检索结果-内蒙古大学图书馆

Comfort with varying levels of human supervision in self-driving cars: Determining factors in Europe

International Journal of Transportation Science and Technology 2023年第3期12卷 809-821页

作者： Daniel Kaszas Adam Charles Roberts ETH Zürich ZürichSwitzerland Singapore-ETH Centre Future Resilient Systems ProgrammeCREATE campusSingapore

While numerous studies have investigated attitudes towards self-driving cars in general,less research attention has been focused on individuals’comfort with the presence(or absence)of third-party human supervision of this automation,and its potential *** the present study we perform a secondary analysis of pre-existing data from The European Commission’s Eurobarometer 92.1,a large-scale European survey(n=27565)of expectations and concerns of connected and automated *** comparing responses to three levels of human supervision in self-driving cars,we aim to identify changes in the importance of predictors of comfort with *** find considerable heterogeneity in both individual attitudes,as well as in country-level attitudes in our descriptive *** find a trend of decreasing comfort as external human supervision is reduced,although this effect differs between *** then investigate potential drivers of self-reported comfort with varying levels of external human supervision in a regression *** differences get stronger with decreasing supervision,suggesting a possible resolution to conflicting evidence in previous *** this,we fit an ordinal random forest model to derive variable importance metrics,which enable us to compare the changing role predictor variables might play in shaping self-reported comfort,depending on varying levels of third-party *** privacy is highlighted as an important variable,regardless of level of *** findings provide confirmation for previous literature in a large sample,while also uncovering a number of novel associations,providing guidance for future policy-making and research efforts.

关键词： Human supervision Third-party supervision Technology acceptance Autonomous vehicles

来源：评论

学校读者我要写书评

暂无评论

EDEntail: An Entailment-based Few-shot Text Classification with Extensional Definition

EDEntail: An Entailment-based Few-shot Text Classification w...

引用

2024 Findings of the Association for Computational Linguistics: NAACL 2024

作者： Zixiao, Zhu Qian, Junlang Feng, Zijian Zhou, Hanzhang Mao, Kezhi Institute of Catastrophe Risk Management Interdisciplinary Graduate Programme Nanyang Technological University Singapore School of Electrical and Electronic Engineering Nanyang Technological University Singapore Future Resilient Systems Programme Singapore-ETH Centre CREATE Campus Singapore

ISBN: (纸本)9798891761193

Few-shot text classification has seen significant advancements, particularly with entailment-based methods, which typically use either class labels or intensional definitions of class labels in hypotheses for label semantics expression. In this paper, we propose EDEntail, a method that employs extensional definition (EDef) of class labels in hypotheses, aiming to express the semantics of class labels more explicitly. To achieve the above goal, we develop an algorithm to gather and select extensional descriptive words of class labels and then order and format them into a sequence to form hypotheses. Our method has been evaluated and compared with state-of-the-art models on five classification datasets. The results demonstrate that our approach surpasses the supervised-learning methods and prompt-based methods under the few-shot setting, which underlines the potential of using an extensional definition of class labels for entailment-based few-shot text classification. Our code is available at https://***/MidiyaZhu/EDEntail. © 2024 Association for Computational Linguistics.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

UNVEILING AND MANIPULATING PROMPT INFLUENCE IN LARGE LANGUAGE MODELS 12

UNVEILING AND MANIPULATING PROMPT INFLUENCE IN LARGE LANGUAG...

引用

12th International Conference on Learning Representations, ICLR 2024

作者： Feng, Zijian Zhou, Hanzhang Zhu, Zixiao Qian, Junlang Mao, Kezhi Institute of Catastrophe Risk Management Interdisciplinary Graduate Programme Nanyang Technological University Singapore School of Electrical and Electronic Engineering Nanyang Technological University Singapore Future Resilient Systems Programme Singapore-ETH Centre CREATE Campus Singapore

Prompts play a crucial role in guiding the responses of Large Language Models (LLMs). However, the intricate role of individual tokens in prompts, known as input saliency, in shaping the responses remains largely underexplored. Existing saliency methods either misalign with LLM generation objectives or rely heavily on linearity assumptions, leading to potential inaccuracies. To address this, we propose Token Distribution Dynamics (TDD), a simple yet effective approach to unveil and manipulate the role of prompts in generating LLM outputs. TDD leverages the robust interpreting capabilities of the language model head (LM head) to assess input saliency. It projects input tokens into the embedding space and then estimates their significance based on distribution dynamics over the vocabulary. We introduce three TDD variants: forward, backward, and bidirectional, each offering unique insights into token relevance. Extensive experiments reveal that the TDD surpasses state-of-the-art baselines with a big margin in elucidating the causal relationships between prompts and LLM outputs. Beyond mere interpretation, we apply TDD to two prompt manipulation tasks for controlled text generation: zero-shot toxic language suppression and sentiment steering. Empirical results underscore TDD's proficiency in identifying both toxic and sentimental cues in prompts, subsequently mitigating toxicity or modulating sentiment in the generated content. © 2024 12th International Conference on Learning Representations, ICLR 2024. All rights reserved.

关键词： Zero-shot learning

来源：评论

学校读者我要写书评

暂无评论

Intrinsic Complexity: Quantifying Task Complexity for Deep Learning-Based City-Wide Traffic Prediction

SSRN

引用

SSRN 2024年

作者： Kumar, Nishant Martin, Henry Raubal, Martin Singapore-ETH Centre Future Resilient Systems Programme CREATE Campus 138602 Singapore ETH Zürich Zurich Switzerland

Deep learning models have been shown to capture the non-linear spatiotemporal dynamics of city-wide traffic forecasting effectively. Evidence of varying deep learning model performance between different cities, different prediction horizons, different scales, specific city regions, and during particular hours of the day abounds in the literature on deep learning-based traffic prediction, yet a unified metric to quantify the complexity of different prediction tasks does not exist. Here, we extend the notion of Critical Samples of deep learning models to propose a novel computationally efficient model complexity metric. Then, we propose an Intrinsic Complexity ($IC$) metric that estimates the underlying complexity of specific prediction tasks using training data. The metric is validated through systematic experimentation using traffic volume data from three cities and across various traffic prediction tasks with varying lengths of input and output windows, prediction horizons, and scales. $IC$ is found to exhibit a linear relationship with the validation error, suggesting it can estimate validation errors before any modelling is attempted. Additionally, we demonstrate how $IC$ can quantify the complexity of different sub-tasks of a given traffic prediction task, thus providing researchers with a priori knowledge of spatial and temporal regions of high complexity in a city-wide traffic prediction task. © 2024, The Authors. All rights reserved.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

The Role of Worker Satisfaction in the Underground Environment with Job Satisfaction and Turnover Intent 18th

The Role of Worker Satisfaction in the Underground Environme...

引用

18th Conference of the Associated Research Centers for the Urban Underground Space, ACUUS 2023

作者： Roberts, Adam C. Soh, Chee-Kiong Yap, Hui Shan Car, Josip Kwok, Kian-Woon Christopoulos, George Singapore-ETH Centre Future Resilient Systems Programme Singapore School of Civil Engineering Southeast University Nanjing China Nanyang Business School Nanyang Technological University Singapore Singapore School of Life Course & Population Sciences King’s College London London United Kingdom School of Humanities and Social Sciences Nanyang Technological University Singapore Singapore

ISBN: (纸本)9789819712564

The work environment plays an important role in shaping the job satisfaction and retention of workers. Previous studies have suggested a link between underground work and worker satisfaction, with underground workers being less satisfied with their workplace. This study examines the relationship between worker satisfaction with work Indoor Environment Quality (IEQ), job satisfaction, and turnover intent in underground and aboveground workplaces. A questionnaire was administered to 401 workers in underground and aboveground workplaces in Singapore. The questionnaire focused on three main factors: IEQ satisfaction, job satisfaction, and turnover intent. Structural equation modelling was used to investigate the relationship between these factors, specifically, whether IEQ satisfaction predicted job satisfaction and turnover intent, and whether there were any differences between underground and aboveground workplaces or worker roles. The results indicated that worker satisfaction with IEQ had a significant positive effect on job satisfaction and a significant negative effect on turnover intent. While there was no significant effect of work location (above or underground), subgroup analyses showed that underground workers showed a stronger link between IEQ satisfaction and job satisfaction. In addition, differences were found between worker roles. Office and control room workers showed significant effects of IEQ satisfaction on job satisfaction and turnover intent. In comparison, workshop workers did not show significant effects. This study highlights the importance of worker satisfaction with their environment in shaping job satisfaction and the willingness to leave a job. The results suggest that companies with underground workspaces should prioritize creating a comfortable indoor environment to improve worker job satisfaction and reduce turnover. The findings also suggest that the effect of the work environment on worker attitudes and intentions is influenced by the w

关键词： Natural environment

来源：评论

学校读者我要写书评

暂无评论

Unveiling and Manipulating Prompt Influence in Large Language Models

arXiv

引用

arXiv 2024年

Prompts play a crucial role in guiding the responses of Large Language Models (LLMs). However, the intricate role of individual tokens in prompts, known as input saliency, in shaping the responses remains largely underexplored. Existing saliency methods either misalign with LLM generation objectives or rely heavily on linearity assumptions, leading to potential inaccuracies. To address this, we propose Token Distribution Dynamics (TDD), a simple yet effective approach to unveil and manipulate the role of prompts in generating LLM outputs. TDD leverages the robust interpreting capabilities of the language model head (LM head) to assess input saliency. It projects input tokens into the embedding space and then estimates their significance based on distribution dynamics over the vocabulary. We introduce three TDD variants: forward, backward, and bidirectional, each offering unique insights into token relevance. Extensive experiments reveal that the TDD surpasses state-of-the-art baselines with a big margin in elucidating the causal relationships between prompts and LLM outputs. Beyond mere interpretation, we apply TDD to two prompt manipulation tasks for controlled text generation: zero-shot toxic language suppression and sentiment steering. Empirical results underscore TDD’s proficiency in identifying both toxic and sentimental cues in prompts, subsequently mitigating toxicity or modulating sentiment in the generated content. Copyright © 2024, The Authors. All rights reserved.

关键词： Computational linguistics

来源：评论

学校读者我要写书评

暂无评论

FreeCtrl: Constructing Control Centers with Feedforward Layers for Learning-Free Controllable Text Generation

arXiv

引用

arXiv 2024年

作者： Feng, Zijian Zhou, Hanzhang Zhu, Zixiao Mao, Kezhi Institute of Catastrophe Risk Management Interdisciplinary Graduate Programme Nanyang Technological University Singapore School of Electrical and Electronic Engineering Nanyang Technological University Singapore Future Resilient Systems Programme Singapore-ETH Centre CREATE Campus Singapore

Controllable text generation (CTG) seeks to craft texts adhering to specific attributes, traditionally employing learning-based techniques such as training, fine-tuning, or prefix-tuning with attribute-specific datasets. These approaches, while effective, demand extensive computational and data resources. In contrast, some proposed learning-free alternatives circumvent learning but often yield inferior results, exemplifying the fundamental machine learning trade-off between computational expense and model efficacy. To overcome these limitations, we propose FreeCtrl, a learning-free approach that dynamically adjusts the weights of selected feedforward neural network (FFN) vectors to steer the outputs of large language models (LLMs). FreeCtrl hinges on the principle that the weights of different FFN vectors influence the likelihood of different tokens appearing in the output. By identifying and adaptively adjusting the weights of attribute-related FFN vectors, FreeCtrl can control the output likelihood of attribute keywords in the generated content. Extensive experiments on single- and multi-attribute control reveal that the learning-free FreeCtrl outperforms other learning-free and learning-based methods, successfully resolving the dilemma between learning costs and model performance. Copyright © 2024, The Authors. All rights reserved.

关键词： Feedforward neural networks

来源：评论

学校读者我要写书评

暂无评论

Logit Separability-Driven Samples and Multiple Class-Related Words Selection for Advancing In-Context Learning

arXiv

引用

arXiv 2024年

作者： Zixiao, Zhu Feng, Zijian Zhou, Hanzhang Qian, Junlang Mao, Kezhi Institute of Catastrophe Risk Management Interdisciplinary Graduate Programme Nanyang Technological University Singapore School of Electrical and Electronic Engineering Nanyang Technological University Singapore Future Resilient Systems Programme Singapore-ETH Centre CREATE campus Singapore

Effective organization of in-context learning (ICL) demonstrations is key to improving the quality of large language model (LLM) responses. To create better sample-label pairs that instruct LLM understanding, we introduce logit separability, a criterion to assess the clarity of both samples and class-related words at the logit level. This facilitates the optimization of sample and label selection, enhancing the precision of information provided in ICL demonstrations. Additionally, we find that incorporating multiple class-related words for each sample, rather than relying on a single class name, improves performance by offering a broader range of label information. Building on these insights, we propose LICL, a logit separability-based method that jointly organizes samples and integrates multiple class-related words into each sample-label pair. Evaluations across seven classification datasets show that this approach significantly improves ICL performance by providing clearer instructions and richer label information. Copyright © 2024, The Authors. All rights reserved.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

UNCERTAINTY QUANTIFICATION FOR IMAGE-BASED TRAFFIC PREDICTION ACROSS CITIES

arXiv

引用

arXiv 2023年

作者： Timans, Alexander Wiedemann, Nina Kumar, Nishant Hong, Ye Raubal, Martin Amsterdam Machine Learning Lab University of Amsterdam Netherlands Institute of Cartography and Geoinformation ETH Zurich Switzerland Future Resilient Systems Programme Singapore-ETH Centre Singapore

Despite the strong predictive performance of deep learning models for traffic prediction, their widespread deployment in real-world intelligent transportation systems has been restrained by a lack of interpretability. Uncertainty quantification (UQ) methods provide an approach to induce probabilistic reasoning, improve decision-making and enhance model deployment potential. To gain a comprehensive picture of the usefulness of existing UQ methods for traffic prediction and the relation between obtained uncertainties and city-wide traffic dynamics, we investigate their application to a large-scale image-based traffic dataset spanning multiple cities and time periods. We compare two epistemic and two aleatoric UQ methods on both temporal and spatio-temporal transfer tasks, and find that meaningful uncertainty estimates can be recovered. We further demonstrate how uncertainty estimates can be employed for unsupervised outlier detection on changes in city traffic dynamics. We find that our approach can capture both temporal and spatial effects on traffic behaviour in a representative case study for the city of Moscow. Our work presents a further step towards boosting uncertainty awareness in traffic prediction tasks, and aims to highlight the value contribution of UQ methods to a better understanding of city traffic dynamics. The code for our experiments is publicly available at https://***/alextimans/traffic4cast-uncertainty. Copyright © 2023, The Authors. All rights reserved.

关键词： Anomaly detection

来源：评论

学校读者我要写书评

暂无评论

SWOT Analysis of Underground Work Spaces 18th

SWOT Analysis of Underground Work Spaces

引用

18th Conference of the Associated Research Centers for the Urban Underground Space, ACUUS 2023

作者： Soh, Chee-Kiong Roberts, Adam C. Marimuther, Vicknaeshwari Car, Josip Kwok, Kian-Woon Christopoulos, George School of Civil Engineering Southeast University Nanjing China Future Resilient Systems Programme Singapore-ETH Centre Singapore Singapore Singapore School of Civil and Environmental Engineering Nanyang Technological University Singapore Singapore Singapore School of Life Course & Population Sciences King’s College London London United Kingdom School of Humanities and Social Sciences Nanyang Technological University Singapore Singapore Nanyang Business School Nanyang Technological University Singapore Singapore Singapore

ISBN: (纸本)9789819712564

Underground work spaces (UWS) are a potential alternative to traditional aboveground office buildings. In addition to potential benefits for tenants, such as being more energy efficient and controllable, UWS may also contribute to a country’s resilience by providing shelter from disasters. This study analyses UWS through the Strength, Weaknesses, Opportunities, and Threats (SWOT) framework. Primary data were collected through questionnaires and interviews with workers in UWS and professionals in fields related to UWS. Secondary data were collected from a review of published literature. The SWOT analysis was conducted by identifying both internal and external factors that may affect UWS, analysing their potential impact, and developing an action plan to improve the acceptance of UWS and mitigate negative impacts. The SWOT analysis reveals that the strengths of UWS include more efficient land use, unique design, and a sense of community. However, UWS often suffer from a lack of natural light and external views through a lack of windows. These could affect employee productivity and morale. Opportunities for UWS include new uses for existing space, the ability to design for connectivity, and protection from climate change. Perceptions of potential users and the general public were highlighted as the biggest threat to the adoption of UWS, alongside technical difficulties in maintenance resulting in poor environmental quality. Our SWOT analysis provides valuable insights for both academics and industry. While UWS have some limitations and potential drawbacks, there are also clear advantages and ways to mitigate limitations and even turn them into strengths. As the most visible threats to UWS adoption relate to potential workers and occupiers, more effort should be concentrated on presenting the benefits of UWS to the public and identifying ways to improve the UWS experience to change worker perceptions. The findings from this study can inform future decisions regarding th

关键词： Economic and social effects

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：