Search

Pursuing postsecondary education is complicated for adult learners with children, also known as parenting adult learners (PALs). PALs must consider family expenses in addition to college expenses and make tough decisions about how to spend their time across their academic, work, and family responsibilities. Nonetheless, PALs are highly motivated, are resilient, and feel that earning a credential will improve the economic outlook for their family. This project leverages new data from AIR's recent survey of adult learners and one-on-one interviews with a subset of PALs. Building on a broad research base on the experiences of parenting adult learners, results from this research point to three ways practitioners and policymakers can better align programs and resources to this student population: (1) consider practices that encourage a family-friendly environment; (2) tailor outreach to parenting adult learners; and (3) collect and examine data on parenting students. [This report was produced by the Center for Applied Research in Postsecondary Education (CARPE).]

Spurred by recent advances in Large Language Models (LLMs), virtual assistants are poised to take a leap forward in terms of their dialogue capabilities. Yet a major bottleneck to achieving genuinely transformative task-oriented dialogue capabilities remains the scarcity of high quality data. Existing datasets, while impressive in scale, have limited domain coverage and contain few genuinely challenging conversational phenomena; those which are present are typically unlabelled, making it difficult to assess the strengths and weaknesses of models without time-consuming and costly human evaluation. Moreover, creating high quality dialogue data has until now required considerable human input, limiting both the scale of these datasets and the ability to rapidly bootstrap data for a new target domain. We aim to overcome these issues with LUCID, a modularised and highly automated LLM-driven data generation system that produces realistic, diverse and challenging dialogues. We use LUCID to generate a seed dataset of 4,277 conversations across 100 intents to demonstrate its capabilities, with a human review finding consistently high quality labels in the generated data., Comment: Accepted at NAACL SRW 2024

Few-shot dialogue state tracking (DST) with Large Language Models (LLM) relies on an effective and efficient conversation retriever to find similar in-context examples for prompt learning. Previous works use raw dialogue context as search keys and queries, and a retriever is fine-tuned with annotated dialogues to achieve superior performance. However, the approach is less suited for scaling to new domains or new annotation languages, where fine-tuning data is unavailable. To address this problem, we handle the task of conversation retrieval based on text summaries of the conversations. A LLM-based conversation summarizer is adopted for query and key generation, which enables effective maximum inner product search. To avoid the extra inference cost brought by LLM-based conversation summarization, we further distill a light-weight conversation encoder which produces query embeddings without decoding summaries for test conversations. We validate our retrieval approach on MultiWOZ datasets with GPT-Neo-2.7B and LLaMA-7B/30B. The experimental results show a significant improvement over relevant baselines in real few-shot DST settings., Comment: NAACL 2024

Colleges across the United States are now placing most or all students directly into college-level courses and providing supplementary, aligned academic support alongside the courses, also known as "corequisite remediation." Developmental education reforms like corequisite remediation could advance racial and ethnic equity in postsecondary education by facilitating early academic progression. However, there is limited evidence available on differential impacts of corequisite models by race and ethnicity. To better understand the potential for differential impacts of English corequisites for Latinx students, this study leverages data from a randomized control trial across five large urban community colleges across Texas. We also utilize student survey data to develop a deeper understanding of how corequisites shape the experiences of Latinx students in their college-level English courses. Latinx students in our study colleges saw larger benefits from taking corequisite English than non-Latinx students in terms of gateway course completion. The survey findings suggest that corequisites provided an environment where Latinx students felt less academically overwhelmed and less bored relative to patterns observed for traditional DE course enrollees. However, Latinx students in corequisites also reported being less likely to participate in class discussions and ask questions relative to their non-Latinx peers.

Over the past 20 years, an estimated 30 to 35 million Americans have left college with some academic credit but no postsecondary credential. Such students are more likely to be students of color, have dependents, face unmet financial needs, or experience other hardships that cause them to struggle in degree programs. A study team drawn from across the College Completion Network conducted a policy and practice scan to better understand how colleges support returning students. The team conducted focus groups with administrators at 59 2-year and broad-access 4-year colleges across the United States. The focus groups centered on strategies colleges use to identify, recruit, and support students who left college without a degree. Based on the information learned, this brief shares five lessons for practitioners interested in supporting returning college students.

Knowledge-based Visual Question Answering (KB-VQA) requires VQA systems to utilize knowledge from external knowledge bases to answer visually-grounded questions. Retrieval-Augmented Visual Question Answering (RA-VQA), a strong framework to tackle KB-VQA, first retrieves related documents with Dense Passage Retrieval (DPR) and then uses them to answer questions. This paper proposes Fine-grained Late-interaction Multi-modal Retrieval (FLMR) which significantly improves knowledge retrieval in RA-VQA. FLMR addresses two major limitations in RA-VQA's retriever: (1) the image representations obtained via image-to-text transforms can be incomplete and inaccurate and (2) relevance scores between queries and documents are computed with one-dimensional embeddings, which can be insensitive to finer-grained relevance. FLMR overcomes these limitations by obtaining image representations that complement those from the image-to-text transforms using a vision model aligned with an existing text-based retriever through a simple alignment network. FLMR also encodes images and questions using multi-dimensional embeddings to capture finer-grained relevance between queries and documents. FLMR significantly improves the original RA-VQA retriever's PRRecall@5 by approximately 8\%. Finally, we equipped RA-VQA with two state-of-the-art large multi-modal/language models to achieve $\sim61\%$ VQA score in the OK-VQA dataset., Comment: To appear at NeurIPS 2023. This is the camera-ready version. We fixed some numbers and added more experiments to address reviewers' comments

Schema-guided dialogue state trackers can generalise to new domains without further training, yet they are sensitive to the writing style of the schemata. Augmenting the training set with human or synthetic schema paraphrases improves the model robustness to these variations but can be either costly or difficult to control. We propose to circumvent these issues by grounding the state tracking model in knowledge-seeking turns collected from the dialogue corpus as well as the schema. Including these turns in prompts during finetuning and inference leads to marked improvements in model robustness, as demonstrated by large average joint goal accuracy and schema sensitivity improvements on SGD and SGD-X., Comment: Best Long Paper of SIGDIAL 2023

This paper analyzes the availability and accuracy of coastal altimetry sea level products in the Strait of Gibraltar. All possible repeats of two sections of the Envisat and AltiKa ground-tracks were used in the eastern and western portions of the strait. For Envisat, along-track sea level anomalies (SLAs) at 18-Hz posting rate were computed using ranges from two sources, namely, the official SGDRs and the outputs of a coastal waveform retracker, the ALES retracker; in addition, SLAs at 1 Hz were obtained from CTOH, France. For AltiKa, along-track SLA at 40 Hz was also computed both from SGDR and ALES ranges. The quality of these altimeter products was validated using two tide gauges located on the southern coast of Spain. We observed an improvement of about 20% in the accuracy of the Envisat SLAs from ALES compared to the standard (SGDR) and the reprocessed CTOH data sets. AltiKa shows higher accuracy, with no significant differences between SGDR and ALES.

The schema-guided paradigm overcomes scalability issues inherent in building task-oriented dialogue (TOD) agents with static ontologies. Instead of operating on dialogue context alone, agents have access to hierarchical schemas containing task-relevant natural language descriptions. Fine-tuned language models excel at schema-guided dialogue state tracking (DST) but are sensitive to the writing style of the schemas. We explore methods for improving the robustness of DST models. We propose a framework for generating synthetic schemas which uses tree-based ranking to jointly optimise lexical diversity and semantic faithfulness. The generalisation of strong baselines is improved when augmenting their training data with prompts generated by our framework, as demonstrated by marked improvements in average joint goal accuracy (JGA) and schema sensitivity (SS) on the SGD-X benchmark., Comment: Accepted at EACL (Findings) 2023

A college credential can expand the range of career opportunities available to young adults. However, existing barriers to enrollment for prospective students pose equity gaps in college access and success, particularly for Black college aspirants. In Florida, racial and ethnic disparities in college enrollment contribute to disparities in educational attainment. Only 31% of Black Floridians hold a college degree. This report examines the attainment gap by exploring various factors contributing to Florida college applicants' decisions to enroll. This paper uses survey and enrollment information from two Florida community colleges (Hillsborough Community College and Miami Dade College) to identify factors related to college applicants' enrollment decisions. In the summer of 2021, nearly 15,000 applicants to the two colleges were surveyed about their pre-college experiences, and roughly 1,200 responded, resulting in a response rate of 8%. Of the respondents, 997 applicants identified as Black (i.e., African American, Black Hispanic, African, or West Indian/Caribbean).

Before the pandemic, millions of college students eligible for the Supplemental Nutritional Assistance Program (SNAP) did not access the benefit. This gap between eligibility for SNAP benefits and the use of those benefits is often a result of confusion around federal eligibility requirements. Even when eligibility guidelines were temporarily changed to include new exemptions during the pandemic to make it easier for more students to access SNAP, many had no idea they were now eligible to receive benefits. College staff around the country are working diligently to inform students of these changes and connect them to SNAP benefits. With support from the Michelson 20MM Foundation, The Hope Center partnered with Compton College to pilot a low-cost intervention that used text-based nudging to provide timely and relevant information about CalFresh, the California SNAP benefits program. In this brief, the authors share lessons from the pilot intervention at Compton College. In addition, they review findings from focus groups with other community college students throughout California and meetings with community organizations that help connect students to SNAP benefits.

Historically Black Colleges and Universities (HBCUs) were established primarily in the post-Civil War era to meet the educational needs of Black Americans. They provide pathways to upward social mobility and have a long-standing commitment to promoting both academic success and students' health and well-being. But persistent funding inequities at both the state and federal level actively undermine those commitments and leave the sector particularly vulnerable during the coronavirus (COVID-19) pandemic. This report examines how students at HBCUs are faring, and to what extent are their basic needs supported so that they can focus on learning and attain degrees? It uses data from the #RealCollege Survey to examine the overlapping pandemic-related and basic needs challenges affecting students attending HBCUs during fall 2020. In total, nearly 5,000 students from 14 public and private HBCUs responded to the survey for a response rate of 8.3% and shared their experiences. Topics covered include: (1) impacts of the pandemic on students' health and employment; (2) students' basic needs security; (3) utilization of public and campus supports, including emergency aid and Supplemental Nutrition Assistance Program (SNAP); and (4) recommendations for federal and state policymakers. [This report is a collaboration with the Center for the Study of Historically Black Colleges and Universities at Virginia Union University.]

Environmental sensors are crucial for monitoring weather conditions and the impacts of climate change. However, it is challenging to place sensors in a way that maximises the informativeness of their measurements, particularly in remote regions like Antarctica. Probabilistic machine learning models can suggest informative sensor placements by finding sites that maximally reduce prediction uncertainty. Gaussian process (GP) models are widely used for this purpose, but they struggle with capturing complex non-stationary behaviour and scaling to large datasets. This paper proposes using a convolutional Gaussian neural process (ConvGNP) to address these issues. A ConvGNP uses neural networks to parameterise a joint Gaussian distribution at arbitrary target locations, enabling flexibility and scalability. Using simulated surface air temperature anomaly over Antarctica as training data, the ConvGNP learns spatial and seasonal non-stationarities, outperforming a non-stationary GP baseline. In a simulated sensor placement experiment, the ConvGNP better predicts the performance boost obtained from new observations than GP baselines, leading to more informative sensor placements. We contrast our approach with physics-based sensor placement methods and propose future steps towards an operational sensor placement recommendation system. Our work could help to realise environmental digital twins that actively direct measurement sampling to improve the digital representation of reality., Comment: Accepted in Environmental Data Science (Climate Informatics 2023 Special Issue)