Abstract

Traditional evaluation metrics like BLEU and ROUGE fall short when capturing the nuanced qualities of generated text, particularly when there is no single ground truth. In this paper, we explore the potential of Large Language Models (LLMs), specifically Google Gemini 1, to serve as automatic evaluators for non-standardized metrics in summarization and dialog-based tasks. We conduct experiments across multiple prompting strategies to examine how LLMs fare as quality evaluators when compared with human judgments on the SummEval and USR datasets, asking the model to generate both a score as well as a justification for the score. Furthermore, we explore the robustness of the LLM evaluator by using perturbed inputs. Our findings suggest that while LLMs show promise, their alignment with human evaluators is limited, they are not robust against perturbations and significant improvements are required for their standalone use as reliable evaluators for subjective metrics.

Abstract

The integration of Vision-Language Models (VLMs) into various applications has highlighted the importance of evaluating these models for inherent biases, especially along gender and racial lines. Traditional bias assessment methods in VLMs typically rely on accuracy metrics, assessing disparities in performance across different demographic groups. These methods, however, often overlook the impact of the model’s disabilities, like lack spatial reasoning, which may skew the bias assessment. In this work, we propose an approach that systematically examines how current bias evaluation metrics account for the model’s limitations. We introduce two methods that circumvent these disabilities by integrating spatial guidance from textual and visual modalities. Our experiments aim to refine bias quantification by effectively mitigating the impact of spatial reasoning limitations, offering a more accurate assessment of biases in VLMs.

Abstract

The implementation of Internet of Things (IoT) as Cyber-Physical Systems (CPS) is essential for the advancement of smart healthcare within a highly digital environment. However, the development of medical CPS (MCPS) poses various challenges, particularly in terms of dependability and security. The complex nature of medical IoT components encompasses both computational and physical aspects including secure reliability. To mitigate these issues, this article introduces a cyber engineering, an innovative approach, that combines principles from complex system design in order to establish a systematic framework for combining the cyber and physical layers of IoT and MCPS together. This integration aims to provide an unified system designing approach that can produce more secure and robust systems. By embracing the cyber engineering ideology, the healthcare industry can overcome reliability challenges and pave the way for the development of secure and dependable smart healthcare solutions, while ensuring security by design perspective. The approach also provides ideal future possibilities to the upcoming researchers.

Abstract

This paper contributes to the expanding research on toxicity in online games, exploring the relationship between game design and player behaviours in India. Using an ethnographic approach and in-depth interviews with gamers, we capture firsthand accounts of toxic gaming experiences, delving into game design, game mechanics, and player interactions intersecting to shape player behaviours. While previous research on toxicity in gaming environments has largely focused on the role of game design, aspects of gender and, to a lesser extent, race, primarily within the context of the Global North, our study shifts the focus to India and, by extension, the Global South. We seek to uncover broader dimensions of gaming toxicity that go beyond design and gender. Our analysis examines the underlying factors contributing to the prevalence and persistence of toxic behaviours in Indian gaming environments by highlighting the influence of cultural and contextual elements in producing toxic behaviours. Expanding the scope of research to include these overlooked aspects, our study aims to provide a more comprehensive understanding of the complexities of toxicity in online gaming spaces within the Indian context.

Abstract

Analysis of foot pressure, also known as plantar pressure analysis, plays a pivotal role in biomedical assessments related to posture and gait analysis. Extensive research has been conducted on leveraging this technique for clinical purposes, leading to the development of flexible pressure sensors. In this letter, we present the use of in-shoe flexible pressure sensor array for determining the nature of the walking surface. The sensor system is fabricated using eight low-cost and robust, in-shoe pressure sensors that leverage the piezoresistivity of velostat. The sensor array was characterized for four different surface types. Random Forest (RF) algorithm was used to classify the surfaces with 86% accuracy. Based on this analysis, we propose a novel method for analyzing various surfaces based on their attributes such as firmness, rigidity, and penetrability. Such a device can be used for ascertaining surface characteristics after construction, or playing surfaces in a stadium.

Abstract

Bidding in a periodic double auction (PDA) is challenging due to its sequential nature, where one needs to consider current as well as future auctions to decide the bids. Monte-Carlo Tree Search (MCTS), which is a state-of-the-art online planning algorithm for tackling sequential problems, seems a perfect fit for bidding in PDAs. However, the success stories of MCTS are largely limited to discrete action spaces, and its efficacy diminishes when dealing with continuous actions. Conventional methods often resort to overly simplistic discretizations that limit exploration and fail to provide valuable insights into unexplored actions. In this work, we propose a novel bidding strategy for PDAs, Regression-MCTS, that is built upon MCTS for a continuous action space of bid prices. Unlike conventional methods, our novel MCTS method leverages information obtained from explored actions to enhance the understanding of the larger action set within the continuous domain to place bids in the auctions, thus generalizing the information about action quality between a wider action space for faster learning. To test the efficacy of our proposed method, we design an efficient PDA simulator that closely resembles real-world PDAs. Our analysis verifies that the increase in the number of rollouts improves its performance. Furthermore, our experimental results demonstrate that our approach outperforms existing MCTS-based bidding strategies and the majority of state-of-the-art PDA bidding strategies, showcasing its superior performance in PDAs.

Abstract

Decision Transformers have demonstrated impressive performance in offline reinforcement learning, but their complex internal mechanisms remain challenging to interpret. This paper presents a comprehensive analysis of Decision Transformers(DT) trained on various MuJoCo environments using different interpretability techniques. Our analysis includes: (a) positional encoding (PE) analysis to assess the impact of temporal information on performance in different types of environments, (b) return-to-go (RTG) analysis to explore the variability in achieving specific returns, (c) embedding analysis revealing the models' ability to learn detailed representations of the environments, (d) attention visualizations to understand the role of attention in behaviour guidance, and (e) perturbation studies to test robustness to input noise. Our analysis reveals the extent to which positional information is necessary for achieving high performance in different MuJoCo environments and provides insights into the model's adaptability to varying task requirements. In addition, the embedding analysis shows hierarchical abstractions captured by the model across layers. Through these analyses, we provide novel insights into the decision-making process of transformers in continuous control domains. These insights are pivotal for designing more interpretable and reliable decision-making systems and deepen our comprehension of the capabilities and limitations of DT in complex tasks.

Abstract

Biological processes like growth, aging, and disease progression are generally studied with follow-up scans taken at different time points, i.e., image time series (TS) based analysis. Image TS represents the evolution of anatomy over time, but different anatomies may have different structural characteristics and temporal paths. Therefore, separating the time-dependent path difference and time-independent basic anatomy/shape changes is important when comparing two image TS to understand the causes of the observed differences better. A method to untangle and quantify the path and shape difference between the TS is presented in this paper. The proposed method is evaluated with simulated and adult and fetal neuro templates. Results show that the metric can separate and quantify the path and shape differences between TS.

Abstract

This article presents ‘Teledrive’, a telepresence robotic system with embodied AI features that empowers an operator to navigate the telerobot in any unknown remote place with minimal human intervention. We conceive Teledrive in the context of democratizing remote ‘care-giving’ for elderly citizens as well as for isolated patients, affected by contagious diseases. In particular, this paper focuses on the problem of navigating to a rough target area (like ‘bedroom’ or ‘kitchen’) rather than pre-specified point destinations. This ushers in a unique ‘AreaGoal’ based navigation feature, which has not been explored in depth in the contemporary solutions. Further, we describe an edge computing-based software system built on a WebRTC-based communication framework to realize the aforementioned scheme through an easy-to-use speech-based human-robot interaction. Moreover, to enhance the ease of operation for the remote caregiver, we incorporate a ‘person following’ feature, whereby a robot follows a person on the move in its premises as directed by the operator. Moreover, the system presented is loosely coupled with specific robot hardware, unlike the existing solutions. We have evaluated the efficacy of the proposed system through baseline experiments, user study, and real-life deployment.

Abstract

Assistive agents performing household tasks such as making the bed or cooking breakfast often compute and execute actions that accomplish one task at a time. However, efficiency can be improved by anticipating upcoming tasks and computing an action sequence that jointly achieves these tasks. State-of-the-art methods for task anticipation use data-driven deep networks and Large Language Models (LLMs), but they do so at the level of high-level tasks and/or require many training examples. Our framework leverages the generic knowledge of LLMs through a small number of prompts to perform high-level task anticipation, using the anticipated tasks as goals in a classical planning system to compute a sequence of finer-granularity actions that jointly achieve these goals. We ground and evaluate our framework’s abilities in realistic scenarios in the VirtualHome environment and demonstrate a 31% reduction in execution time compared with a system that does not consider upcoming tasks.