Publications
Our teams aspire to make discoveries that impact everyone, and core to our approach is sharing our research and tools to fuel progress in the field.
Our teams aspire to make discoveries that impact everyone, and core to our approach is sharing our research and tools to fuel progress in the field.
Sort By
1 - 15 of 10128 publications
Now You See Me, Now You Don't: 'Poverty of the Stimulus' Problems and Arbitrary Correspondences in End-to-End Speech Models
Proceedings of the Second Workshop on Computation and Written Language (CAWL) 2024
Preview abstract
End-to-end models for speech recognition and speech synthesis have many benefits, but we argue they also face a unique set of challenges not encountered in conventional multi-stage hybrid systems, which relied on the explicit injection of linguistic knowledge through resources such as phonemic dictionaries and verbalization grammars. These challenges include handling words with unusual grapheme-to-phoneme correspondences, converting between written forms like ‘12’ and spoken forms such as ‘twelve’, and contextual disambiguation of homophones or homographs. We describe the mitigation strategies that have been used for these problems in end-to-end systems, either implicitly or explicitly, and call out that the most commonly used mitigation techniques are likely incompatible with newly emerging approaches that use minimal amounts of supervised audio training data. We review best-of-both-world approaches that allow the use of end-to-end models combined with traditional linguistic resources, which we show are increasingly straightforward to create at scale, and close with an optimistic outlook for bringing speech technologies to many more languages by combining these strands of research.
View details
Preview abstract
To tackle the challenge of optimizing middle-mile logistics, the crucial link between warehouses and final deliveries, we introduce a novel instance generator that aims to create a rich and adaptable dataset of diverse instances to empower researchers and developers. The instance defines a logistics network with hubs, vehicles, routes, lines, and rotations. Additionally, it specifies a list of shipments that need to be transported through this network. To customize the instance, the user can adjust various parameters, such as the number of hubs, density of the space graphs, distribution of shipment weights, or the maximum number of vehicles.
The generator reflects real-world complexities through variations in network size and structure. We developed a random graph generator to mimic real-world middle mile networks, by generating space graphs for hubs. Subsequently, lines and routes are randomly constructed on the generated space graphs, while adhering to user-defined constraints.
The tool is in the form of an optimized C++ library that enables the generation of instances with a large number of hubs and shipments. It offers the immense potential for advancing middle-mile logistics optimization by providing a comprehensive and adaptable dataset for benchmarking optimization approaches, training machine learning models, and analyzing the impact of network configurations and shipments characteristics on overall efficiency.
View details
On the Benefits of Traffic “Reprofiling” The Single Hop Case
Henry Sariowan
Jiaming Qiu
Jiayi Song
Roch Guerin
IEEE/ACM Transactions on Networking (2024)
Preview abstract
Datacenters have become a significant source of traffic, much of which is carried over private networks. The operators of those networks commonly have access to detailed traffic profiles and performance goals, which they seek to meet as efficiently as possible. Of interest are solutions that guarantee latency while minimizing network bandwidth. The paper explores a basic building block towards realizing such solutions, namely, a single hop configuration. The main results are in the form of optimal solutions for meeting local deadlines under schedulers of varying complexity and therefore cost. The results demonstrate how judiciously modifying flows’ traffic profiles, i.e., reprofiling them, can help simple schedulers reduce the bandwidth they require, often performing nearly as well as more complex ones.
View details
Rapid initial state preparation for the quantum simulation of strongly correlated molecules and materials
Dominic Berry
Yu Tong
Alec White
Tae In Kim
Lin Lin
Seunghoon Lee
Garnet Chan
arXiv:2409.11748 (2024)
Preview abstract
Studies on quantum algorithms for ground state energy estimation often assume perfect ground state preparation; however, in reality the initial state will have imperfect overlap with the true ground state. Here we address that problem in two ways: by faster preparation of matrix product state (MPS) approximations, and more efficient filtering of the prepared state to find the ground state energy. We show how to achieve unitary synthesis with a Toffoli complexity about $7 \times$ lower than that in prior work, and use that to derive a more efficient MPS preparation method. For filtering we present two different approaches: sampling and binary search. For both we use the theory of window functions to avoid large phase errors and minimise the complexity. We find that the binary search approach provides better scaling with the overlap at the cost of a larger constant factor, such that it will be preferred for overlaps less than about 0.003. Finally, we estimate the total resources to perform ground state energy estimation of FeMoco and Iron cluster systems by estimating ground state overlap on an MPS initial state through extrapolation. With a modest bond dimension of 4000 we estimate a 0.96 overlap squared value producing total resources of $7.5 \times 10^{10}$ Toffoli gates; validating naive estimates where we assume perfect ground state overlap. These extrapolations allay practical concerns of exponential overlap decay in challenging-to-compute chemical systems.
View details
Limoncello: Prefetchers for Scale
Carlos Villavieja
Baris Kasikci
Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Association for Computing Machinery, New York, NY, United States (2024)
Preview abstract
This paper presents Limoncello, a novel software system that dynamically configures data prefetching for high utilization systems. We demonstrate that in resource-constrained environments, such as large data centers, traditional methods of hardware prefetching can increase memory latency and decrease available memory bandwidth. To address this, Limoncello dynamically configures data prefetching, disabling hardware prefetchers when memory bandwidth utilization is high and leveraging targeted software prefetching to reduce cache misses when hardware prefetchers are disabled. Limoncello is software-centric and does not require any modifications to hardware. Our evaluation of the deployment on a real-world hyperscale system reveals that Limoncello unlocks significant performance gains for high-utilization systems: it improves application throughput by 10%, due to a 15% reduction in memory latency, while maintaining minimal change in cache miss rate for targeted library functions.
View details
Towards a Complete Benchmark on Video Moment Localization
Jinyeong Chae
Donghwa Kim
Kwanseok Kim
Doyeon Lee
Sangho Lee
Seongsu Ha
Jonghwan Mun
Wooyoung Kang
Byungseok Roh
(2024)
Preview abstract
In this paper, we propose and conduct a comprehensive benchmark on moment localization task, which aims to retrieve a segment that corresponds to a text query from a single untrimmed video. Our study starts from an observation that most moment localization papers report experimental results only on a few datasets in spite of availability of far more benchmarks. Thus, we conduct an extensive benchmark study to measure the performance of representative methods on widely used 7 datasets. Looking further into the details, we pose additional research questions and empirically verify them, including if they rely on unintended biases introduced by specific training data, if advanced visual features trained on classification task transfer well to this task, and if computational cost of each model pays off. With a series of these experiments, we provide multifaceted evaluation of state-of-the-art moment localization models. Codes are available at https://github.com/snuviplab/MoLEF.
View details
Batch Calibration: Rethinking Calibration For In-Context Learning And Prompt Engineering
Lev Proleev
Diana Mincu
International Conference on Learning Representations (ICLR) (2024)
Preview abstract
Prompting and in-context learning (ICL) have become efficient learning paradigms for large language models (LLMs). However, LLMs suffer from prompt brittleness and various bias factors in the prompt, including but not limited to the formatting, the choice verbalizers, and the ICL examples. To address this problem that results in unexpected performance degradation, calibration methods have been developed to mitigate the effects of these biases while recovering LLM performance. In this work, we first conduct a systematic analysis of the existing calibration methods, where we both provide a unified view and reveal the failure cases. Inspired by these analyses, we propose Batch Calibration (BC), a simple yet intuitive method that controls the contextual bias from the batched input, unifies various prior approaches, and effectively addresses the aforementioned issues. BC is zero-shot, inference-only, and incurs negligible additional costs. In the few-shot setup, we further extend BC to allow it to learn the contextual bias from labeled data. We validate the effectiveness of BC with PaLM 2-(S, M, L) and CLIP models and demonstrate state-of-the-art performance over previous calibration baselines across more than 10 natural language understanding and image classification tasks.
View details
The Task-oriented Queries Benchmark (ToQB)
arXiv:2406.02943 (2024)
Preview abstract
Task-oriented queries (e.g., one-shot queries to play videos, order food, or call a taxi) are crucial for assessing the quality of virtual assistants, chatbots, and other large language model (LLM)-based services. However, a standard benchmark for task-oriented queries is not yet available, as existing benchmarks in the relevant NLP (Natural Language Processing) fields have primarily focused on task-oriented dialogues. Thus, we present a new methodology for efficiently generating the Task-oriented Queries Benchmark (ToQB) using existing task-oriented dialogue datasets and an LLM service. Our methodology involves formulating the underlying NLP task to summarize the original intent of a speaker in each dialogue, detailing the key steps to perform the devised NLP task using an LLM service, and outlining a framework for automating a major part of the benchmark generation process. Through a case study encompassing three domains (i.e., two single-task domains and one multi-task domain), we demonstrate how to customize the LLM prompts (e.g., omitting system utterances or speaker labels) for those three domains and characterize the generated task-oriented queries. The generated ToQB dataset is made available to the public.We further discuss new domains that can be added to ToQB by community contributors and its practical applications.
View details
Taming Self-Training for Open-Vocabulary Object Detection
Shiyu Zhao
Samuel Schulter
Zhixing Zhang
Vijay Kumar B G
Yumin Suh
Manmohan Chandraker
Dimitris N. Metaxas
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2024)
Preview abstract
Recent studies have shown promising performance in open-vocabulary object detection (OVD) by utilizing pseudo labels (PLs) from pretrained vision and language models (VLMs). However, teacher-student self-training, a powerful and widely used paradigm to leverage PLs, is rarely explored for OVD. This work identifies two challenges of using self-training in OVD: noisy PLs from VLMs and frequent distribution changes of PLs. To address these challenges, we propose SAS-Det that tames self-training for OVD from two key perspectives. First, we present a split-and-fusion (SAF) head that splits a standard detection into an open-branch and a closed-branch. This design can reduce noisy supervision from pseudo boxes. Moreover, the two branches learn complementary knowledge from different training data, significantly enhancing performance when fused together. Second, in our view, unlike in closed-set tasks, the PL distributions in OVD are solely determined by the teacher model. We introduce a periodic update strategy to decrease the number of updates to the teacher, thereby decreasing the frequency of changes in PL distributions, which stabilizes the training process. Extensive experiments demonstrate SAS-Det is both efficient and effective. SAS-Det outperforms recent models of the same scale by a clear margin and achieves 37.4 AP50 and 29.1 APr on novel categories of the COCO and LVIS benchmarks, respectively.
View details
Open Se Cura: First Silicon Results of an Auditable and Transparent Hardware Root of Trust System using Open EDA in 16-nm
Guanchen Tao
Ming-Hung Chen
Bangfei Pan
Kai Yick
Dennis Sylvester
Mehdi Saligane
IEEE Solid-State Circuits Magazine, 16(2024), pp. 58-66
Preview abstract
Hardware Root of Trust (HRoT) is essential for Internet-of-Things (IoT) devices as it provides critical user data protection. However, each novel use case significantly lengthens the development time for an HRoT system. Furthermore, most HRoT solutions are proprietary, and users lack permission to inspect and audit such systems [1-2]. This paper introduces Open Se Cura, which is an open-source framework designed to expedite the implementation of secure and transparent HRoT systems. It utilizes open-source Electronic Design Automation (EDA) tools like OpenROAD [3-4] and OpenFASOC [5-6], along with open-source Process Design Kits (PDKs), to present a transparent and auditable approach to hardware-software co-design platforms. This approach enables fast and trustworthy HRoT system implementation and is made openly available to reproduce its results and security efficacy [7]. Our reference design is showcased through FPGA emulation, and the first measurement results of a silicon implementation in 16nm of Open Se Cura security domain subsets integrated using open-source EDA are presented.
View details
Preview abstract
For Extended Reality (XR) headsets, a key aim is the natural interaction in 3D space beyond what traditional methods of keyboard, mouse, and touchscreen can offer. With the release of the Apple Vision Pro, a novel interaction paradigm is now widely available where users seamlessly navigate content through the combined use of their eyes and hands. However, blending these modalities poses unique design challenges due to their dynamic nature and the absence of established principles and standards.
In this article, we present five design principles and issues for the Gaze + Pinch interaction technique, informed by eye-hand research in the human-computer interaction field. The design principles encompass mechanisms like division of labor and minimalistic timing, which are crucial for usability, alongside enhancements for the manipulation of objects, indirect interactions, and drag & drop. Whether in design, technology, or research domains, this exploration offers valuable perspectives for navigating the evolving landscape of 3D interaction.
View details
Rich Human Feedback for Text to Image Generation
Katherine Collins
Nicholas Carolan
Youwei Liang
Peizhao Li
Dj Dvijotham
Gang Li
Sarah Young
Jiao Sun
Arseniy Klimovskiy
Preview abstract
Recent Text-to-Image (T2I) generation models such as Stable Diffusion and Imagen have made significant progress in generating high-resolution images based on text descriptions. However, many generated images still suffer from issues such as artifacts/implausibility, misalignment with text descriptions, and low aesthetic quality.
Inspired by the success of Reinforcement Learning with Human Feedback (RLHF) for large language models, prior work collected human-provided scores as feedback on generated images and trained a reward model to improve the T2I generation.
In this paper, we enrich the feedback signal by (i) marking image regions that are implausible or misaligned with the text, and (ii) annotating which keywords in the text prompt are not represented in the image.
We collect such rich human feedback on 18K generated images and train a multimodal transformer to predict these rich feedback automatically.
We show that the predicted rich human feedback can be leveraged to improve image generation, for example, by selecting high-quality training data to finetune and improve the generative models, or by creating masks with predicted heatmaps to inpaint the problematic regions.
Notably, the improvements generalize to models (Muse) beyond those used to generate the images on which human feedback data were collected (Stable Diffusion variants).
View details
Preview abstract
Recent significant advances in text-to-image models unlock the possibility of training vision systems using synthetic images, potentially overcoming the difficulty of collecting curated data at scale. It is unclear, however, how these models behave at scale, as more synthetic data is added to the training set. In this paper we study the scaling laws of synthetic images generated by state of the art text-to-image models, for the training of supervised models: image classifiers with label supervision, and CLIP with language supervision. We identify several factors, including text prompts, classifier-free guidance scale, and types of text-to-image models, that significantly affect scaling behavior. After tuning these factors, we observe that synthetic images demonstrate a scaling trend similar to, but slightly less effective than, real images in CLIP training, while they significantly underperform in scaling when training supervised image classifiers. Our analysis indicates that the main reason for this underperformance is the inability of off-the-shelf text-to-image models to generate certain concepts, a limitation that significantly impairs the training of image classifiers. Our findings also suggest that scaling synthetic data can be particularly effective in scenarios such as: (1) when there is a limited supply of real images for a supervised problem (e.g., fewer than 0.5 million images in ImageNet), (2) when the evaluation dataset diverges significantly from the training data, indicating the out-of-distribution scenario, or (3) when synthetic data is used in conjunction with real images, as demonstrated in the training of CLIP models.
View details
Learning to Rewrite Prompts for Personalized Text Generation
Qiaozhu Mei
Proceedings of the ACM Web Conference 2024
Preview abstract
Facilitated by large language models (LLMs), personalized text generation has become a rapidly growing research direction. Most existing studies focus on designing specialized models for a particular domain, or they require fine-tuning the LLMs to generate personalized text. We consider a typical scenario in which the large language model, which generates personalized output, is frozen and can only be accessed through APIs. Under this constraint, all one can do is to improve the input text (i.e., text prompts) sent to the LLM, a procedure that is usually done manually. In this paper, we propose a novel method to automatically revise prompts for personalized text generation. The proposed method takes the initial prompts generated by a state-of-the-art, multistage framework for personalized generation and rewrites a few critical components that summarize and synthesize the personal context. The prompt rewriter employs a training paradigm that chains together supervised learning (SL) and reinforcement learning (RL), where SL reduces the search space of RL and RL facilitates end-to-end training of the rewriter. Using datasets from three representative domains, we demonstrate that the rewritten prompts outperform both the original prompts and the prompts optimized via supervised learning or reinforcement learning alone. In-depth analysis of the rewritten prompts shows that they are not only human readable, but also able to guide manual revision of prompts when there is limited resource to employ reinforcement learning to train the prompt rewriter, or when it is costly to deploy an automatic prompt rewriter for inference.
View details
Preview abstract
It’s no secret that generative artificial intelligence (GenAI) is rapidly changing the landscape of software development, with discussions about best practices for applying this transformative technology dominating the popular press [cite cite cite]. Perhaps nowhere on Earth have these discussions been more frequent and passionate than inside the organizations dedicated to making GenAI accessible and useful to developers, including at Google. During one such discussion between researchers on our DevOps Research and Assessment (DORA) and Engineering Productivity Research (EPR) teams, we were struck by a recurring finding common to development professionals both inside and outside of Google:
Using GenAI makes developers feel more productive, and developers who trust GenAI use it more.
On the surface, this finding may seem somewhat… obvious. But, for us, it highlighted the deep need to better understand the factors that impact developers’ trust in GenAI systems and ways to foster that trust, so that developers and development firms can yield the most benefit from their investment in GenAI development tools.
Here, we share findings from seven studies conducted at Google, regarding the productivity gains of GenAI use in development, the impacts of developers’ trust on GenAI use, and the factors we’ve observed which positively impact developers’ trust in GenAI. We conclude with five suggested strategies that organizations engaged in software development might employ to foster their developers’ trust in GenAI, thereby increasing their GenAI use and maximizing GenAI-related productivity gains.
View details
