Streaming services have reshaped how we discover and engage with digital entertainment. Despite these advancements, effectively understanding the wide spectrum of user search queries continues to pose a significant challenge. An accurate query understanding system that can handle a variety of entities that represent different user intents is essential for delivering an enhanced user experience. We can build such a system by training a natural language understanding (NLU) model; however, obtaining high-quality labeled training data in this specialized domain is a substantial obstacle. Manual annotation is costly and impractical for capturing users' vast vocabulary variations. To address this, we introduce a novel approach that leverages large language models (LLMs) through weak supervision to automatically annotate a vast collection of user search queries. Using prompt engineering and a diverse set of LLM personas, we generate training data that matches human annotator expectations. By incorporating domain knowledge via Chain of Thought and In-Context Learning, our approach leverages the labeled data to train low-latency models optimized for real-time inference. Extensive evaluations demonstrated that our approach outperformed the baseline with an average relative gain of 113% in recall. Furthermore, our novel prompt engineering framework yields higher quality LLM-generated data to be used for weak supervision; we observed 47.60% improvement over baseline in agreement rate between LLM predictions and human annotations with respect to F1 score, weighted according to the distribution of occurrences of the search queries. Our persona selection routing mechanism further adds an additional 3.67% increase in weighted F1 score on top of our novel prompt engineering framework.
Large pre-trained models (LPMs) have demonstrated exceptional performance in diverse natural language processing and computer vision tasks. However, fully fine-tuning these models poses substantial memory challenges, particularly in resource-constrained environments. Parameter-efficient fine-tuning (PEFT) methods, such as LoRA, mitigate this issue by adjusting only a small subset of parameters. Nevertheless, these methods typically employ random initialization for low-rank matrices, which can lead to inefficiencies in gradient descent and diminished generalizability due to suboptimal starting points. To address these limitations, we propose SVFit, a novel PEFT approach that leverages singular value decomposition (SVD) to initialize low-rank matrices using critical singular values as trainable parameters. Specifically, SVFit performs SVD on the pre-trained weight matrix to obtain the best rank-r approximation matrix, emphasizing the most critical singular values that capture over 99% of the matrix's information. These top-r singular values are then used as trainable parameters to scale the fundamental subspaces of the matrix, facilitating rapid domain adaptation. Extensive experiments across various pre-trained models in natural language understanding, text-to-image generation, and image classification tasks reveal that SVFit outperforms LoRA while requiring 16 times fewer trainable parameters.
Author profiling is the task of inferring characteristics about individuals by analyzing content they share. Supervised machine learning still dominates automatic systems that perform this task, despite the popularity of prompting large language models to address natural language understanding tasks. One reason is that the classification instances consist of large amounts of posts, potentially a whole user profile, which may exceed the input length of Transformers. Even if a model can use a large context window, the entirety of posts makes the application of API-accessed black box systems costly and slow, next to issues which come with such "needle-in-the-haystack" tasks. To mitigate this limitation, we propose a new method for author profiling which aims at distinguishing relevant from irrelevant content first, followed by the actual user profiling only with relevant data. To circumvent the need for relevance-annotated data, we optimize this relevance filter via reinforcement learning with a reward function that utilizes the zero-shot capabilities of large language models. We evaluate our method for Big Five personality trait prediction on two Twitter corpora. On publicly available real-world data with a skewed label distribution, our method shows similar efficacy to using all posts in a user profile, but with a substantially shorter context. An evaluation on a version of these data balanced with artificial posts shows that the filtering to relevant posts leads to a significantly improved accuracy of the predictions.
Relation classification (RC) plays a pivotal role in both natural language understanding and knowledge graph completion. It is generally formulated as a task to recognize the relationship between two entities of interest appearing in a free-text sentence. Conventional approaches on RC, regardless of feature engineering or deep learning based, can obtain promising performance on categorizing common types of relation leaving a large proportion of unrecognizable long-tail relations due to insufficient labeled instances for training. In this paper, we consider few-shot learning is of great practical significance to RC and thus improve a modern framework of metric learning for few-shot RC. Specifically, we adopt the large-margin ProtoNet with fine-grained features, expecting they can generalize well on long-tail relations. Extensive experiments were conducted by FewRel, a large-scale supervised few-shot RC dataset, to evaluate our framework: LM-ProtoNet (FGF). The results demonstrate that it can achieve substantial improvements over many baseline approaches.
Recent breakthroughs in large language models (LLMs) offer unprecedented natural language understanding and generation capabilities. However, existing surveys on LLMs in biomedicine often focus on specific applications or model architectures, lacking a comprehensive analysis that integrates the latest advancements across various biomedical domains. This review, based on an analysis of 484 publications sourced from databases including PubMed, Web of Science, and arXiv, provides an in-depth examination of the current landscape, applications, challenges, and prospects of LLMs in biomedicine, distinguishing itself by focusing on the practical implications of these models in real-world biomedical contexts. Firstly, we explore the capabilities of LLMs in zero-shot learning across a broad spectrum of biomedical tasks, including diagnostic assistance, drug discovery, and personalized medicine, among others, with insights drawn from 137 key studies. Then, we discuss adaptation strategies of LLMs, including fine-tuning methods for both uni-modal and multi-modal LLMs to enhance their performance in specialized biomedical contexts where zero-shot fails to achieve, such as medical question answering and efficient processing of biomedical literature. Finally, we discuss the challenges that LLMs face in the biomedicine domain including data privacy concerns, limited model interpretability, issues with dataset quality, and ethics due to the sensitive nature of biomedical data, the need for highly reliable model outputs, and the ethical implications of deploying AI in healthcare. To address these challenges, we also identify future research directions of LLM in biomedicine including federated learning methods to preserve data privacy and integrating explainable AI methodologies to enhance the transparency of LLMs.
Explainable fake news detection predicts the authenticity of news items with annotated explanations. Today, Large Language Models (LLMs) are known for their powerful natural language understanding and explanation generation abilities. However, presenting LLMs for explainable fake news detection remains two main challenges. Firstly, fake news appears reasonable and could easily mislead LLMs, leaving them unable to understand the complex news-faking process. Secondly, utilizing LLMs for this task would generate both correct and incorrect explanations, which necessitates abundant labor in the loop. In this paper, we propose LLM-GAN, a novel framework that utilizes prompting mechanisms to enable an LLM to become Generator and Detector and for realistic fake news generation and detection. Our results demonstrate LLM-GAN's effectiveness in both prediction performance and explanation quality. We further showcase the integration of LLM-GAN to a cloud-native AI platform to provide better fake news detection service in the cloud.
Large Language Models (LLMs) have demonstrated great capabilities in natural language understanding and generation, largely attributed to the intricate alignment process using human feedback. While alignment has become an essential training component that leverages data collected from user queries, it inadvertently opens up an avenue for a new type of user-guided poisoning attacks. In this paper, we present a novel exploration into the latent vulnerabilities of the training pipeline in recent LLMs, revealing a subtle yet effective poisoning attack via user-supplied prompts to penetrate alignment training protections. Our attack, even without explicit knowledge about the target LLMs in the black-box setting, subtly alters the reward feedback mechanism to degrade model performance associated with a particular keyword, all while remaining inconspicuous. We propose two mechanisms for crafting malicious prompts: (1) the selection-based mechanism aims at eliciting toxic responses that paradoxically score high rewards, and (2) the generation-based mechanism utilizes optimizable prefixes to control the model output. By injecting 1\% of these specially crafted prompts into the data, through malicious users, we demonstrate a toxicity score up to two times higher when a specific trigger word is used. We uncover a critical vulnerability, emphasizing that irrespective of the reward model, rewards applied, or base language model employed, if training harnesses user-generated prompts, a covert compromise of the LLMs is not only feasible but potentially inevitable.
In this article, we propose the R2GQA system, a Retriever-Reader-Generator Question Answering system, consisting of three main components: Document Retriever, Machine Reader, and Answer Generator. The Retriever module employs advanced information retrieval techniques to extract the context of articles from a dataset of legal regulation documents. The Machine Reader module utilizes state-of-the-art natural language understanding algorithms to comprehend the retrieved documents and extract answers. Finally, the Generator module synthesizes the extracted answers into concise and informative responses to questions of students regarding legal regulations. Furthermore, we built the ViRHE4QA dataset in the domain of university training regulations, comprising 9,758 question-answer pairs with a rigorous construction process. This is the first Vietnamese dataset in the higher regulations domain with various types of answers, both extractive and abstractive. In addition, the R2GQA system is the first system to offer abstractive answers in Vietnamese. This paper discusses the design and implementation of each module within the R2GQA system on the ViRHE4QA dataset, highlighting their functionalities and interactions. Furthermore, we present experimental results demonstrating the effectiveness and utility of the proposed system in supporting the comprehension of students of legal regulations in higher education settings. In general, the R2GQA system and the ViRHE4QA dataset promise to contribute significantly to related research and help students navigate complex legal documents and regulations, empowering them to make informed decisions and adhere to institutional policies effectively. Our dataset is available for research purposes.
This thesis provides methods and analysis of models which make progress on this goal. The techniques outlined are task agnostic, and should provide benefit when used with nearly any transformer LM. We introduce two new finetuning methods which add new capabilities to the models they are used on. The first adds a recurrence mechanism, which removes the fixed-window sized constraint and improves the efficiency of a transformer decoder. The second allows masked language models (MLMs) to be used for initialization of both the encoder and decoder of a non-autoregressive sequence-to-sequence transformer, opening up generative applications of models which were previously only used for natural language understanding tasks. We also introduce two new techniques for improving the quality of predictions of any transformer decoder without additional finetuning. One, hidden state optimization, can be applied to any transformer decoder to improve the quality of predictions at inference time, especially for few-shot classification. The other, conditional beam search, allows practitioners to search for natural language generation (NLG) model outputs with high likelihood while conditioning on the event that the output is not degenerate (e.g. empty, repetitive, etc.). Finally, we provide theoretical and empirical insights on the divergence of model-likelihood and output quality which has widely been observed in prior work. These insights apply to any model which represents a distribution over text, and apply to language models which are not transformers or even autoregressive. We argue that the NLP community has, to some extent, misunderstood the implications of these findings, and encourage a point of view which has more nuance.
Recently, large language and vision models have shown strong performance, but due to high pre-training and fine-tuning costs, research has shifted towards faster training via dataset pruning. Previous methods used sample loss as an evaluation criterion, aiming to select the most "difficult" samples for training. However, when the pruning rate increases, the number of times each sample is trained becomes more evenly distributed, which causes many critical or general samples to not be effectively fitted. We refer to this as Low-Frequency Learning (LFL). In other words, LFL prevents the model from remembering most samples. In our work, we decompose the scoring function of LFL, provide a theoretical explanation for the inefficiency of LFL, and propose adding a memory term to the scoring function to enhance the model's memory capability, along with an approximation of this memory term. Similarly, we explore memory in Self-Supervised Learning (SSL), marking the first discussion on SSL memory. Using contrastive learning, we derive the memory term both theoretically and experimentally. Finally, we propose Enhance Memory Pruning (EMP), which addresses the issue of insufficient memory under high pruning rates by enhancing the model's memory of data, thereby improving its performance. We evaluated the performance of EMP in tasks such as image classification, natural language understanding, and model pre-training. The results show that EMP can improve model performance under extreme pruning rates. For example, in the CIFAR100-ResNet50 pre-training task, with 70\% pruning, EMP outperforms current methods by 2.2\%.
Detecting biases in natural language understanding (NLU) for African American Vernacular English (AAVE) is crucial to developing inclusive natural language processing (NLP) systems. To address dialect-induced performance discrepancies, we introduce AAVENUE ({AAVE} {N}atural Language {U}nderstanding {E}valuation), a benchmark for evaluating large language model (LLM) performance on NLU tasks in AAVE and Standard American English (SAE). AAVENUE builds upon and extends existing benchmarks like VALUE, replacing deterministic syntactic and morphological transformations with a more flexible methodology leveraging LLM-based translation with few-shot prompting, improving performance across our evaluation metrics when translating key tasks from the GLUE and SuperGLUE benchmarks. We compare AAVENUE and VALUE translations using five popular LLMs and a comprehensive set of metrics including fluency, BARTScore, quality, coherence, and understandability. Additionally, we recruit fluent AAVE speakers to validate our translations for authenticity. Our evaluations reveal that LLMs consistently perform better on SAE tasks than AAVE-translated versions, underscoring inherent biases and highlighting the need for more inclusive NLP models. We have open-sourced our source code on GitHub and created a website to showcase our work at https://aavenue.live.
Fine-tuning large language models (LLMs) on downstream tasks requires substantial computational resources. A class of parameter-efficient fine-tuning (PEFT) aims to mitigate these computational challenges by selectively fine-tuning only a small fraction of the model parameters. Although computationally efficient, these techniques often fail to match the performance of fully fine-tuned models, primarily due to inherent biases introduced during parameter selection. Traditional selective PEFT techniques use a fixed set of parameters based on a predefined budget (a process also known as unmasking), failing to capture parameter importance dynamically and often ending up exceeding the budget. We introduce $\text{ID}^3$, a novel selective PEFT method that calculates parameter importance continually and dynamically unmasks parameters by balancing exploration and exploitation in parameter selection. Our empirical study on 15 tasks spanning natural language understanding and generative tasks demonstrates the effectiveness of our method compared to fixed-masking-based PEFT techniques. We analytically show that $\text{ID}^3$ reduces the number of gradient updates by a factor of two, enhancing computational efficiency. $\text{ID}^3$ is robust to random initialization of neurons and, therefore, can be seamlessly integrated into existing additive and reparametrization-based PEFT modules such as adapters and LoRA for dynamic sparsification.
The emergence of large language models such as ChatGPT, Gemini, and others highlights the importance of evaluating their diverse capabilities, ranging from natural language understanding to code generation. However, their performance on spatial tasks has not been thoroughly assessed. This study addresses this gap by introducing a new multi-task spatial evaluation dataset designed to systematically explore and compare the performance of several advanced models on spatial tasks. The dataset includes twelve distinct task types, such as spatial understanding and simple route planning, each with verified and accurate answers. We evaluated multiple models, including OpenAI's gpt-3.5-turbo, gpt-4-turbo, gpt-4o, ZhipuAI's glm-4, Anthropic's claude-3-sonnet-20240229, and MoonShot's moonshot-v1-8k, using a two-phase testing approach. First, we conducted zero-shot testing. Then, we categorized the dataset by difficulty and performed prompt-tuning tests. Results show that gpt-4o achieved the highest overall accuracy in the first phase, with an average of 71.3%. Although moonshot-v1-8k slightly underperformed overall, it outperformed gpt-4o in place name recognition tasks. The study also highlights the impact of prompt strategies on model performance in specific tasks. For instance, the Chain-of-Thought (CoT) strategy increased gpt-4o's accuracy in simple route planning from 12.4% to 87.5%, while a one-shot strategy improved moonshot-v1-8k's accuracy in mapping tasks from 10.1% to 76.3%.
The natural language understanding (NLU) performance of large language models (LLMs) has been evaluated across various tasks and datasets. The existing evaluation methods, however, do not take into account the variance in scores due to differences in prompts, which leads to unfair evaluation and comparison of NLU performance. Moreover, evaluation designed for specific prompts is inappropriate for instruction tuning, which aims to perform well with any prompt. It is therefore necessary to find a way to measure NLU performance in a fair manner, considering score variance between different instruction templates. In this study, we provide English and Japanese cross-lingual datasets for evaluating the NLU performance of LLMs, which include multiple instruction templates for fair evaluation of each task, along with regular expressions to constrain the output format. Furthermore, we propose the Sharpe score as an evaluation metric that takes into account the variance in scores between templates. Comprehensive analysis of English and Japanese LLMs reveals that the high variance among templates has a significant impact on the fair evaluation of LLMs.
In human reading and communication, individuals tend to engage in geospatial reasoning, which involves recognizing geographic entities and making informed inferences about their interrelationships. To mimic such cognitive process, current methods either utilize conventional natural language understanding toolkits, or directly apply models pretrained on geo-related natural language corpora. However, these methods face two significant challenges: i) they do not generalize well to unseen geospatial scenarios, and ii) they overlook the importance of integrating geospatial context from geographical databases with linguistic information from the Internet. To handle these challenges, we propose GeoReasoner, a language model capable of reasoning on geospatially grounded natural language. Specifically, it first leverages Large Language Models (LLMs) to generate a comprehensive location description based on linguistic and geospatial information. It also encodes direction and distance information into spatial embedding via treating them as pseudo-sentences. Consequently, the model is trained on both anchor-level and neighbor-level inputs to learn geo-entity representation. Extensive experimental results demonstrate GeoReasoner's superiority in three tasks: toponym recognition, toponym linking, and geo-entity typing, compared to the state-of-the-art baselines.
Although Large Language Models(LLMs) can generate coherent and contextually relevant text, they often struggle to recognise the intent behind the human user's query. Natural Language Understanding (NLU) models, however, interpret the purpose and key information of user's input to enable responsive interactions. Existing NLU models generally map individual utterances to a dual-level semantic frame, involving sentence-level intent and word-level slot labels. However, real-life conversations primarily consist of multi-turn conversations, involving the interpretation of complex and extended dialogues. Researchers encounter challenges addressing all facets of multi-turn dialogue conversations using a unified single NLU model. This paper introduces a novel approach, MIDAS, leveraging a multi-level intent, domain, and slot knowledge distillation for multi-turn NLU. To achieve this, we construct distinct teachers for varying levels of conversation knowledge, namely, sentence-level intent detection, word-level slot filling, and conversation-level domain classification. These teachers are then fine-tuned to acquire specific knowledge of their designated levels. A multi-teacher loss is proposed to facilitate the combination of these multi-level teachers, guiding a student model in multi-turn dialogue tasks. The experimental results demonstrate the efficacy of our model in improving the overall multi-turn conversation understanding, showcasing the potential for advancements in NLU models through the incorporation of multi-level dialogue knowledge distillation techniques.
Fine-tuning Large Language Models (LLMs) incurs considerable training costs, driving the need for data-efficient training with optimised data ordering. Human-inspired strategies offer a solution by organising data based on human learning practices. This study evaluates the fine-tuning efficiency of five human-inspired strategies across four language models, three datasets, and both human- and LLM-labelled data in the context of medical question answering. These strategies achieve the best accuracy gain of 1.81% and an average gain of 1.02% across datasets, with interleaved strategies delivering the best average results. However, the best strategy varies across model-dataset combinations, limiting the generalisability of the effects of any single strategy. Additionally, LLM-defined question difficulty outperforms human-defined labels in curriculum-based learning, showing the potential of model-generated data as a cost-effective alternative for optimising fine-tuning.
To enable non-experts to specify long-horizon, multi-robot collaborative tasks, language models are increasingly used to translate natural language commands into formal specifications. However, because translation can occur in multiple ways, such translations may lack accuracy or lead to inefficient multi-robot planning. Our key insight is that concise hierarchical specifications can simplify planning while remaining straightforward to derive from human instructions. We propose Nl2Hltl2Plan, a framework that translates natural language commands into hierarchical Linear Temporal Logic (LTL) and solves the corresponding planning problem. The translation involves two steps leveraging Large Language Models (LLMs). First, an LLM transforms instructions into a Hierarchical Task Tree, capturing logical and temporal relations. Next, a fine-tuned LLM converts sub-tasks into flat LTL formulas, which are aggregated into hierarchical specifications, with the lowest level corresponding to ordered robot actions. These specifications are then used with off-the-shelf planners. Our Nl2Hltl2Plan demonstrates the potential of LLMs in hierarchical reasoning for multi-robot task planning. Evaluations in simulation and real-world experiments with human participants show that Nl2Hltl2Plan outperforms existing methods, handling more complex instructions while achieving higher success rates and lower costs in task allocation and planning. Additional details are available at https://nl2hltl2plan.github.io .
Event Temporal Relation Extraction (ETRE) aims to identify the temporal relationship between two events, which plays an important role in natural language understanding. Most previous works follow a single-label classification style, classifying an event pair into either a specific temporal relation (e.g., \textit{Before}, \textit{After}), or a special label \textit{Vague} when there may be multiple possible temporal relations between the pair. In our work, instead of directly making predictions on \textit{Vague}, we propose a multi-label classification solution for ETRE (METRE) to infer the possibility of each temporal relation independently, where we treat \textit{Vague} as the cases when there is more than one possible relation between two events. We design a speculation mechanism to explore the possible relations hidden behind \textit{Vague}, which enables the latent information to be used efficiently. Experiments on TB-Dense, MATRES and UDS-T show that our method can effectively utilize the \textit{Vague} instances to improve the recognition for specific temporal relations and outperforms most state-of-the-art methods.
Vision-Language Models (VLMs) building upon the foundation of powerful large language models have made rapid progress in reasoning across visual and textual data. While VLMs perform well on vision tasks that they are trained on, our results highlight key challenges in abstract pattern recognition. We present GlyphPattern, a 954 item dataset that pairs 318 human-written descriptions of visual patterns from 40 writing systems with three visual presentation styles. GlyphPattern evaluates abstract pattern recognition in VLMs, requiring models to understand and judge natural language descriptions of visual patterns. GlyphPattern patterns are drawn from a large-scale cognitive science investigation of human writing systems; as a result, they are rich in spatial reference and compositionality. Our experiments show that GlyphPattern is challenging for state-of-the-art VLMs (GPT-4o achieves only 55% accuracy), with marginal gains from few-shot prompting. Our detailed error analysis reveals challenges at multiple levels, including visual processing, natural language understanding, and pattern generalization.