Title
Topic
-
‘Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking’
“Fine-tuning on generalized tasks such as instruction following, code generation, and mathematics has been shown to enhance language models’ performance on a range of tasks. Nevertheless, explanations of how such fine-tuning influences the internal computations in these models remain elusive. We study how fine-tuning affects the internal mechanisms implemented in language models. As a case study, we explore the property of entity tracking, a crucial facet of language comprehension, where models fine-tuned on mathematics have substantial performance gains.” Find the paper and full list of authors at ArXiv.
-
‘Measuring and Controlling Persona Drift in Language Model Dialogs’
“System-prompting is a standard tool for customizing language-model chatbots, enabling them to follow a specific instruction. An implicit assumption in the use of system prompts is that they will be stable, so the chatbot will continue to generate text according to the stipulated instructions for the duration of a conversation. We propose a quantitative benchmark to test this assumption, evaluating instruction stability via self-chats between two instructed chatbots. Testing popular models like LLaMA2-chat-70B and GPT-3.5, we reveal a significant instruction drift within eight rounds of conversations.” Find the paper and full list of authors at ArXiv.
-
‘Fully Dynamic Matching: (2-√2)-Approximation in Polylog Update Time’
“We study maximum matchings in fully dynamic graphs, … graphs that undergo both edge insertions and deletions. Our focus is on algorithms that estimate the size of maximum matching after each update while spending a small time. … We show that for any fixed ɛ > 0, a (2 — √2— ɛ) approximation can be maintained in poly(log n) time per update even in general graphs. Our techniques also lead to the same approximation for general graphs in two passes of the semi-streaming setting, removing a similar gap.” Find the paper and authors list in the 2024 Annual ACM-SIAM Symposium…
-
‘A Retrospective Study of One Decade of Artifact Evaluations’
“Reproducibility is a vital property of experimental and empirical research, without whichit is difficult to establish trust in derived conclusions. If results cannot be independently confirmed, they may be affected by observer bias or other confounding factors. As the full-scale reproduction of scientific results from a study takes significant time, which does not match well with the conference-focused publication in computer science, a lighter quality assurance mechanism for scientific work has been established. … After a decade of artifact evaluations, we analyze the impact they have had on published articles and artifacts.” Find the paper authors list at Software Engineering…
-
‘The Arrangement of Marks Impacts Afforded Messages: Ordering, Partitioning, Spacing and Coloring in Bar Charts’
“Data visualizations present a massive number of potential messages to an observer. … The message that a viewer tends to notice — the message that a visualization ‘affords’ — is strongly affected by how values are arranged in a chart, e.g., how the values are colored or positioned. … We present a set of empirical evaluations of how different messages … are afforded by variations in ordering, partitioning, spacing and coloring of values, within the ubiquitous case study of bar graphs.” Find the paper and full list of authors at Transactions on Visualization and Computer Graphics.
-
‘Giant Mobile Coralliths From the Florida Keys, USA’
“Coralliths are spherical, free-living (motile), scleractinian colonies inhabiting present day and ancient coral reefs. They form by the coral rolling on the seabed which can occur through biological and/or physical processes. While diving and snorkeling in nearshore environments in the upper Florida Keys, we observed hundreds of coralliths of varying sizes and species. … The largest coralliths we observed were all [Solenastrea] bournoni and ranged between 0.5 and 1.4 m in diameter. The exceptionally large colonies identified (>1 m) may be the largest and oldest spherical coralliths described to date.” Find the paper and authors list at the Bulletin of Marine…
-
‘Twisty-Puzzle-Inspired Approach to Clifford Synthesis’
“The problem of decomposing an arbitrary Clifford element into a sequence of Clifford gates is known as Clifford synthesis. Drawing inspiration from similarities between this and the famous Rubik’s cube twisty puzzle, we develop a machine learning approach for Clifford synthesis based on learning an approximation to the distance to the identity. This approach is probabilistic and computationally intensive. However, when a decomposition is successfully found, it often involves fewer gates than the decomposition methods used in the Qiskit decomposition protocol, which uses a combination of several well-known Clifford decomposition schemes.” Find the paper and authors list at Physical Review A.
-
‘Interpretation Issues With “Genomic Vulnerability” Arise From Conceptual Issues in Local Adaptation and Maladaptation’
“As climate change causes the environment to shift away from the local optimum that populations have adapted to, fitness declines are predicted to occur. Recently, methods known as genomic offsets (GOs) have become a popular tool to predict population responses to climate change from landscape genomic data. Populations with a high GO have been interpreted to have a high “genomic vulnerability” to climate change. … This study uses hypothetical and empirical data to explore situations in which different types of fitness offsets may or may not be correlated with each other or with a GO.”
-
Advancing wireless with ‘Deep learning-based polymorphic platform’
“Electrical and computer engineering William Lincoln Smith Professor Tommaso Melodia and assistant professor Francesco Restuccia were awarded a patent for ‘Deep learning-based polymorphic platform.'”
-
‘In Situ Polymer Gelation in Confined Flow Controls Intermittent Dynamics’
“Polymer flows through pores, nozzles and other small channels govern engineered and naturally occurring dynamics in many processes. … The crosslinking of polymers can change their material properties dramatically, and it is advantageous to know a priori whether or not crosslinking polymers will lead to clogged channels or cessation of flow. In this study, we investigate the flow of a common biopolymer, alginate, while it undergoes crosslinking by the addition of a crosslinker, calcium, driven through a microfluidic channel at constant flow rate.” Find the paper and full list of authors at Soft Matter.
-
Mid-Atlantic Topology Conference held at Northeastern
The 2024 Mid-Atlantic Topology Conference was held at Northeastern University in late March, 2024. The conference was sponsored by professors Ben Knudsen, Iva Halacheva and Jose Perea in the department of mathematics.
-
NIH funding for nanomedicine co-ops fighting cancer
“CaNCURE: Cancer Nanomedicine Co-ops for Undergraduate Research Experiences Program seeks to provide training and education in research at the interface of nanotechnology, cancer biology and medicine to attract, retain and encourage young scientists and engineers, particularly those from underrepresented minorities, to pursue careers in cancer research. The program will achieve this central aim by creating meaningful and rigorous research experiences mentored by outstanding academic and clinical researchers and supplemented with high-impact learning and professional development opportunities. The program combines Northeastern University’s model of co-op education with the outstanding cancer nanomedicine research infrastructure at Dana-Farber/Harvard Cancer Center.”
-
Hillman receives funding to study ‘Persistence of Neurocognitive Benefits of Exercise’
“In this 5-year follow-up to the IGNITE exercise intervention study we will examine whether exercise has a delayed or protracted benefit to neurocognitive outcomes including Alzheimer’s disease neuropathology. We will also examine whether we can predict long-term exercise engagement by utilizing the rich biological and psychological data from IGNITE.”
-
NSF funding secured for ‘Mobility Data for Communities: Uncovering Segregation, Climate Resilience and Economic Development from Cell-Phone Records’
“In this project, we plan to build a public mobility data platform for the Boston area that will help neighborhoods and communities use cellphone-generated data to address issues of social equity, racial and socioeconomic segregation, economic development and climate resilience.”
-
‘FANCJ Promotes PARP1 Activity during DNA replication that Is Essential in BRCA1 Deficient Cells’
“The effectiveness of poly (ADP-ribose) polymerase inhibitors (PARPi) in creating single-stranded DNA gaps and inducing sensitivity requires the FANCJ DNA helicase. Yet, how FANCJ relates to PARP1 inhibition or trapping, which contribute to PARPi toxicity, remains unclear. Here, we find PARPi effectiveness hinges on S-phase PARP1 activity, which is reduced in FANCJ deficient cells as G-quadruplexes sequester PARP1 and MSH2.” Find the paper and full list of authors at Nature Communications.
-
‘Ensemble2: Scenarios Ensembling for Communication and Performance Analysis’
“Throughout the COVID-19 pandemic, scenario modeling played a crucial role in shaping the decision-making process of public health policies. Unlike forecasts, scenario projections rely on specific assumptions about the future that consider different plausible states-of-the-world that may or may not be realized. … As a consequence, long-term scenario projections require different evaluation criteria than the ones used for traditional short-term epidemic forecasts. Here, we propose a novel ensemble procedure for assessing pandemic scenario projections using the results of the Scenario Modeling Hub (SMH) for COVID-19 in the United States (US).” Find the paper and full list of authors in Epidemics.
-
‘AI Fairness in Action: A Human-Computer Perspective on AI Fairness in Organizations and Society’
“As fairness is acutely subjective in nature, and it is those subjective experiences that determine behavioral outcomes, it follows that the use of AI systems in decision-making raises important concerns about the fairness of processes, treatment and outcomes. Therefore, we need to situate the psychology of how humans perceive fairness into the realm of algorithmic decision-making. … In this special issue, we aim to fill these voids [in the literature] by presenting theoretical and empirical work that contribute to our understanding of what AI fairness encapsulates.” Find the paper and full list of authors at the International Journal of Human–Computer…
-
Weng receives Grantham funding to engineer carbon sequestering plant roots
“This project aims to develop engineered plants that produce chemically inert sporopollenin in non-anther tissues for long-term carbon storage. By reconstituting sporopollenin biosynthesis and secretion in roots, inert polymers can be injected into soil. Progress includes identifying sporopollenin biosynthetic genes, developing analytical methods and exploring yeast expression systems. Future work will assess sporopollenin durability in soil and investigate sporopollenin co-polymers in plants. This scalable, self-repairable approach offers a biological solution for atmospheric carbon sequestration.”
-
‘Accounting for Hydroclimatic Properties in Flood Frequency Analysis Procedures’
“Flood hazard is typically evaluated by computing extreme flood probabilities from a flood frequency distribution following nationally defined procedures. … These procedures, also known as flood frequency analysis, typically recommend only one probability distribution family for all watersheds within a country or region. However, large uncertainties associated with extreme flood probability estimates (>50-year flood or Q50) can be further biased when fit to an inappropriate distribution model. … Here, we demonstrate that hydroclimatic parameters can aid in the selection of a parametric flood frequency distribution.” Find the paper and full list of authors at Hydrology and Earth System Sciences.
-
‘Physical Activity-Related Individual Differences in Functional Human Connectome Are Linked to Fluid Intelligence in Older Adults’
“The study examined resting state functional connectivity (rs-FC) associated with moderate-to-vigorous physical activity (MV-PA), sedentary time (ST), TV viewing, computer use and their relationship to cognitive performance in older adults. We used pre-intervention data from 119 participants from the Fit & Active Seniors trial. … Our findings suggest that greater functional integration within networks implementing top-down control and within those supporting visuospatial abilities, paired with segregation between networks critical and those not critical to top-down control, may help promote cognitive reserve in more physically active seniors.” Find the paper and full list of authors at Neurobiology of Aging.
-
‘Spin–Orbit Exciton–Induced Phonon Chirality in a Quantum Magnet’
“Phonons are known as an important energy carrier and their interaction with electrons is critical for understanding exotic properties in correlated materials. … Chiral phonons arising from circular ionic motions alone would only exhibit a small magnetic moment. In contrast, chiral phonons reported recently in several classes of materials including the pseudogap phase of cuprates and topological semimetals exhibit surprisingly large moments. Here, we report studies of chiral phonons in CoTiO3, a quantum magnet that hosts interesting bosonic excitations such as topological magnons and spin–orbit excitons.” Find the paper and full list of authors at PNAS.
-
Wales will target ‘SHP-1 through a newfound metabolite-regulated cysteine activation site’
“This new project is focused on chemical manipulation and characterization of SHP-1 based on recent insights, gained in large part from the development of mass spectrometry approaches to map small molecule covalent interactions proteome-wide done in the Chouchani laboratory. My research group will apply hydrogen-deuterium exchange (HDX) mass spectrometry (MS) to investigate conformational changes of SHP-1 upon binding to itaconate and itaconate-mimetics to ascertain if SHP-1 activation leads to conformational rearrangement and then determine where those rearrangements take place within SHP-1. These HDX MS studies will provide solution-based measurements that will complement the X-ray crystallography studies of the same system.”
-
‘Effect of Emotion Naming on Emotion Regulation in Younger and Older Adults’
“Prior research has shown that naming one’s emotions before regulation has a detrimental effect on regulation success (Nook et al., 2020). … The aim of the current replication-extension study was to evaluate whether this effect would replicate in an older adult sample. Due to age-related changes in emotion regulation knowledge and behaviors, we hypothesized that age would moderate this effect. … However, contrary to our hypothesis, this effect was larger in older adults (d=.82) than younger adults (d=.41). These findings suggest that affect crystallization may be stronger or faster in older adulthood.” Find the paper and list of authors at Innovation…
-
‘Associations Between Cognitive Resources and Emotion Regulation Tactics in an Adult Lifespan Sample’
“The current study investigated how trait-level cognitive capacity relates to emotion regulation tactic preferences in everyday life in adulthood and old age. 51 younger adults (ages 18-39), 53 middle-aged adults (ages 40-59), and 55 older adults (ages 60+) completed measures of working memory and verbal fluency, as well as 21 days of experience sampling. On each survey, participants indicated if they had regulated since the last survey and if so, what emotion regulation strategies they used and how they implemented that strategy through specific emotion regulation tactics.” Find the paper and full list of authors at Innovation in Aging.
-
‘Modeling the Role of Immune Cell Conversion in the Tumor-Immune Microenvironment’
“Tumors develop in a complex physical, biochemical, and cellular milieu, referred to as the tumor microenvironment. Of special interest is the set of immune cells that reciprocally interact with the tumor, the tumor-immune microenvironment (TIME). … While tumor cells are known to induce immune cells to switch from anti-tumor to pro-tumor phenotypes, this type of ecological interaction has been largely overlooked. To address this gap in cancer modeling, we develop a minimal, ecological model of the TIME with immune cell conversion, to highlight this important interaction and explore its consequences.” Find the paper and authors list at Bulletin of Mathematical…
-
‘Switching the Spin Cycloid in BiFeO3 With an Electric Field’
“Bismuth ferrite (BiFeO3) is a multiferroic material that exhibits both ferroelectricity and canted antiferromagnetism at room temperature, making it a unique candidate in the development of electric-field controllable magnetic devices. The magnetic moments in BiFeO3 are arranged into a spin cycloid, resulting in unique magnetic properties which are tied to the ferroelectric order. Previous understanding of this coupling has relied on average, mesoscale measurements. Using nitrogen vacancy-based diamond magnetometry, we observe the magnetic spin cycloid structure of BiFeO3 in real space.” Find the paper and full list of authors at Nature Communications.