Understanding temporal relationships and accurately reconstructing the event timeline is important for case law analysis, compliance monitoring, and legal summarization. However, existing benchmarks lack specialized language evaluation, leaving a gap in understanding how LLMs handle event ordering in legal contexts. We introduce LexTime, a dataset designed to evaluate LLMs' event ordering capabilities in legal language, consisting of 512 instances from U.S. Federal Complaints with annotated event pairs and their temporal relations. Our findings show that (1) LLMs are more accurate on legal event ordering than on narrative texts (up to +10.5%); (2) longer input contexts and implicit events boost accuracy, reaching 80.8% for implicit-explicit event pairs; (3) legal linguistic complexities and nested clauses remain a challenge. While performance is promising, specific features of legal texts remain a bottleneck for legal temporal event reasoning, and we propose concrete modeling directions to better address them.

Hallucination, the generation of factually incorrect content, is a growing challenge in Large Language Models (LLMs). Existing detection and mitigation methods are often isolated and insufficient for domain-specific needs, lacking a standardized pipeline. This paper introduces THaMES (Tool for Hallucination Mitigations and EvaluationS), an integrated framework and library addressing this gap. THaMES offers an end-to-end solution for evaluating and mitigating hallucinations in LLMs, featuring automated test set generation, multifaceted benchmarking, and adaptable mitigation strategies. It automates test set creation from any corpus, ensuring high data quality, diversity, and cost-efficiency through techniques like batch processing, weighted sampling, and counterfactual validation. THaMES assesses a model's ability to detect and reduce hallucinations across various tasks, including text generation and binary classification, applying optimal mitigation strategies like In-Context Learning (ICL), Retrieval Augmented Generation (RAG), and Parameter-Efficient Fine-tuning (PEFT). Evaluations of state-of-the-art LLMs using a knowledge base of academic papers, political news, and Wikipedia reveal that commercial models like GPT-4o benefit more from RAG than ICL, while open-weight models like Llama-3.1-8B-Instruct and Mistral-Nemo gain more from ICL. Additionally, PEFT significantly enhances the performance of Llama-3.1-8B-Instruct in both evaluation tasks.

Despite the recent proliferation of large language models (LLMs), their training recipes -- model architecture, pre-training data and optimization algorithm -- are often very similar. This naturally raises the question of the similarity among the resulting models. In this paper, we propose a novel setting, imaginary question answering (IQA), to better understand model similarity. In IQA, we ask one model to generate purely imaginary questions (e.g., on completely made-up concepts in physics) and prompt another model to answer. Surprisingly, despite the total fictionality of these questions, all models can answer each other's questions with remarkable success, suggesting a "shared imagination space" in which these models operate during such hallucinations. We conduct a series of investigations into this phenomenon and discuss implications on model homogeneity, hallucination, and computational creativity.

There is vivid research on adapting Large Language Models (LLMs) to perform a variety of tasks in high-stakes domains such as healthcare. Despite their popularity, there is a lack of understanding of the extent and contributing factors that allow LLMs to recall relevant knowledge and combine it with presented information in the clinical and biomedical domain: a fundamental pre-requisite for success on down-stream tasks. Addressing this gap, we use Multiple Choice and Abstractive Question Answering to conduct a large-scale empirical study on 22 datasets in three generalist and three specialist biomedical sub-domains. Our multifaceted analysis of the performance of 15 LLMs, further broken down by sub-domain, source of knowledge and model architecture, uncovers success factors such as instruction tuning that lead to improved recall and comprehension. We further show that while recently proposed domain-adapted models may lack adequate knowledge, directly fine-tuning on our collected medical knowledge datasets shows encouraging results, even generalising to unseen specialist sub-domains. We complement the quantitative results with a skill-oriented manual error analysis, which reveals a significant gap between the models' capabilities to simply recall necessary knowledge and to integrate it with the presented context. To foster research and collaboration in this field we share M-QALM, our resources, standardised methodology, and evaluation results, with the research community to facilitate further advancements in clinical knowledge representation learning within language models.

This paper introduces UQA, a novel dataset for question answering and text comprehension in Urdu, a low-resource language with over 70 million native speakers. UQA is generated by translating the Stanford Question Answering Dataset (SQuAD2.0), a large-scale English QA dataset, using a technique called EATS (Enclose to Anchor, Translate, Seek), which preserves the answer spans in the translated context paragraphs. The paper describes the process of selecting and evaluating the best translation model among two candidates: Google Translator and Seamless M4T. The paper also benchmarks several state-of-the-art multilingual QA models on UQA, including mBERT, XLM-RoBERTa, and mT5, and reports promising results. For XLM-RoBERTa-XL, we have an F1 score of 85.99 and 74.56 EM. UQA is a valuable resource for developing and testing multilingual NLP systems for Urdu and for enhancing the cross-lingual transferability of existing models. Further, the paper demonstrates the effectiveness of EATS for creating high-quality datasets for other languages and domains. The UQA dataset and the code are publicly available at www.github.com/sameearif/UQA.

Recent models have achieved human level performance on the Stanford Question Answering Dataset when using F1 scores to evaluate the reading comprehension task. Yet, teaching machines to comprehend text has not been solved in the general case. By appending one adversarial sentence to the context paragraph, past research has shown that the F1 scores from reading comprehension models drop almost in half. In this paper, I replicate past adversarial research with a new model, ELECTRA-Small, and demonstrate that the new model's F1 score drops from 83.9% to 29.2%. To improve ELECTRA-Small's resistance to this attack, I finetune the model on SQuAD v1.1 training examples with one to five adversarial sentences appended to the context paragraph. Like past research, I find that the finetuned model on one adversarial sentence does not generalize well across evaluation datasets. However, when finetuned on four or five adversarial sentences the model attains an F1 score of more than 70% on most evaluation datasets with multiple appended and prepended adversarial sentences. The results suggest that with enough examples we can make models robust to adversarial attacks.

Machine question answering is an essential yet challenging task in natural language processing. Recently, Pre-trained Contextual Embeddings (PCE) models like Bidirectional Encoder Representations from Transformers (BERT) and A Lite BERT (ALBERT) have attracted lots of attention due to their great performance in a wide range of NLP tasks. In our Paper, we utilized the fine-tuned ALBERT models and implemented combinations of additional layers (e.g. attention layer, RNN layer) on top of them to improve model performance on Stanford Question Answering Dataset (SQuAD 2.0). We implemented four different models with different layers on top of ALBERT-base model, and two other models based on ALBERT-xlarge and ALBERT-xxlarge. We compared their performance to our baseline model ALBERT-base-v2 + ALBERT-SQuAD-out with details. Our best-performing individual model is ALBERT-xxlarge + ALBERT-SQuAD-out, which achieved an F1 score of 88.435 on the dev set. Furthermore, we have implemented three different ensemble algorithms to boost overall performance. By passing in several best-performing models' results into our weighted voting ensemble algorithm, our final result ranks first on the Stanford CS224N Test PCE SQuAD Leaderboard with F1 = 90.123.

Question generation has recently shown impressive results in customizing question answering (QA) systems to new domains. These approaches circumvent the need for manually annotated training data from the new domain and, instead, generate synthetic question-answer pairs that are used for training. However, existing methods for question generation rely on large amounts of synthetically generated datasets and costly computational resources, which render these techniques widely inaccessible when the text corpora is of limited size. This is problematic as many niche domains rely on small text corpora, which naturally restricts the amount of synthetic data that can be generated. In this paper, we propose a novel framework for domain adaptation called contrastive domain adaptation for QA (CAQA). Specifically, CAQA combines techniques from question generation and domain-invariant learning to answer out-of-domain questions in settings with limited text corpora. Here, we train a QA system on both source data and generated data from the target domain with a contrastive adaptation loss that is incorporated in the training objective. By combining techniques from question generation and domain-invariant learning, our model achieved considerable improvements compared to state-of-the-art baselines.

When answering a question, people often draw upon their rich world knowledge in addition to the particular context. While recent works retrieve supporting facts/evidence from commonsense knowledge bases to supply additional information to each question, there is still ample opportunity to advance it on the quality of the evidence. It is crucial since the quality of the evidence is the key to answering commonsense questions, and even determines the upper bound on the QA systems performance. In this paper, we propose a recursive erasure memory network (REM-Net) to cope with the quality improvement of evidence. To address this, REM-Net is equipped with a module to refine the evidence by recursively erasing the low-quality evidence that does not explain the question answering. Besides, instead of retrieving evidence from existing knowledge bases, REM-Net leverages a pre-trained generative model to generate candidate evidence customized for the question. We conduct experiments on two commonsense question answering datasets, WIQA and CosmosQA. The results demonstrate the performance of REM-Net and show that the refined evidence is explainable.

State-of-the-art unsupervised multilingual models (e.g., multilingual BERT) have been shown to generalize in a zero-shot cross-lingual setting. This generalization ability has been attributed to the use of a shared subword vocabulary and joint training across multiple languages giving rise to deep multilingual abstractions. We evaluate this hypothesis by designing an alternative approach that transfers a monolingual model to new languages at the lexical level. More concretely, we first train a transformer-based masked language model on one language, and transfer it to a new language by learning a new embedding matrix with the same masked language modeling objective--freezing parameters of all other layers. This approach does not rely on a shared vocabulary or joint training. However, we show that it is competitive with multilingual BERT on standard cross-lingual classification benchmarks and on a new Cross-lingual Question Answering Dataset (XQuAD). Our results contradict common beliefs of the basis of the generalization ability of multilingual models and suggest that deep monolingual models learn some abstractions that generalize across languages. We also release XQuAD as a more comprehensive cross-lingual benchmark, which comprises 240 paragraphs and 1190 question-answer pairs from SQuAD v1.1 translated into ten languages by professional translators.