Most existing retrieval-augmented language models (LMs) assume a naive dichotomy within a retrieved document set: query-relevance and irrelevance. Our work investigates a more challenging scenario in which even the “relevant” documents may contain misleading or incorrect information, causing conflict among the retrieved documents and thereby negatively influencing model decisions as noise. We observe that existing LMs are highly brittle to the presence of conflicting information in both the fine-tuning and in-context few-shot learning scenarios. We propose approaches for handling knowledge conflicts among retrieved documents by explicitly fine-tuning a discriminator or prompting GPT-3.5 to elicit its discriminative capability. Our empirical results on open-domain QA show that these approaches significantly enhance model robustness. We also provide our findings on incorporating the fine-tuned discriminator’s decision into the in-context learning process, proposing a way to exploit the benefits of two disparate learning schemes. Alongside our findings, we provide MacNoise, a machine-generated, conflict-induced dataset to further encourage research in this direction.
Recent works have demonstrated the effectiveness of self-alignment in which a large language model is aligned to follow general instructions using instructional data generated from the model itself starting from a handful of human-written seeds. Instead of general alignment, in this work, we focus on self-alignment for expert domain specialization (e.g., biomedicine, finance). As a preliminary, we quantitively show the marginal effect that generic instruction-following training has on downstream expert domains’ performance. To remedy this, we propose self-specialization - allowing for effective model specialization while achieving cross-task generalization by leveraging only a few labeled seeds. Self-specialization offers a data- and parameter-efficient way of “carving out” an expert model out of a generalist pre-trained LLM. Exploring a variety of popular open large models as a base for specialization, our experimental results in both biomedical and financial domains show that our self-specialized models outperform their base models by a large margin, and even larger models that are generally instruction-tuned or that have been adapted to the target domain by other means.
Fine-tuning large models is highly effective, however, inference can be expensive and produces carbon emissions. Knowledge distillation has been shown to be a practical solution to reduce inference costs, but the distillation process itself requires significant computational resources. Rather than buying or renting GPUs to fine-tune, then distill a large model, an NLP practitioner might instead choose to allocate the available budget to hire annotators and manually label additional fine-tuning data. In this paper, we investigate how to most efficiently use a fixed budget to build a compact model. Through extensive experiments on six diverse tasks, we show that distilling from T5-XXL (11B) to T5-Small (60M) is almost always a cost-efficient strategy compared to annotating more data to directly train a compact model (T5-Small). We further investigate how the optimal budget allocated towards computation varies across scenarios. We will make our code, datasets, annotation cost estimates, and baseline models available as a benchmark to support further work on cost-efficient training of compact models.
A graph is a suitable data structure to represent the structural information of text. Recently, multi-hop question answering (MHQA) tasks, which require inter-paragraph/sentence linkages, have come to exploit such properties of a graph. Previous approaches to MHQA relied on leveraging the graph information along with the pre-trained language model (PLM) encoders. However, this trend exhibits the following drawbacks: (i) sample inefficiency while training in a low-resource setting; (ii) lack of reusability due to changes in the model structure or input. Our work proposes the Graph-Induced Transformer (GIT) that applies graph-derived attention patterns directly into a PLM, without the need to employ external graph modules. GIT can leverage the useful inductive bias of graphs while retaining the unperturbed Transformer structure and parameters. Our experiments on HotpotQA successfully demonstrate both the sample efficient characteristic of GIT and its capacity to replace the graph modules while preserving model performance.
Numerical reasoning over text is a challenging subtask in question answering (QA) that requires both the understanding of texts and numbers. However, existing language models in these numerical reasoning QA models tend to overly rely on the pre-existing parametric knowledge at inference time, which commonly causes hallucination in interpreting numbers. Our work proposes a novel attention masked reasoning model, the NC-BERT, that learns to leverage the number-related contextual knowledge to alleviate the over-reliance on parametric knowledge and enhance the numerical reasoning capabilities of the QA model. The empirical results suggest that understanding of numbers in their context by reducing the parametric knowledge influence, and refining numerical information in the number embeddings lead to improved numerical reasoning accuracy and performance in DROP, a numerical QA dataset.
Customer reviews are useful in providing an indirect, secondhand experience of a product. People often use reviews written by other customers as a guideline prior to purchasing a product. Such behavior signifies the authenticity of reviews in e-commerce platforms. However, fake reviews are increasingly becoming a hassle for both consumers and product owners. To address this issue, we propose You Only Need Gold (YONG), an essential information mining tool for detecting fake reviews and augmenting user discretion. Our experimental results show the poor human performance on fake review detection, substantially improved user capability given our tool, and the ultimate need for user reliance on the tool.
The semantic matching capabilities of neural information retrieval can ameliorate synonymy and polysemy problems of symbolic approaches. However, neural models’ dense representations are more suitable for re-ranking, due to their inefficiency. Sparse representations, either in symbolic or latent form, are more efficient with an inverted index. Taking the merits of the sparse and dense representations, we propose an ultra-high dimensional (UHD) representation scheme equipped with directly controllable sparsity. UHD’s large capacity and minimal noise and interference among the dimensions allow for binarized representations, which are highly efficient for storage and search. Also proposed is a bucketing method, where the embeddings from multiple layers of BERT are selected/merged to represent diverse linguistic aspects. We test our models with MS MARCO and TREC CAR, showing that our models outperforms other sparse models.
Tag recommendation relies on either a ranking function for top-k tags or an autoregressive generation method. However, the previous methods neglect one of two seemingly conflicting yet desirable characteristics of a tag set: orderlessness and inter-dependency. While the ranking approach fails to address the inter-dependency among tags when they are ranked, the autoregressive approach fails to take orderlessness into account because it is designed to utilize sequential relations among tokens. We propose a sequence-oblivious generation method for tag recommendation, in which the next tag to be generated is independent of the order of the generated tags and the order of the ground truth tags occurring in training data. Empirical results on two different domains, Instagram and Stack Overflow, show that our method is significantly superior to the previous approaches.
Numerical reasoning in machine reading comprehension (MRC) has shown drastic improvements over the past few years. While the previous models for numerical MRC are able to interpolate the learned numerical reasoning capabilities, it is not clear whether they can perform just as well on numbers unseen in the training dataset. Our work rigorously tests state-of-the-art models on DROP, a numerical MRC dataset, to see if they can handle passages that contain out-of-range numbers. One of the key findings is that the models fail to extrapolate to unseen numbers. Presenting numbers as digit-by-digit input to the model, we also propose the E-digit number form that alleviates the lack of extrapolation in models and reveals the need to treat numbers differently from regular words in the text. Our work provides a valuable insight into the numerical MRC models and the way to represent number forms in MRC.
Unsupervised question answering (UQA) has been proposed to avoid the high cost of creating high-quality datasets for QA. One approach to UQA is to train a QA model with questions generated automatically. However, the generated questions are either too similar to a word sequence in the context or too drifted from the semantics of the context, thereby making it difficult to train a robust QA model. We propose a novel regularization method based on teacher-student architecture to avoid bias toward a particular question generation strategy and modulate the process of generating individual words when a question is generated. Our experiments demonstrate that we have achieved the goal of generating higher-quality questions for UQA across diverse QA datasets and tasks. We also show that this method can be useful for creating a QA model with few-shot learning.
Advances in Question Answering (QA) research require additional datasets for new domains, languages, and types of questions, as well as for performance increases. Human creation of a QA dataset like SQuAD, however, is expensive. As an alternative, an unsupervised QA approach has been proposed so that QA training data can be generated automatically. However, the performance of unsupervised QA is much lower than that of supervised QA models. We identify two anomalies in the automatically generated questions and propose how they can be mitigated. We show our approach helps improve unsupervised QA significantly across a number of QA tasks.
Question Generation (QG) is a Natural Language Processing (NLP) task that aids advances in Question Answering (QA) and conversational assistants. Existing models focus on generating a question based on a text and possibly the answer to the generated question. They need to determine the type of interrogative word to be generated while having to pay attention to the grammar and vocabulary of the question. In this work, we propose Interrogative-Word-Aware Question Generation (IWAQG), a pipelined system composed of two modules: an interrogative word classifier and a QG model. The first module predicts the interrogative word that is provided to the second module to create the question. Owing to an increased recall of deciding the interrogative words to be used for the generated questions, the proposed model achieves new state-of-the-art results on the task of QG in SQuAD, improving from 46.58 to 47.69 in BLEU-1, 17.55 to 18.53 in BLEU-4, 21.24 to 22.33 in METEOR, and from 44.53 to 46.94 in ROUGE-L.