In recent years, large language models (LLMs) and generative AI have revolutionized natural language processing (NLP), offering unprecedented capabilities in education. This chapter explores the transformative potential of LLMs in automated question generation and answer assessment. It begins by examining the mechanisms behind LLMs, emphasizing their ability to comprehend and generate human-like text. The chapter then discusses methodologies for creating diverse, contextually relevant questions, enhancing learning through tailored, adaptive strategies. Key prompting techniques, such as zero-shot and chain-of-thought prompting, are evaluated for their effectiveness in generating high-quality questions, including open-ended and multiple-choice formats in various languages. Advanced NLP methods like fine-tuning and prompt-tuning are explored for their role in generating task-specific questions, despite associated costs. The chapter also covers the human evaluation of generated questions, highlighting quality variations across different methods and areas for improvement. Furthermore, it delves into automated answer assessment, demonstrating how LLMs can accurately evaluate responses, provide constructive feedback, and identify nuanced understanding or misconceptions. Examples illustrate both successful assessments and areas needing improvement. The discussion underscores the potential of LLMs to replace costly, time-consuming human assessments when appropriately guided, showcasing their advanced understanding and reasoning capabilities in streamlining educational processes.

With recent advances in large language models (LLMs), there has been emerging numbers of research in developing Semantic IDs based on LLMs to enhance the performance of recommendation systems. However, the dimension of these embeddings needs to match that of the ID embedding in recommendation, which is usually much smaller than the original length. Such dimension compression results in inevitable losses in discriminability and dimension robustness of the LLM embeddings, which motivates us to scale up the semantic representation. In this paper, we propose Mixture-of-Codes, which first constructs multiple independent codebooks for LLM representation in the indexing stage, and then utilizes the Semantic Representation along with a fusion module for the downstream recommendation stage. Extensive analysis and experiments demonstrate that our method achieves superior discriminability and dimension robustness scalability, leading to the best scale-up performance in recommendations. An intuitive practice is to simply project the LLM embeddings to low-dimension embeddings via only MLPs into the recommendation systems for feature interactions.

Large Language Models (LLMs) have revolutionized natural language understanding and generation tasks but suffer from high memory consumption and slow inference times due to their large parameter sizes. Traditional model compression techniques, such as quantization and pruning, mitigate these issues but often require retraining to maintain accuracy, which is computationally expensive. Our method reduces quantization error while leveraging sparse representations compatible with accelerated hardware architectures. Additionally, we propose a parameter-efficient fine-tuning recipe that significantly reduces overhead compared to conventional quantizationaware training. M achieves up to a 5.4% improvement in model accuracy for sparsity patterns like 2:4, and the fine-tuning step further enhances accuracy by up to 5.8%, demonstrating state-of-the-art performance. This work provides a pathway for efficiently deploying large models in memory-constrained environments without compromising accuracy. Large Language Models (LLMs) (Brown et al., 2020; Radford et al., 2019) are transformative for natural language understanding and generation (Suzgun et al., 2022; Zhou et al., 2023); however, their extensive parameter count leads to large memory footprints and longer inference times, making them expensive to execute. Model compression methods, such as sparsity and quantization, have shown promising results in reducing the inference costs of LLMs. However, these methods often require costly retraining on large amounts of data to restore the original model accuracy (Sanh et al., 2020; Park et al., 2018), while facing numerical and optimization stability challenges when dealing with quantized weights in quantization-aware-training (Gholami et al., 2022).

Vocabulary adaptation, which integrates new vocabulary into pre-trained language models (LMs), enables expansion to new languages and mitigates token overfragmentation. However, existing approaches are limited by their reliance on heuristic or external embeddings. We propose VocADT, a novel method for vocabulary adaptation using adapter modules that are trained to learn the optimal linear combination of existing embeddings while keeping the model's weights fixed. VocADT offers a flexible and scalable solution without requiring external resources or language constraints. Across 11 languages--with various scripts, resource availability, and fragmentation--we demonstrate that VocADT outperforms the original Mistral model (Jiang et al., 2023) and other baselines across various multilingual tasks. We find that Latin-script languages and highly fragmented languages benefit the most from vocabulary adaptation. We further finetune the adapted model on the generative task of machine translation and find that vocabulary adaptation is still beneficial after fine-tuning and that VocADT is the most effective method. Vocabulary adaptation (or transfer)--a process of modifying a pre-trained language model (LM) to use a new vocabulary--offers several key advantages.

Understanding the training dynamics of transformers is important to explain the impressive capabilities behind large language models. In this work, we study the dynamics of training a shallow transformer on a task of recognizing co-occurrence of two designated words. In the literature of studying training dynamics of transformers, several simplifications are commonly adopted such as weight reparameterization, attention linearization, special initialization, and lazy regime. In contrast, we analyze the gradient flow dynamics of simultaneously training three attention matrices and a linear MLP layer from random initialization, and provide a framework of analyzing such dynamics via a coupled dynamical system. We establish near minimum loss and characterize the attention model after training. We discover that gradient flow serves as an inherent mechanism that naturally divide the training process into two phases. In Phase 1, the linear MLP quickly aligns with the two target signals for correct classification, whereas the softmax attention remains almost unchanged. In Phase 2, the attention matrices and the MLP evolve jointly to enlarge the classification margin and reduce the loss to a near minimum value. Technically, we prove a novel property of the gradient flow, termed \textit{automatic balancing of gradients}, which enables the loss values of different samples to decrease almost at the same rate and further facilitates the proof of near minimum training loss. We also conduct experiments to verify our theoretical results.

In the evolving NLP landscape, benchmarks serve as yardsticks for gauging progress. However, existing Legal NLP benchmarks only focus on predictive tasks, overlooking generative tasks. This work curates LexSumm, a benchmark designed for evaluating legal summarization tasks in English. It comprises eight English legal summarization datasets, from diverse jurisdictions, such as the US, UK, EU and India. Additionally, we release LexT5, legal oriented sequence-to-sequence model, addressing the limitation of the existing BERT-style encoder-only models in the legal domain. We assess its capabilities through zero-shot probing on LegalLAMA and fine-tuning on LexSumm. Our analysis reveals abstraction and faithfulness errors even in summaries generated by zero-shot LLMs, indicating opportunities for further improvements. LexSumm benchmark and LexT5 model are available at https://github.com/TUMLegalTech/LexSumm-LexT5.

This paper proposes a zero-shot speech emotion recognition (SER) method that estimates emotions not previously defined in the SER model training. Conventional methods are limited to recognizing emotions defined by a single word. Moreover, we have the motivation to recognize unknown bipolar emotions such as ``I want to buy - I do not want to buy.'' In order to allow the model to define classes using sentences freely and to estimate unknown bipolar emotions, our proposed method expands upon the contrastive language-audio pre-training (CLAP) framework by introducing multi-class and multi-task settings. We also focus on purchase intention as a bipolar emotion and investigate the model's performance to zero-shot estimate it. This study is the first attempt to estimate purchase intention from speech directly. Experiments confirm that the results of zero-shot estimation by the proposed method are at the same level as those of the model trained by supervised learning.

Visual Question Answering (VQA) is the task of answering a question about an image and requires processing multimodal input and reasoning to obtain the answer. Modular solutions that use declarative representations within the reasoning component have a clear advantage over end-to-end trained systems regarding interpretability. The downside is that crafting the rules for such a component can be an additional burden on the developer. We address this challenge by presenting an approach for declarative knowledge distillation from Large Language Models (LLMs). Our method is to prompt an LLM to extend an initial theory on VQA reasoning, given as an answer-set program, to meet the requirements of the VQA task. Examples from the VQA dataset are used to guide the LLM, validate the results, and mend rules if they are not correct by using feedback from the ASP solver. We demonstrate that our approach works on the prominent CLEVR and GQA datasets. Our results confirm that distilling knowledge from LLMs is in fact a promising direction besides data-driven rule learning approaches.

Large language models (LLMs) sometimes demonstrate poor performance on knowledge-intensive tasks, commonsense reasoning is one of them. Researchers typically address these issues by retrieving related knowledge from knowledge graphs or employing self-enhancement methods to elicit knowledge in LLMs. However, noisy knowledge and invalid reasoning issues hamper their ability to answer questions accurately. To this end, we propose a novel method named eliciting, filtering and integrating knowledge in large language model (LINKED). In it, we design a reward model to filter out the noisy knowledge and take the marginal consistent reasoning module to reduce invalid reasoning. With our comprehensive experiments on two complex commonsense reasoning benchmarks, our method outperforms SOTA baselines (up to 9.0% improvement of accuracy). Besides, to measure the positive and negative impact of the injected knowledge, we propose a new metric called effectiveness-preservation score for the knowledge enhancement works. Finally, through extensive experiments, we conduct an in-depth analysis and find many meaningful conclusions about LLMs in commonsense reasoning tasks.

Challenge sets such as the Winograd Schema Challenge (WSC) are used to benchmark systems' ability to resolve ambiguities in natural language. If one assumes as in existing work that solving a given challenge set is at least as difficult as solving some more general task, then high performance on the challenge set should indicate high performance on the general task overall. However, we show empirically that this assumption of difficulty does not always hold. In particular, we demonstrate that despite the strong performance of prompted language models (LMs) on the WSC and its variants, these same modeling techniques perform relatively poorly at resolving certain pronominal ambiguities attested in OntoNotes and related datasets that are perceived to be easier. Motivated by these findings, we propose a method for ensembling a prompted LM with a supervised, task-specific system that is overall more accurate at resolving pronominal coreference across datasets. Finally, we emphasize that datasets involving the same linguistic phenomenon draw on distinct, but overlapping, capabilities, and evaluating on any one dataset alone does not provide a complete picture of a system's overall capability.