Recent prompt optimisation approaches use the generative nature of language models to produce prompts – even rivaling the performance of human-curated prompts. In this paper, we demonstrate that randomly sampling tokens from the model vocabulary as “separators” can be as effective as language models for prompt-style text classification. Our experiments show that random separators are competitive baselines, having less than a 1% difference compared to previous self-optimisation methods and showing a 12% average relative improvement over strong human baselines across nine text classification tasks and eight language models. We further analyse this phenomenon in detail using three different random generation strategies, establishing that the language space is rich with potentially good separators, with a greater than 40% average chance that a randomly drawn separator performs better than human-curated separators. These observations challenge the common assumption that an effective prompt should be human readable or task relevant and establish a strong baseline for prompt optimisation research.

Large language models (LLMs) exhibit positional bias in how they use context, which especially affects listwise ranking. To address this, we propose permutation self-consistency, a form of self-consistency over the ranking list outputs of black-box LLMs. Our key idea is to marginalize out different list orders in the prompt to produce an order-independent ranking with less positional bias. First, given some input prompt, we repeatedly shuffle the list in the prompt and pass it through the LLM while holding the instructions the same. Next, we aggregate the resulting sample of rankings by computing the central ranking closest in distance to all of them, marginalizing out prompt order biases in the process. Theoretically, we prove the robustness of our method, showing convergence to the true ranking under random perturbations.Empirically, on five datasets in sorting and passage reranking, our approach improves scores from conventional inference by up to 34-52% for Mistral, 7-18% for GPT-3.5, 8-16% for LLaMA v2 (70B). Our code is at https://github.com/castorini/perm-sc.

There exists a wide variety of efficiency methods for natural language processing (NLP) tasks, such as pruning, distillation, dynamic inference, quantization, etc. From a different perspective, we can consider an efficiency method as an operator applied on a model. Naturally, we may construct a pipeline of operators, i.e., to apply multiple efficiency methods on the model sequentially. In this paper, we study the plausibility of this idea, and more importantly, the commutativity and cumulativeness of efficiency operators. We make two interesting observations from our experiments: (1) The operators are commutative—the order of efficiency methods within the pipeline has little impact on the final results; (2) The operators are also cumulative—the final results of combining several efficiency methods can be estimated by combining the results of individual methods. These observations deepen our understanding of efficiency operators and provide useful guidelines for building them in real-world applications.

Deep neural networks (DNNs) are often used for text classification due to their high accuracy. However, DNNs can be computationally intensive, requiring millions of parameters and large amounts of labeled data, which can make them expensive to use, to optimize, and to transfer to out-of-distribution (OOD) cases in practice. In this paper, we propose a non-parametric alternative to DNNs that’s easy, lightweight, and universal in text classification: a combination of a simple compressor like gzip with a k-nearest-neighbor classifier. Without any training parameters, our method achieves results that are competitive with non-pretrained deep learning methods on six in-distribution datasets.It even outperforms BERT on all five OOD datasets, including four low-resource languages. Our method also excels in the few-shot setting, where labeled data are too scarce to train DNNs effectively.

Diffusion models are a milestone in text-to-image generation, but they remain poorly understood, lacking interpretability analyses. In this paper, we perform a text-image attribution analysis on Stable Diffusion, a recently open-sourced model. To produce attribution maps, we upscale and aggregate cross-attention maps in the denoising module, naming our method DAAM. We validate it by testing its segmentation ability on nouns, as well as its generalized attribution quality on all parts of speech, rated by humans. On two generated datasets, we attain a competitive 58.8-64.8 mIoU on noun segmentation and fair to good mean opinion scores (3.4-4.2) on generalized attribution. Then, we apply DAAM to study the role of syntax in the pixel space across head–dependent heat map interaction patterns for ten common dependency relations. We show that, for some relations, the head map consistently subsumes the dependent, while the opposite is true for others. Finally, we study several semantic phenomena, focusing on feature entanglement; we find that the presence of cohyponyms worsens generation quality by 9%, and descriptive adjectives attend too broadly. We are the first to interpret large diffusion models from a visuolinguistic perspective, which enables future research. Our code is at https://github.com/castorini/daam.

End-to-end automatic speech recognition systems represent the state of the art, but they rely on thousands of hours of manually annotated speech for training, as well as heavyweight computation for inference. Of course, this impedes commercialization since most companies lack vast human and computational resources. In this paper, we explore training and deploying an ASR system in the label-scarce, compute-limited setting. To reduce human labor, we use a third-party ASR system as a weak supervision source, supplemented with labeling functions derived from implicit user feedback. To accelerate inference, we propose to route production-time queries across a pool of CUDA graphs of varying input lengths, the distribution of which best matches the traffic’s. Compared to our third-party ASR, we achieve a relative improvement in word-error rate of 8% and a speedup of 600%. Our system, called SpeechNet, currently serves 12 million queries per day on our voice-enabled smart television. To our knowledge, this is the first time a large-scale, Wav2vec-based deployment has been described in the academic literature.

Query auto completion (QAC) is the task of predicting a search engine user’s final query from their intermediate, incomplete query. In this paper, we extend QAC to the streaming voice search setting, where automatic speech recognition systems produce intermediate transcriptions as users speak. Naively applying existing methods fails because the intermediate transcriptions often don’t form prefixes or even substrings of the final transcription. To address this issue, we propose to condition QAC approaches on intermediate transcriptions to complete voice queries. We evaluate our models on a speech-enabled smart television with real-life voice search traffic, finding that this ASR-aware conditioning improves the completion quality. Our best method obtains an 18% relative improvement in mean reciprocal rank over previous methods.

The slow speed of BERT has motivated much research on accelerating its inference, and the early exiting idea has been proposed to make trade-offs between model quality and efficiency. This paper aims to address two weaknesses of previous work: (1) existing fine-tuning strategies for early exiting models fail to take full advantage of BERT; (2) methods to make exiting decisions are limited to classification tasks. We propose a more advanced fine-tuning strategy and a learning-to-exit module that extends early exiting to tasks other than classification. Experiments demonstrate improved early exiting for BERT, with better trade-offs obtained by the proposed fine-tuning strategy, successful application to regression tasks, and the possibility to combine it with other acceleration methods. Source code can be found at https://github.com/castorini/berxit.

In selective prediction, a classifier is allowed to abstain from making predictions on low-confidence examples. Though this setting is interesting and important, selective prediction has rarely been examined in natural language processing (NLP) tasks. To fill this void in the literature, we study in this paper selective prediction for NLP, comparing different models and confidence estimators. We further propose a simple error regularization trick that improves confidence estimation without substantially increasing the computation budget. We show that recent pre-trained transformer models simultaneously improve both model accuracy and confidence estimation effectiveness. We also find that our proposed regularization improves confidence estimation and can be applied to other relevant scenarios, such as using classifier cascades for accuracy–efficiency trade-offs. Source code for this paper can be found at https://github.com/castorini/transformers-selective.

Fine-tuned pre-trained transformers achieve the state of the art in passage reranking. Unfortunately, how they make their predictions remains vastly unexplained, especially at the end-to-end, input-to-output level. Little known is how tokens, layers, and passages precisely contribute to the final prediction. In this paper, we address this gap by leveraging the recently developed information bottlenecks for attribution (IBA) framework. On BERT-based models for passage reranking, we quantitatively demonstrate the framework’s veracity in extracting attribution maps, from which we perform detailed, token-wise analysis about how predictions are made. Overall, we find that BERT still cares about exact token matching for reranking; the [CLS] token mainly gathers information for predictions at the last layer; top-ranked passages are robust to token removal; and BERT fine-tuned on MSMARCO has positional bias towards the start of the passage.

Large-scale pre-trained language models such as BERT have brought significant improvements to NLP applications. However, they are also notorious for being slow in inference, which makes them difficult to deploy in real-time applications. We propose a simple but effective method, DeeBERT, to accelerate BERT inference. Our approach allows samples to exit earlier without passing through the entire model. Experiments show that DeeBERT is able to save up to ~40% inference time with minimal degradation in model quality. Further analyses show different behaviors in the BERT transformer layers and also reveal their redundancy. Our work provides new ideas to efficiently apply deep transformer-based models to downstream tasks. Code is available at https://github.com/castorini/DeeBERT.

In natural language processing, a recently popular line of work explores how to best report the experimental results of neural networks. One exemplar publication, titled “Show Your Work: Improved Reporting of Experimental Results” (Dodge et al., 2019), advocates for reporting the expected validation effectiveness of the best-tuned model, with respect to the computational budget. In the present work, we critically examine this paper. As far as statistical generalizability is concerned, we find unspoken pitfalls and caveats with this approach. We analytically show that their estimator is biased and uses error-prone assumptions. We find that the estimator favors negative errors and yields poor bootstrapped confidence intervals. We derive an unbiased alternative and bolster our claims with empirical evidence from statistical simulation. Our codebase is at https://github.com/castorini/meanmax.

We describe Howl, an open-source wake word detection toolkit with native support for open speech datasets such as Mozilla Common Voice (MCV) and Google Speech Commands (GSC). We report benchmark results of various models supported by our toolkit on GSC and our own freely available wake word detection dataset, built from MCV. One of our models is deployed in Firefox Voice, a plugin enabling speech interactivity for the Firefox web browser. Howl represents, to the best of our knowledge, the first fully productionized, open-source wake word detection toolkit with a web browser deployment target. Our codebase is at howl.ai.

We present Covidex, a search engine that exploits the latest neural ranking models to provide information access to the COVID-19 Open Research Dataset curated by the Allen Institute for AI. Our system has been online and serving users since late March 2020. The Covidex is the user application component of our three-pronged strategy to develop technologies for helping domain experts tackle the ongoing global pandemic. In addition, we provide robust and easy-to-use keyword search infrastructure that exploits mature fusion-based methods as well as standalone neural ranking models that can be incorporated into other applications. These techniques have been evaluated in the multi-round TREC-COVID challenge: Our infrastructure and baselines have been adopted by many participants, including some of the best systems. In round 3, we submitted the highest-scoring run that took advantage of previous training data and the second-highest fully automatic run. In rounds 4 and 5, we submitted the highest-scoring fully automatic runs.

Pretrained transformers achieve the state of the art across tasks in natural language processing, motivating researchers to investigate their inner mechanisms. One common direction is to understand what features are important for prediction. In this paper, we apply information bottlenecks to analyze the attribution of each feature for prediction on a black-box model. We use BERT as the example and evaluate our approach both quantitatively and qualitatively. We show the effectiveness of our method in terms of attribution and the ability to provide insight into how information flows through layers. We demonstrate that our technique outperforms two competitive methods in degradation tests on four datasets. Code is available at https://github.com/bazingagin/IBA.

Fine-tuned variants of BERT are able to achieve state-of-the-art accuracy on many natural language processing tasks, although at significant computational costs. In this paper, we verify BERT’s effectiveness for document classification and investigate the extent to which BERT-level effectiveness can be obtained by different baselines, combined with knowledge distillation—a popular model compression method. The results show that BERT-level effectiveness can be achieved by a single-layer LSTM with at least 40× fewer FLOPS and only ∼3% parameters. More importantly, this study analyzes the limits of knowledge distillation as we distill BERT’s knowledge all the way down to linear models—a relevant baseline for the task. We report substantial improvement in effectiveness for even the simplest models, as they capture the knowledge learnt by BERT.

Semantic similarity modeling is central to many NLP problems such as natural language inference and question answering. Syntactic structures interact closely with semantics in learning compositional representations and alleviating long-range dependency issues. How-ever, such structure priors have not been well exploited in previous work for semantic mod-eling. To examine their effectiveness, we start with the Pairwise Word Interaction Model, one of the best models according to a recent reproducibility study, then introduce components for modeling context and structure using multi-layer BiLSTMs and TreeLSTMs. In addition, we introduce residual connections to the deep convolutional neural network component of the model. Extensive evaluations on eight benchmark datasets show that incorporating structural information contributes to consistent improvements over strong baselines.

Used for simple commands recognition on devices from smart speakers to mobile phones, keyword spotting systems are everywhere. Ubiquitous as well are web applications, which have grown in popularity and complexity over the last decade. However, despite their obvious advantages in natural language interaction, voice-enabled web applications are still few and far between. We attempt to bridge this gap with Honkling, a novel, JavaScript-based keyword spotting system. Purely client-side and cross-device compatible, Honkling can be deployed directly on user devices. Our in-browser implementation enables seamless personalization, which can greatly improve model quality; in the presence of underrepresented, non-American user accents, we can achieve up to an absolute 10% increase in accuracy in the personalized model with only a few examples.

Knowledge distillation can effectively transfer knowledge from BERT, a deep language representation model, to traditional, shallow word embedding-based neural networks, helping them approach or exceed the quality of other heavyweight language representation models. As shown in previous work, critical to this distillation procedure is the construction of an unlabeled transfer dataset, which enables effective knowledge transfer. To create transfer set examples, we propose to sample from pretrained language models fine-tuned on task-specific text. Unlike previous techniques, this directly captures the purpose of the transfer set. We hypothesize that this principled, general approach outperforms rule-based techniques. On four datasets in sentiment classification, sentence similarity, and linguistic acceptability, we show that our approach improves upon previous methods. We outperform OpenAI GPT, a deep pretrained transformer, on three of the datasets, while using a single-layer bidirectional LSTM that runs at least ten times faster.

Neural network models for many NLP tasks have grown increasingly complex in recent years, making training and deployment more difficult. A number of recent papers have questioned the necessity of such architectures and found that well-executed, simpler models are quite effective. We show that this is also the case for document classification: in a large-scale reproducibility study of several recent neural models, we find that a simple BiLSTM architecture with appropriate regularization yields accuracy and F1 that are either competitive or exceed the state of the art on four standard benchmark datasets. Surprisingly, our simple model is able to achieve these results without attention mechanisms. While these regularization techniques, borrowed from language modeling, are not novel, to our knowledge we are the first to apply them in this context. Our work provides an open-source platform and the foundation for future work in document classification.