Can Liu

Amazon AGI, Aaron Langford, Aayush Shah et al. · amazon-science

We present Amazon Nova, a new generation of state-of-the-art foundation models that deliver frontier intelligence and industry-leading price performance. Amazon Nova Pro is a highly-capable multimodal model with the best combination of accuracy, speed, and cost for a wide range of tasks. Amazon Nova Lite is a low-cost multimodal model that is lightning fast for processing images, video, documents and text. Amazon Nova Micro is a text-only model that delivers our lowest-latency responses at very low cost. Amazon Nova Canvas is an image generation model that creates professional grade images with rich customization controls. Amazon Nova Reel is a video generation model offering high-quality outputs, customization, and motion control. Our models were built responsibly and with a commitment to customer trust, security, and reliability. We report benchmarking results for core capabilities, agentic performance, long context, functional adaptation, runtime performance, and human evaluation.

Jonathan Pilault, Can Liu, Mohit Bansal et al. · amazon-science, mila

Prompts have been shown to be an effective method to adapt a frozen Pretrained Language Model (PLM) to perform well on downstream tasks. Prompts can be represented by a human-engineered word sequence or by a learned continuous embedding. In this work, we investigate conditional and compositional differentiable prompting. We propose a new model, Prompt Production System (PRopS), which learns to transform task instructions or input metadata, into continuous prompts that elicit task-specific outputs from the PLM. Our model uses a modular network structure based on our neural formulation of Production Systems, which allows the model to learn discrete rules -- neural functions that learn to specialize in transforming particular prompt input patterns, making it suitable for compositional transfer learning and few-shot learning. We present extensive empirical and theoretical analysis and show that PRopS consistently surpasses other PLM adaptation techniques, and often improves upon fully fine-tuned models, on compositional generalization tasks, controllable summarization and multilingual translation, while needing fewer trainable parameters.

Yong Zhao, Runxin He, Nicholas Kersting et al.

Understanding why a neural network model makes certain decisions can be as important as the inference performance. Various methods have been proposed to help practitioners explain the prediction of a neural network model, of which Shapley values are most popular. SHAP package is a leading implementation of Shapley values to explain neural networks implemented in TensorFlow or PyTorch but lacks cross-platform support, one-shot deployment and is highly inefficient. To address these problems, we present the ONNXExplainer, which is a generic framework to explain neural networks using Shapley values in the ONNX ecosystem. In ONNXExplainer, we develop its own automatic differentiation and optimization approach, which not only enables One-Shot Deployment of neural networks inference and explanations, but also significantly improves the efficiency to compute explanation with less memory consumption. For fair comparison purposes, we also implement the same optimization in TensorFlow and PyTorch and measure its performance against the current state of the art open-source counterpart, SHAP. Extensive benchmarks demonstrate that the proposed optimization approach improves the explanation latency of VGG19, ResNet50, DenseNet201, and EfficientNetB0 by as much as 500%.

Can Liu, Wenjie Jiang, Shaolun Ruan et al.

Pitch-based sonification of quantitative data increases the accessibility of data visualizations that are otherwise inaccessible for blind and low-vision (BLV) individuals. We argue that, although pitch representations can reveal the coarse-grained information of data, such as data trend and value comparison, they cannot effectively convey the fine-grained details like the sign and exact value of individual data points. Informed by existing sound perception research, we propose a spatial audio-based approach by representing data values as the sound direction in the azimuth plane to achieve accessible fine-grained data representation. We conducted a user study with 26 participants (including 10 BLV participants) on four data perception tasks. The results show our approach significantly outperforms pitch representation on fine-grained data perception tasks like recognizing data signs and exact values, and performs similarly on data trend identification, despite its inferior accuracy on data value comparison.

Jiyao Wang, Xiao Yang, Qingyong Hu et al. · tsinghua

Robust and unobtrusive in-vehicle physiological monitoring is crucial for ensuring driving safety and user experience. While remote physiological measurement (RPM) offers a promising non-invasive solution, its translation to real-world driving scenarios is critically constrained by the scarcity of comprehensive datasets. Existing resources are often limited in scale, modality diversity, the breadth of biometric annotations, and the range of captured conditions, thereby omitting inherent real-world challenges in driving. Here, we present PhysDrive, the first large-scale multimodal dataset for contactless in-vehicle physiological sensing with dedicated consideration on various modality settings and driving factors. PhysDrive collects data from 48 drivers, including synchronized RGB, near-infrared camera, and raw mmWave radar data, accompanied with six synchronized ground truths (ECG, BVP, Respiration, HR, RR, and SpO2). It covers a wide spectrum of naturalistic driving conditions, including driver motions, dynamic natural light, vehicle types, and road conditions. We extensively evaluate both signal-processing and deep-learning methods on PhysDrive, establishing a comprehensive benchmark across all modalities, and release full open-source code with compatibility for mainstream public toolboxes. We envision PhysDrive will serve as a foundational resource and accelerate research on multimodal driver monitoring and smart-cockpit systems.

Peiling Jiang, Li Feng, Fuling Sun et al.

Existing text selection techniques on touchscreen focus on improving the control for moving the carets. Coarse-grained text selection on word and phrase levels has not received much support beyond word-snapping and entity recognition. We introduce 1D-Touch, a novel text selection method that complements the carets-based sub-word selection by facilitating the selection of semantic units of words and above. This method employs a simple vertical slide gesture to expand and contract a selection area from a word. The expansion can be by words or by semantic chunks ranging from sub-phrases to sentences. This technique shifts the concept of text selection, from defining a range by locating the first and last words, towards a dynamic process of expanding and contracting a textual semantic entity. To understand the effects of our approach, we prototyped and tested two variants: WordTouch, which offers a straightforward word-by-word expansion, and ChunkTouch, which leverages NLP to chunk text into syntactic units, allowing the selection to grow by semantically meaningful units in response to the sliding gesture. Our evaluation, focused on the coarse-grained selection tasks handled by 1D-Touch, shows a 20% improvement over the default word-snapping selection method on Android.

Zhaohui Liang, Yonglin Chen, Naser Al Madi et al.

Transcripts displayed on dictation interfaces can be hard to read due to recognition errors and disfluencies. LLM-based text auto-correction could help, but changing the text during production could lead to distraction and unintended phrasing. To understand how to balance readability, attention, and accuracy, we conducted an eye-tracking experiment with 20 participants to compare five dictation interfaces: PLAIN (real-time transcription), AOC (periodic corrections), RAKE (keyword highlights), GP-TSM (grammar-preserving highlights), and SUMMARY (LLM-generated abstractive summary). By analyzing participants' gaze patterns during speech composition and reviewing processes, we found that during composition, participants spent only 7-11% of their time in active reading regardless of the interface. Although SUMMARY introduced unfamiliar words and phrasing during composition, it was easier to read and more preferred by participants. Our findings suggest a high user tolerance for altering spoken words in LLM-enabled diction interfaces.

Can Liu, Chunlin Da, Xiaoxiao Long et al.

Current multimodal large language models (MLLMs), while effective in natural image understanding, struggle with visualization understanding due to their inability to decode the data-to-visual mapping and extract structured information. To address these challenges, we propose SimVec, a novel simplified vector format that encodes chart elements such as mark type, position, and size. The effectiveness of SimVec is demonstrated by using MLLMs to reconstruct chart information from SimVec formats. Then, we build a new visualization dataset, SimVecVis, to enhance the performance of MLLMs in visualization understanding, which consists of three key dimensions: bitmap images of charts, their SimVec representations, and corresponding data-centric question-answering (QA) pairs with explanatory chain-of-thought (CoT) descriptions. We finetune state-of-the-art MLLMs (e.g., MiniCPM and Qwen-VL), using SimVecVis with different dataset dimensions. The experimental results show that it leads to substantial performance improvements of MLLMs with good spatial perception capabilities (e.g., MiniCPM) in data-centric QA tasks. Our dataset and source code are available at: https://github.com/VIDA-Lab/SimVecVis.

Xiaolin Wen, Changlin Li, Manusha Karunathilaka et al.

Many expressive visualizations are shared online only as bitmap images, making them difficult to redesign or adapt to new data. Reusing such image-based visualizations requires substantial expertise and is often time-consuming, even for experienced visualization practitioners. Existing work on reproducing visualizations often relies on structured SVG or specifications, supports limited visualization types, and offers limited flexibility for customization. To address these challenges, we present ReVis, a human-AI collaboration approach that enables flexible reuse of image-based visualizations. First, a generic Domain-Specific language (DSL) is proposed to model complex visualizations and support both visualization decomposition and reproduction. Then, ReVis employs an MLLM-based pipeline to parse an image-based visualization into the DSL, delineating its core visual structures and data-to-encoding mappings, and further reproduces the visualization from the DSL. Finally, ReVis includes an interactive interface to allow users to upload visualization images, inspect reproduced results, update the underlying data, and customize visual encodings. A gallery of 40 visualizations demonstrates the expressiveness of the DSL, and a quantitative study evaluates the reproduction quality of ReVis on these examples. Two usage scenarios and user interviews with 16 visualization practitioners demonstrate the effectiveness of ReVis.

Xiang Li, Cara Li, Emily Kuang et al.

Knowledge workers face increasing challenges in synthesizing information from multiple documents into structured conceptual understanding. This process is inherently iterative: users explore content, identify relationships between concepts, and continuously reorganize their mental models. However, current approaches offer limited support. LLM-based systems let users query information but not shape how knowledge is organized; manual tools like mind maps support structure creation but lack intelligent assistance. This leaves an open opportunity: supporting collaborative construction where users and AI jointly develop an evolving knowledge representation. We present MindTrellis, an interactive visual system where users and AI collaboratively build a dynamic knowledge graph. Users can query the graph to retrieve document-grounded information, and contribute by introducing new concepts, modifying relationships, and reorganizing the hierarchy to reflect their developing understanding. In a user study where 12 participants created slide decks, MindTrellis outperformed retrieval-only baselines in knowledge organization and cognitive load, as measured by expert ratings of content coverage and structural quality.

Can Liu, Sizhe Cheng, Feng Liang et al.

With the rise of mobile-first consumption, users increasingly engage with data visualizations on mobile devices. However, the vast majority of existing visualizations are originally authored for desktop environments. Due to significant differences in viewport size and interaction paradigms, directly scaling desktop charts often results in illegible text, information loss, and interaction failures. To bridge this gap, we propose an automated framework to adapt desktop-based visualizations for mobile screens. By systematically categorizing the operations involved in the adaptation process, we establish a multi-level design space. This space defines evolution rules spanning from the global topology level, through the reference frame level, down to the visual elements level. Guided by this theoretical framework, we developed Proteus, a large language model-driven multi-agent system that automatically parses online visualizations, predicts optimal transformation strategies within the design space, and generates equivalent, highly readable visualizations for mobile devices. Case studies and an in-depth user study with 12 participants demonstrate the effectiveness and usability of Proteus.

Xiao Yang, Jiyao Wang, Yuxuan Fan et al.

Remote physiological measurement (RPM) has emerged as a promising non-invasive method for monitoring physiological signals using the non-contact device. Although various domain adaptation and generalization methods were proposed to promote the adaptability of deep-based RPM models in unseen deployment environments, considerations in aspects such as privacy concerns and real-time adaptation restrict their application in real-world deployment. Thus, we aim to propose a novel fully Test-Time Adaptation (TTA) strategy tailored for RPM tasks in this work. Specifically, based on prior knowledge in physiology and our observations, we noticed not only there is spatio-temporal consistency in the frequency domain of BVP signals, but also that inconsistency in the time domain was significant. Given this, by leveraging both consistency and inconsistency priors, we introduce an innovative expert knowledge-based self-supervised \textbf{C}onsistency-\textbf{i}n\textbf{C}onsistency-\textbf{i}ntegration (\textbf{CiCi}) framework to enhances model adaptation during inference. Besides, our approach further incorporates a gradient dynamic control mechanism to mitigate potential conflicts between priors, ensuring stable adaptation across instances. Through extensive experiments on five diverse datasets under the TTA protocol, our method consistently outperforms existing techniques, presenting state-of-the-art performance in real-time self-supervised adaptation without accessing source data. The code will be released later.

Can Liu, Jin Wang

As a new distributed computing framework that can protect data privacy, federated learning (FL) has attracted more and more attention in recent years. It receives gradients from users to train the global model and releases the trained global model to working users. Nonetheless, the gradient inversion (GI) attack reflects the risk of privacy leakage in federated learning. Attackers only need to use gradients through hundreds of thousands of simple iterations to obtain relatively accurate private data stored on users' local devices. For this, some works propose simple but effective strategies to obtain user data under a single-label dataset. However, these strategies induce a satisfactory visual effect of the inversion image at the expense of higher time costs. Due to the semantic limitation of a single label, the image obtained by gradient inversion may have semantic errors. We present a novel gradient inversion strategy based on canny edge detection (MGIC) in both the multi-label and single-label datasets. To reduce semantic errors caused by a single label, we add new convolution layers' blocks in the trained model to obtain the image's multi-label. Through multi-label representation, serious semantic errors in inversion images are reduced. Then, we analyze the impact of parameters on the difficulty of input image reconstruction and discuss how image multi-subjects affect the inversion performance. Our proposed strategy has better visual inversion image results than the most widely used ones, saving more than 78% of time costs in the ImageNet dataset.

Can Liu, Jin Wang, and Yipeng Zhou et al.

Federated learning (FL) empowers privacypreservation in model training by only exposing users' model gradients. Yet, FL users are susceptible to gradient inversion attacks (GIAs) which can reconstruct ground-truth training data such as images based on model gradients. However, reconstructing high-resolution images by existing GIAs faces two challenges: inferior accuracy and slow-convergence, especially when duplicating labels exist in the training batch. To address these challenges, we present an Accurate and Fast-convergent Gradient Inversion attack algorithm, called AFGI, with two components: Label Recovery Block (LRB) which can accurately restore duplicating labels of private images based on exposed gradients; VME Regularization Term, which includes the total variance of reconstructed images, the discrepancy between three-channel means and edges, between values from exposed gradients and reconstructed images, respectively. The AFGI can be regarded as a white-box attack strategy to reconstruct images by leveraging labels recovered by LRB. In particular, AFGI is efficient that accurately reconstruct ground-truth images when users' training batch size is up to 48. Our experimental results manifest that AFGI can diminish 85% time costs while achieving superb inversion quality in the ImageNet dataset. At last, our study unveils the shortcomings of FL in privacy-preservation, prompting the development of more advanced countermeasure strategies.

Can Liu, Jaeuk Lee, Tianhe Chen et al.

Interactivity is crucial for effective data visualizations. However, it is often challenging to implement interactions for existing static visualizations, since the underlying code and data for existing static visualizations are often not available, and it also takes significant time and effort to enable interactions for them even if the original code and data are available. To fill this gap, we propose Athanor, a novel approach to transform existing static visualizations into interactive ones using multimodal large language models (MLLMs) and natural language instructions. Our approach introduces three key innovations: (1) an action-modification interaction design space that maps visualization interactions into user actions and corresponding adjustments, (2) a multi-agent requirement analyzer that translates natural language instructions into an actionable operational space, and (3) a visualization abstraction transformer that converts static visualizations into flexible and interactive representations regardless of their underlying implementation. Athanor allows users to effortlessly author interactions through natural language instructions, eliminating the need for programming. We conducted two case studies and in-depth interviews with target users to evaluate our approach. The results demonstrate the effectiveness and usability of our approach in allowing users to conveniently enable flexible interactions for static visualizations.

Yen Nhi Truong Vu, Richard Wang, Niranjan Balachandar et al.

Self-supervised contrastive learning between pairs of multiple views of the same image has been shown to successfully leverage unlabeled data to produce meaningful visual representations for both natural and medical images. However, there has been limited work on determining how to select pairs for medical images, where availability of patient metadata can be leveraged to improve representations. In this work, we develop a method to select positive pairs coming from views of possibly different images through the use of patient metadata. We compare strategies for selecting positive pairs for chest X-ray interpretation including requiring them to be from the same patient, imaging study or laterality. We evaluate downstream task performance by fine-tuning the linear layer on 1% of the labeled dataset for pleural effusion classification. Our best performing positive pair selection strategy, which involves using images from the same patient from the same study across all lateralities, achieves a performance increase of 14.4% in mean AUC from the ImageNet pretrained baseline. Our controlled experiments show that the keys to improving downstream performance on disease classification are (1) using patient metadata to appropriately create positive pairs from different images with the same underlying pathologies, and (2) maximizing the number of different images used in query pairing. In addition, we explore leveraging patient metadata to select hard negative pairs for contrastive learning, but do not find improvement over baselines that do not use metadata. Our method is broadly applicable to medical image interpretation and allows flexibility for incorporating medical insights in choosing pairs for contrastive learning.

Claudia Pérez-D'Arpino, Can Liu, Patrick Goebel et al.

Navigating fluently around pedestrians is a necessary capability for mobile robots deployed in human environments, such as buildings and homes. While research on social navigation has focused mainly on the scalability with the number of pedestrians in open spaces, typical indoor environments present the additional challenge of constrained spaces such as corridors and doorways that limit maneuverability and influence patterns of pedestrian interaction. We present an approach based on reinforcement learning (RL) to learn policies capable of dynamic adaptation to the presence of moving pedestrians while navigating between desired locations in constrained environments. The policy network receives guidance from a motion planner that provides waypoints to follow a globally planned trajectory, whereas RL handles the local interactions. We explore a compositional principle for multi-layout training and find that policies trained in a small set of geometrically simple layouts successfully generalize to more complex unseen layouts that exhibit composition of the structural elements available during training. Going beyond walls-world like domains, we show transfer of the learned policy to unseen 3D reconstructions of two real environments. These results support the applicability of the compositional principle to navigation in real-world buildings and indicate promising usage of multi-agent simulation within reconstructed environments for tasks that involve interaction.

Weixin Liang, Zixuan Liu, Can Liu

Training a Generative Adversarial Networks (GAN) for a new domain from scratch requires an enormous amount of training data and days of training time. To this end, we propose DAWSON, a Domain Adaptive FewShot Generation FrameworkFor GANs based on meta-learning. A major challenge of applying meta-learning GANs is to obtain gradients for the generator from evaluating it on development sets due to the likelihood-free nature of GANs. To address this challenge, we propose an alternative GAN training procedure that naturally combines the two-step training procedure of GANs and the two-step training procedure of meta-learning algorithms. DAWSON is a plug-and-play framework that supports a broad family of meta-learning algorithms and various GANs with architectural-variants. Based on DAWSON, We also propose MUSIC MATINEE, which is the first few-shot music generation model. Our experiments show that MUSIC MATINEE could quickly adapt to new domains with only tens of songs from the target domains. We also show that DAWSON can learn to generate new digits with only four samples in the MNIST dataset. We release source codes implementation of DAWSON in both PyTorch and Tensorflow, generated music samples on two genres and the lightning video.

Can Liu, Shridatt Sugrim, Gradeigh D. Clark et al.

Gesture and signature passwords are two-dimensional figures created by drawing on the surface of a touchscreen with one or more fingers. Prior results about their security have used resilience to either shoulder surfing, a human observation attack, or dictionary attacks. These evaluations restrict generalizability since the results are: non-comparable to other password systems (e.g. PINs), harder to reproduce, and attacker-dependent. Strong statements about the security of a password system use an analysis of the statistical distribution of the password space, which models a best-case attacker who guesses passwords in order of most likely to least likely. Estimating the distribution of recognition passwords is challenging because many different trials need to map to one password. In this paper, we solve this difficult problem by: (1) representing a recognition password of continuous data as a discrete alphabet set, and (2) estimating the password distribution through modeling the unseen passwords. We use Symbolic Aggregate approXimation (SAX) to represent time series data as symbols and develop Markov chains to model recognition passwords. We use a partial guessing metric, which demonstrates how many guesses an attacker needs to crack a percentage of the entire space, to compare the security of the distributions for gestures, signatures, and Android unlock patterns. We found the lower bounds of the partial guessing metric of gestures and signatures are much higher than the upper bound of the partial guessing metric of Android unlock patterns.

Guoxiong Xu, Zhengfei Wang, Hongshi Huang et al.

The process of determining which disease or condition explains a person's symptoms and signs can be very complicated and may be inaccurate in some cases. The general belief is that diagnosing diseases relies on doctors' keen intuition, rich experience and professional equipment. In this work, we employ ideas from recent advances in plantar pressure research and from the powerful capacity of the convolutional neural network for learning representations. Here, we propose a model using convolutional neural network based on plantar pressure for medical diagnosis. Our model learns a network that maps plantar pressure data to its corresponding medical diagnostic label. We then apply our model to make the medical diagnosis on datasets we collected from cooperative hospital and achieve an accuracy of 98.36%. We demonstrate that the model base on the convolutional neural network is competitive in medical diagnosis.