The emergence of large language models (LLMs) has sparked enormous interest due to their potential application across a range of educational tasks. For example, recent work in programming education has used LLMs to generate learning resources, improve error messages, and provide feedback on code. However, one factor that limits progress within the field is that much of the research uses bespoke datasets and different evaluation metrics, making direct comparisons between results unreliable. Thus, there is a pressing need for standardization and benchmarks that facilitate the equitable comparison of competing approaches. One task where LLMs show great promise is program repair, which can be used to provide debugging support and next-step hints to students. In this article, we propose a novel educational program repair benchmark. We curate two high-quality publicly available programming datasets, present a unified evaluation procedure introducing a novel evaluation metric rouge@k for approximating the quality of repairs, and evaluate a set of five recent models to establish baseline performance.

Evaluating open-ended written examination responses from students is an essential yet time-intensive task for educators, requiring a high degree of effort, consistency, and precision. Recent developments in Large Language Models (LLMs) present a promising opportunity to balance the need for thorough evaluation with efficient use of educators' time. In our study, we explore the effectiveness of LLMs ChatGPT-3.5, ChatGPT-4, Claude-3, and Mistral-Large in assessing university students' open-ended answers to questions made about reference material they have studied. Each model was instructed to evaluate 54 answers repeatedly under two conditions: 10 times (10-shot) with a temperature setting of 0.0 and 10 times with a temperature of 0.5, expecting a total of 1,080 evaluations per model and 4,320 evaluations across all models. The RAG (Retrieval Augmented Generation) framework was used as the framework to make the LLMs to process the evaluation of the answers. As of spring 2024, our analysis revealed notable variations in consistency and the grading outcomes provided by studied LLMs. There is a need to comprehend strengths and weaknesses of LLMs in educational settings for evaluating open-ended written responses. Further comparative research is essential to determine the accuracy and cost-effectiveness of using LLMs for educational assessments.

Large pretrained language models (LLMs) have shown surprising In-Context Learning (ICL) ability. An important application in deploying large language models is to augment LLMs with a private database for some specific task. The main problem with this promising commercial use is that LLMs have been shown to memorize their training data and their prompt data are vulnerable to membership inference attacks (MIA) and prompt leaking attacks. In order to deal with this problem, we treat LLMs as untrusted in privacy and propose a locally differentially private framework of in-context learning (LDP-ICL) in the settings where labels are sensitive. Considering the mechanisms of in-context learning in Transformers by gradient descent, we provide an analysis of the trade-off between privacy and utility in such LDP-ICL for classification. Moreover, we apply LDP-ICL to the discrete distribution estimation problem. In the end, we perform several experiments to demonstrate our analysis results.

Unlike the majority of works that only opensourced the model weights and architecture, we have made everything needed to train a model available, including the training data, the training process, and the evaluation code. Our goal is to empower and strengthen the open research community, fostering transparency and enabling a new wave of innovation in the field of language modeling. Furthermore, we extend the context length to 1M tokens through lightweight continual pretraining and demonstrate strong needlein-a-haystack retrieval performance. Countless models have been opensourced on AI communities like HuggingFace to facilitate their use by researchers (Bai et al., 2023; Singer et al., 2024; Zhang et al., 2024). These models can broadly be divided into two categories: 1) Open source model weights and data sources, which constitute the vast majority.

Large language models (LLMs) have emerged as an integral part of modern societies, powering user-facing applications such as personal assistants and enterprise applications like recruitment tools. Despite their utility, research indicates that LLMs perpetuate systemic biases. Yet, prior works on LLM harms predominantly focus on Western concepts like race and gender, often overlooking cultural concepts from other parts of the world. Additionally, these studies typically investigate "harm" as a singular dimension, ignoring the various and subtle forms in which harms manifest. To address this gap, we introduce the Covert Harms and Social Threats (CHAST), a set of seven metrics grounded in social science literature. We utilize evaluation models aligned with human assessments to examine the presence of covert harms in LLM-generated conversations, particularly in the context of recruitment. Our experiments reveal that seven out of the eight LLMs included in this study generated conversations riddled with CHAST, characterized by malign views expressed in seemingly neutral language unlikely to be detected by existing methods. Notably, these LLMs manifested more extreme views and opinions when dealing with non-Western concepts like caste, compared to Western ones such as race.

As large language models (LLMs) continue to make significant strides, their better integration into agent-based simulations offers a transformational potential for understanding complex social systems. However, such integration is not trivial and poses numerous challenges. Based on this observation, in this paper, we explore architectures and methods to systematically develop LLM-augmented social simulations and discuss potential research directions in this field. We conclude that integrating LLMs with agent-based simulations offers a powerful toolset for researchers and scientists, allowing for more nuanced, realistic, and comprehensive models of complex systems and human behaviours.

As the popularity of Large Language Models (LLMs) grow, combining model safety with utility becomes increasingly important. The challenge is making sure that LLMs can recognize and decline dangerous prompts without sacrificing their ability to be helpful. The problem of "exaggerated safety" demonstrates how difficult this can be. To reduce excessive safety behaviours -- which was discovered to be 26.1% of safe prompts being misclassified as dangerous and refused -- we use a combination of XSTest dataset prompts as well as interactive, contextual, and few-shot prompting to examine the decision bounds of LLMs such as Llama2, Gemma Command R+, and Phi-3. We find that few-shot prompting works best for Llama2, interactive prompting works best Gemma, and contextual prompting works best for Command R+ and Phi-3. Using a combination of these prompting strategies, we are able to mitigate exaggerated safety behaviors by an overall 92.9% across all LLMs. Our work presents a multiple prompting strategies to jailbreak LLMs' decision-making processes, allowing them to navigate the tight line between refusing unsafe prompts and remaining helpful.

Evaluating free-text explanations is a multifaceted, subjective, and labor-intensive task. Large language models (LLMs) present an appealing alternative due to their potential for consistency, scalability, and cost-efficiency. In this work, we present ACORN, a new dataset of 3,500 free-text explanations and aspect-wise quality ratings, and use it to gain insights into how LLMs evaluate explanations. We observed that replacing one of the human ratings sometimes maintained, but more often lowered the inter-annotator agreement across different settings and quality aspects, suggesting that their judgments are not always consistent with human raters. We further quantified this difference by comparing the correlation between LLM-generated ratings with majority-voted human ratings across different quality aspects. With the best system, Spearman's rank correlation ranged between 0.53 to 0.95, averaging 0.72 across aspects, indicating moderately high but imperfect alignment. Finally, we considered the alternative of using an LLM as an additional rater when human raters are scarce, and measured the correlation between majority-voted labels with a limited human pool and LLMs as an additional rater, compared to the original gold labels. While GPT-4 improved the outcome when there were only two human raters, in all other observed cases, LLMs were neutral to detrimental when there were three or more human raters. We publicly release the dataset to support future improvements in LLM-in-the-loop evaluation here: https://github.com/a-brassard/ACORN.

A variety of techniques have been developed, e.g., evolutionary computation[60, 133], methods incorporating templated mutation operators[71], semantic inference techniques[79] targeting single-cause defects, and methods designed to handle multi-hunk bugs[100]. Increasingly, researchers have applied ML-based methods to APR tasks (Section 3), but data leakage is a concern(Section 4). Each new technique, or modification of an existing technique, tends to be developed by an independent research team, without reference to a common, formal definition of APR. Benchmarks are not enough to standardize evaluation on their own (Section 5). As motivating examples, consider the following inconsistencies in the published literature: Correctness. VFix [123] identifies correct patches that pass all test cases and are semantically or syntactically equivalent to the original bug-fix, while VRepair[26] reports repair accuracy in terms of semantic equivalence to the original bug-fix, and SynFix [10] defines correctness simply as passing the test cases. Each of these is a reasonable definition, but collectively, their differences make it difficult to compare results.

Given the increasing demand for mental health assistance, artificial intelligence (AI), particularly large language models (LLMs), may be valuable for integration into automated clinical support systems. In this work, we leverage a decision transformer architecture for topic recommendation in counseling conversations between patients and mental health professionals. The architecture is utilized for offline reinforcement learning, and we extract states (dialogue turn embeddings), actions (conversation topics), and rewards (scores measuring the alignment between patient and therapist) from previous turns within a conversation to train a decision transformer model. We demonstrate an improvement over baseline reinforcement learning methods, and propose a novel system of utilizing our model's output as synthetic labels for fine-tuning a large language model for the same task. Although our implementation based on LLaMA-2 7B has mixed results, future work can undoubtedly build on the design.