since solving higher-depth problems requires long chains of reasoning rather than additional memorized knowledge. We release the dataset for further research in this area.

Roopal Garg, Andrea Burns, Burcu Karagol Ayan et al.

Despite the longstanding adage "an image is worth a thousand words," generating accurate hyper-detailed image descriptions remains unsolved. Trained on short web-scraped image text, vision-language models often generate incomplete descriptions with visual inconsistencies. We address this via a novel data-centric approach with ImageInWords (IIW), a carefully designed human-in-the-loop framework for curating hyper-detailed image descriptions. Human evaluations on IIW data show major gains compared to recent datasets (+66%) and GPT4V (+48%) across comprehensiveness, specificity, hallucinations, and more. We also show that fine-tuning with IIW data improves these metrics by +31% against models trained with prior work, even with only 9k samples. Lastly, we evaluate IIW models with text-to-image generation and vision-language reasoning tasks. Our generated descriptions result in the highest fidelity images, and boost compositional reasoning by up to 6% on ARO, SVO-Probes, and Winoground datasets. We release the IIW Eval benchmark with human judgement labels, object and image-level annotations from our framework, and existing image caption datasets enriched via IIW-model.

Lijie Fan, Luming Tang, Siyang Qin et al. · deepmind

We present UniFluid, a unified autoregressive framework for joint visual generation and understanding leveraging continuous visual tokens. Our unified autoregressive architecture processes multimodal image and text inputs, generating discrete tokens for text and continuous tokens for image. We find though there is an inherent trade-off between the image generation and understanding task, a carefully tuned training recipe enables them to improve each other. By selecting an appropriate loss balance weight, the unified model achieves results comparable to or exceeding those of single-task baselines on both tasks. Furthermore, we demonstrate that employing stronger pre-trained LLMs and random-order generation during training is important to achieve high-fidelity image generation within this unified framework. Built upon the Gemma model series, UniFluid exhibits competitive performance across both image generation and understanding, demonstrating strong transferability to various downstream tasks, including image editing for generation, as well as visual captioning and question answering for understanding.

Gemini Team, Rohan Anil, Sebastian Borgeaud et al.

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of the Gemini family in cross-modal reasoning and language understanding will enable a wide variety of use cases. We discuss our approach toward post-training and deploying Gemini models responsibly to users through services including Gemini, Gemini Advanced, Google AI Studio, and Cloud Vertex AI.

Xi Chen, Josip Djolonga, Piotr Padlewski et al.

We present the training recipe and results of scaling up PaLI-X, a multilingual vision and language model, both in terms of size of the components and the breadth of its training task mixture. Our model achieves new levels of performance on a wide-range of varied and complex tasks, including multiple image-based captioning and question-answering tasks, image-based document understanding and few-shot (in-context) learning, as well as object detection, video question answering, and video captioning. PaLI-X advances the state-of-the-art on most vision-and-language benchmarks considered (25+ of them). Finally, we observe emerging capabilities, such as complex counting and multilingual object detection, tasks that are not explicitly in the training mix.

Mert İnan, Piyush Sharma, Baber Khalid et al.

Developers of text generation models rely on automated evaluation metrics as a stand-in for slow and expensive manual evaluations. However, image captioning metrics have struggled to give accurate learned estimates of the semantic and pragmatic success of output text. We address this weakness by introducing the first discourse-aware learned generation metric for evaluating image descriptions. Our approach is inspired by computational theories of discourse for capturing information goals using coherence. We present a dataset of image$\unicode{x2013}$description pairs annotated with coherence relations. We then train a coherence-aware metric on a subset of the Conceptual Captions dataset and measure its effectiveness$\unicode{x2014}$its ability to predict human ratings of output captions$\unicode{x2014}$on a test set composed of out-of-domain images. We demonstrate a higher Kendall Correlation Coefficient for our proposed metric with the human judgments for the results of a number of state-of-the-art coherence-aware caption generation models when compared to several other metrics including recently proposed learned metrics such as BLEURT and BERTScore.

Zhenhai Zhu, Radu Soricut

We describe an efficient hierarchical method to compute attention in the Transformer architecture. The proposed attention mechanism exploits a matrix structure similar to the Hierarchical Matrix (H-Matrix) developed by the numerical analysis community, and has linear run time and memory complexity. We perform extensive experiments to show that the inductive bias embodied by our hierarchical attention is effective in capturing the hierarchical structure in the sequences typical for natural language and vision tasks. Our method is superior to alternative sub-quadratic proposals by over +6 points on average on the Long Range Arena benchmark. It also sets a new SOTA test perplexity on One-Billion Word dataset with 5x fewer model parameters than that of the previous-best Transformer-based models.

Nan Ding, Xi Chen, Tomer Levinboim et al.

Despite recent advances in its theoretical understanding, there still remains a significant gap in the ability of existing PAC-Bayesian theories on meta-learning to explain performance improvements in the few-shot learning setting, where the number of training examples in the target tasks is severely limited. This gap originates from an assumption in the existing theories which supposes that the number of training examples in the observed tasks and the number of training examples 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 theories. Furthermore, we derive a new computationally-efficient PACMAML algorithm, and show it outperforms existing meta-learning algorithms on several few-shot benchmark datasets.

Soravit Changpinyo, Piyush Sharma, Nan Ding et al.

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 and benchmark its effectiveness against CC3M on multiple downstream tasks with an emphasis on long-tail visual recognition. Our 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.

Soravit Changpinyo, Jordi Pont-Tuset, Vittorio Ferrari et al.

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 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.

Edwin G. Ng, Bo Pang, Piyush Sharma et al.

Image captioning models generally lack the capability to take into account user interest, and usually default to global descriptions that try to balance readability, informativeness, and information overload. On the other hand, VQA models generally lack the ability to provide long descriptive answers, while expecting the textual question to be quite precise. We present a method to control the concepts that an image caption should focus on, using an additional input called the guiding text that refers to either groundable or ungroundable concepts in the image. Our model consists of a Transformer-based multimodal encoder that uses the guiding text together with global and object-level image features to derive early-fusion representations used to generate the guided caption. While models trained on Visual Genome data have an in-domain advantage of fitting well when guided with automatic object labels, we find that guided captioning models trained on Conceptual Captions generalize better on out-of-domain images and guiding texts. Our human-evaluation results indicate that attempting in-the-wild guided image captioning requires access to large, unrestricted-domain training datasets, and that increased style diversity (even without increasing the number of unique tokens) is a key factor for improved performance.

Gabriel Huang, Bo Pang, Zhenhai Zhu et al.

Learning specific hands-on skills such as cooking, car maintenance, and home repairs increasingly happens via instructional videos. The user experience with such videos is known to be improved by meta-information such as time-stamped annotations for the main steps involved. Generating such annotations automatically is challenging, and we describe here two relevant contributions. First, we construct and release a new dense video captioning dataset, Video Timeline Tags (ViTT), featuring a variety of instructional videos together with time-stamped annotations. Second, we explore several multimodal sequence-to-sequence pretraining strategies that leverage large unsupervised datasets of videos and caption-like texts. We pretrain and subsequently finetune dense video captioning models using both YouCook2 and ViTT. We show that such models generalize well and are robust over a wide variety of instructional videos.

Xi Chen, Nan Ding, Tomer Levinboim et al.

Recent advances in automatic evaluation metrics for text have shown that deep contextualized word representations, such as those generated by BERT encoders, are helpful for designing metrics that correlate well with human judgements. At the same time, it has been argued that contextualized word representations exhibit sub-optimal statistical properties for encoding the true similarity between words or sentences. In this paper, we present two techniques for improving encoding representations for similarity metrics: a batch-mean centering strategy that improves statistical properties; and a computationally efficient tempered Word Mover Distance, for better fusion of the information in the contextualized word representations. We conduct numerical experiments that demonstrate the robustness of our techniques, reporting results over various BERT-backbone learned metrics and achieving state of the art correlation with human ratings on several benchmarks.

Sebastian Goodman, Nan Ding, Radu Soricut

Sequence generation models trained with teacher-forcing suffer from issues related to exposure bias and lack of differentiability across timesteps. Our proposed method, Teacher-Forcing with N-grams (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. 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 generation quality on one Machine Translation benchmark, WMT 2014 English-French, and two News Summarization benchmarks, CNN/Dailymail and Gigaword.

Nan Ding, Xinjie Fan, Zhenzhong Lan et al.

Models based on the Transformer architecture have achieved better accuracy than the ones based on competing architectures for a large set of tasks. A unique feature of the Transformer is its universal application of a self-attention mechanism, which allows for free information flow at arbitrary distances. Following a probabilistic view of the attention via the Gaussian mixture model, we find empirical evidence that the Transformer attention tends to "explain away" certain input neurons. To compensate for this, we propose a doubly-normalized attention scheme that is simple to implement and provides theoretical guarantees for avoiding the "explaining away" effect without introducing significant computational or memory cost. Empirically, we show that the new attention schemes result in improved performance on several well-known benchmarks.

Khyathi Raghavi Chandu, Piyush Sharma, Soravit Changpinyo et al.

Training large-scale image captioning (IC) models demands access to a rich and diverse set of training examples, gathered from the wild, often from noisy alt-text data. However, recent modeling approaches to IC often fall short in terms of performance in this case, because they assume a clean annotated dataset (as opposed to the noisier alt-text--based annotations), and employ an end-to-end generation approach, which often lacks both controllability and interpretability. We address these problems by breaking down the task into two simpler, more controllable tasks -- skeleton prediction and skeleton-based caption generation. Specifically, we show that selecting content words as skeletons} helps in generating improved and denoised captions when leveraging rich yet noisy alt-text--based uncurated datasets. We also show that the predicted English skeletons can be further cross-lingually leveraged to generate non-English captions, and present experimental results covering caption generation in French, Italian, German, Spanish and Hindi. We also show that skeleton-based prediction allows for better control of certain caption properties, such as length, content, and gender expression, providing a handle to perform human-in-the-loop semi-automatic corrections.

Nicholas Trieu, Sebastian Goodman, Pradyumna Narayana et al.

Multi-sentence summarization is a well studied problem in NLP, while generating image descriptions for a single image is a well studied problem in Computer Vision. However, for applications such as image cluster labeling or web page summarization, summarizing a set of images is also a useful and challenging task. This paper proposes the new task of multi-image summarization, which aims to generate a concise and descriptive textual summary given a coherent set of input images. We propose a model that extends the image-captioning Transformer-based architecture for single image to multi-image. A dense average image feature aggregation network allows the model to focus on a coherent subset of attributes across the input images. We explore various input representations to the Transformer network and empirically show that aggregated image features are superior to individual image embeddings. We additionally show that the performance of the model is further improved by pretraining the model parameters on a single-image captioning task, which appears to be particularly effective in eliminating hallucinations in the output.

Malihe Alikhani, Piyush Sharma, Shengjie Li et al.

We use coherence relations inspired by computational models of discourse to study the information needs and goals of image captioning. Using an annotation protocol specifically devised for capturing image--caption coherence relations, we annotate 10,000 instances from publicly-available image--caption pairs. We introduce a new task for learning inferences in imagery and text, coherence relation prediction, and show that these coherence annotations can be exploited to learn relation classifiers as an intermediary step, and also train coherence-aware, controllable image captioning models. The results show a dramatic improvement in the consistency and quality of the generated captions with respect to information needs specified via coherence relations.

Ashish V. Thapliyal, Radu Soricut

Cross-modal language generation tasks such as image captioning are directly hurt in their ability to support non-English languages by the trend of data-hungry models combined with the lack of non-English annotations. We investigate potential solutions for combining existing language-generation annotations in English with translation capabilities in order to create solutions at web-scale in both domain and language coverage. We describe an approach called Pivot-Language Generation Stabilization (PLuGS), which leverages directly at training time both existing English annotations (gold data) as well as their machine-translated versions (silver data); at run-time, it generates first an English caption and then a corresponding target-language caption. We show that PLuGS models outperform other candidate solutions in evaluations performed over 5 different target languages, under a large-domain testset using images from the Open Images dataset. Furthermore, we find an interesting effect where the English captions generated by the PLuGS models are better than the captions generated by the original, monolingual English model.

Jack Hessel, Zhenhai Zhu, Bo Pang et al.

Pretraining from unlabelled web videos has quickly become the de-facto means of achieving high performance on many video understanding tasks. Features are learned via prediction of grounded relationships between visual content and automatic speech recognition (ASR) tokens. However, prior pretraining work has been limited to only instructional videos; a priori, we expect this domain to be relatively "easy:" speakers in instructional videos will often reference the literal objects/actions being depicted. We ask: can similar models be trained on more diverse video corpora? And, if so, what types of videos are "grounded" and what types are not? We fit a representative pretraining model to the diverse YouTube8M dataset, and study its success and failure cases. We find that visual-textual grounding is indeed possible across previously unexplored video categories, and that pretraining on a more diverse set results in representations that generalize to both non-instructional and instructional domains.

Jordi Pont-Tuset, Jasper Uijlings, Soravit Changpinyo et al.

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.

Paul Hongsuck Seo, Piyush Sharma, Tomer Levinboim et al.

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 improve 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, where policy gradients are estimated by samples from a distribution that focuses on the captions in a caption ratings dataset. Our empirical evidence indicates that the proposed method learns to generalize the human raters' judgments 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.

Jack Hessel, Bo Pang, Zhenhai Zhu et al.

Instructional videos get high-traffic on video sharing platforms, and prior work suggests that providing time-stamped, subtask annotations (e.g., "heat the oil in the pan") improves user experiences. However, current automatic annotation methods based on visual features alone perform only slightly better than constant prediction. Taking cues from prior work, we show that we can improve performance significantly by considering automatic speech recognition (ASR) tokens as input. Furthermore, jointly modeling ASR tokens and visual features results in higher performance compared to training individually on either modality. We find that unstated background information is better explained by visual features, whereas fine-grained distinctions (e.g., "add oil" vs. "add olive oil") are disambiguated more easily via ASR tokens.

Sebastian Goodman, Zhenzhong Lan, Radu Soricut

Neural models for abstractive summarization tend to achieve the best performance in the presence of highly specialized, summarization specific modeling add-ons such as pointer-generator, coverage-modeling, and inferencetime heuristics. We show here that pretraining can complement such modeling advancements to yield improved results in both short-form and long-form abstractive summarization using two key concepts: full-network initialization and multi-stage pretraining. Our method allows the model to transitively benefit from multiple pretraining tasks, from generic language tasks to a specialized summarization task to an even more specialized one such as bullet-based summarization. Using this approach, we demonstrate improvements of 1.05 ROUGE-L points on the Gigaword benchmark and 1.78 ROUGE-L points on the CNN/DailyMail benchmark, compared to a randomly-initialized baseline.

Tomer Levinboim, Ashish V. Thapliyal, Piyush Sharma et al.

Automatic image captioning has improved significantly over the last few years, but the problem is far from being solved, with state of the art models still often producing low quality captions when used in the wild. In this paper, we focus on the task of Quality Estimation (QE) for image captions, which attempts to model the caption quality from a human perspective and without access to ground-truth references, so that it can be applied at prediction time to detect low-quality captions produced on previously unseen images. For this task, we develop a human evaluation process that collects coarse-grained caption annotations from crowdsourced users, which is then used to collect a large scale dataset spanning more than 600k caption quality ratings. We then carefully validate the quality of the collected ratings and establish baseline models for this new QE task. Finally, we further collect fine-grained caption quality annotations from trained raters, and use them to demonstrate that QE models trained over the coarse ratings can effectively detect and filter out low-quality image captions, thereby improving the user experience from captioning systems.

Soravit Changpinyo, Bo Pang, Piyush Sharma et al.

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.

Sanqiang Zhao, Piyush Sharma, Tomer Levinboim et al.

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.

Ye Zhang, Nan Ding, Radu Soricut

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.

Nan Ding, Radu Soricut

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 value function 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.

Nan Ding, Sebastian Goodman, Fei Sha et al.

We introduce a new multi-modal task for computer systems, posed as a combined vision-language comprehension challenge: identifying the most suitable 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. Instead, it 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-studied metric: the accuracy in detecting the true target among the decoys. The paper makes several contributions: an effective and extensible 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; human evaluation results on this dataset, informing a performance upper-bound; and several baseline and competitive learning approaches that illustrate the utility of the proposed task and dataset in advancing both image and language comprehension. We also show that, in a multi-task learning setting, the performance on the proposed task is positively correlated with the end-to-end task of image captioning.