-
LionGuard: Building a Contextualized Moderation Classifier to Tackle Localized Unsafe Content
Authors:
Jessica Foo,
Shaun Khoo
Abstract:
As large language models (LLMs) become increasingly prevalent in a wide variety of applications, concerns about the safety of their outputs have become more significant. Most efforts at safety-tuning or moderation today take on a predominantly Western-centric view of safety, especially for toxic, hateful, or violent speech. In this paper, we describe LionGuard, a Singapore-contextualized moderatio…
▽ More
As large language models (LLMs) become increasingly prevalent in a wide variety of applications, concerns about the safety of their outputs have become more significant. Most efforts at safety-tuning or moderation today take on a predominantly Western-centric view of safety, especially for toxic, hateful, or violent speech. In this paper, we describe LionGuard, a Singapore-contextualized moderation classifier that can serve as guardrails against unsafe LLM outputs. When assessed on Singlish data, LionGuard outperforms existing widely-used moderation APIs, which are not finetuned for the Singapore context, by 14% (binary) and up to 51% (multi-label). Our work highlights the benefits of localization for moderation classifiers and presents a practical and scalable approach for low-resource languages.
△ Less
Submitted 18 July, 2024; v1 submitted 24 June, 2024;
originally announced July 2024.
-
A Comprehensive Survey on Graph Summarization with Graph Neural Networks
Authors:
Nasrin Shabani,
Jia Wu,
Amin Beheshti,
Quan Z. Sheng,
Jin Foo,
Venus Haghighi,
Ambreen Hanif,
Maryam Shahabikargar
Abstract:
As large-scale graphs become more widespread, more and more computational challenges with extracting, processing, and interpreting large graph data are being exposed. It is therefore natural to search for ways to summarize these expansive graphs while preserving their key characteristics. In the past, most graph summarization techniques sought to capture the most important part of a graph statisti…
▽ More
As large-scale graphs become more widespread, more and more computational challenges with extracting, processing, and interpreting large graph data are being exposed. It is therefore natural to search for ways to summarize these expansive graphs while preserving their key characteristics. In the past, most graph summarization techniques sought to capture the most important part of a graph statistically. However, today, the high dimensionality and complexity of modern graph data are making deep learning techniques more popular. Hence, this paper presents a comprehensive survey of progress in deep learning summarization techniques that rely on graph neural networks (GNNs). Our investigation includes a review of the current state-of-the-art approaches, including recurrent GNNs, convolutional GNNs, graph autoencoders, and graph attention networks. A new burgeoning line of research is also discussed where graph reinforcement learning is being used to evaluate and improve the quality of graph summaries. Additionally, the survey provides details of benchmark datasets, evaluation metrics, and open-source tools that are often employed in experimentation settings, along with a detailed comparison, discussion, and takeaways for the research community focused on graph summarization. Finally, the survey concludes with a number of open research challenges to motivate further study in this area.
△ Less
Submitted 3 January, 2024; v1 submitted 13 February, 2023;
originally announced February 2023.
-
Deployment of the Saddle Space Transformation in Tracking the Base of Support
Authors:
Carlo Tiseo,
Ming Jeat Foo,
Kalyana C Veluvolu,
Wei Tech Ang
Abstract:
Balance is the fundamental skill behind human locomotion, and its impairment is the principal indicator of self-perceived disability. Despite significant improvements in balance assessment, there is still large incidence of fall related injuries among elderlies. The Base of Support (BoS) is a popular method for bipedal stability assessment, but its accuracy depends on the accuracy the BoS geometry…
▽ More
Balance is the fundamental skill behind human locomotion, and its impairment is the principal indicator of self-perceived disability. Despite significant improvements in balance assessment, there is still large incidence of fall related injuries among elderlies. The Base of Support (BoS) is a popular method for bipedal stability assessment, but its accuracy depends on the accuracy the BoS geometry measurement. This work presents a method to ease the BoS tracking by the identification of a reference frame that allows to define postural models of the BoS geometry. Although we also propose a geometry based on the geometry determined from centre of pressure range of motion within the foot obtained from literature, this methodology can be used with other models (i.e., the feasible base of support). The model has been tested with 12 healthy subjects, which have been asked to explore their stability in six different postures. The results show that the model can accurate deform the geometry of the BoS to adapt its shape to the different postures, which can remove the necessity of force/torque sensors in some application. Potentially the proposed method can be also applied to describe any posture dependent attribute (e.g., gravitational forces), and it can be also applied to bipedal robots. Therefore, it constitutes a novel mathematical tool that can be deployed to develop both better sensors and models for bipeds. For example, it can be used with the Extrapolated CoM model to evaluate dynamic stability from the body kinematics.
△ Less
Submitted 3 September, 2018; v1 submitted 23 May, 2018;
originally announced May 2018.
-
The Strange Attractor of Bipedal Locomotion and Consequences on Motor Control
Authors:
Carlo Tiseo,
Ming Jeat Foo,
Kalyana C Veluvolu,
Wei Tech Ang
Abstract:
Despite decades of study, the mechanisms that determine human locomotion are still unknown, available models and motor control theories can only partially capture the phenomenon. This is probably the principal cause of the reduced efficacy of lower limbs rehabilitation therapies. Recently, it has been proposed that human locomotion may be planned in the task-space by taking advantage of the gravit…
▽ More
Despite decades of study, the mechanisms that determine human locomotion are still unknown, available models and motor control theories can only partially capture the phenomenon. This is probably the principal cause of the reduced efficacy of lower limbs rehabilitation therapies. Recently, it has been proposed that human locomotion may be planned in the task-space by taking advantage of the gravitational pull acting on the Centre of Mass (CoM) that we have used to formulate a task-space planner for straight locomotion at a constant speed. The proposed model represents the CoM transversal trajectory as simple harmonic oscillator moving forward at a constant speed. On the other hand, the vertical trajectory of the CoM is controlled through the ankle strategies. Our solution is composed of closed-form equations which can plan human-like trajectories for both the CoM and the foot swing. The model output can be seen as the optimal trajectory determined based on the average behaviour of 12 healthy subjects walking at three self-selected speeds. Furthermore, the planner formulation is compatible with an extended formulation of the Passive Motion Paradigm which enables us to design a hierarchical architecture of semi-autonomous controllers. The final architecture can also describe the motor primitives as a particular case of dynamic primitives, shows strong parallels with the nervous system organization, and is compatible with the optimal feedback control theory.
△ Less
Submitted 18 October, 2019; v1 submitted 9 February, 2018;
originally announced February 2018.
