Research on neural networks has focused on understanding a single model trained on a single dataset. However, relatively little is known about the relationships between different models, particularly those trained or tested on different datasets. We address this by studying how the weight space and the underlying loss landscape of different models are interconnected. Specifically, we demonstrate that finetuned models that were optimized for high performance, reside in well-defined regions in weight space, and vice versa – that any model that resides anywhere in those regions also exhibits high performance. Notably, we show that language models that have been finetuned on the same dataset form a tight cluster in the weight space, while models finetuned on different datasets from the same underlying task form a looser cluster. Moreover, traversing around the region between the models leads to new models that perform comparably or even better than models obtained via finetuning, even on tasks that the original models were not finetuned on. Our findings provide insight into the relationships between models, demonstrating that a model positioned between two similar models can acquire the knowledge of both. We leverage this and design a method for selecting a better model for efficient finetuning. Specifically, we show that starting from the center of the region is as effective, if not more, than using the pretrained model in 11 out of 12 datasets, resulting in an average accuracy improvement of 3.06.

Previous studies observed that finetuned models may be better base models than the vanilla pretrained model. Such a model, finetuned on some source dataset, may provide a better starting point for a new finetuning process on a desired target dataset. Here, we perform a systematic analysis of this intertraining scheme, over a wide range of English classification tasks. Surprisingly, our analysis suggests that the potential intertraining gain can be analyzed independently for the target dataset under consideration, and for a base model being considered as a starting point. Hence, a performant model is generally strong, even if its training data was not aligned with the target dataset. Furthermore, we leverage our analysis to propose a practical and efficient approach to determine if and how to select a base model in real-world settings. Last, we release an updating ranking of best models in the HuggingFace hub per architecture.

Pretraining has been shown to scale well with compute, data size and data diversity. Multitask learning trains on a mixture of supervised datasets and produces improved performance compared to self-supervised pretraining. Until now, massively multitask learning required simultaneous access to all datasets in the mixture and heavy compute resources that are only available to well-resourced teams. In this paper, we propose ColD Fusion, a method that provides the benefits of multitask learning but leverages distributed computation and requires limited communication and no sharing of data. Consequentially, ColD Fusion can create a synergistic loop, where finetuned models can be recycled to continually improve the pretrained model they are based on. We show that ColD Fusion yields comparable benefits to multitask training by producing a model that (a) attains strong performance on all of the datasets it was multitask trained on and (b) is a better starting point for finetuning on unseen datasets. We find ColD Fusion outperforms RoBERTa and even previous multitask models. Specifically, when training and testing on 35 diverse datasets, ColD Fusion-based model outperforms RoBERTa by 2.19 points on average without any changes to the architecture.

The tutorial focuses on Debating Technologies, a sub-field of computational argumentation defined as “computational technologies developed directly to enhance, support, and engage with human debating” (Gurevych et al., 2016). A recent milestone in this field is Project Debater, which was revealed in 2019 as the first AI system that can debate human experts on complex topics. Project Debater is the third in the series of IBM Research AI’s grand challenges, following Deep Blue and Watson. It has been developed for over six years by a large team of researchers and engineers, and its live demonstration in February 2019 received massive media attention. This research effort has resulted in more than 50 scientific papers to date, and many datasets freely available for research purposes. We discuss the scientific challenges that arise when building such a system, including argument mining, argument quality assessment, stance classification, principled argument detection, narrative generation, and rebutting a human opponent. Many of the underlying capabilities of Project Debater have been made freely available for academic research, and the tutorial will include a detailed explanation of how to use and leverage these tools. In addition to discussing individual components, the tutorial also provides a holistic view of a debating system. Such a view is largely missing in the academic literature, where each paper typically addresses a specific problem in isolation. We present a complete pipeline of a debating system, and discuss the information flow and the interaction between the various components. Finally, we discuss practical applications and future challenges of debating technologies.

Project Debater was revealed in 2019 as the first AI system that can debate human experts on complex topics. Engaging in a live debate requires a diverse set of skills, and Project Debater has been developed accordingly as a collection of components, each designed to perform a specific subtask. Project Debater APIs provide access to many of these capabilities, as well as to more recently developed ones. This diverse set of web services, publicly available for academic use, includes core NLP services, argument mining and analysis capabilities, and higher-level services for content summarization. We describe these APIs and their performance, and demonstrate how they can be used for building practical solutions. In particular, we will focus on Key Point Analysis, a novel technology that identifies the main points and their prevalence in a collection of texts such as survey responses and user reviews.

We explore the task of automatic assessment of argument quality. To that end, we actively collected 6.3k arguments, more than a factor of five compared to previously examined data. Each argument was explicitly and carefully annotated for its quality. In addition, 14k pairs of arguments were annotated independently, identifying the higher quality argument in each pair. In spite of the inherent subjective nature of the task, both annotation schemes led to surprisingly consistent results. We release the labeled datasets to the community. Furthermore, we suggest neural methods based on a recently released language model, for argument ranking as well as for argument-pair classification. In the former task, our results are comparable to state-of-the-art; in the latter task our results significantly outperform earlier methods.

In this paper we suggest to leverage the partition of articles into sections, in order to learn thematic similarity metric between sentences. We assume that a sentence is thematically closer to sentences within its section than to sentences from other sections. Based on this assumption, we use Wikipedia articles to automatically create a large dataset of weakly labeled sentence triplets, composed of a pivot sentence, one sentence from the same section and one from another section. We train a triplet network to embed sentences from the same section closer. To test the performance of the learned embeddings, we create and release a sentence clustering benchmark. We show that the triplet network learns useful thematic metrics, that significantly outperform state-of-the-art semantic similarity methods and multipurpose embeddings on the task of thematic clustering of sentences. We also show that the learned embeddings perform well on the task of sentence semantic similarity prediction.