生成式大型语言模型（LLM）得益于文本生成的普适性，已成为少样本和零样本学习的主流选择。然而，许多用户在仅希望自动化分类任务时，并不需要生成式LLM的广泛能力。较小的类BERT模型也能学习通用任务，使其无需微调即可执行任意文本分类任务（零样本分类），或仅需少量样本即可学习新任务（少样本），同时效率显著高于生成式LLM。本文（1）解释了如何将自然语言推理（NLI）用作通用分类任务，其遵循与生成式LLM指令微调相似的原理，（2）提供了分步指南及可复用的Jupyter notebook用于构建通用分类器，（3）分享了在33个数据集和389个多样化类别上训练得到的通用分类器。我们分享的部分代码曾用于训练早期的零样本分类器，截至2023年12月，该分类器在Hugging Face Hub上的下载量已超过5500万次。我们的新分类器将零样本性能提升了9.4%。
英文摘要：
Generative Large Language Models (LLMs) have become the mainstream choice for fewshot and zeroshot learning thanks to the universality of text generation. Many users, however, do not need the broad capabilities of generative LLMs when they only want to automate a classification task. Smaller BERT-like models can also learn universal tasks, which allow them to do any text classification task without requiring fine-tuning (zeroshot classification) or to learn new tasks with only a few examples (fewshot), while being significantly more efficient than generative LLMs. This paper (1) explains how Natural Language Inference (NLI) can be used as a universal classification task that follows similar principles as instruction fine-tuning of generative LLMs, (2) provides a step-by-step guide with reusable Jupyter notebooks for building a universal classifier, and (3) shares the resulting universal classifier that is trained on 33 datasets with 389 diverse classes. Parts of the code we share has been used to train our older zeroshot classifiers that have been downloaded more than 55 million times via the Hugging Face Hub as of December 2023. Our new classifier improves zeroshot performance by 9.4%.
University of Illinois at Chicago（伊利诺伊大学芝加哥分校）；Salesforce Research（Salesforce研究中心）；Huazhong University of Science and Technology（华中科技大学）
Intent detection is one of the core components of goal-oriented dialog systems, and detecting out-of-scope (OOS) intents is also a practically important skill. Few-shot learning is attracting much attention to mitigate data scarcity, but OOS detection becomes even more challenging. In this paper, we present a simple yet effective approach, discriminative nearest neighbor classification with deep self-attention. Unlike softmax classifiers, we leverage BERT-style pairwise encoding to train a binary classifier that estimates the best matched training example for a user input. We propose to boost …
University of Chicago（芝加哥大学）；Toyota Technological Institute at Chicago（芝加哥丰田理工学院）；Peking University（北京大学）；Peng Cheng Laboratory（鹏城实验室）
While discriminative neural network classifiers are generally preferred, recent work has shown advantages of generative classifiers in term of data efficiency and robustness. In this paper, we focus on natural language inference (NLI). We propose GenNLI, a generative classifier for NLI tasks, and empirically characterize its performance by comparing it to five baselines, including discriminative models and large-scale pretrained language representation models like BERT. We explore training objectives for discriminative fine-tuning of our generative classifiers, showing improvements over log lo…
CID GmbH（CID有限公司）；Technical University of Darmstadt（达姆施塔特工业大学）
In many cases of machine learning, research suggests that the development of training data might have a higher relevance than the choice and modelling of classifiers themselves. Thus, data augmentation methods have been developed to improve classifiers by artificially created training data. In NLP, there is the challenge of establishing universal rules for text transformations which provide new linguistic patterns. In this paper, we present and evaluate a text generation method suitable to increase the performance of classifiers for long and short texts. We achieved promising improvements when…
Recent work has demonstrated the vulnerability of modern text classifiers to universal adversarial attacks, which are input-agnostic sequences of words added to text processed by classifiers. Despite being successful, the word sequences produced in such attacks are often ungrammatical and can be easily distinguished from natural text. We develop adversarial attacks that appear closer to natural English phrases and yet confuse classification systems when added to benign inputs. We leverage an adversarially regularized autoencoder (ARAE) to generate triggers and propose a gradient-based search t…
Jožef Stefan Institute（约瑟夫·斯蒂芬研究所）；Jožef Stefan International Postgraduate School（约瑟夫·斯蒂芬国际研究生院）
Natural Language Inference is an important task for Natural Language Understanding. It is concerned with classifying the logical relation between two sentences. In this paper, we propose several text generative neural networks for generating text hypothesis, which allows construction of new Natural Language Inference datasets. To evaluate the models, we propose a new metric -- the accuracy of the classifier trained on the generated dataset. The accuracy obtained by our best generative model is only 2.7% lower than the accuracy of the classifier trained on the original, human crafted dataset. F…
University of Maryland（马里兰大学）；University of Maryland, Baltimore County（马里兰大学巴尔的摩县分校）
Given a state-of-the-art deep neural network text classifier, we show the existence of a universal and very small perturbation vector (in the embedding space) that causes natural text to be misclassified with high probability. Unlike images on which a single fixed-size adversarial perturbation can be found, text is of variable length, so we define the "universality" as "token-agnostic", where a single perturbation is applied to each token, resulting in different perturbations of flexible sizes at the sequence level. We propose an algorithm to compute universal adversarial perturbations, and sh…
Facebook AI Research（Facebook人工智能研究院）；LIUM, Université Le Mans（勒芒大学LIUM实验室）
Many modern NLP systems rely on word embeddings, previously trained in an unsupervised manner on large corpora, as base features. Efforts to obtain embeddings for larger chunks of text, such as sentences, have however not been so successful. Several attempts at learning unsupervised representations of sentences have not reached satisfactory enough performance to be widely adopted. In this paper, we show how universal sentence representations trained using the supervised data of the Stanford Natural Language Inference datasets can consistently outperform unsupervised methods like SkipThought ve…
ShanghaiTech University（上海科技大学）；Shanghai Engineering Research Center of Intelligent Vision and Imaging（上海智能视觉成像工程研究中心）；DAMO Academy, Alibaba Group（阿里巴巴集团达摩院）
This paper presents the system used in our submission to the \textit{IWPT 2020 Shared Task}. Our system is a graph-based parser with second-order inference. For the low-resource Tamil corpus, we specially mixed the training data of Tamil with other languages and significantly improved the performance of Tamil. Due to our misunderstanding of the submission requirements, we submitted graphs that are not connected, which makes our system only rank \textbf{6th} over 10 teams. However, after we fixed this problem, our system is 0.6 ELAS higher than the team that ranked \textbf{1st} in the official …
Causal understanding is essential for many kinds of decision-making, but causal inference from observational data has typically only been applied to structured, low-dimensional datasets. While text classifiers produce low-dimensional outputs, their use in causal inference has not previously been studied. To facilitate causal analyses based on language data, we consider the role that text classifiers can play in causal inference through established modeling mechanisms from the causality literature on missing data and measurement error. We demonstrate how to conduct causal analyses using text cl…
We study the problem of combining the outcomes of several different classifiers in a way that provides a coherent inference that satisfies some constraints. In particular, we develop two general approaches for an important subproblem-identifying phrase structure. The first is a Markovian approach that extends standard HMMs to allow the use of a rich observation structure and of general classifiers to model state-observation dependencies. The second is an extension of constraint satisfaction formalisms. We develop efficient combination algorithms under both models and study them experimentally …
Discovering the existence of universal adversarial perturbations had large theoretical and practical impacts on the field of adversarial learning. In the text domain, most universal studies focused on adversarial prefixes which are added to all texts. However, unlike the vision domain, adding the same perturbation to different inputs results in noticeably unnatural inputs. Therefore, we introduce a new universal adversarial setup - a universal adversarial policy, which has many advantages of other universal attacks but also results in valid texts - thus making it relevant in practice. We achie…
Beihang University（北京航空航天大学）；Microsoft Research Asia（微软亚洲研究院）；University of California, San Diego（加利福尼亚大学圣地亚哥分校）
In this paper, we propose Patience-based Early Exit, a straightforward yet effective inference method that can be used as a plug-and-play technique to simultaneously improve the efficiency and robustness of a pretrained language model (PLM). To achieve this, our approach couples an internal-classifier with each layer of a PLM and dynamically stops inference when the intermediate predictions of the internal classifiers remain unchanged for a pre-defined number of steps. Our approach improves inference efficiency as it allows the model to make a prediction with fewer layers. Meanwhile, experimen…
