Title
Topic
-
‘Will the Prince Get True Love’s Kiss? On the Model Sensitivity to Gender Perturbation Over Fairytale Texts’
“Recent studies show that traditional fairytales are rife with harmful gender biases. To help mitigate these gender biases in fairytales, this work aims to assess learned biases of language models by evaluating their robustness against gender perturbations. Specifically, we focus on Question Answering (QA) tasks in fairytales. Using counterfactual data augmentation to the FairytaleQA dataset, we evaluate model robustness against swapped gender character information, and then mitigate learned biases by introducing counterfactual gender stereotypes during training time.” Find the paper and full list of authors at ArXiv.
-
‘Sample Complexity of Opinion Formation on Networks’
“Consider public health officials aiming to spread awareness about a new vaccine in a community interconnected by a social network. How can they distribute information with minimal resources, ensuring community-wide understanding that aligns with the actual facts? … In this paper, we initialize the study of sample complexity in opinion formation to solve this problem. … Intriguingly, we discover optimal strategies often allocate samples inversely to the degree, hinting at vital policy implications.” Find the paper and full list of authors at ArXiv.
-
‘Inclusive Portraits: Race-Aware Human-in-the-Loop Technology’
“AI has revolutionized the … automatic facial verification of people. Automated approaches … can face challenges when processing content from certain communities, including communities of people of color. This challenge has prompted the adoption of ‘human-in-the-loop’ (HITL) approaches, where human workers collaborate with the AI to minimize errors. However, most HITL approaches do not consider workers’ individual characteristics and backgrounds. This paper proposes a new approach … that connects with social theories around race to design a racially-aware human-in-the-loop system.” Find the paper and authors list at the 3rd ACM Conference on Equity and Access in Algorithms, Mechanisms and Optimization…
-
‘User Inference Attacks on Large Language Models’
“In this paper, we study the privacy implications of fine-tuning LLMs on user data. To this end, we define a realistic threat model, called user inference, wherein an attacker infers whether or not a user’s data was used for fine-tuning. We implement attacks for this threat model that require only a small set of samples from a user (possibly different from the samples used for training). … We find that LLMs are susceptible to user inference attacks across a variety of fine-tuning datasets, at times with near perfect attack success rates.” Find the paper and full list of authors at…
-
‘What’s Next in Affective Modeling? Large Language Models’
“Large Language Models (LLM) have recently been shown to perform well at various tasks from language understanding, reasoning, storytelling, and information search to theory of mind. In an extension of this work, we explore the ability of GPT-4 to solve tasks related to emotion prediction. GPT-4 performs well across multiple emotion tasks; it can distinguish emotion theories and come up with emotional stories. We show that by prompting GPT-4 to identify key factors of an emotional experience, it is able to manipulate the emotional intensity of its own stories.” Find the paper and full list of authors at ArXiv.
-
‘FibeRobo: Fabricating 4D Fiber Interfaces by Continuous Drawing of Temperature Tunable Liquid Crystal Elastomers’
“We present FibeRobo, a thermally-actuated liquid crystal elastomer (LCE) fiber that can be embedded or structured into textiles and enable silent and responsive interactions with shape-changing, fiber-based interfaces. … This paper contributes to developing temperature-responsive LCE fibers, a facile and scalable fabrication pipeline with optional heating element integration for digital control, mechanical characterization and the establishment of higher hierarchical textile structures and design space. Finally, we introduce a set of demonstrations that illustrate the design space FibeRobo enables.” Find the paper and full list of authors in the 36th Annual ACM Symposium on User Interface Software and Technology proceedings.
-
‘Exploring Question Decomposition for Zero-Shot VQA’
“Visual question answering (VQA) has traditionally been treated as a single-step task where each question receives the same amount of effort, unlike natural human question-answering strategies. We explore a question decomposition strategy for VQA to overcome this limitation. We probe the ability of recently developed large vision-language models to use human-written decompositions and produce their own decompositions of visual questions, finding they are capable of learning both tasks from demonstrations alone. However, we show that naive application of model-written decompositions can hurt performance.” Find the paper and full list of authors at ArXiv.
-
‘Rethinking Neighborhood Consistency Learning on Unsupervised Domain Adaptation’
“Unsupervised domain adaptation (UDA) involves predicting unlabeled data in a target domain by using labeled data from the source domain. However, recent advances in pseudo-labeling (PL) methods have been hampered by noisy pseudo-labels. … Although neighborhood-based PL can help preserve the local structure, it also risks assigning the whole local neighborhood to the wrong semantic category. To address this issue, we propose a novel framework called neighborhood consistency learning (NCL) that operates at both the semantic and instance levels and features a new consistency objective function.” Find the paper and authors list in the 31st ACM International Conference on Multimedia…
-
‘GRASP: Accelerating Shortest Path Attacks via Graph Attention’
“Recent advances in machine learning (ML) have shown promise in aiding and accelerating classical combinatorial optimization algorithms. ML-based speed ups that aim to learn in an end to end manner (i.e., directly output the solution) tend to trade off run time with solution quality. … We consider an APX-hard problem, where an adversary aims to attack shortest paths in a graph by removing the minimum number of edges. We propose the GRASP algorithm: Graph Attention Accelerated Shortest Path Attack, an ML aided optimization algorithm that achieves run times up to 10x faster.” Find the paper and full authors list at…
-
‘In the Room Where It Happens: Characterizing Local Communication and Threats in Smart Homes’
“The network communication between Internet of Things (IoT) devices on the same local network has significant implications for platform and device interoperability, security, privacy and correctness. Yet, the analysis of local home Wi-Fi network traffic and its associated security and privacy threats have been largely ignored by prior literature. … In this paper, we present a comprehensive and empirical measurement study to shed light on the local communication within a smart home deployment and its threats.” Find the paper and full list of authors at the 2023 ACM on Internet Measurement Conference proceedings.
-
‘Tracking, Profiling and Ad Targeting in the Alexa Echo Smart Speaker Ecosystem’
“Smart speakers collect voice commands, which can be used to infer sensitive information about users. Given the potential for privacy harms, there is a need for greater transparency and control over the data collected, used and shared by smart speaker platforms as well as third party skills supported on them. To bridge this gap, we build a framework to measure data collection, usage, and sharing by the smart speaker platforms. … Our results show that Amazon and third parties … collect smart speaker interaction data.” Find the paper and full list of authors at the 2023 ACM on Internet Measurement…
-
‘Medical Imaging RPA System Design’
“Robotic Process Automation (RPA) can minimize human errors, improve efficiency and create a seamless operational environment in the healthcare industry. This paper examines the existing radiology imaging requisition system, which requires human labourers to perform medical request processing and classification. To improve this slow, error-prone and hard-to-scale process, we design an RPA approach that significantly improves efficiency.” Find the paper and full list of authors in the Canadian Conference on Electrical and Computer Engineering proceedings.
-
‘AI-Based General Visual Inspection of Aircrafts Based on YOLOv5’
“Safety is the cornerstone on which the commercial airline industry is built. … The time required for general visual inspections of aircraft can be drastically reduced by using deep learning and remotely piloted aircraft systems (RPAS). Deep learning techniques can be used in aircraft maintenance thanks to the availability of Graphic Processing Units. In our proof of concept study, we use YOLOv5 to build a model that uses high-quality data to find five different aircraft flaws.” Find the paper and full list of authors at the 2023 Canadian Conference on Electrical and Computer Engineering proceedings.
-
‘Robust Client and Server State Synchronisation Framework For React Applications: React-State-Sync’
“As the front-end web frameworks ecosystem evolves, we have encountered problems managing client data. Not only are the solutions for this problem diverse, but the problem too has devolved into two parts — client-side state and server-side state. … Our goal is to provide a consolidated architecture that ensures a full sync between the two states while being performant and developer friendly. Based on our tests against React Context API, we increased the dispatch performance by over 400%.” Find the paper and full list of authors at the 2023 Canadian Conference on Electrical and Computer Engineering proceedings.
-
Function Vectors in Large Language Models
“We report the presence of a simple neural mechanism that represents an input-output function as a vector within autoregressive transformer language models (LMs). Using causal mediation analysis on a diverse range of in-context-learning (ICL) tasks, we find that a small number attention heads transport a compact representation of the demonstrated task, which we call a function vector (FV). … We test FVs across a range of tasks, models and layers and find strong causal effects across settings in middle layers.” Find the paper and full list of authors at ArXiv.
-
‘The Effects of Computational Resources on Flaky Tests’
“Flaky tests are tests that nondeterministically pass and fail in unchanged code. These tests can be detrimental to developers’ productivity. Particularly when tests run in continuous integration environments, the tests may be competing for access to limited computational resources (CPUs, memory etc.) and we hypothesize that resource (in)availability may be a significant factor in the failure rate of flaky tests. We present the first assessment of the impact that computational resources have on flaky tests, including a total of 52 projects written in Java, JavaScript and Python and 27 different resource configurations.” Find the paper and authors list at ArXiv.
-
‘Graph-SCP: Accelerating Set Cover Problems With Graph Neural Networks’
“Machine learning (ML) approaches are increasingly being used to accelerate combinatorial optimization (CO) problems. We look specifically at the Set Cover Problem (SCP) and propose Graph-SCP, a graph neural network method that can augment existing optimization solvers by learning to identify a much smaller sub-problem that contains the solution space. We evaluate the performance of Graph-SCP on synthetic weighted and unweighted SCP instances with diverse problem characteristics and complexities, and on instances from the OR Library, a canonical benchmark for SCP.” Find the paper and full list of authors at ArXiv.
-
To combat counterfeiting, organizations must employ a ‘multilayered strategy’
Anand Nair, professor of supply chain and information management in the D’Amore-McKim School of Business, writes — with Thomas Choi and Robert Handfield — about the “counterfeiting epidemic” affecting many companies, “whether their leaders know it or not.” Combating this crisis, they write, “works best with a multilayered strategy encompassing diverse methods and engaging the entire organization and its partners.” This strategy includes “routinely keeping tabs on contract manufacturers and charting how products move through the supply chain … scoping out what is for sale in consumer markets, deploying covert markings, reviewing warranty claims, educating customers and partnering with key…
-
‘Biochemical Activity of 17 Cancer-Associated Variants of DNA Polymerase Kappa Predicted by Electrostatic Properties’
“DNA damage and repair have been widely studied in relation to cancer and therapeutics. Y-family DNA polymerases can bypass DNA lesions, which may result from external or internal DNA damaging agents, including some chemotherapy agents. Overexpression of the Y-family polymerase human pol kappa can result in tumorigenesis and drug resistance in cancer. This report describes the use of computational tools to predict the effects of single nucleotide polymorphism variants on pol kappa activity.” Find the paper and full list of authors at Chemical Research in Toxicology.
-
‘Linguistic Illusions Guide Eye Movement: Evidence From Doubling’
“Across languages, certain phonological patterns are preferred to others (e.g., blog > lbog). But whether such preferences arise from abstract linguistic constraints or sensorimotor pressures is controversial. We address this debate by examining the constraints on doubling (e.g., slaflaf, generally, XX). Doubling demonstrably elicits conflicting responses (aversion or preference), depending on the linguistic level of analysis (phonology vs. morphology). Since the stimulus remains unchanged, the shifting responses imply abstract constraints. Here, we ask whether these constraints apply online, in eye movements.” Find the paper and full list of authors in the Journal of Psycholinguistic Research.
-
‘Hyper-cores Promote Localization and Efficient Seeding in Higher-Order Processes’
“Going beyond networks, to include higher-order interactions of arbitrary sizes, is a major step to better describe complex systems. In the resulting hypergraph representation, tools to identify structures and central nodes are scarce. We consider the decomposition of a hypergraph in hyper-cores, subsets of nodes connected by at least a certain number of hyperedges of at least a certain size. We show that this provides a fingerprint for data described by hypergraphs and suggests a novel notion of centrality, the hypercoreness.” Find the paper and full list of authors at Nature Communications.
-
Northeastern welcomes researchers from around the globe for Genome Interpretation Workshop
“After years of involvement with the Critical Assessment of Genome Interpretation (CAGI), and after co-chairing the group’s most recent conference in 2022, Predrag Radivojac, professor of computer science and associate dean of research, finally got his chance to welcome the group to Boston and to Northeastern. Beginning on Friday, September 29 and continuing through the weekend, nearly 100 computational genomics and biomedical researchers flocked to the top floor of Northeastern’s East Village to catch up on the latest research and evaluate the challenges that remain.”
-
‘Recent Advances in the Neuroscience of Spontaneous and Off-Task Thought: Implications for Mental Health’
“People spend a remarkable 30–50% of their awake life thinking about something other than what they are currently doing. These experiences of being ‘off-task’ can be described as spontaneous thought when mental dynamics are relatively flexible. Here we review recent neuroscience developments in this area and consider implications for mental well-being and illness. We provide updated overviews of the roles of the default mode network and large-scale network dynamics and we discuss emerging candidate mechanisms involving hippocampal memory (sharp-wave ripples, replay) and neuromodulatory (noradrenergic and serotonergic) systems.” Find the paper and full list of authors at Nature Mental Health.
-
‘DNA Damage Alters Binding Conformations of E. coli Single-Stranded DNA-Binding Protein’
“Single-stranded DNA-binding proteins (SSBs) are essential cellular components, binding to transiently exposed regions of single-stranded DNA (ssDNA) with high affinity and sequence non-specificity to coordinate DNA repair and replication. … We introduce non-canonical DNA bases that mimic naturally occurring DNA damage, synthetic abasic sites, as well as a non-DNA linker into our experimental constructs at sites predicted to interact with [Escherichia coli SSB (EcSSB)]. … Changes in the binding and cooperative behaviors of EcSSB across these constructs can inform how genomic repair and replication processes may change as environmental damage accumulates in DNA.” Find the paper and authors list at…
-
Professor Analia Albuja receives SAGE Emerging Scholar Award for work on identities ‘that don’t neatly fit’ society’s expectations
Assistant professor of psychology Analia Albuja is among a cohort of eight scientists who have received a SAGE Emerging Scholar Award. Her work focuses on belonging, social identity and how subjects assert their identities in different situations.