Search

In the field of heavy-duty vehicles, fuel efficiency and environmental protection are factors that need to be focused on, while the aerodynamic drag generated during vehicle travelling is one of the most influential aspects. This thesis delves into the aerodynamic simulation of Scania trucks using the open-source Computational Fluid Dynamics (CFD) tool, OpenFOAM v2206. This study rigorously investigates the aerodynamics of two Scania truck models under different operating conditions, including scenarios with different crosswind environments at high speeds.The core of this study is to compare and analyse the computational results of OpenFOAM v2206 and its predecessor OpenFOAM v3.0+ in a number of aspects, in order to elucidate the evolution and improvement of CFD techniques and their practical impact on vehicle simulation performance. In order to save computational resources, the RANS method was used for the steady-state simulations. Preliminary comparisons were also made with results from PowerFLOW, another CFD software widely used within the Scania group.Another important part of this thesis is the exploration of an alternative meshing method (ANSA Hextreme Mesh) in CFD simulations. As a widely used pre-processing software in the Scania group today, analysing and comparing the advantages and disadvantages of ANSA and OpenFOAM in terms of meshing, such as the ease of meshing and the accuracy of aerodynamic predictions, can help to provide valuable guidance for the application of truck shape design and aerodynamic simulation.The results indicate that OpenFOAM v2206 excels in predicting aerodynamics and has utility in optimising truck design. Compared to OpenFOAM v3.0+, OpenFOAM v2206 shows smaller discrepancies in results with PowerFLOW. Further exploration is required regarding transient simulations using OpenFOAM. In terms of meshing methods, a simplified model (Allan Body) was investigated, and there is further research to be done on meshing the complete truck.In

In the field of heavy-duty vehicles, fuel efficiency and environmental protection are factors that need to be focused on, while the aerodynamic drag generated during vehicle travelling is one of the most influential aspects. This thesis delves into the aerodynamic simulation of Scania trucks using the open-source Computational Fluid Dynamics (CFD) tool, OpenFOAM v2206. This study rigorously investigates the aerodynamics of two Scania truck models under different operating conditions, including scenarios with different crosswind environments at high speeds.The core of this study is to compare and analyse the computational results of OpenFOAM v2206 and its predecessor OpenFOAM v3.0+ in a number of aspects, in order to elucidate the evolution and improvement of CFD techniques and their practical impact on vehicle simulation performance. In order to save computational resources, the RANS method was used for the steady-state simulations. Preliminary comparisons were also made with results from PowerFLOW, another CFD software widely used within the Scania group.Another important part of this thesis is the exploration of an alternative meshing method (ANSA Hextreme Mesh) in CFD simulations. As a widely used pre-processing software in the Scania group today, analysing and comparing the advantages and disadvantages of ANSA and OpenFOAM in terms of meshing, such as the ease of meshing and the accuracy of aerodynamic predictions, can help to provide valuable guidance for the application of truck shape design and aerodynamic simulation.The results indicate that OpenFOAM v2206 excels in predicting aerodynamics and has utility in optimising truck design. Compared to OpenFOAM v3.0+, OpenFOAM v2206 shows smaller discrepancies in results with PowerFLOW. Further exploration is required regarding transient simulations using OpenFOAM. In terms of meshing methods, a simplified model (Allan Body) was investigated, and there is further research to be done on meshing the complete truck.In

The Constrained Shortest Path (CSP) problem seeks to identify the shortest path between two vertices in a road network while adhering to a specific constraint on another criterion. Solving the CSP problem frequently entails navigating the two-criteria skyline path problem, which incurs a substantial computational expense in large road networks. The primary challenge lies in handling a vast quantity of partial skyline paths, which often hinders index-based solutions from accurately determining the skyline paths. This paper introduces a-FHL, a practical approximation method designed to circumvent the costly skyline path search and hasten computation on skyline path indexing. a-FHL uses tree decomposition to hierarchically assign approximation ratios, thereby facilitating effective pruning within the labelling index. Moreover, we devise various strategies to allocate approximation ratios and an efficient approximation concatenation method to respond to the approximate CSP queries via the a-FHL index. Our method culminates in swift index construction and efficient query response. Comprehensive exper-iments conducted on real-world road networks substantiate the superiority of our approach over contemporary solutions

Weijian Zhu,1 Ying He,1 Mufang Huang,1 Shezhu Fu,1 Ziyi Liu,1 Xiaoqi Wang,1 Zhixin Li,1 Xiaoliang Li,1 Jiangming Chen,2 Yangqiu Li3 1Department of Hematology, Zhuhai Clinical Medical College of Jinan University (Zhuhai People’s Hospital), Zhuahai, 519050, People’s Republic of China; 2Department of Haematology, Wuzhou Gongren Hospital, Wuzhou, 543001, People’s Republic of China; 3Institute of Hematology, School of Medicine, Jinan University, Guangzhou, 10632, People’s Republic of ChinaCorrespondence: Yangqiu Li, Institute of Hematology, School of Medicine, Jinan University, Guangzhou, 510632, People’s Republic of China, Email yangqiuli@hotmail.comObjective: We evaluated the long-term safety and efficacy of thalidomide in the treatment of transfusion-dependent β-thalassemia (TDT).Methods: Fifty patients with TDT were treated with thalidomide and followed-up for 5 years. Thalidomide at a 50 mg dose was administered once a day after dinner. The dose was increased to 150 mg/d after 3 d if well tolerated. After 1 year of treatment, the hemoglobin (Hb) level was stabilized at its maximum, and thalidomide was gradually reduced and maintained at the minimum dose. The hematological response, transfusion dependence, and haemolytic indicators were assessed.Results: At 9 month of follow-up, 38 (76%) patients achieved an excellent response, 1 (2%) a good response, 4(8%) a minor response, and 7(14%) did not show a response. The overall response rate was 86%. At 9 months, the Hb level increased from 79.0 ± 13.2 g/L at baseline to 99.0 ± 13.7g/L (P< 0.001). Patients who achieved excellent response continued to show an increase in Hb levels during follow-up. At 48 months, the mean Hb level was 98.99 ± 10.3g/L; 21 patients (84.0%) became transfusion independent. Thalidomide was reduced and maintained to 25 mg/d in three of these patients. Moreover, five patients completed 60 months of follow-up, and with a mean Hb level of 99.8 ± 6.7g/L. During follow-up, grade 1–

The development of large language models (LLMs) has brought unprecedented possibilities for artificial intelligence (AI) based medical diagnosis. However, the application perspective of LLMs in real diagnostic scenarios is still unclear because they are not adept at collecting patient data proactively. This study presents a LLM-based diagnostic system that enhances planning capabilities by emulating doctors. Our system involves two external planners to handle planning tasks. The first planner employs a reinforcement learning approach to formulate disease screening questions and conduct initial diagnoses. The second planner uses LLMs to parse medical guidelines and conduct differential diagnoses. By utilizing real patient electronic medical record data, we constructed simulated dialogues between virtual patients and doctors and evaluated the diagnostic abilities of our system. We demonstrated that our system obtained impressive performance in both disease screening and differential diagnoses tasks. This research represents a step towards more seamlessly integrating AI into clinical settings, potentially enhancing the accuracy and accessibility of medical diagnostics., Comment: Work in Progress

The wide deployment of Face Recognition (FR) systems poses risks of privacy leakage. One countermeasure to address this issue is adversarial attacks, which deceive malicious FR searches but simultaneously interfere the normal identity verification of trusted authorizers. In this paper, we propose the first Double Privacy Guard (DPG) scheme based on traceable adversarial watermarking. DPG employs a one-time watermark embedding to deceive unauthorized FR models and allows authorizers to perform identity verification by extracting the watermark. Specifically, we propose an information-guided adversarial attack against FR models. The encoder embeds an identity-specific watermark into the deep feature space of the carrier, guiding recognizable features of the image to deviate from the source identity. We further adopt a collaborative meta-optimization strategy compatible with sub-tasks, which regularizes the joint optimization direction of the encoder and decoder. This strategy enhances the representation of universal carrier features, mitigating multi-objective optimization conflicts in watermarking. Experiments confirm that DPG achieves significant attack success rates and traceability accuracy on state-of-the-art FR models, exhibiting remarkable robustness that outperforms the existing privacy protection methods using adversarial attacks and deep watermarking, or simple combinations of the two. Our work potentially opens up new insights into proactive protection for FR privacy.

Weakly-supervised temporal action localization (WTAL) aims to recognize and localize action instances with only video-level labels. Despite the significant progress, existing methods suffer from severe performance degradation when transferring to different distributions and thus may hardly adapt to real-world scenarios . To address this problem, we propose the Generalizable Temporal Action Localization task (GTAL), which focuses on improving the generalizability of action localization methods. We observed that the performance decline can be primarily attributed to the lack of generalizability to different action scales. To address this problem, we propose STAT (Self-supervised Temporal Adaptive Teacher), which leverages a teacher-student structure for iterative refinement. Our STAT features a refinement module and an alignment module. The former iteratively refines the model's output by leveraging contextual information and helps adapt to the target scale. The latter improves the refinement process by promoting a consensus between student and teacher models. We conduct extensive experiments on three datasets, THUMOS14, ActivityNet1.2, and HACS, and the results show that our method significantly improves the Baseline methods under the cross-distribution evaluation setting, even approaching the same-distribution evaluation performance., Comment: 14 pages, LaTeX

Simultaneously optimizing the processing parameters of functional thin films remains a challenge. The design and utilization of a fully automated platform called SPINBOT is presented for the engineering of solution‐processed functional thin films. The SPINBOT is capable of performing experiments with high sampling variability through the unsupervised processing of hundreds of substrates with exceptional experimental control. Through the iterative optimization process enabled by the Bayesian optimization (BO) algorithm, the SPINBOT explores an intricate parameter space, continuously improving the quality and reproducibility of the produced thin films. This machine learning (ML)‐guided reliable SPINBOT platform enables the acceleration of the optimization process of perovskite solar cells via a simple photoluminescence characterization of films. As a result, this study arrives at an optimal film that, when processed into a solar cell in an ambient atmosphere, immediately yields a champion power conversion efficiency (PCE) of 21.6% with satisfactory performance reproducibility. The unsealed devices retain 90% of their initial efficiency after 1100 h of continuous operation at 60–65 °C under metal‐halide lamps. It is anticipated that the integration of robotic platforms with the intelligent algorithm will facilitate the widespread adoption of effective autonomous experimentation to address the evolving needs and constraints within the materials science research community.

Fang Rong,1,2,&ast; Ziyi Liu,3,4,&ast; Pengbin Yang,3,4 Feng Wu,5 Yu Sun,5 Xuewei Sun,5 Jun Zhou5 1Department of General Practice, The Affiliated Hospital of Yangzhou University, Yangzhou, Jiangsu, People’s Republic of China; 2Graduate School Department of Dalian Medical University, Dalian, Liaoning, People’s Republic of China; 3Jiangsu Co-Innovation Center for Prevention and Control of Important Animal Infectious Diseases and Zoonoses, College of Veterinary Medicine, Yangzhou University, Yangzhou, Jiangsu, People’s Republic of China; 4Institute of Comparative Medicine, Yangzhou University, Yangzhou, Jiangsu, People’s Republic of China; 5Department of Respiratory and Critical Care Medicine, The Affiliated Hospital of Yangzhou University, Yangzhou University, Yangzhou, Jiangsu, People’s Republic of China&ast;These authors contributed equally to this workCorrespondence: Jun Zhou, Department of Respiratory and Critical Care Medicine, The Affiliated Hospital of Yangzhou University, Yangzhou University, 368 Hanjiang Middle Road, Yangzhou, Jiangsu, 225009, People’s Republic of China, Email zj662121@163.comObjective: Carbapenem-resistant Klebsiella pneumoniae (CRKP) has emerged and spread worldwide. It can usually cause a serious threat complicating treatment options in clinical settings. However, treatment options are limited. The present study investigates the prevalence and genetic characteristics of blaNDM-1 and blaKPC-2 co-harboring clinical isolates of Klebsiella pneumoniae.Methods: In this study, Multiplex polymerase chain reaction (PCR) was performed to detect the carbapenem-resistant genes, and the broth microdilution method was used to determine the minimum inhibitory concentrations (MICs) of antibacterial drugs. The transferability of carbapenem-resistant phenotypes was examined using filter mating assays. Overall, we used Illumina sequencing to evaluate the epidemiological and molecular characteristics of blaNDM-1 and blaKPC-2 (genes

Graph partitioning is a common solution to scale up the graph algorithms, and shortest path (SP) computation is one of them. However, the existing solutions typically have a fixed partition method with a fixed path index and fixed partition structure, so it is unclear how the partition method and path index influence the pathfinding performance. Moreover, few studies have explored the index maintenance of partitioned SP (PSP) on dynamic graphs. To provide a deeper insight into the dynamic PSP indexes, we systematically deliberate on the existing works and propose a universal scheme to analyze this problem theoretically. Specifically, we first propose two novel partitioned index strategies and one optimization to improve index construction, query answering, or index maintenance of PSP index. Then we propose a path-oriented graph partitioning classification criteria for easier partition method selection. After that, we re-couple the dimensions in our scheme (partitioned index strategy, path index, and partition structure) to propose five new partitioned SP indexes that are more efficient either in the query or update on different networks. Finally, we demonstrate the effectiveness of our new indexes by comparing them with state-of-the-art PSP indexes through comprehensive evaluations.

In general-sum games, the interaction of self-interested learning agents commonly leads to socially worse outcomes, such as defect-defect in the iterated stag hunt (ISH). Previous works address this challenge by sharing rewards or shaping their opponents' learning process, which require too strong assumptions. In this paper, we demonstrate that agents trained to optimize expected returns are more likely to choose a safe action that leads to guaranteed but lower rewards. However, there typically exists a risky action that leads to higher rewards in the long run only if agents cooperate, e.g., cooperate-cooperate in ISH. To overcome this, we propose using action value distribution to characterize the decision's risk and corresponding potential payoffs. Specifically, we present Adaptable Risk-Sensitive Policy (ARSP). ARSP learns the distributions over agent's return and estimates a dynamic risk-seeking bonus to discover risky coordination strategies. Furthermore, to avoid overfitting training opponents, ARSP learns an auxiliary opponent modeling task to infer opponents' types and dynamically alter corresponding strategies during execution. Empirically, agents trained via ARSP can achieve stable coordination during training without accessing opponent's rewards or learning process, and can adapt to non-cooperative opponents during execution. To the best of our knowledge, it is the first method to learn coordination strategies between agents both in iterated prisoner's dilemma (IPD) and iterated stag hunt (ISH) without shaping opponents or rewards, and can adapt to opponents with distinct strategies during execution. Furthermore, we show that ARSP can be scaled to high-dimensional settings., Comment: arXiv admin note: substantial text overlap with arXiv:2205.15859

Among the remarkable emergent capabilities of large language models (LMs) is free-text rationalization; beyond a certain scale, large LMs are capable of generating seemingly useful rationalizations, which in turn, can dramatically enhance their performances on leaderboards. This phenomenon raises a question: can machine generated rationales also be useful for humans, especially when lay humans try to answer questions based on those machine rationales? We observe that human utility of existing rationales is far from satisfactory, and expensive to estimate with human studies. Existing metrics like task performance of the LM generating the rationales, or similarity between generated and gold rationales are not good indicators of their human utility. While we observe that certain properties of rationales like conciseness and novelty are correlated with their human utility, estimating them without human involvement is challenging. We show that, by estimating a rationale's helpfulness in answering similar unseen instances, we can measure its human utility to a better extent. We also translate this finding into an automated score, GEN-U, that we propose, which can help improve LMs' ability to generate rationales with better human utility, while maintaining most of its task performance. Lastly, we release all code and collected data with this project., Comment: Accepted at ACL 2023

Machine learning (ML) based time series forecasting models often require and assume certain degrees of stationarity in the data when producing forecasts. However, in many real-world situations, the data distributions are not stationary and they can change over time while reducing the accuracy of the forecasting models, which in the ML literature is known as concept drift. Handling concept drift in forecasting is essential for many ML methods in use nowadays, however, the prior work only proposes methods to handle concept drift in the classification domain. To fill this gap, we explore concept drift handling methods in particular for Global Forecasting Models (GFM) which recently have gained popularity in the forecasting domain. We propose two new concept drift handling methods, namely: Error Contribution Weighting (ECW) and Gradient Descent Weighting (GDW), based on a continuous adaptive weighting concept. These methods use two forecasting models which are separately trained with the most recent series and all series, and finally, the weighted average of the forecasts provided by the two models are considered as the final forecasts. Using LightGBM as the underlying base learner, in our evaluation on three simulated datasets, the proposed models achieve significantly higher accuracy than a set of statistical benchmarks and LightGBM baselines across four evaluation metrics., Comment: 23 pages, 3 figures, 2 tables

The Multi-Constraint Shortest Path (MCSP) problem aims to find the shortest path between two nodes in a network subject to a given constraint set. It is typically processed as a skyline path problem. However, the number of intermediate skyline paths becomes larger as the network size increases and the constraint number grows, which brings about the dramatical growth of computational cost and further makes the existing index-based methods hardly capable of obtaining the complete exact results. In this paper, we propose a novel high-dimensional skyline path concatenation method to avoid the expensive skyline path search, which supports the efficient construction of hop labeling index for MCSP queries. Specifically, a set of insightful observations and techniques are proposed to improve the efficiency of concatenating two skyline path set, a n-Cube technique is designed to prune the concatenation space among multiple hops, and a constraint pruning method is used to avoid the unnecessary computation. Furthermore, to scale up to larger networks, we propose a novel forest hop labeling which enables the parallel label construction from different network partitions. Our approach is the first method that can achieve accuracy and efficiency for MCSP query. Extensive experiments on real-life road networks demonstrate the superiority of our method over state-of-the-art solutions. © 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

In a plethora of recent work, large language models (LLMs) demonstrated impressive reasoning ability, but many proposed downstream reasoning tasks focus on performance-wise evaluation. Two fundamental questions persist: 1) how reliable is the quality of reasoning, and 2) can models detect unreliable reasoning? In this paper, we investigate self-contradictory (Self-Contra) reasoning, where the model reasoning does not support predictions. To address 1), we assess the Self-Contra rate across four datasets and delve into finer-grained categories of Self-Contra reasoning. We find that LLMs often contradict themselves when performing reasoning tasks that involve contextual information understanding or commonsense. Importantly, a higher accuracy does not necessarily correspond to a lower Self-Contra rate. The model may appear to generate correct answers but it may take shortcuts in reasoning or skip over contextual evidence, thereby displaying Self-Contra behaviors with compromised reasoning. As for 2), we task GPT-4 with identifying Self-Contra reasoning and finer-grained fallacies. We observe that GPT-4 struggles to effectively detect Self-Contra reasoning, with significantly low performance compared with human judgment. Our results indicate that the current LLMs lack robustness necessary for reliable reasoning and we emphasize the urgent need for establishing best practices in comprehensive reasoning evaluations beyond accuracy-based metrics.

The malicious applications of deep forgery, represented by face swapping, have introduced security threats such as misinformation dissemination and identity fraud. While some research has proposed the use of robust watermarking methods to trace the copyright of facial images for post-event traceability, these methods cannot effectively prevent the generation of forgeries at the source and curb their dissemination. To address this problem, we propose a novel comprehensive active defense mechanism that combines traceability and adversariality, called Dual Defense. Dual Defense invisibly embeds a single robust watermark within the target face to actively respond to sudden cases of malicious face swapping. It disrupts the output of the face swapping model while maintaining the integrity of watermark information throughout the entire dissemination process. This allows for watermark extraction at any stage of image tracking for traceability. Specifically, we introduce a watermark embedding network based on original-domain feature impersonation attack. This network learns robust adversarial features of target facial images and embeds watermarks, seeking a well-balanced trade-off between watermark invisibility, adversariality, and traceability through perceptual adversarial encoding strategies. Extensive experiments demonstrate that Dual Defense achieves optimal overall defense success rates and exhibits promising universality in anti-face swapping tasks and dataset generalization ability. It maintains impressive adversariality and traceability in both original and robust settings, surpassing current forgery defense methods that possess only one of these capabilities, including CMUA-Watermark, Anti-Forgery, FakeTagger, or PGD methods.

We propose BOSS, an approach that automatically learns to solve new long-horizon, complex, and meaningful tasks by growing a learned skill library with minimal supervision. Prior work in reinforcement learning require expert supervision, in the form of demonstrations or rich reward functions, to learn long-horizon tasks. Instead, our approach BOSS (BOotStrapping your own Skills) learns to accomplish new tasks by performing "skill bootstrapping," where an agent with a set of primitive skills interacts with the environment to practice new skills without receiving reward feedback for tasks outside of the initial skill set. This bootstrapping phase is guided by large language models (LLMs) that inform the agent of meaningful skills to chain together. Through this process, BOSS builds a wide range of complex and useful behaviors from a basic set of primitive skills. We demonstrate through experiments in realistic household environments that agents trained with our LLM-guided bootstrapping procedure outperform those trained with naive bootstrapping as well as prior unsupervised skill acquisition methods on zero-shot execution of unseen, long-horizon tasks in new environments. Website at clvrai.com/boss., Comment: CoRL 2023 (Oral); 24 pages, 11 figures

Shortest path (SP) computation is the fundamental operation in various networks such as urban networks, logistic networks, communication networks, social networks, etc. With the development of technology and societal expansions, those networks tend to be massive. This, in turn, causes deteriorated performance of SP computation, and graph partitioning is commonly leveraged to scale up the SP algorithms. However, the partitioned shortest path (PSP) index has never been systematically investigated and theoretically analyzed, and there is a lack of experimental comparison among different PSP indexes. Moreover, few studies have explored PSP index maintenance in dynamic networks. Therefore, in this paper, we systematically analyze the dynamic PSP index by proposing a universal scheme for it. Specifically, we first propose two novel partitioned shortest path strategies (No-boundary and Post-boundary strategies) to improve the performance of PSP indexes and design the corresponding index maintenance approaches to deal with dynamic scenarios. Then we categorize the partition methods from the perspective of partition structure to facilitate the selection of partition methods in the PSP index. Furthermore, we propose a universal scheme for designing the PSP index by coupling its three dimensions (i.e. PSP strategy, partition structure, and SP algorithm). Based on this scheme, we propose five new PSP indexes with prominent performance in either query or update efficiency. Lastly, extensive experiments are implemented to demonstrate the effectiveness of the proposed PSP scheme, with valuable guidance provided on the PSP index design.

Deep learning models have a risk of utilizing spurious clues to make predictions, such as recognizing actions based on the background scene. This issue can severely degrade the open-set action recognition performance when the testing samples have different scene distributions from the training samples. To mitigate this problem, we propose a novel method, called Scene-debiasing Open-set Action Recognition (SOAR), which features an adversarial scene reconstruction module and an adaptive adversarial scene classification module. The former prevents the decoder from reconstructing the video background given video features, and thus helps reduce the background information in feature learning. The latter aims to confuse scene type classification given video features, with a specific emphasis on the action foreground, and helps to learn scene-invariant information. In addition, we design an experiment to quantify the scene bias. The results indicate that the current open-set action recognizers are biased toward the scene, and our proposed SOAR method better mitigates such bias. Furthermore, our extensive experiments demonstrate that our method outperforms state-of-the-art methods, and the ablation studies confirm the effectiveness of our proposed modules., Comment: Accepted to ICCV 2023, code:https://github.com/yhZhai/SOAR

This study investigates clustered federated learning (FL), one of the formulations of FL with non-i.i.d. data, where the devices are partitioned into clusters and each cluster optimally fits its data with a localized model. We propose a clustered FL framework that incorporates a nonconvex penalty to pairwise differences of parameters. Without a priori knowledge of the set of devices in each cluster and the number of clusters, this framework can autonomously estimate cluster structures. To implement the proposed framework, we introduce a novel clustered FL method called Fusion Penalized Federated Clustering (FPFC). Building upon the standard alternating direction method of multipliers (ADMM), FPFC can perform partial updates at each communication round and allows parallel computation with variable workload. These strategies significantly reduce the communication cost while ensuring privacy, making it practical for FL. We also propose a new warmup strategy for hyperparameter tuning in FL settings and explore the asynchronous variant of FPFC (asyncFPFC). Theoretical analysis provides convergence guarantees for FPFC with general losses and establishes the statistical convergence rate under a linear model with squared loss. Extensive experiments have demonstrated the superiority of FPFC compared to current methods, including robustness and generalization capability., Comment: 46 pages, 9 figures

NLP models are susceptible to learning spurious biases (i.e., bugs) that work on some datasets but do not properly reflect the underlying task. Explanation-based model debugging aims to resolve spurious biases by showing human users explanations of model behavior, asking users to give feedback on the behavior, then using the feedback to update the model. While existing model debugging methods have shown promise, their prototype-level implementations provide limited practical utility. Thus, we propose XMD: the first open-source, end-to-end framework for explanation-based model debugging. Given task- or instance-level explanations, users can flexibly provide various forms of feedback via an intuitive, web-based UI. After receiving user feedback, XMD automatically updates the model in real time, by regularizing the model so that its explanations align with the user feedback. The new model can then be easily deployed into real-world applications via Hugging Face. Using XMD, we can improve the model's OOD performance on text classification tasks by up to 18%., Comment: 6 pages, 7 figures. Project page: https://inklab.usc.edu/xmd

Benefiting from large-scale pretrained vision language models (VLMs), the performance of visual question answering (VQA) has approached human oracles. However, finetuning such models on limited data often suffers from overfitting and poor generalization issues, leading to a lack of model robustness. In this paper, we aim to improve input robustness from an information bottleneck perspective when adapting pretrained VLMs to the downstream VQA task. Input robustness refers to the ability of models to defend against visual and linguistic input variations, as well as shortcut learning involved in inputs. Generally, the representations obtained by pretrained VLMs inevitably contain irrelevant and redundant information for a specific downstream task, resulting in statistically spurious correlations and insensitivity to input variations. To encourage representations to converge to a minimal sufficient statistic in multimodal learning, we propose Correlation Information Bottleneck (CIB), which seeks a tradeoff between compression and redundancy in representations by minimizing the mutual information (MI) between inputs and representations while maximizing the MI between outputs and representations. Moreover, we derive a tight theoretical upper bound for the mutual information between multimodal inputs and representations, incorporating different internal correlations that guide models to learn more robust representations and facilitate modality alignment. Extensive experiments consistently demonstrate the effectiveness and superiority of the proposed CIB in terms of input robustness and accuracy., Comment: 20 pages, 6 figures, 13 tables

To automatically correct handwritten assignments, the traditional approach is to use an OCR model to recognize characters and compare them to answers. The OCR model easily gets confused on recognizing handwritten Chinese characters, and the textual information of the answers is missing during the model inference. However, teachers always have these answers in mind to review and correct assignments. In this paper, we focus on the Chinese cloze tests correction and propose a multimodal approach (named AiM). The encoded representations of answers interact with the visual information of students' handwriting. Instead of predicting 'right' or 'wrong', we perform the sequence labeling on the answer text to infer which answer character differs from the handwritten content in a fine-grained way. We take samples of OCR datasets as the positive samples for this task, and develop a negative sample augmentation method to scale up the training data. Experimental results show that AiM outperforms OCR-based methods by a large margin. Extensive studies demonstrate the effectiveness of our multimodal approach., Comment: Accepted to COLING 2022

By explaining how humans would solve a given task, human rationales can provide strong learning signal for neural language models (LMs). Explanation regularization (ER) aims to improve LM generalization by pushing the LM's machine rationales (Which input tokens did the LM focus on?) to align with human rationales (Which input tokens would humans focus on?). Though prior works primarily study ER via in-distribution (ID) evaluation, out-of-distribution (OOD) generalization is often more critical in real-world scenarios, yet ER's effect on OOD generalization has been underexplored. In this paper, we introduce ER-Test, a framework for evaluating ER models' OOD generalization along three dimensions: unseen dataset tests, contrast set tests, and functional tests. Using ER-Test, we extensively analyze how ER models' OOD generalization varies with different ER design choices. Across two tasks and six datasets, ER-Test shows that ER has little impact on ID performance but can yield large OOD performance gains. Also, we find that ER can improve OOD performance even with limited rationale supervision. ER-Test's results help demonstrate ER's utility and establish best practices for using ER effectively., Comment: Findings of EMNLP 2022

While various multi-agent reinforcement learning methods have been proposed in cooperative settings, few works investigate how self-interested learning agents achieve mutual coordination in decentralized general-sum games and generalize pre-trained policies to non-cooperative opponents during execution. In this paper, we present Generalizable Risk-Sensitive Policy (GRSP). GRSP learns the distributions over agent's return and estimate a dynamic risk-seeking bonus to discover risky coordination strategies. Furthermore, to avoid overfitting to training opponents, GRSP learns an auxiliary opponent modeling task to infer opponents' types and dynamically alter corresponding strategies during execution. Empirically, agents trained via GRSP can achieve mutual coordination during training stably and avoid being exploited by non-cooperative opponents during execution. To the best of our knowledge, it is the first method to learn coordination strategies between agents both in iterated prisoner's dilemma (IPD) and iterated stag hunt (ISH) without shaping opponents or rewards, and firstly consider generalization during execution. Furthermore, we show that GRSP can be scaled to high-dimensional settings., Comment: the paper will be updated soon

Multi-Constraint Shortest Path (MCSP) generalizes the classic shortest path from single to multiple criteria such that more personalized needs can be satisfied. However, MCSP query is essentially a high-dimensional skyline problem and thus time-consuming to answer. Although the current Forest Hop Labeling (FHL) index can answer MCSP efficiently, it takes a long time to construct and lacks the flexibility to handle arbitrary criteria combinations. In this paper, we propose a skyline-cube-based FHL index that can handle the flexible MCSP efficiently. Firstly, we analyze the relation between low and high-dimensional skyline paths theoretically and use a cube to organize them hierarchically. After that, we propose methods to derive the high-dimensional path from the lower ones, which can adapt to the flexible scenario naturally and reduce the expensive high dimensional path concatenation. Then we introduce efficient methods for both single and multi-hop cube concatenations and propose pruning methods to further alleviate the computation. Finally, we improve the FHL structure with lower height for faster construction and query. Experiments on real-life road networks demonstrate the superiority of our method over the state-of-the-art. © 2022, VLDB Endowment. All rights reserved.

The Constrained Shortest Path (CSP) problem aims to find the shortest path between two nodes in a road network subject to a given constraint on another attribute. It is typically processed as a skyline path problem on the two attributes, resulting in very high computational cost which can be prohibitive for large road networks. The main bottleneck is to deal with a large amount of partial skyline paths, which further makes the existing index-based methods incapable to obtain the complete exact skyline paths. In this paper, we propose a novel skyline path concatenation approach to avoid the expensive skyline path search, which is then used to efficiently construct a 2-hop labeling index for the CSP queries. Specifically, a rectangle-based technique is designed to prune the concatenation space from multiple hops, and a constraint pruning method is used to further speed up the CSP query processing. To further scale up to larger networks, we propose a novel forest hop labeling that constructs labels from different partitions in parallel. Our approach is the first method that can achieve both accuracy and efficiency for CSP query answering. Extensive experiments on real-life road networks demonstrate that our method outperforms the state-of-the-art CSP solutions by several orders of magnitude. © 2021 IEEE.

The Constrained Shortest Path (CSP) problem aims to find the shortest path between two nodes in a road network subject to a given constraint on another attribute. It is typically processed as a skyline path problem on the two attributes, resulting in very high computational cost which can be prohibitive for large road networks. The main bottleneck is to deal with a large amount of partial skyline paths, which further makes the existing index-based methods incapable to obtain the complete exact skyline paths. In this paper, we propose a novel skyline path concatenation approach to avoid the expensive skyline path search, which is then used to efficiently construct a 2-hop labeling index for the CSP queries. Specifically, a rectangle-based technique is designed to prune the concatenation space from multiple hops, and a constraint pruning method is used to further speed up the CSP query processing. To further scale up to larger networks, we propose a novel forest hop labeling that constructs labels from different partitions in parallel. Our approach is the first method that can achieve both accuracy and efficiency for CSP query answering. Extensive experiments on real-life road networks demonstrate that our method outperforms the state-of-the-art CSP solutions by several orders of magnitude. © 2021 IEEE.

With the penetration of distributed energy resources, peer-to-peer (P2P) energy trading is becoming a promising way to harmonize the decarbonization and decentralization transformations in the energy sector. P2P markets give peers the autonomy to make individual decisions and cooperative behaviors between the peers may emerge. However, existing studies on cooperative behaviors in P2P markets focus mostly on the electricity sector, P2P multi-energy markets are rarely studied. In fact, other energy carriers not only constitute a large part of the total energy demand, but their coupling can potentially benefit the system as well as the end-users. In this paper, we propose a P2P multi-energy market mechanism that allows peers to trade both electricity and heat, where the peers can join two predefined trading coalitions. The proposed system model thus explores the integrated effects of the multi-energy coupling and the cooperative behaviors in the P2P market. In order to maximize the net benefits of the peers, price conditions are derived, based on which the peers will join either of the coalitions and determine their trading volumes. Then, the electricity and the heat markets are cleared separately by a market operator. Lastly, the market mechanism is illustrated by a case study on a neighborhood in the Netherlands using realistic data., Complex Systems Engineering and Management (CoSEM)

Video Question Answering (VideoQA), aiming to correctly answer the given question based on understanding multi-modal video content, is challenging due to the rich video content. From the perspective of video understanding, a good VideoQA framework needs to understand the video content at different semantic levels and flexibly integrate the diverse video content to distill question-related content. To this end, we propose a Lightweight Visual-Linguistic Reasoning framework named LiVLR. Specifically, LiVLR first utilizes the graph-based Visual and Linguistic Encoders to obtain multi-grained visual and linguistic representations. Subsequently, the obtained representations are integrated with the devised Diversity-aware Visual-Linguistic Reasoning module (DaVL). The DaVL considers the difference between the different types of representations and can flexibly adjust the importance of different types of representations when generating the question-related joint representation, which is an effective and general representation integration method. The proposed LiVLR is lightweight and shows its performance advantage on two VideoQA benchmarks, MRSVTT-QA and KnowIT VQA. Extensive ablation studies demonstrate the effectiveness of LiVLR key components., Comment: 11 pages, 5 figures, Code: https://github.com/jingjing12110/LiVLR-VideoQA

Background: Electronic Health Records (EHRs) contain rich information of patients' health history, which usually include both structured and unstructured data. There have been many studies focusing on distilling valuable information from structured data, such as disease codes, laboratory test results, and treatments. However, relying on structured data only might be insufficient in reflecting patients' comprehensive information and such data may occasionally contain erroneous records. Objective: With the recent advances of machine learning (ML) and deep learning (DL) techniques, an increasing number of studies seek to obtain more accurate results by incorporating unstructured free-text data as well. This paper reviews studies that use multimodal data, i.e. a combination of structured and unstructured data, from EHRs as input for conventional ML or DL models to address the targeted tasks. Materials and Methods: We searched in the Institute of Electrical and Electronics Engineers (IEEE) Digital Library, PubMed, and Association for Computing Machinery (ACM) Digital Library for articles related to ML-based multimodal EHR studies. Results and Discussion: With the final 94 included studies, we focus on how data from different modalities were combined and interacted using conventional ML and DL techniques, and how these algorithms were applied in EHR-related tasks. Further, we investigate the advantages and limitations of these fusion methods and indicate future directions for ML-based multimodal EHR research., Comment: 37 pages, 3 tables, 2 figures, to be submitted to JAMIA

Single-cell sequencing has a significant role to explore biological processes such as embryonic development, cancer evolution, and cell differentiation. These biological properties can be presented by a two-dimensional scatter plot. However, single-cell sequencing data generally has very high dimensionality. Therefore, dimensionality reduction should be used to process the high dimensional sequencing data for 2D visualization and subsequent biological analysis. The traditional dimensionality reduction methods, which do not consider the structure characteristics of single-cell sequencing data, are difficult to reveal the data structure in the 2D representation. In this paper, we develop a 2D feature representation method based on graph convolutional networks (GCN) for the visualization of single-cell data, termed single-cell sequencing graph embedding networks (SGEN). This method constructs the graph by the similarity relationship between cells and adopts GCN to analyze the neighbor embedding information of samples, which makes the similar cell closer to each other on the 2D scatter plot. The results show SGEN achieves obvious 2D distribution and preserves the high-dimensional relationship of different cells. Meanwhile, similar cell clusters have spatial continuity rather than relying heavily on random initialization, which can reflect the trajectory of cell development in this scatter plot., Comment: 6 pages body + 2 pages reference

The Constrained Shortest Path (CSP) problem aims to find the shortest path between two nodes in a road network subject to a given constraint on another attribute. It is typically processed as a skyline path problem on the two attributes, resulting in very high computational cost which can be prohibitive for large road networks. The main bottleneck is to deal with a large amount of partial skyline paths, which further makes the existing index-based methods incapable to obtain the complete exact skyline paths. In this paper, we propose a novel skyline path concatenation approach to avoid the expensive skyline path search, which is then used to efficiently construct a 2-hop labeling index for the CSP queries. Specifically, a rectangle-based technique is designed to prune the concatenation space from multiple hops, and a constraint pruning method is used to further speed up the CSP query processing. To further scale up to larger networks, we propose a novel forest hop labeling that constructs labels from different partitions in parallel. Our approach is the first method that can achieve both accuracy and efficiency for CSP query answering. Extensive experiments on real-life road networks demonstrate that our method outperforms the state-of-the-art CSP solutions by several orders of magnitude.

With the penetration of distributed energy resources, peer-to-peer (P2P) energy trading is becoming a promising way to harmonize the decarbonization and decentralization transformations in the energy sector. P2P markets give peers the autonomy to make individual decisions and cooperative behaviors between the peers may emerge. However, existing studies on cooperative behaviors in P2P markets focus mostly on the electricity sector, P2P multi-energy markets are rarely studied. In fact, other energy carriers not only constitute a large part of the total energy demand, but their coupling can potentially benefit the system as well as the end-users. In this paper, we propose a P2P multi-energy market mechanism that allows peers to trade both electricity and heat, where the peers can join two predefined trading coalitions. The proposed system model thus explores the integrated effects of the multi-energy coupling and the cooperative behaviors in the P2P market. In order to maximize the net benefits of the peers, price conditions are derived, based on which the peers will join either of the coalitions and determine their trading volumes. Then, the electricity and the heat markets are cleared separately by a market operator. Lastly, the market mechanism is illustrated by a case study on a neighborhood in the Netherlands using realistic data., Complex Systems Engineering and Management (CoSEM)

The \textit{Multi-Constraint Shortest Path (MCSP)} problem aims to find the shortest path between two nodes in a network subject to a given constraint set. It is typically processed as a \textit{skyline path} problem. However, the number of intermediate skyline paths becomes larger as the network size increases and the constraint number grows, which brings about the dramatical growth of computational cost and further makes the existing index-based methods hardly capable of obtaining the complete exact results. In this paper, we propose a novel high-dimensional skyline path concatenation method to avoid the expensive skyline path search, which then supports the efficient construction of hop labeling index for \textit{MCSP} queries. Specifically, a set of insightful observations and techniques are proposed to improve the efficiency of concatenating two skyline path set, a \textit{n-Cube} technique is designed to prune the concatenation space among multiple hops, and a \textit{constraint pruning} method is used to avoid the unnecessary computation. Furthermore, to scale up to larger networks, we propose a novel \textit{forest hop labeling} which enables the parallel label construction from different network partitions. Our approach is the first method that can achieve both accuracy and efficiency for \textit{MCSP} query answering. Extensive experiments on real-life road networks demonstrate the superiority of our method over the state-of-the-art solutions.

Encouraging progress has been made towards Visual Question Answering (VQA) in recent years, but it is still challenging to enable VQA models to adaptively generalize to out-of-distribution (OOD) samples. Intuitively, recompositions of existing visual concepts (\ie, attributes and objects) can generate unseen compositions in the training set, which will promote VQA models to generalize to OOD samples. In this paper, we formulate OOD generalization in VQA as a compositional generalization problem and propose a graph generative modeling-based training scheme (X-GGM) to implicitly model the problem. X-GGM leverages graph generative modeling to iteratively generate a relation matrix and node representations for the predefined graph that utilizes attribute-object pairs as nodes. Furthermore, to alleviate the unstable training issue in graph generative modeling, we propose a gradient distribution consistency loss to constrain the data distribution with adversarial perturbations and the generated distribution. The baseline VQA model (LXMERT) trained with the X-GGM scheme achieves state-of-the-art OOD performance on two standard VQA OOD benchmarks, \ie, VQA-CP v2 and GQA-OOD. Extensive ablation studies demonstrate the effectiveness of X-GGM components. Code is available at \url{https://github.com/jingjing12110/x-ggm}., Comment: Accepted by ACM MM2021

Deep convolutional neural networks (DCNNs) have aided high dynamic range (HDR) imaging recently and have received a lot of attention. The quality of DCNN-generated HDR images has overperformed the traditional counterparts. However, DCNNs are prone to be computationally intensive and power-hungry, and hence cannot be implemented on various embedded computing platforms with limited power and hardware resources. Embedded systems have a huge market, and utilizing DCNNs' powerful functionality into them will further reduce human intervention. To address the challenge, we propose LightFuse, a lightweight CNN-based algorithm for extreme dual-exposure image fusion, which achieves better functionality than a conventional DCNN and can be deployed in embedded systems. Two sub-networks are utilized: a GlobalNet (G) and a DetailNet (D). The goal of G is to learn the global illumination information on the spatial dimension, whereas D aims to enhance local details on the channel dimension. Both G and D are based solely on depthwise convolution (D_Conv) and pointwise convolution (P_Conv) to reduce required parameters and computations. Experimental results show that this proposed technique could generate HDR images in extremely exposed regions with sufficient details to be legible. Our model outperforms other state-of-the-art approaches in peak signal-to-noise ratio (PSNR) score by 0.9 to 8.7 while achieving 16.7 to 306.2 times parameter reduction.

Weakly-supervised Temporal Action Localization (WS-TAL) methods learn to localize temporal starts and ends of action instances in a video under only video-level supervision. Existing WS-TAL methods rely on deep features learned for action recognition. However, due to the mismatch between classification and localization, these features cannot distinguish the frequently co-occurring contextual background, i.e., the context, and the actual action instances. We term this challenge action-context confusion, and it will adversely affect the action localization accuracy. To address this challenge, we introduce a framework that learns two feature subspaces respectively for actions and their context. By explicitly accounting for action visual elements, the action instances can be localized more precisely without the distraction from the context. To facilitate the learning of these two feature subspaces with only video-level categorical labels, we leverage the predictions from both spatial and temporal streams for snippets grouping. In addition, an unsupervised learning task is introduced to make the proposed module focus on mining temporal information. The proposed approach outperforms state-of-the-art WS-TAL methods on three benchmarks, i.e., THUMOS14, ActivityNet v1.2 and v1.3 datasets., Comment: Accepted by the 35th AAAI Conference on Artificial Intelligence (AAAI 2021)

The object of Weakly-supervised Temporal Action Localization (WS-TAL) is to localize all action instances in an untrimmed video with only video-level supervision. Due to the lack of frame-level annotations during training, current WS-TAL methods rely on attention mechanisms to localize the foreground snippets or frames that contribute to the video-level classification task. This strategy frequently confuse context with the actual action, in the localization result. Separating action and context is a core problem for precise WS-TAL, but it is very challenging and has been largely ignored in the literature. In this paper, we introduce an Action-Context Separation Network (ACSNet) that explicitly takes into account context for accurate action localization. It consists of two branches (i.e., the Foreground-Background branch and the Action-Context branch). The Foreground- Background branch first distinguishes foreground from background within the entire video while the Action-Context branch further separates the foreground as action and context. We associate video snippets with two latent components (i.e., a positive component and a negative component), and their different combinations can effectively characterize foreground, action and context. Furthermore, we introduce extended labels with auxiliary context categories to facilitate the learning of action-context separation. Experiments on THUMOS14 and ActivityNet v1.2/v1.3 datasets demonstrate the ACSNet outperforms existing state-of-the-art WS-TAL methods by a large margin., Comment: Accepted by the 35th AAAI Conference on Artificial Intelligence (AAAI 2021)

Wide dynamic range (WDR) image tone mapping is in high demand in many applications like film production, security monitoring, and photography. It is especially crucial for mobile devices because most of the images taken today are from mobile phones, hence such technology is highly demanded in the consumer market of mobile devices and is essential for a good customer experience. However, high-quality and high-performance WDR image tone mapping implementations are rarely found in the mobile-end. In this paper, we introduce a high performance, mobile-end WDR image tone mapping implementation. It leverages the tone mapping results of multiple receptive fields and calculates a suitable value for each pixel. The utilization of integral image and integral histogram significantly reduce the required computation. Moreover, GPU parallel computation is used to increase the processing speed. The experimental results indicate that our implementation can process a high-resolution WDR image within a second on mobile devices and produce appealing image quality.

In this paper, we present a novel tone mapping algorithm that can be used for displaying wide dynamic range (WDR) images on low dynamic range (LDR) devices. The proposed algorithm is mainly motivated by the logarithmic response and local adaptation features of the human visual system (HVS). HVS perceives luminance differently when under different adaptation levels, and therefore our algorithm uses functions built upon different scales to tone map pixels to different values. Functions of large scales are used to maintain image brightness consistency and functions of small scales are used to preserve local detail and contrast. An efficient method using local variance has been proposed to fuse the values of different scales and to remove artifacts. The algorithm utilizes integral images and integral histograms to reduce computation complexity and processing time. Experimental results show that the proposed algorithm can generate high brightness, good contrast, and appealing images that surpass the performance of many state-of-the-art tone mapping algorithms. This project is available at https://github.com/jieyang1987/ToneMapping-Based-on-Multi-scale-Histogram-Synthesis.

Wide dynamic range (WDR) images contain more scene details and contrast when compared to common images. However, it requires tone mapping to process the pixel values in order to display properly. The details of WDR images can diminish during the tone mapping process. In this work, we address the problem by combining a novel reformulated Laplacian pyramid and deep learning. The reformulated Laplacian pyramid always decompose a WDR image into two frequency bands where the low-frequency band is global feature-oriented, and the high-frequency band is local feature-oriented. The reformulation preserves the local features in its original resolution and condenses the global features into a low-resolution image. The generated frequency bands are reconstructed and fine-tuned to output the final tone mapped image that can display on the screen with minimum detail and contrast loss. The experimental results demonstrate that the proposed method outperforms state-of-the-art WDR image tone mapping methods. The code is made publicly available at https://github.com/linmc86/Deep-Reformulated-Laplacian-Tone-Mapping.

Currently, face detection approaches focus on facial information by varying specific parameters including pose, occlusion, lighting, background, race, and gender. These studies only utilized the information obtained from low dynamic range images, however, face detection in wide dynamic range (WDR) scenes has received little attention. To our knowledge, there is no publicly available WDR database for face detection research. To facilitate and support future face detection research in the WDR field, we propose the first WDR database for face detection, called WDR FACE, which contains a total of 398 16-bit megapixel grayscale wide dynamic range images collected from 29 subjects. These WDR images (WDRIs) were taken in eight specific WDR scenes. The dynamic range of 90% images surpasses 60,000:1, and that of 70% images exceeds 65,000:1. Furthermore, we show the effect of different face detection procedures on the WDRIs in our database. This is done with 25 different tone mapping operators and five different face detectors. We provide preliminary experimental results of face detection on this unique WDR database.

The dynamic range of our normal life can exceeds 120 dB, however, the smart-phone cameras and the conventional digital cameras can only capture a dynamic range of 90 dB, which sometimes leads to loss of details for the recorded image. Now, some professional hardware applications and image fusion algorithms have been devised to take wide dynamic range (WDR), but unfortunately existing devices cannot display WDR image. Tone mapping (TM) thus becomes an essential step for exhibiting WDR image on our ordinary screens, which convert the WDR image into low dynamic range (LDR) image. More and more researchers are focusing on this topic, and give their efforts to design an excellent tone mapping operator (TMO), showing detailed images as the same as the perception that human eyes could receive. Therefore, it is important for us to know the history, development, and trend of TM before proposing a practicable TMO. In this paper, we present a comprehensive study of the most well-known TMOs, which divides TMOs into traditional and machine learning-based category.

Health departments have been deploying text classification systems for the early detection of foodborne illness complaints in social media documents such as Yelp restaurant reviews. Current systems have been successfully applied for documents in English and, as a result, a promising direction is to increase coverage and recall by considering documents in additional languages, such as Spanish or Chinese. Training previous systems for more languages, however, would be expensive, as it would require the manual annotation of many documents for each new target language. To address this challenge, we consider cross-lingual learning and train multilingual classifiers using only the annotations for English-language reviews. Recent zero-shot approaches based on pre-trained multi-lingual BERT (mBERT) have been shown to effectively align languages for aspects such as sentiment. Interestingly, we show that those approaches are less effective for capturing the nuances of foodborne illness, our public health application of interest. To improve performance without extra annotations, we create artificial training documents in the target language through machine translation and train mBERT jointly for the source (English) and target language. Furthermore, we show that translating labeled documents to multiple languages leads to additional performance improvements for some target languages. We demonstrate the benefits of our approach through extensive experiments with Yelp restaurant reviews in seven languages. Our classifiers identify foodborne illness complaints in multilingual reviews from the Yelp Challenge dataset, which highlights the potential of our general approach for deployment in health departments., Comment: Accepted for the 11th International Workshop on Health Text Mining and Information Analysis (LOUHI@EMNLP 2020)