Public figures receive a disproportionate amount of abuse on social media, impacting their active participation in public life. Automated systems can identify abuse at scale but labelling training data is expensive, complex and potentially harmful. So, it is desirable that systems are efficient and generalisable, handling both shared and specific aspects of online abuse. We explore the dynamics of cross-group text classification in order to understand how well classifiers trained on one domain or demographic can transfer to others, with a view to building more generalisable abuse classifiers. We fine-tune language models to classify tweets targeted at public figures across DOmains (sport and politics) and DemOgraphics (women and men) using our novel DODO dataset, containing 28,000 labelled entries, split equally across four domain-demographic pairs. We find that (i) small amounts of diverse data are hugely beneficial to generalisation and model adaptation; (ii) models transfer more easily across demographics but models trained on cross-domain data are more generalisable; (iii) some groups contribute more to generalisability than others; and (iv) dataset similarity is a signal of transferability.
相關內容
The unparalleled performance of closed-sourced ChatGPT has sparked efforts towards its democratization, with notable strides made by leveraging real user and ChatGPT conversations, as evidenced by Vicuna. However, due to challenges in gathering conversations involving human participation, current endeavors like Baize and UltraChat aim to automatically generate conversational data. They primarily rely on ChatGPT conducting roleplay to simulate human behaviors based on instructions rather than genuine learning from humans, resulting in limited scope, diminished diversity, and an absence of genuine multi-round conversational dynamics. To address the above issues, we target human questions extracted from genuine human-machine conversations as a learning goal and train a user simulator called `Socratic' to produce a high-quality human-centric synthetic conversation dataset. Subsequently, this dataset was used to train our assistant model, named `PlatoLM'. Experimentally, PlatoLM outpaces baseline models in both Vicuna-Bench and MT-Bench by pairwise comparison when considering equivalent training set sizes, and manual evaluation also shows that our model is highly competitive. Impressively, when fine-tuned with the latest LLaMA 2 model, PlatoLM achieves the SOTA performance among 7B models (including LLaMA-2-7B-chat and Vicuna-7B) in MT-Bench benchmark and in Alpaca-Eval benchmark, it ranks second among 7B models, even beating some larger scale models (including LLaMA-2-13B-chat and GPT-3.5). Further in-depth analysis demonstrates the scalability and transferability of our approach. The code is available at //github.com/FreedomIntelligence/PlatoLM.
While avatars have grown in popularity in social settings, their use in the workplace is still debatable. We conducted a large-scale survey to evaluate knowledge worker sentiment towards avatars, particularly the effects of realism on their acceptability for work meetings. Our survey of 2509 knowledge workers from multiple countries rated five avatar styles for use by managers, known colleagues and unknown colleagues. In all scenarios, participants favored higher realism, but fully realistic avatars were sometimes perceived as uncanny. Less realistic avatars were rated worse when interacting with an unknown colleague or manager, as compared to a known colleague. Avatar acceptability varied by country, with participants from the United States and South Korea rating avatars more favorably. We supplemented our quantitative findings with a thematic analysis of open-ended responses to provide a comprehensive understanding of factors influencing work avatar choices. In conclusion, our results show that realism had a significant positive correlation with acceptability. Non-realistic avatars were seen as fun and playful, but only suitable for occasional use.
Disagreements over peering fees have risen to the level of potential government regulation. ISPs assert that content providers should pay them based on the volume of downstream traffic. Transit providers and content providers assert that consumers have already paid ISPs to transmit the content they request and that peering agreements should be settlement-free. Our goal is to determine the fair payment between an ISP and an interconnecting network. We consider fair cost sharing between two Tier-1 ISPs, and derive the peering fee that equalizes their net backbone transportation costs. We then consider fair cost sharing between an ISP and a transit provider. We derive the peering fee that equalizes their net backbone transportation costs, and illustrate how it depends on the traffic ratio and the amount of localization of that content. Finally, we consider the fair peering fee between an ISP and a content provider. We derive the peering fee that results in the same net cost to the ISP, and illustrate how the peering fee depends on the number of interconnection points and the amount of localization of that content. We dispense with the ISP argument that it should be paid regardless of the amount of localization of content.
Practitioners and academics have long appreciated the benefits of covariate balancing when they conduct randomized experiments. For web-facing firms running online A/B tests, however, it still remains challenging in balancing covariate information when experimental subjects arrive sequentially. In this paper, we study an online experimental design problem, which we refer to as the "Online Blocking Problem." In this problem, experimental subjects with heterogeneous covariate information arrive sequentially and must be immediately assigned into either the control or the treated group. The objective is to minimize the total discrepancy, which is defined as the minimum weight perfect matching between the two groups. To solve this problem, we propose a randomized design of experiment, which we refer to as the "Pigeonhole Design." The pigeonhole design first partitions the covariate space into smaller spaces, which we refer to as pigeonholes, and then, when the experimental subjects arrive at each pigeonhole, balances the number of control and treated subjects for each pigeonhole. We analyze the theoretical performance of the pigeonhole design and show its effectiveness by comparing against two well-known benchmark designs: the match-pair design and the completely randomized design. We identify scenarios when the pigeonhole design demonstrates more benefits over the benchmark design. To conclude, we conduct extensive simulations using Yahoo! data to show a 10.2% reduction in variance if we use the pigeonhole design to estimate the average treatment effect.
The recent progress in large language models (LLMs), especially the invention of chain-of-thought prompting, has made it possible to automatically answer questions by stepwise reasoning. However, when faced with more complicated problems that require non-linear thinking, even the strongest LLMs make mistakes. To address this, we explore whether LLMs are able to recognize errors in their own step-by-step reasoning, without resorting to external resources. To this end, we propose SelfCheck, a general-purpose zero-shot verification schema for recognizing such errors. We then use the results of these checks to improve question-answering performance by conducting weighted voting on multiple solutions to the question. We test SelfCheck on three datasets (GSM8K, MathQA, and MATH) and find that it successfully recognizes errors and, in turn, increases final answer accuracies.
Diagnosing lung inflammation, particularly pneumonia, is of paramount importance for effectively treating and managing the disease. Pneumonia is a common respiratory infection caused by bacteria, viruses, or fungi and can indiscriminately affect people of all ages. As highlighted by the World Health Organization (WHO), this prevalent disease tragically accounts for a substantial 15% of global mortality in children under five years of age. This article presents a comparative study of the Inception-ResNet deep learning model's performance in diagnosing pneumonia from chest radiographs. The study leverages Mendeleys chest X-ray images dataset, which contains 5856 2D images, including both Viral and Bacterial Pneumonia X-ray images. The Inception-ResNet model is compared with seven other state-of-the-art convolutional neural networks (CNNs), and the experimental results demonstrate the Inception-ResNet model's superiority in extracting essential features and saving computation runtime. Furthermore, we examine the impact of transfer learning with fine-tuning in improving the performance of deep convolutional models. This study provides valuable insights into using deep learning models for pneumonia diagnosis and highlights the potential of the Inception-ResNet model in this field. In classification accuracy, Inception-ResNet-V2 showed superior performance compared to other models, including ResNet152V2, MobileNet-V3 (Large and Small), EfficientNetV2 (Large and Small), InceptionV3, and NASNet-Mobile, with substantial margins. It outperformed them by 2.6%, 6.5%, 7.1%, 13%, 16.1%, 3.9%, and 1.6%, respectively, demonstrating its significant advantage in accurate classification.
Reasoning is a fundamental aspect of human intelligence that plays a crucial role in activities such as problem solving, decision making, and critical thinking. In recent years, large language models (LLMs) have made significant progress in natural language processing, and there is observation that these models may exhibit reasoning abilities when they are sufficiently large. However, it is not yet clear to what extent LLMs are capable of reasoning. This paper provides a comprehensive overview of the current state of knowledge on reasoning in LLMs, including techniques for improving and eliciting reasoning in these models, methods and benchmarks for evaluating reasoning abilities, findings and implications of previous research in this field, and suggestions on future directions. Our aim is to provide a detailed and up-to-date review of this topic and stimulate meaningful discussion and future work.
In pace with developments in the research field of artificial intelligence, knowledge graphs (KGs) have attracted a surge of interest from both academia and industry. As a representation of semantic relations between entities, KGs have proven to be particularly relevant for natural language processing (NLP), experiencing a rapid spread and wide adoption within recent years. Given the increasing amount of research work in this area, several KG-related approaches have been surveyed in the NLP research community. However, a comprehensive study that categorizes established topics and reviews the maturity of individual research streams remains absent to this day. Contributing to closing this gap, we systematically analyzed 507 papers from the literature on KGs in NLP. Our survey encompasses a multifaceted review of tasks, research types, and contributions. As a result, we present a structured overview of the research landscape, provide a taxonomy of tasks, summarize our findings, and highlight directions for future work.
Deep neural models in recent years have been successful in almost every field, including extremely complex problem statements. However, these models are huge in size, with millions (and even billions) of parameters, thus demanding more heavy computation power and failing to be deployed on edge devices. Besides, the performance boost is highly dependent on redundant labeled data. To achieve faster speeds and to handle the problems caused by the lack of data, knowledge distillation (KD) has been proposed to transfer information learned from one model to another. KD is often characterized by the so-called `Student-Teacher' (S-T) learning framework and has been broadly applied in model compression and knowledge transfer. This paper is about KD and S-T learning, which are being actively studied in recent years. First, we aim to provide explanations of what KD is and how/why it works. Then, we provide a comprehensive survey on the recent progress of KD methods together with S-T frameworks typically for vision tasks. In general, we consider some fundamental questions that have been driving this research area and thoroughly generalize the research progress and technical details. Additionally, we systematically analyze the research status of KD in vision applications. Finally, we discuss the potentials and open challenges of existing methods and prospect the future directions of KD and S-T learning.
Image segmentation is still an open problem especially when intensities of the interested objects are overlapped due to the presence of intensity inhomogeneity (also known as bias field). To segment images with intensity inhomogeneities, a bias correction embedded level set model is proposed where Inhomogeneities are Estimated by Orthogonal Primary Functions (IEOPF). In the proposed model, the smoothly varying bias is estimated by a linear combination of a given set of orthogonal primary functions. An inhomogeneous intensity clustering energy is then defined and membership functions of the clusters described by the level set function are introduced to rewrite the energy as a data term of the proposed model. Similar to popular level set methods, a regularization term and an arc length term are also included to regularize and smooth the level set function, respectively. The proposed model is then extended to multichannel and multiphase patterns to segment colourful images and images with multiple objects, respectively. It has been extensively tested on both synthetic and real images that are widely used in the literature and public BrainWeb and IBSR datasets. Experimental results and comparison with state-of-the-art methods demonstrate that advantages of the proposed model in terms of bias correction and segmentation accuracy.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191