The impressive capabilities of recent language models can be largely attributed to the multi-trillion token pretraining datasets that they are trained on. However, model developers fail to disclose their construction methodology which has lead to a lack of open information on how to develop effective pretraining sets. To address this issue, we perform the first systematic study across the entire pipeline of pretraining set construction. First, we run ablations on existing techniques for pretraining set development to identify which methods translate to the largest gains in model accuracy on downstream evaluations. Then, we categorize the most widely used data source, web crawl snapshots, across the attributes of toxicity, quality, type of speech, and domain. Finally, we show how such attribute information can be used to further refine and improve the quality of a pretraining set. These findings constitute an actionable set of steps that practitioners can use to develop high quality pretraining sets.

The advancement of large language models (LLMs) has extended their use to dynamic and interactive real-world applications, where models engage continuously with their environment and potentially enhance their performance over time. Most existing LLM benchmarks evaluate LLMs on i.i.d. tasks, overlooking their ability to learn iteratively from past experiences. Our paper bridges this evaluation gap by proposing a novel framework, LLM-Evolve, which extends established benchmarks to sequential problem-solving settings. LLM-Evolve evaluates LLMs over multiple rounds, providing feedback after each round to build a demonstration memory that the models can query in future tasks. We applied LLM-Evolve to the MMLU, GSM8K, and AgentBench benchmarks, testing 8 state-of-the-art open-source and closed-source models. Results show that LLMs can achieve performance improvements of up to 17% by learning from past interactions, with the quality of retrieval algorithms and feedback significantly influencing this capability. These insights advocate for more understanding and benchmarks for LLMs’ performance in evolving interactive scenarios.

Closed-book question answering (QA) requires a model to directly answer an open-domain question without access to any external knowledge. Prior work on closed-book QA either directly finetunes or prompts a pretrained language model (LM) to leverage the stored knowledge. However, they do not fully exploit the parameterized knowledge. To address this inefficiency, we propose a two-stage, closed-book QA framework which employs a coarse-to-fine approach to extract the relevant knowledge and answer a question. We first generate a related context for a given question by prompting a pretrained LM. We then prompt the same LM to generate an answer using the generated context and the question. Additionally, we marginalize over the generated contexts to improve the accuracies and reduce context uncertainty. Experimental results on three QA benchmarks show that our method significantly outperforms previous closed-book QA methods. For example on TriviaQA, our method improves exact match accuracy from 55.3% to 68.6%, and is on par with open-book QA methods (68.6% vs. 68.0%). Our results show that our new methodology is able to better exploit the stored knowledge in pretrained LMs without adding extra learnable parameters or needing finetuning, and paves the way for hybrid models that integrate pretrained LMs with external knowledge.

Pretrained large language models have become indispensable for solving various natural language processing (NLP) tasks. However, safely deploying them in real world applications is challenging because they generate toxic content. To address this challenge, we propose two novel pretraining data augmentation strategies that significantly reduce model toxicity without compromising its utility. Our two strategies are: (1) MEDA: adds raw toxicity score as meta-data to the pretraining samples, and (2) INST: adds instructions to those samples indicating their toxicity. Our results indicate that our best performing strategy (INST) substantially reduces the toxicity probability up to 61% while preserving the accuracy on five benchmark NLP tasks as well as improving AUC scores on four bias detection tasks by 1.3%. We also demonstrate the generalizability of our techniques by scaling the number of training samples and the number of model parameters.

Parameter efficient learning methods (PERMs)have recently gained significant attention asthey provide an efficient way for pre-trainedlanguage models (PLMs) to adapt to a downstream task. However, these conclusions aremostly drawn from in-domain evaluations overthe full training set. In this paper, we presentcomparisons between PERMs and finetuningfrom three new perspectives: (1) the effect ofsample and model size to in-domain evaluations, (2) generalization to unseen domains andnew datasets, and (3) the faithfulness of generations. Our results show that for in-domainsettings (a) there is a cross point of samplesize for which PERMs will perform better thanfinetuning when training with fewer samples,and (b) larger PLMs have larger cross points.For cross-domain and cross-dataset cases, weshow that (a) Adapter (Houlsby et al., 2019)performs the best amongst all the PERMs studied here, and (b) it outperforms finetuning ifthe task dataset is below a certain size. Wealso compare the faithfulness of generationsand show that PERMs can achieve better faithfulness score than finetuning, especially forsmall training set, by as much as 6%. Finally,we apply Adapter to MT-NLG 530b (Smithet al., 2022) and achieve new state-of-the-artresults on Xsum (Narayan et al., 2018) for allROUGE scores (ROUGE-1 49.17, ROUGE-227.20, ROUGE-L 40.98).

Existing knowledge-grounded dialogue systems typically use finetuned versions of a pretrained language model (LM) and large-scale knowledge bases. These models typically fail to generalize on topics outside of the knowledge base, and require maintaining separate potentially large checkpoints each time finetuning is needed. In this paper, we aim to address these limitations by leveraging the inherent knowledge stored in the pretrained LM as well as its powerful generation ability. We propose a multi-stage prompting approach to generate knowledgeable responses from a single pretrained LM. We first prompt the LM to generate knowledge based on the dialogue context. Then, we further prompt it to generate responses based on the dialogue context and the previously generated knowledge. Results show that our knowledge generator outperforms the state-of-the-art retrieval-based model by 5.8% when combining knowledge relevance and correctness. In addition, our multi-stage prompting outperforms the finetuning-based dialogue model in terms of response knowledgeability and engagement by up to 10% and 5%, respectively. Furthermore, we scale our model up to 530 billion parameters and demonstrate that larger LMs improve the generation correctness score by up to 10%, and response relevance, knowledgeability and engagement by up to 10%. Our code is available at: https://github.com/NVIDIA/Megatron-LM.

Predicting the effects of unexpected situations is an important reasoning task, e.g., would cloudy skies help or hinder plant growth? Given a context, the goal of such situational reasoning is to elicit the consequences of a new situation (st) that arises in that context. We propose CURIE, a method to iteratively build a graph of relevant consequences explicitly in a structured situational graph (st graph) using natural language queries over a finetuned language model. Across multiple domains, CURIE generates st graphs that humans find relevant and meaningful in eliciting the consequences of a new situation (75% of the graphs were judged correct by humans). We present a case study of a situation reasoning end task (WIQA-QA), where simply augmenting their input with st graphs improves accuracy by 3 points. We show that these improvements mainly come from a hard subset of the data, that requires background knowledge and multi-hop reasoning.

Recent work in natural language processing (NLP) has focused on ethical challenges such as understanding and mitigating bias in data and algorithms; identifying objectionable content like hate speech, stereotypes and offensive language; and building frameworks for better system design and data handling practices. However, there has been little discussion about the ethical foundations that underlie these efforts. In this work, we study one ethical theory, namely deontological ethics, from the perspective of NLP. In particular, we focus on the generalization principle and the respect for autonomy through informed consent. We provide four case studies to demonstrate how these principles can be used with NLP systems. We also recommend directions to avoid the ethical issues in these systems.

Document grounded generation is the task of using the information provided in a document to improve text generation. This work focuses on two different document grounded generation tasks: Wikipedia Update Generation task and Dialogue response generation. Our work introduces two novel adaptations of large scale pre-trained encoder-decoder models focusing on building context driven representation of the document and enabling specific attention to the information in the document. Additionally, we provide a stronger BART baseline for these tasks. Our proposed techniques outperform existing methods on both automated (at least 48% increase in BLEU-4 points) and human evaluation for closeness to reference and relevance to the document. Furthermore, we perform comprehensive manual inspection of the generated output and categorize errors to provide insights into future directions in modeling these tasks.

Neural controllable text generation is an important area gaining attention due to its plethora of applications. Although there is a large body of prior work in controllable text generation, there is no unifying theme. In this work, we provide a new schema of the pipeline of the generation process by classifying it into five modules. The control of attributes in the generation process requires modification of these modules. We present an overview of different techniques used to perform the modulation of these modules. We also provide an analysis on the advantages and disadvantages of these techniques. We further pave ways to develop new architectures based on the combination of the modules described in this paper.

This paper introduces a new task of politeness transfer which involves converting non-polite sentences to polite sentences while preserving the meaning. We also provide a dataset of more than 1.39 instances automatically labeled for politeness to encourage benchmark evaluations on this new task. We design a tag and generate pipeline that identifies stylistic attributes and subsequently generates a sentence in the target style while preserving most of the source content. For politeness as well as five other transfer tasks, our model outperforms the state-of-the-art methods on automatic metrics for content preservation, with a comparable or better performance on style transfer accuracy. Additionally, our model surpasses existing methods on human evaluations for grammaticality, meaning preservation and transfer accuracy across all the six style transfer tasks. The data and code is located at https://github.com/tag-and-generate.

Sentence ordering is the task of arranging the sentences of a given text in the correct order. Recent work using deep neural networks for this task has framed it as a sequence prediction problem. In this paper, we propose a new framing of this task as a constraint solving problem and introduce a new technique to solve it. Additionally, we propose a human evaluation for this task. The results on both automatic and human metrics across four different datasets show that this new technique is better at capturing coherence in documents.

Recent work in neural generation has attracted significant interest in controlling the form of text, such as style, persona, and politeness. However, there has been less work on controlling neural text generation for content. This paper introduces the notion of Content Transfer for long-form text generation, where the task is to generate a next sentence in a document that both fits its context and is grounded in a content-rich external textual source such as a news story. Our experiments on Wikipedia data show significant improvements against competitive baselines. As another contribution of this paper, we release a benchmark dataset of 640k Wikipedia referenced sentences paired with the source articles to encourage exploration of this new task.

Visual storytelling is the task of generating stories based on a sequence of images. Inspired by the recent works in neural generation focusing on controlling the form of text, this paper explores the idea of generating these stories in different personas. However, one of the main challenges of performing this task is the lack of a dataset of visual stories in different personas. Having said that, there are independent datasets for both visual storytelling and annotated sentences for various persona. In this paper we describe an approach to overcome this by getting labelled persona data from a different task and leveraging those annotations to perform persona based story generation. We inspect various ways of incorporating personality in both the encoder and the decoder representations to steer the generation in the target direction. To this end, we propose five models which are incremental extensions to the baseline model to perform the task at hand. In our experiments we use five different personas to guide the generation process. We find that the models based on our hypotheses perform better at capturing words while generating stories in the target persona.

We critique recent work on ethics in natural language processing. Those discussions have focused on data collection, experimental design, and interventions in modeling. But we argue that we ought to first understand the frameworks of ethics that are being used to evaluate the fairness and justice of algorithmic systems. Here, we begin that discussion by outlining deontological and consequentialist ethics, and make predictions on the research agenda prioritized by each.

There is a long record of research on equity in schools. As machine learning researchers begin to study fairness and bias in earnest, language technologies in education have an unusually strong theoretical and applied foundation to build on. Here, we introduce concepts from culturally relevant pedagogy and other frameworks for teaching and learning, identifying future work on equity in NLP. We present case studies in a range of topics like intelligent tutoring systems, computer-assisted language learning, automated essay scoring, and sentiment analysis in classrooms, and provide an actionable agenda for research.

Style transfer is the task of rephrasing the text to contain specific stylistic properties without changing the intent or affect within the context. This paper introduces a new method for automatic style transfer. We first learn a latent representation of the input sentence which is grounded in a language translation model in order to better preserve the meaning of the sentence while reducing stylistic properties. Then adversarial generation techniques are used to make the output match the desired style. We evaluate this technique on three different style transformations: sentiment, gender and political slant. Compared to two state-of-the-art style transfer modeling techniques we show improvements both in automatic evaluation of style transfer and in manual evaluation of meaning preservation and fluency.

This paper introduces a document grounded dataset for conversations. We define “Document Grounded Conversations” as conversations that are about the contents of a specified document. In this dataset the specified documents were Wikipedia articles about popular movies. The dataset contains 4112 conversations with an average of 21.43 turns per conversation. This positions this dataset to not only provide a relevant chat history while generating responses but also provide a source of information that the models could use. We describe two neural architectures that provide benchmark performance on the task of generating the next response. We also evaluate our models for engagement and fluency, and find that the information from the document helps in generating more engaging and fluent responses.

There has been a long standing interest in understanding ‘Social Influence’ both in Social Sciences and in Computational Linguistics. In this paper, we present a novel approach to study and measure interpersonal influence in daily interactions. Motivated by the basic principles of influence, we attempt to identify indicative linguistic features of the posts in an online knitting community. We present the scheme used to operationalize and label the posts as influential or non-influential. Experiments with the identified features show an improvement in the classification accuracy of influence by 3.15%. Our results illustrate the important correlation between the structure of the language and its potential to influence others.