论文检索

Zhe Yang, Damai Dai, Peiyi Wang, Zhifang Sui

Large Language Models (LLMs) have recently gained the In-Context Learning (ICL) ability with the models scaling up, allowing them to quickly adapt to downstream tasks with only a few demonstration examples prepended in the input sequence. Nonetheless, the current practice of ICL treats all demonstration examples equally, which still warrants improvement, as the quality of examples is usually uneven. In this paper, we investigate how to determine approximately optimal weights for demonstration examples and how to apply them during ICL. To assess the quality of weights in the absence of additional validation data, we design a masked self-prediction (MSP) score that exhibits a strong correlation with the final ICL performance. To expedite the weight-searching process, we discretize the continuous weight space and adopt beam search. With approximately optimal weights obtained, we further propose two strategies to apply them to demonstrations at different model positions. Experimental results on 8 text classification tasks show that our approach outperforms conventional ICL by a large margin. Our code are publicly available at https:github.com/Zhe-Young/WICL.

Max Müller-Eberstein, Rob van der Goot, Barbara Plank, Ivan Titov

Representational spaces learned via language modeling are fundamental to Natural Language Processing (NLP), however there has been limited understanding regarding how and when during training various types of linguistic information emerge and interact. Leveraging a novel information theoretic probing suite, which enables direct comparisons of not just task performance, but their representational subspaces, we analyze nine tasks covering syntax, semantics and reasoning, across 2M pre-training steps and five seeds. We identify critical learning phases across tasks and time, during which subspaces emerge, share information, and later disentangle to specialize. Across these phases, syntactic knowledge is acquired rapidly after 0.5% of full training. Continued performance improvements primarily stem from the acquisition of open-domain knowledge, while semantics and reasoning tasks benefit from later boosts to long-range contextualization and higher specialization. Measuring cross-task similarity further reveals that linguistically related tasks share information throughout training, and do so more during the critical phase of learning than before or after. Our findings have implications for model interpretability, multi-task learning, and learning from limited data.

Viet Dac Lai, Nghia Ngo, Amir Pouran Ben Veyseh, Hieu Man, Franck Dernoncourt, Trung Bui, Thien Huu Nguyen

Over the last few years, large language models (LLMs) have emerged as the most important breakthroughs in natural language processing (NLP) that fundamentally transform research and developments in the field. ChatGPT represents one of the most exciting LLM systems developed recently to showcase impressive skills for language generation and highly attract public attention. Among various exciting applications discovered for ChatGPT in English, the model can process and generate texts for multiple languages due to its multilingual training data. Given the broad adoption of ChatGPT for English in different problems and areas, a natural question is whether ChatGPT can also be applied effectively for other languages or it is necessary to develop more language-specific technologies. The answer to this question requires a thorough evaluation of ChatGPT over multiple tasks with diverse languages and large datasets (i.e., beyond reported anecdotes), which is still missing or limited in current research. Our work aims to fill this gap for the evaluation of ChatGPT and similar LLMs to provide more comprehensive information for multilingual NLP applications. In particular, we evaluate ChatGPT on 7 different tasks, covering 37 diverse languages with high, medium, low, and extremely low resources. Compared to the performance of previous models, our extensive experiments demonstrate the worse performance of ChatGPT for different NLP tasks and languages, calling for further research to develop better models and understanding for multilingual learning.

Chen Shani, Jilles Vreeken, Dafna Shahaf

Concepts play a pivotal role in various human cognitive functions, including learning, reasoning and communication. However, there is very little work on endowing machines with the ability to form and reason with concepts. In particular, state-of-the-art large language models (LLMs) work at the level of tokens, not concepts. In this work, we analyze how well contemporary LLMs capture human concepts and their structure. We then discuss ways to develop concept-aware LLMs, taking place at different stages of the pipeline. We sketch a method for pretraining LLMs using concepts, and also explore the simpler approach that uses the output of existing LLMs. Despite its simplicity, our proof-of-concept is shown to better match human intuition, as well as improve the robustness of predictions. These preliminary results underscore the promise of concept-aware LLMs.

Xin Sun, Qiang Liu, Shu Wu, Zilei Wang, Liang Wang

Distantly supervised relation extraction (DSRE) aims to extract relational facts from texts but suffers from noisy instances. To mitigate the influence of noisy labels, current methods typically use the Multi-Instance-Learning framework to extract relations for each bag. However, these approaches are not capable of extracting relation labels for individual sentences. Several studies have focused on sentence-level DSRE to solve the above problem. These studies primarily aim to develop methods for identifying noisy samples and filtering them out to mitigate the impact of noise. However, discarding noisy samples directly leads to the loss of useful information. To this end, we propose SSLRE, a novel Semi-Supervised-Learning Relation Extraction framework for sentence-level DSRE. We discard only the labels of the noisy samples and utilize these instances without labels as unlabeled samples. Our SSLRE framework utilizes a weighted K-NN graph to select confident samples as labeled data and the rest as unlabeled. We then design a robust semi-supervised learning framework that can efficiently handle remaining label noise present in the labeled dataset, while also making effective use of unlabeled samples. Based on our experiments on two real-world datasets, the SSLRE framework we proposed has achieved significant enhancements in sentence-level relation extraction performance compared to the existing state-of-the-art methods. Moreover, it has also attained a state-of-the-art level of performance in bag-level relation extraction with ONE aggregation strategy.

Taejun Yun, Jinhyeon Kim, Deokyeong Kang, Seonghoon Lim, Jihoon Kim, Taeuk Kim

Cross-lingual transfer (XLT) is an emergent ability of multilingual language models that preserves their performance on a task to a significant extent when evaluated in languages that were not included in the fine-tuning process. While English, due to its widespread usage, is typically regarded as the primary language for model adaption in various tasks, recent studies have revealed that the efficacy of XLT can be amplified by selecting the most appropriate source languages based on specific conditions. In this work, we propose the utilization of sub-network similarity between two languages as a proxy for predicting the compatibility of the languages in the context of XLT. Our approach is model-oriented, better reflecting the inner workings of foundation models. In addition, it requires only a moderate amount of raw text from candidate languages, distinguishing it from the majority of previous methods that rely on external resources. In experiments, we demonstrate that our method is more effective than baselines across diverse tasks. Specifically, it shows proficiency in ranking candidates for zero-shot XLT, achieving an improvement of 4.6% on average in terms of NDCG@3. We also provide extensive analyses that confirm the utility of sub-networks for XLT prediction.

Weng Tam, Xiao Liu, Kaixuan Ji, Lilong Xue, Jiahua Liu, Tao Li, Yuxiao Dong, Jie Tang

Prompt tuning attempts to update few task-specific parameters in pre-trained models. It has achieved comparable performance to fine-tuning of the full parameter set on both language understanding and generation tasks. In this work, we study the problem of prompt tuning for neural text retrievers. We introduce parameter-efficient prompt tuning for text retrieval across in-domain, cross-domain, and cross-topic settings. Through an extensive analysis, we show that the strategy can mitigate the two issues—parameter-inefficiency and weak generalizability—faced by fine-tuning based retrieval methods. Notably, it can significantly improve the out-of-domain zero-shot generalization of the retrieval models. By updating only 0.1% of the model parameters, the prompt tuning strategy can help retrieval models achieve better generalization performance than traditional methods in which all parameters are updated. Finally, to facilitate research on retrievers’ cross-topic generalizability, we curate and release an academic retrieval dataset with 18K query-results pairs in 87 topics, making it the largest topic-specific one to date.

Rituraj Singh, Rahul Kumar, Vivek Sridhar

Negation is the fundamental component in a natural language that reverses the semantic meaning of a sentence. It plays an extremely important role across a wide range of applications, yet they are underrepresented in pre-trained language models (LMs), resulting often in wrong inferences. In this work, we try to improve the underlying understanding of the negation in the pre-trained LMs. To augment negation understanding, we propose a language model objective with a weighted cross-entropy loss and elastic weight consolidation regularization. We reduce the mean top 1 error rate for BERT-base to 1.1%, BERT-large to 0.78%, RoBERTA-base to 3.74%, RoBERTA-large to 0.01% on the negated LAMA dataset. It minimizes the BERT error rate by a margin of 8% and also outperform the existing negation models. We also provide empirical evidences that negated augmented models outperform the classical models on original as well as negation benchmarks on natural language inference tasks.

Ruoyu Zhang, Yanzeng Li, Yongliang Ma, Ming Zhou, Lei Zou

Prevalent supervised learning methods in natural language processing (NLP) are notoriously data-hungry, which demand large amounts of high-quality annotated data. In practice, acquiring such data is a costly endeavor. Recently, the superior few-shot performance of large language models (LLMs) has propelled the development of dataset generation, where the training data are solely synthesized from LLMs. However, such an approach usually suffers from low-quality issues, and requires orders of magnitude more labeled data to achieve satisfactory performance. To fully exploit the potential of LLMs and make use of massive unlabeled data, we propose LLMaAA, which takes LLMs as annotators and puts them into an active learning loop to determine what to annotate efficiently. To learn robustly with pseudo labels, we optimize both the annotation and training processes: (1) we draw k-NN examples from a small demonstration pool as in-context examples, and (2) we adopt the example reweighting technique to assign training samples with learnable weights. Compared with previous approaches, LLMaAA features both efficiency and reliability. We conduct experiments and analysis on two classic NLP tasks, named entity recognition and relation extraction. With LLMaAA, task-specific models trained from LLM-generated labels can outperform the teacher within only hundreds of annotated examples, which is much more cost-effective than other baselines.

Jichen Yang, Kai Fan, Minpeng Liao, Boxing Chen, Zhongqiang Huang

One of the main problems in speech translation is the mismatches between different modalities. The second problem, scarcity of parallel data covering multiple modalities, means that the end-to-end multi-modal models tend to perform worse than cascade models, although there are exceptions under favorable conditions. To address these problems, we propose an end-to-end zero-shot speech translation model, connecting two pre-trained uni-modality modules via word rotator’s distance. The model retains the ability of zero-shot, which is like cascade models, and also can be trained in an end-to-end style to avoid error propagation. Our comprehensive experiments on the MuST-C benchmarks show that our end-to-end zero-shot approach performs better than or as well as those of the CTC-based cascade models and that our end-to-end model with supervised training also matches the latest baselines.

Han Zhou, Xingchen Wan, Ivan Vulić, Anna Korhonen

Prompt-based learning has been an effective paradigm for large pretrained language models (LLM), enabling few-shot or even zero-shot learning. Black-box prompt search has received growing interest recently for its distinctive properties of gradient-free optimization, proven particularly useful and powerful for model-as-a-service usage. However, the discrete nature and the complexity of combinatorial optimization hinder the efficiency of modern black-box approaches. Despite extensive research on search algorithms, the crucial aspect of search space design and optimization has been largely overlooked. In this paper, we first conduct a sensitivity analysis by prompting LLM, revealing that only a small number of tokens exert a disproportionate amount of influence on LLM predictions. Leveraging this insight, we propose the Clustering and Pruning for Efficient Black-box Prompt Search (ClaPS), a simple black-box search method that first clusters and prunes the search space to focus exclusively on influential prompt tokens. By employing even simple search methods within the pruned search space, ClaPS achieves state-of-the-art performance across various tasks and LLMs, surpassing the performance of complex approaches while significantly reducing search costs. Our findings underscore the critical role of search space design and optimization in enhancing both the usefulness and the efficiency of black-box prompt-based learning.

Zhiyuan Fan, Shizhu He

Open Information Extraction (OpenIE) is a fundamental yet challenging task in Natural Language Processing, which involves extracting all triples (subject, predicate, object) from a given sentence. While labelling-based methods have their merits, generation-based techniques offer unique advantages, such as the ability to generate tokens not present in the original sentence. However, these generation-based methods often require a significant amount of training data to learn the task form of OpenIE and substantial training time to overcome slow model convergence due to the order penalty. In this paper, we introduce a novel framework, OK-IE, that ingeniously transforms the task form of OpenIE into the pre-training task form of the T5 model, thereby reducing the need for extensive training data. Furthermore, we introduce an innovative concept of ‘anchors’ to control the sequence of model outputs, effectively eliminating the impact of order penalty on model convergence and significantly reducing training time. Experimental results indicate that, compared to previous SOTA methods, OK-IE requires only 1/100 of the training data (900 instances) and 1/120 of the training time (3 minutes) to achieve comparable results.

Changjiang Gao, Shujian Huang, Jixing Li, Jiajun Chen

Recent large language models (LLMs) have revealed strong abilities to understand natural language. Since most of them share the same basic structure, i.e. the transformer block, possible contributors to their success in the training process are scaling and instruction tuning. However, how these factors affect the models’ language perception is unclear. This work compares the self-attention of several existing LLMs (LLaMA, Alpaca and Vicuna) in different sizes (7B, 13B, 30B, 65B), together with eye saccade, an aspect of human reading attention, to assess the effect of scaling and instruction tuning on language perception. Results show that scaling enhances the human resemblance and improves the effective attention by reducing the trivial pattern reliance, while instruction tuning does not. However, instruction tuning significantly enhances the models’ sensitivity to instructions. We also find that current LLMs are consistently closer to non-native than native speakers in attention, suggesting a sub-optimal language perception of all models. Our code and data used in the analysis is available on GitHub.

Osman İnce, Tanin Zeraati, Semih Yagcioglu, Yadollah Yaghoobzadeh, Erkut Erdem, Aykut Erdem

Neural networks have revolutionized language modeling and excelled in various downstream tasks. However, the extent to which these models achieve compositional generalization comparable to human cognitive abilities remains a topic of debate. While existing approaches in the field have mainly focused on novel architectures and alternative learning paradigms, we introduce a pioneering method harnessing the power of dataset cartography (Swayamdipta et al., 2020). By strategically identifying a subset of compositional generalization data using this approach, we achieve a remarkable improvement in model accuracy, yielding enhancements of up to 10% on CFQ and COGS datasets. Notably, our technique incorporates dataset cartography as a curriculum learning criterion, eliminating the need for hyperparameter tuning while consistently achieving superior performance. Our findings highlight the untapped potential of dataset cartography in unleashing the full capabilities of compositional generalization within Transformer models.

Yangyang Luo, Shiyu Tian, Caixia Yuan, Xiaojie Wang

Conversational Machine Reading (CMR) requires answering a user’s initial question through multi-turn dialogue interactions based on a given document. Although there exist many effective methods, they largely neglected the alignment between the \textit{document} and the \textit{user-provided information}, which significantly affects the intermediate decision-making and subsequent follow-up question generation. To address this issue, we propose a pipeline framework that (1) aligns the aforementioned two sides in an explicit way, (2) makes decisions using a lightweight many-to-many entailment reasoning module, and (3) directly generates follow-up questions based on the document and previously asked questions. Our proposed method achieves state-of-the-art in micro-accuracy and ranks the first place on the public leaderboard of the CMR benchmark dataset ShARC.

Zhisong Zhang, Emma Strubell, Eduard Hovy

In this work we propose a pragmatic method that reduces the annotation cost for structured label spaces using active learning. Our approach leverages partial annotation, which reduces labeling costs for structured outputs by selecting only the most informative sub-structures for annotation. We also utilize self-training to incorporate the current model’s automatic predictions as pseudo-labels for un-annotated sub-structures. A key challenge in effectively combining partial annotation with self-training to reduce annotation cost is determining which sub-structures to select to label. To address this challenge, we adopt an error estimator to adaptively decide the partial selection ratio according to the current model’s capability. In evaluations spanning four structured prediction tasks, we show that our combination of partial annotation and self-training using an adaptive selection ratio reduces annotation cost over strong full annotation baselines under a fair comparison scheme that takes reading time into consideration.

Peter Jansen

In this work, we show that contemporary language models have a previously unknown skill – the capacity for electronic circuit design from high-level textual descriptions, akin to code generation. We introduce two benchmarks: PINS100, assessing model knowledge of electrical components, and MICRO25, evaluating a model’s capability to design common microcontroller circuits and code in the Arduino ecosystem that involve input, output, sensors, motors, protocols, and logic – with models such as GPT-4 and Claude-V1 achieving between 60% to 96% Pass@1 on generating full devices. We include six case studies of using language models as a design assistant for moderately complex devices, such as a radiation-powered random number generator, an emoji keyboard, a visible spectrometer, and several assistive devices, while offering a qualitative analysis performance, outlining evaluation challenges, and suggesting areas of development to improve complex circuit design and practical utility. With this work, we aim to spur research at the juncture of natural language processing and electronic design.

Dohee Kim, Ra Yoo, Soyoung Yang, Hee Yang, Jaegul Choo

Event extraction (EE), as a crucial information extraction (IE) task, aims to identify event triggers and their associated arguments from unstructured text, subsequently classifying them into pre-defined types and roles. In the biomedical domain, EE is widely used to extract complex structures representing biological events from literature. Due to the complicated semantics and specialized domain knowledge, it is challenging to construct biomedical event extraction datasets. Additionally, most existing biomedical EE datasets primarily focus on cell experiments or the overall experimental procedures. Therefore, we introduce AniEE, an event extraction dataset concentrated on the animal experiment stage. We establish a novel animal experiment customized entity and event scheme in collaboration with domain experts. We then create an expert-annotated high-quality dataset containing discontinuous entities and nested events and evaluate our dataset on the recent outstanding NER and EE models.

William Gantt, Reno Kriz, Yunmo Chen, Siddharth Vashishtha, Aaron Steven White

As information extraction (IE) systems have grown more adept at processing whole documents, the classic task of *template filling* has seen renewed interest as a benchmark for document-level IE. In this position paper, we call into question the suitability of template filling for this purpose. We argue that the task demands definitive answers to thorny questions of *event individuation* — the problem of distinguishing distinct events — about which even human experts disagree. Through an annotation study and error analysis, we show that this raises concerns about the usefulness of template filling metrics, the quality of datasets for the task, and the ability of models to learn it. Finally, we consider possible solutions.

Sahithya Ravi, Raymond Ng, Vered Shwartz

Understanding the speaker’s intended meaning often involves drawing commonsense inferences to reason about what is not stated explicitly. In multi-event sentences, it requires understanding the relationships between events based on contextual knowledge. We propose COMET-M (Multi-Event), an event-centric commonsense model capable of generating commonsense inferences for a target event within a complex sentence. COMET-M builds upon COMET (Bosselut et al., 2019), which excels at generating event-centric inferences for simple sentences, but struggles with the complexity of multi-event sentences prevalent in natural text. To overcome this limitation, we curate a Multi-Event Inference (MEI) dataset of 35K human-written inferences. We train COMET-M on the human-written inferences and also create baselines using automatically labeled examples. Experimental results demonstrate the significant performance improvement of COMET-M over COMET in generating multi-event inferences. Moreover, COMET-M successfully produces distinct inferences for each target event, taking the complete context into consideration. COMET-M holds promise for downstream tasks involving natural text such as coreference resolution, dialogue, and story understanding.