Large Language Models (LLMs) are known to have limited extrapolation ability beyond their pre-trained context window, constraining their application in downstream tasks with lengthy inputs. Recent studies have sought to extend LLMs' context window by modifying rotary position embedding (RoPE), a popular position encoding method adopted by well-known LLMs such as LLaMA, PaLM, and GPT-NeoX. However, prior works like Position Interpolation (PI) and YaRN are resource-intensive and lack comparative experiments to assess their applicability. In this work, we identify the inherent need for LLMs' attention entropy (i.e. the information entropy of attention scores) to maintain stability and introduce a novel extension to RoPE which combines adjusting RoPE's base frequency and scaling the attention logits to help LLMs efficiently adapt to a larger context window. We validate the superiority of our method in both fine-tuning performance and robustness across different context window sizes on various context-demanding tasks. Notably, our method extends the context window of LLaMA-2-7B-Chat to 16,384 with only 100 samples and 6 training steps, showcasing extraordinary efficiency. Finally, we also explore how data compositions and training curricula affect context window extension for specific downstream tasks, suggesting fine-tuning LLMs with lengthy conversations as a good starting point. We release our code and SFT data at https://github.com/GAIR-NLP/Entropy-ABF.

The release of ChatGPT in November 2022 prompted a massive uptake of generative artificial intelligence (GenAI) across higher education institutions (HEIs). HEIs scrambled to respond to its use, especially by students, looking first to regulate it and then arguing for its productive integration within teaching and learning. In the year since the release, HEIs have increasingly provided policies and guidelines to direct GenAI. In this paper we examined documents produced by 116 US universities categorized as high research activity or R1 institutions to comprehensively understand GenAI related advice and guidance given to institutional stakeholders. Through an extensive analysis, we found the majority of universities (N=73, 63%) encourage the use of GenAI and many provide detailed guidance for its use in the classroom (N=48, 41%). More than half of all institutions provided sample syllabi (N=65, 56%) and half (N=58, 50%) provided sample GenAI curriculum and activities that would help instructors integrate and leverage GenAI in their classroom. Notably, most guidance for activities focused on writing, whereas code and STEM-related activities were mentioned half the time and vaguely even when they were (N=58, 50%). Finally, more than one half of institutions talked about the ethics of GenAI on a range of topics broadly, including Diversity, Equity and Inclusion (DEI) (N=60, 52%). Overall, based on our findings we caution that guidance for faculty can become burdensome as extensive revision of pedagogical approaches is often recommended in the policies.

Large-scale administrative data is a common input in early warning systems for college dropout in higher education. Still, the terminology and methodology vary significantly across existing studies, and the implications of different modeling decisions are not fully understood. This study provides a systematic evaluation of contributing factors and predictive performance of machine learning models over time and across different student groups. Drawing on twelve years of administrative data at a large public university in the US, we find that dropout prediction at the end of the second year has a 20% higher AUC than at the time of enrollment in a Random Forest model. Also, most predictive factors at the time of enrollment, including demographics and high school performance, are quickly superseded in predictive importance by college performance and in later stages by enrollment behavior. Regarding variability across student groups, college GPA has more predictive value for students from traditionally disadvantaged backgrounds than their peers. These results can help researchers and administrators understand the comparative value of different data sources when building early warning systems and optimizing decisions under specific policy goals.

Mr. Khan's vision of tutoring bots tapped into a decades-old Silicon Valley dream: automated teaching platforms that instantly customize lessons for each student. Proponents argue that developing such systems would help close achievement gaps in schools by delivering relevant, individualized instruction to children faster and more efficiently than human teachers ever could. In pursuit of such ideals, tech companies and philanthropists over the years have urged schools to purchase a laptop for each child, championed video tutorial platforms and financed learning apps that customize students' lessons. Some online math and literacy interventions have reported positive effects. But many education technology efforts have not proved to significantly close academic achievement gaps or improve student results like high school graduation rates.

Harvard University offers a behemoth of courses that teach its students topics including "Queering Education," "Black Radicalism" and sexual fetishes. However, its course catalog – while offering many topics some would consider strongly critical of America – shows it does not offer significant courses focusing on American patriotism in depth despite taking in hundreds of millions of taxpayer dollars every year. In 2021, Harvard received 625 million from American taxpayers, all the while the Ivy League boasts over 50 billion in its endowment. Some companies and prospective students are starting to question their interest in Harvard, particularly after scandals relating to alleged pervasive antisemitism and pro-Hamas sentiment on its campus – prompting legal action and a civil rights investigation from the U.S. Department of Education. Harvard's education department for prospective K-12 teachers elaborates on how one can bring queerness and transgenderism into schools.

Question generation (QG) is a natural language processing task with an abundance of potential benefits and use cases in the educational domain. In order for this potential to be realized, QG systems must be designed and validated with pedagogical needs in mind. However, little research has assessed or designed QG approaches with the input from real teachers or students. This paper applies a large language model-based QG approach where questions are generated with learning goals derived from Bloom's taxonomy. The automatically generated questions are used in multiple experiments designed to assess how teachers use them in practice. The results demonstrate that teachers prefer to write quizzes with automatically generated questions, and that such quizzes have no loss in quality compared to handwritten versions. Further, several metrics indicate that automatically generated questions can even improve the quality of the quizzes created, showing the promise for large scale use of QG in the classroom setting.

Recent work demonstrates that, after being fine-tuned on a high-quality instruction dataset, the resulting model can obtain impressive capabilities to address a wide range of tasks. However, existing methods for instruction data generation often produce duplicate data and are not controllable enough on data quality. In this paper, we extend the generalization of instruction tuning by classifying the instruction data to 4 code-related tasks and propose a LLM-based Generator-Discriminator data process framework to generate diverse, high-quality instruction data from open source code. Hence, we introduce CodeOcean, a dataset comprising 20,000 instruction instances across 4 universal code-related tasks,which is aimed at augmenting the effectiveness of instruction tuning and improving the generalization ability of fine-tuned model. Subsequently, we present WaveCoder, a fine-tuned Code LLM with Widespread And Versatile Enhanced instruction tuning. This model is specifically designed for enhancing instruction tuning of Code Language Models (LLMs). Our experiments demonstrate that Wavecoder models outperform other open-source models in terms of generalization ability across different code-related tasks at the same level of fine-tuning scale. Moreover, Wavecoder exhibits high efficiency in previous code generation tasks. This paper thus offers a significant contribution to the field of instruction data generation and fine-tuning models, providing new insights and tools for enhancing performance in code-related tasks.

Analyzing and evaluating students' progress in any learning environment is stressful and time consuming if done using traditional analysis methods. This is further exasperated by the increasing number of students due to the shift of focus toward integrating the Internet technologies in education and the focus of academic institutions on moving toward e-Learning, blended, or online learning models. As a result, the topic of student performance prediction has become a vibrant research area in recent years. To address this, machine learning and data mining techniques have emerged as a viable solution. To that end, this work proposes the use of deep learning techniques (CNN and RNN-LSTM) to predict the students' performance at the midpoint stage of the online course delivery using three distinct datasets collected from three different regions of the world. Experimental results show that deep learning models have promising performance as they outperform other optimized traditional ML models in two of the three considered datasets while also having comparable performance for the third dataset.

We study and quantify the problem of forgetting when fine-tuning pre-trained large language models (LLMs) on a downstream task. We find that parameter-efficient fine-tuning (PEFT) strategies, such as Low-Rank Adapters (LoRA), still suffer from catastrophic forgetting. In particular, we identify a strong inverse linear relationship between the fine-tuning performance and the amount of forgetting when fine-tuning LLMs with LoRA. We further obtain precise scaling laws that show forgetting increases as a shifted power law in the number of parameters fine-tuned and the number of update steps. We also examine the impact of forgetting on knowledge, reasoning, and the safety guardrails trained into Llama 2 7B chat. Our study suggests that forgetting cannot be avoided through early stopping or by varying the number of parameters fine-tuned. We believe this opens up an important safety-critical direction for future research to evaluate and develop fine-tuning schemes which mitigate forgetting.

Sequences of labeled events observed at irregular intervals in continuous time are ubiquitous across various fields. Temporal Point Processes (TPPs) provide a mathematical framework for modeling these sequences, enabling inferences such as predicting the arrival time of future events and their associated label, called mark. However, due to model misspecification or lack of training data, these probabilistic models may provide a poor approximation of the true, unknown underlying process, with prediction regions extracted from them being unreliable estimates of the underlying uncertainty. This paper develops more reliable methods for uncertainty quantification in neural TPP models via the framework of conformal prediction. A primary objective is to generate a distribution-free joint prediction region for the arrival time and mark, with a finite-sample marginal coverage guarantee. A key challenge is to handle both a strictly positive, continuous response and a categorical response, without distributional assumptions. We first consider a simple but overly conservative approach that combines individual prediction regions for the event arrival time and mark. Then, we introduce a more effective method based on bivariate highest density regions derived from the joint predictive density of event arrival time and mark. By leveraging the dependencies between these two variables, this method exclude unlikely combinations of the two, resulting in sharper prediction regions while still attaining the pre-specified coverage level. We also explore the generation of individual univariate prediction regions for arrival times and marks through conformal regression and classification techniques. Moreover, we investigate the stronger notion of conditional coverage. Finally, through extensive experimentation on both simulated and real-world datasets, we assess the validity and efficiency of these methods.