Large Language Models (LLMs) have been the subject of active research, significantly advancing the field of Natural Language Processing (NLP). From BERT to BLOOM, LLMs have surpassed state-of-the-art results in various natural language tasks such as question answering, summarization, and text generation. Many ongoing efforts focus on understanding LLMs' capabilities, including their knowledge of the world, syntax, and semantics. However, extending the textual prowess of LLMs to symbolic reasoning has been slow and predominantly focused on tackling problems related to the mathematical field. In this paper, we explore the use of LLMs for automated planning - a branch of AI concerned with the realization of action sequences (plans) to achieve a goal, typically executed by intelligent agents, autonomous robots, and unmanned vehicles. We introduce Plansformer; an LLM fine-tuned on planning problems and capable of generating plans with favorable behavior in terms of correctness and length with reduced knowledge-engineering efforts. We also demonstrate the adaptability of Plansformer in solving different planning domains with varying complexities, owing to the transfer learning abilities of LLMs. For one configuration of Plansformer, we achieve ~97% valid plans, out of which ~95% are optimal for Towers of Hanoi - a puzzle-solving domain.
Stance detection deals with the identification of an author's stance towards a target and is applied on various text domains like social media and news. In many cases, inferring the stance is challenging due to insufficient access to contextual information. Complementary context can be found in knowledge bases but integrating the context into pretrained language models is non-trivial due to their graph structure. In contrast, we explore an approach to integrate contextual information as text which aligns better with transformer architectures. Specifically, we train a model consisting of dual encoders which exchange information via cross-attention. This architecture allows for integrating contextual information from heterogeneous sources. We evaluate context extracted from structured knowledge sources and from prompting large language models. Our approach is able to outperform competitive baselines (1.9pp on average) on a large and diverse stance detection benchmark, both (1) in-domain, i.e. for seen targets, and (2) out-of-domain, i.e. for targets unseen during training. Our analysis shows that it is able to regularize for spurious label correlations with target-specific cue words.
Large Language Model (LLM) trained on the mixture of text and code has demonstrated impressive capability in translating natural language (NL) into structured code. In this work, we propose Code4Struct to leverage such text-to-structure translation capability to tackle structured prediction tasks in NLP. For example, Event Argument Extraction (EAE) aims to convert text into event-argument structures that can be represented as a class object using code. This alignment between structures and code enables us to take advantage of Programming Language (PL) features such as inheritance and type annotation to introduce external knowledge or add constraints with ease. We exploit the analogy between PL and NLP problems, and, as a case study, we use Code4Struct to tackle the EAE task using code generation. We ask a LLM to generate code to instantiate an event class with predicted arguments given a NL sentence. Despite only using 50 training instances for each event type, Code4Struct is comparable to fully-supervised models trained on 4,202 event instances and, when given the same 50-shot data, outperforms current state-of-the-art (SOTA) by 20.8% absolute F1. When prompted with hierarchical event types implemented using inheritance, Code4Struct can predict arguments for low-resource event types using 10-shot training instances from its sibling event type and outperforms zero-shot baseline by 12% absolute F1.
Stressors are related to depression, but this relationship is complex. We investigate the relationship between open-ended text responses about stressors and depressive symptoms across gender and racial/ethnic groups. First, we use topic models and other NLP tools to find thematic and vocabulary differences when reporting stressors across demographic groups. We train language models using self-reported stressors to predict depressive symptoms, finding a relationship between stressors and depression. Finally, we find that differences in stressors translate to downstream performance differences across demographic groups.
Masked language models like BERT can perform text classification in a zero-shot fashion by reformulating downstream tasks as text infilling. However, this approach is highly sensitive to the template used to prompt the model, yet practitioners are blind when designing them in strict zero-shot settings. In this paper, we propose an alternative mining-based approach for zero-shot learning. Instead of prompting language models, we use regular expressions to mine labeled examples from unlabeled corpora, which can optionally be filtered through prompting, and used to finetune a pretrained model. Our method is more flexible and interpretable than prompting, and outperforms it on a wide range of tasks when using comparable templates. Our results suggest that the success of prompting can partly be explained by the model being exposed to similar examples during pretraining, which can be directly retrieved through regular expressions.
We introduce two data augmentation techniques, which, used with a Resnet-BiLSTM-CTC network, significantly reduce Word Error Rate (WER) and Character Error Rate (CER) beyond best-reported results on handwriting text recognition (HTR) tasks. We apply a novel augmentation that simulates strikethrough text (HandWritten Blots) and a handwritten text generation method based on printed text (StackMix), which proved to be very effective in HTR tasks. StackMix uses weakly-supervised framework to get character boundaries. Because these data augmentation techniques are independent of the network used, they could also be applied to enhance the performance of other networks and approaches to HTR. Extensive experiments on ten handwritten text datasets show that HandWritten Blots augmentation and StackMix significantly improve the quality of HTR models
Very large language models such as GPT-3 have shown impressive performance across a wide variety of tasks, including text summarization. In this paper, we show that this strong performance extends to opinion summarization. We explore several pipeline methods for applying GPT-3 to summarize a large collection of user reviews in a zero-shot fashion, notably approaches based on recursive summarization and selecting salient content to summarize through supervised clustering or extraction. On two datasets, an aspect-oriented summarization dataset of hotel reviews and a generic summarization dataset of Amazon and Yelp reviews, we show that the GPT-3 models achieve very strong performance in human evaluation. We argue that standard evaluation metrics do not reflect this, and evaluate against several new measures targeting faithfulness, factuality, and genericity to contrast these different methods.
State space models (SSMs) have demonstrated state-of-the-art sequence modeling performance in some modalities, but underperform attention in language modeling. Moreover, despite scaling nearly linearly in sequence length instead of quadratically, SSMs are still slower than Transformers due to poor hardware utilization. In this paper, we make progress on understanding the expressivity gap between SSMs and attention in language modeling, and on reducing the hardware barrier between SSMs and attention. First, we use synthetic language modeling tasks to understand the gap between SSMs and attention. We find that existing SSMs struggle with two capabilities: recalling earlier tokens in the sequence and comparing tokens across the sequence. To understand the impact on language modeling, we propose a new SSM layer, H3, that is explicitly designed for these abilities. H3 matches attention on the synthetic languages and comes within 0.4 PPL of Transformers on OpenWebText. Furthermore, a hybrid 125M-parameter H3-attention model that retains two attention layers surprisingly outperforms Transformers on OpenWebText by 1.0 PPL. Next, to improve the efficiency of training SSMs on modern hardware, we propose FlashConv. FlashConv uses a fused block FFT algorithm to improve efficiency on sequences up to 8K, and introduces a novel state passing algorithm that exploits the recurrent properties of SSMs to scale to longer sequences. FlashConv yields 2$\times$ speedup on the long-range arena benchmark and allows hybrid language models to generate text 1.6$\times$ faster than Transformers. Using FlashConv, we scale hybrid H3-attention language models up to 1.3B parameters on the Pile and find promising initial results, achieving lower perplexity than Transformers and outperforming Transformers in zero- and few-shot learning on a majority of tasks in the SuperGLUE benchmark.
Pixel-level labels are particularly expensive to acquire. Hence, pretraining is a critical step to improve models on a task like semantic segmentation. However, prominent algorithms for pretraining neural networks use image-level objectives, e.g. image classification, image-text alignment a la CLIP, or self-supervised contrastive learning. These objectives do not model spatial information, which might be suboptimal when finetuning on downstream tasks with spatial reasoning. In this work, we propose to pretrain networks for semantic segmentation by predicting the relative location of image parts. We formulate this task as a classification problem where each patch in a query view has to predict its position relatively to another reference view. We control the difficulty of the task by masking a subset of the reference patch features visible to those of the query. Our experiments show that this location-aware (LOCA) self-supervised pretraining leads to representations that transfer competitively to several challenging semantic segmentation benchmarks.
Solving the challenges of automatic machine translation of Building Automation System text metadata is a crucial first step in efficiently deploying smart building applications. The vocabulary used to describe building metadata appears small compared to general natural languages, but each term has multiple commonly used abbreviations. Conventional machine learning techniques are inefficient since they need to learn many different forms for the same word, and large amounts of data must be used to train these models. It is also difficult to apply standard techniques such as tokenisation since this commonly results in multiple output tags being associated with a single input token, something traditional sequence labelling models do not allow. Finite State Transducers can model sequence-to-sequence tasks where the input and output sequences are different lengths, and they can be combined with language models to ensure a valid output sequence is generated. We perform a preliminary analysis into the use of transducer-based language models to parse and normalise building point metadata.