The landscape of available textual adversarial attacks keeps growing, posing severe threats and raising concerns regarding the deep NLP system's integrity. However, the crucial problem of defending against malicious attacks has only drawn the attention of the NLP community. The latter is nonetheless instrumental in developing robust and trustworthy systems. This paper makes two important contributions in this line of search: (i) we introduce LAROUSSE, a new framework to detect textual adversarial attacks and (ii) we introduce STAKEOUT, a new benchmark composed of nine popular attack methods, three datasets, and two pre-trained models. LAROUSSE is ready-to-use in production as it is unsupervised, hyperparameter-free, and non-differentiable, protecting it against gradient-based methods. Our new benchmark STAKEOUT allows for a robust evaluation framework: we conduct extensive numerical experiments which demonstrate that LAROUSSE outperforms previous methods, and which allows to identify interesting factors of detection rate variations.

Link prediction is an important task in addressing the incompleteness problem of knowledge graphs (KG). Previous link prediction models suffer from issues related to either performance or explanatory capability. Furthermore, models that are capable of generating explanations, often struggle with erroneous paths or reasoning leading to the correct answer. To address these challenges, we introduce a novel Neural-Symbolic model named FaSt-FLiP (stands for Fast and Slow Thinking with Filtered rules for Link Prediction task), inspired by two distinct aspects of human cognition: "commonsense reasoning" and "thinking, fast and slow." Our objective is to combine a logical and neural model for enhanced link prediction. To tackle the challenge of dealing with incorrect paths or rules generated by the logical model, we propose a semi-supervised method to convert rules into sentences. These sentences are then subjected to assessment and removal of incorrect rules using an NLI (Natural Language Inference) model. Our approach to combining logical and neural models involves first obtaining answers from both the logical and neural models. These answers are subsequently unified using an Inference Engine module, which has been realized through both algorithmic implementation and a novel neural model architecture. To validate the efficacy of our model, we conducted a series of experiments. The results demonstrate the superior performance of our model in both link prediction metrics and the generation of more reliable explanations.

Bilingual Lexicon Induction (BLI) is a core task in multilingual NLP that still, to a large extent, relies on calculating cross-lingual word representations. Inspired by the global paradigm shift in NLP towards Large Language Models (LLMs), we examine the potential of the latest generation of LLMs for the development of bilingual lexicons. We ask the following research question: Is it possible to prompt and fine-tune multilingual LLMs (mLLMs) for BLI, and how does this approach compare against and complement current BLI approaches? To this end, we systematically study 1) zero-shot prompting for unsupervised BLI and 2) few-shot in-context prompting with a set of seed translation pairs, both without any LLM fine-tuning, as well as 3) standard BLI-oriented fine-tuning of smaller LLMs. We experiment with 18 open-source text-to-text mLLMs of different sizes (from 0.3B to 13B parameters) on two standard BLI benchmarks covering a range of typologically diverse languages. Our work is the first to demonstrate strong BLI capabilities of text-to-text mLLMs. The results reveal that few-shot prompting with in-context examples from nearest neighbours achieves the best performance, establishing new state-of-the-art BLI scores for many language pairs. We also conduct a series of in-depth analyses and ablation studies, providing more insights on BLI with (m)LLMs, also along with their limitations.

With language technology increasingly affecting individuals' lives, many recent works have investigated the ethical aspects of NLP. Among other topics, researchers focused on the notion of morality, investigating, for example, which moral judgements language models make. However, there has been little to no discussion of the terminology and the theories underpinning those efforts and their implications. This lack is highly problematic, as it hides the works' underlying assumptions and hinders a thorough and targeted scientific debate of morality in NLP. In this work, we address this research gap by (a) providing an overview of some important ethical concepts stemming from philosophy and (b) systematically surveying the existing literature on moral NLP w.r.t. their philosophical foundation, terminology, and data basis. For instance, we analyse what ethical theory an approach is based on, how this decision is justified, and what implications it entails. Our findings surveying 92 papers show that, for instance, most papers neither provide a clear definition of the terms they use nor adhere to definitions from philosophy. Finally, (c) we give three recommendations for future research in the field. We hope our work will lead to a more informed, careful, and sound discussion of morality in language technology.

Social media has become a major driver of social change, by facilitating the formation of online social movements. Automatically understanding the perspectives driving the movement and the voices opposing it, is a challenging task as annotated data is difficult to obtain. We propose a weakly supervised graph-based approach that explicitly models perspectives in #BackLivesMatter-related tweets. Our proposed approach utilizes a social-linguistic representation of the data. We convert the text to a graph by breaking it into structured elements and connect it with the social network of authors, then structured prediction is done over the elements for identifying perspectives. Our approach uses a small seed set of labeled examples. We experiment with large language models for generating artificial training examples, compare them to manual annotation, and find that it achieves comparable performance. We perform quantitative and qualitative analyses using a human-annotated test set. Our model outperforms multitask baselines by a large margin, successfully characterizing the perspectives supporting and opposing #BLM.

The Intensive Care Unit (ICU) is one of the most important parts of a hospital, which admits critically ill patients and provides continuous monitoring and treatment. Various patient outcome prediction methods have been attempted to assist healthcare professionals in clinical decision-making. Existing methods focus on measuring the similarity between patients using deep neural networks to capture the hidden feature structures. However, the higher-order relationships are ignored, such as patient characteristics (e.g., diagnosis codes) and their causal effects on downstream clinical predictions. In this paper, we propose a novel Hypergraph Convolutional Network that allows the representation of non-pairwise relationships among diagnosis codes in a hypergraph to capture the hidden feature structures so that fine-grained patient similarity can be calculated for personalized mortality risk prediction. Evaluation using a publicly available eICU Collaborative Research Database indicates that our method achieves superior performance over the state-of-the-art models on mortality risk prediction. Moreover, the results of several case studies demonstrated the effectiveness and robustness of the model decisions.

Despite the remarkable ability of large language models (LMs) to comprehend and generate language, they have a tendency to hallucinate and create factually inaccurate output. Augmenting LMs by retrieving information from external knowledge resources is one promising solution. Most existing retrieval augmented LMs employ a retrieve-and-generate setup that only retrieves information once based on the input. This is limiting, however, in more general scenarios involving generation of long texts, where continually gathering information throughout generation is essential. In this work, we provide a generalized view of active retrieval augmented generation, methods that actively decide when and what to retrieve across the course of the generation. We propose Forward-Looking Active REtrieval augmented generation (FLARE), a generic method which iteratively uses a prediction of the upcoming sentence to anticipate future content, which is then utilized as a query to retrieve relevant documents to regenerate the sentence if it contains low-confidence tokens. We test FLARE along with baselines comprehensively over 4 long-form knowledge-intensive generation tasks/datasets. FLARE achieves superior or competitive performance on all tasks, demonstrating the effectiveness of our method. Code and datasets are available at https://github.com/jzbjyb/FLARE.

Large Language Models (LLMs) have made remarkable strides in various tasks. Whether LLMs are competitive few-shot solvers for information extraction (IE) tasks, however, remains an open problem. In this work, we aim to provide a thorough answer to this question. Through extensive experiments on nine datasets across four IE tasks, we demonstrate that current advanced LLMs consistently exhibit inferior performance, higher latency, and increased budget requirements compared to fine-tuned SLMs under most settings. Therefore, we conclude that LLMs are not effective few-shot information extractors in general. Nonetheless, we illustrate that with appropriate prompting strategies, LLMs can effectively complement SLMs and tackle challenging samples that SLMs struggle with. And moreover, we propose an adaptive filter-then-rerank paradigm to combine the strengths of LLMs and SLMs. In this paradigm, SLMs serve as filters and LLMs serve as rerankers. By prompting LLMs to rerank a small portion of difficult samples identified by SLMs, our preliminary system consistently achieves promising improvements (2.4% F1-gain on average) on various IE tasks, with an acceptable time and cost investment.

Large-scale pre-training has shown remarkable performance in building open-domain dialogue systems. However, previous works mainly focus on showing and evaluating the conversational performance of the released dialogue model, ignoring the discussion of some key factors towards a powerful human-like chatbot, especially in Chinese scenarios. In this paper, we conduct extensive experiments to investigate these under-explored factors, including data quality control, model architecture designs, training approaches, and decoding strategies. We propose EVA2.0, a large-scale pre-trained open-domain Chinese dialogue model with 2.8 billion parameters, and will make our models and codes publicly available. Automatic and human evaluations show that EVA2.0 significantly outperforms other open-source counterparts. We also discuss the limitations of this work by presenting some failure cases and pose some future research directions on large-scale Chinese open-domain dialogue systems.

Novice pilots find it difficult to operate and land unmanned aerial vehicles (UAVs), due to the complex UAV dynamics, challenges in depth perception, lack of expertise with the control interface and additional disturbances from the ground effect. Therefore we propose a shared autonomy approach to assist pilots in safely landing a UAV under conditions where depth perception is difficult and safe landing zones are limited. Our approach comprises of two modules: a perception module that encodes information onto a compressed latent representation using two RGB-D cameras and a policy module that is trained with the reinforcement learning algorithm TD3 to discern the pilot's intent and to provide control inputs that augment the user's input to safely land the UAV. The policy module is trained in simulation using a population of simulated users. Simulated users are sampled from a parametric model with four parameters, which model a pilot's tendency to conform to the assistant, proficiency, aggressiveness and speed. We conduct a user study (n = 28) where human participants were tasked with landing a physical UAV on one of several platforms under challenging viewing conditions. The assistant, trained with only simulated user data, improved task success rate from 51.4% to 98.2% despite being unaware of the human participants' goal or the structure of the environment a priori. With the proposed assistant, regardless of prior piloting experience, participants performed with a proficiency greater than the most experienced unassisted participants.