Pretrained language models (LMs) are susceptible to generate text with nonfactual information. In this work, we measure and improve the factual accuracy of large-scale LMs for open-ended text generation. We design the FactualityPrompts test set and metrics to measure the factuality of LM generations. Based on that, we study the factual accuracy of LMs with parameter sizes ranging from 126M to 530B. Interestingly, we find that larger LMs are more factual than smaller ones, although a previous study suggests that larger LMs can be less truthful in terms of misconceptions. In addition, popular sampling algorithms (e.g., top-p) in open-ended text generation can harm the factuality due to the ''uniform randomness'' introduced at every sampling step. We propose the factual-nucleus sampling algorithm that dynamically adapts the randomness to improve the factuality of generation while maintaining quality. Furthermore, we analyze the inefficiencies of the standard training method in learning correct associations between entities from factual text corpus (e.g., Wikipedia). We propose a factuality-enhanced training method that uses TopicPrefix for better awareness of facts and sentence completion as the training objective, which can vastly reduce the factual errors. We release our code and FactualityPrompts benchmark at: https://github.com/nayeon7lee/FactualityPrompt.
One critical challenge in deploying highly performant machine learning models in real-life applications is out of distribution (OOD) detection. Given a predictive model which is accurate on in distribution (ID) data, an OOD detection system will further equip the model with the option to defer prediction when the input is novel and the model has little confidence in prediction. There has been some recent interest in utilizing the gradient information in pre-trained models for OOD detection. While these methods have shown competitive performance, there are misconceptions about the true mechanism underlying them, which conflate their performance with the necessity of gradients. In this work, we provide an in-depth analysis and comparison of gradient based methods and elucidate the key components that warrant their OOD detection performance. We further propose a general, non-gradient based method of OOD detection which improves over previous baselines in both performance and computational efficiency.
The COVID-19 pandemic has been accompanied by an `infodemic' -- of accurate and inaccurate health information across social media. Detecting misinformation amidst dynamically changing information landscape is challenging; identifying relevant keywords and posts is arduous due to the large amount of human effort required to inspect the content and sources of posts. We aim to reduce the resource cost of this process by introducing a weakly-supervised iterative graph-based approach to detect keywords, topics, and themes related to misinformation, with a focus on COVID-19. Our approach can successfully detect specific topics from general misinformation-related seed words in a few seed texts. Our approach utilizes the BERT-based Word Graph Search (BWGS) algorithm that builds on context-based neural network embeddings for retrieving misinformation-related posts. We utilize Latent Dirichlet Allocation (LDA) topic modeling for obtaining misinformation-related themes from the texts returned by BWGS. Furthermore, we propose the BERT-based Multi-directional Word Graph Search (BMDWGS) algorithm that utilizes greater starting context information for misinformation extraction. In addition to a qualitative analysis of our approach, our quantitative analyses show that BWGS and BMDWGS are effective in extracting misinformation-related content compared to common baselines in low data resource settings. Extracting such content is useful for uncovering prevalent misconceptions and concerns and for facilitating precision public health messaging campaigns to improve health behaviors.
Block-based visual programming environments are increasingly used to introduce computing concepts to beginners. Given that programming tasks are open-ended and conceptual, novice students often struggle when learning in these environments. AI-driven programming tutors hold great promise in automatically assisting struggling students, and need several components to realize this potential. We investigate the crucial component of student modeling, in particular, the ability to automatically infer students' misconceptions for predicting (synthesizing) their behavior. We introduce a novel benchmark, StudentSyn, centered around the following challenge: For a given student, synthesize the student's attempt on a new target task after observing the student's attempt on a fixed reference task. This challenge is akin to that of program synthesis; however, instead of synthesizing a {solution} (i.e., program an expert would write), the goal here is to synthesize a {student attempt} (i.e., program that a given student would write). We first show that human experts (TutorSS) can achieve high performance on the benchmark, whereas simple baselines perform poorly. Then, we develop two neuro/symbolic techniques (NeurSS and SymSS) in a quest to close this gap with TutorSS.
NLP-powered automatic question generation (QG) techniques carry great pedagogical potential of saving educators' time and benefiting student learning. Yet, QG systems have not been widely adopted in classrooms to date. In this work, we aim to pinpoint key impediments and investigate how to improve the usability of automatic QG techniques for educational purposes by understanding how instructors construct questions and identifying touch points to enhance the underlying NLP models. We perform an in-depth need finding study with 11 instructors across 7 different universities, and summarize their thought processes and needs when creating questions. While instructors show great interests in using NLP systems to support question design, none of them has used such tools in practice. They resort to multiple sources of information, ranging from domain knowledge to students' misconceptions, all of which missing from today's QG systems. We argue that building effective human-NLP collaborative QG systems that emphasize instructor control and explainability is imperative for real-world adoption. We call for QG systems to provide process-oriented support, use modular design, and handle diverse sources of input.
We investigate the optimal model size and number of tokens for training a transformer language model under a given compute budget. We find that current large language models are significantly undertrained, a consequence of the recent focus on scaling language models whilst keeping the amount of training data constant. By training over \nummodels language models ranging from 70 million to over 16 billion parameters on 5 to 500 billion tokens, we find that for compute-optimal training, the model size and the number of training tokens should be scaled equally: for every doubling of model size the number of training tokens should also be doubled. We test this hypothesis by training a predicted compute-optimal model, \chinchilla, that uses the same compute budget as \gopher but with 70B parameters and 4$\times$ more more data. \chinchilla uniformly and significantly outperforms \Gopher (280B), GPT-3 (175B), Jurassic-1 (178B), and Megatron-Turing NLG (530B) on a large range of downstream evaluation tasks. This also means that \chinchilla uses substantially less compute for fine-tuning and inference, greatly facilitating downstream usage. As a highlight, \chinchilla reaches a state-of-the-art average accuracy of 67.5\% on the MMLU benchmark, greater than a 7\% improvement over \gopher.
Although there has been a dramatically reduction on the prices of capturing devices and an increase on computing power in the last decade, it seems that biometric systems are still far from massive adoption for civilian applications. This paper deals with the causes of this phenomenon, as well as some misconceptions regarding biometric identification.
Although the existing max-value entropy search (MES) is based on the widely celebrated notion of mutual information, its empirical performance can suffer due to two misconceptions whose implications on the exploration-exploitation trade-off are investigated in this paper. These issues are essential in the development of future acquisition functions and the improvement of the existing ones as they encourage an accurate measure of the mutual information such as the rectified MES (RMES) acquisition function we develop in this work. Unlike the evaluation of MES, we derive a closed-form probability density for the observation conditioned on the max-value and employ stochastic gradient ascent with reparameterization to efficiently optimize RMES. As a result of a more principled acquisition function, RMES shows a consistent improvement over MES in several synthetic function benchmarks and real-world optimization problems.
Phylogenetic placement refers to a family of tools and methods to analyze, visualize, and interpret the tsunami of metagenomic sequencing data generated by high-throughput sequencing. Compared to alternative (e. g., similarity-based) methods, it puts metabarcoding sequences into a phylogenetic context using a set of known reference sequences and taking evolutionary history into account. Thereby, one can increase the accuracy of metagenomic surveys and eliminate the requirement for having exact or close matches with existing sequence databases. Phylogenetic placement constitutes a valuable analysis tool per se, but also entails a plethora of downstream tools to interpret its results. A common use case is to analyze species communities obtained from metagenomic sequencing, for example via taxonomic assignment, diversity quantification, sample comparison, and identification of correlations with environmental variables. In this review, we provide an overview over the methods developed during the first ten years. In particular, the goals of this review are (i) to motivate the usage of phylogenetic placement and illustrate some of its use cases, (ii) to outline the full workflow, from raw sequences to publishable figures, including best practices, (iii) to introduce the most common tools and methods and their capabilities, (iv) to point out common placement pitfalls and misconceptions,(v) to showcase typical placement-based analyses, and how they can help to analyze, visualize, and interpret phylogenetic placement data.
In this paper, we discuss the conceptual underpinnings of Modern Coexistence Theory (MCT), a quantitative framework for understanding ecological coexistence. In order to use MCT to infer how species are coexisting, one must relate a complex model (which simulates coexistence in the real world) to simple models in which previously proposed explanations for coexistence have been codified. This can be accomplished in three steps: 1) relating the construct of coexistence to invasion growth rates, 2) mathematically partitioning the invasion growth rates into coexistence mechanisms (i.e., classes of explanations for coexistence), and 3) relating coexistence mechanisms to simple explanations for coexistence. Previous research has primarily focused on step 2. Here, we discuss the other crucial steps and their implications for inferring the mechanisms of coexistence in real communities. Our discussion of step 3 -- relating coexistence mechanisms to simple explanations for coexistence -- serves a heuristic guide for hypothesizing about the causes of coexistence in new models; but also addresses misconceptions about coexistence mechanisms. For example, the storage effect has little to do with bet-hedging or "storage" via a robust life-history stage; relative nonlinearity is more likely to promote coexistence than originally thought; and fitness-density covariance is an amalgam of a large number of previously proposed explanations for coexistence (e.g., the competition-colonization trade-off, heteromyopia, spatially-varying resource supply ratios). Additionally, we review a number of topics in MCT, including the role of "scaling factors"; whether coexistence mechanisms are approximations; whether the magnitude or sign of invasion growth rates matters more; whether Hutchinson solved the paradox of the plankton; the scale-dependence of coexistence mechanisms; and much more.
Language modelling provides a step towards intelligent communication systems by harnessing large repositories of written human knowledge to better predict and understand the world. In this paper, we present an analysis of Transformer-based language model performance across a wide range of model scales -- from models with tens of millions of parameters up to a 280 billion parameter model called Gopher. These models are evaluated on 152 diverse tasks, achieving state-of-the-art performance across the majority. Gains from scale are largest in areas such as reading comprehension, fact-checking, and the identification of toxic language, but logical and mathematical reasoning see less benefit. We provide a holistic analysis of the training dataset and model's behaviour, covering the intersection of model scale with bias and toxicity. Finally we discuss the application of language models to AI safety and the mitigation of downstream harms.
Databases covering all individuals of a population are increasingly used for research and decision-making. The massive size of such databases is often mistaken as a guarantee for valid inferences. However, population data have characteristics that make them challenging to use. Various assumptions on population coverage and data quality are commonly made, including how such data were captured and what types of processing have been applied to them. Furthermore, the full potential of population data can often only be unlocked when such data are linked to other databases. Record linkage often implies subtle technical problems, which are easily missed. We discuss a diverse range of misconceptions relevant for anybody capturing, processing, linking, or analysing population data. Remarkably many of these misconceptions are due to the social nature of data collections and are therefore missed by purely technical accounts of data processing. Many of these misconceptions are also not well documented in scientific publications. We conclude with a set of recommendations for using population data.
Artificial intelligence (AI) is paving the way towards the fourth industrial revolution with the fire domain (Fire 4.0). As a matter of fact, the next few years will be elemental to how this technology will shape our academia, practice, and entrepreneurship. Despite the growing interest between fire research groups, AI remains absent of our curriculum, and we continue to lack a methodical framework to adopt, apply and create AI solutions suitable for our problems. The above is also true for parallel engineering domains (i.e., civil/mechanical engineering), and in order to negate the notion of history repeats itself (e.g., look at the continued debate with regard to modernizing standardized fire testing, etc.), it is the motivation behind this letter to the Editor to demystify some of the big ideas behind AI to jump-start prolific and strategic discussions on the front of AI & Fire. In addition, this letter intends to explain some of the most fundamental concepts and clear common misconceptions specific to the adoption of AI in fire engineering. This short letter is a companion to the Smart Systems in Fire Engineering special issue sponsored by Fire Technology. An in-depth review of AI algorithms [1] and success stories to the proper implementations of such algorithms can be found in the aforenoted special issue and collection of papers. This letter comprises two sections. The first section outlines big ideas pertaining to AI, and answers some of the burning questions with regard to the merit of adopting AI in our domain. The second section presents a set of rules or technical recommendations an AI user may deem helpful to practice whenever AI is used as an investigation methodology. The presented set of rules are complementary to the big ideas.
Sentiment analysis is often a crowdsourcing task prone to subjective labels given by many annotators. It is not yet fully understood how the annotation bias of each annotator can be modeled correctly with state-of-the-art methods. However, resolving annotator bias precisely and reliably is the key to understand annotators' labeling behavior and to successfully resolve corresponding individual misconceptions and wrongdoings regarding the annotation task. Our contribution is an explanation and improvement for precise neural end-to-end bias modeling and ground truth estimation, which reduces an undesired mismatch in that regard of the existing state-of-the-art. Classification experiments show that it has potential to improve accuracy in cases where each sample is annotated only by one single annotator. We provide the whole source code publicly and release an own domain-specific sentiment dataset containing 10,000 sentences discussing organic food products. These are crawled from social media and are singly labeled by 10 non-expert annotators.
We propose a benchmark to measure whether a language model is truthful in generating answers to questions. The benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics. We crafted questions that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. We tested GPT-3, GPT-Neo/J, GPT-2 and a T5-based model. The best model was truthful on 58% of questions, while human performance was 94%. Models generated many false answers that mimic popular misconceptions and have the potential to deceive humans. The largest models were generally the least truthful. This contrasts with other NLP tasks, where performance improves with model size. However, this result is expected if false answers are learned from the training distribution. We suggest that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web.
While recent works have indicated that federated learning (FL) may be vulnerable to poisoning attacks by compromised clients, their real impact on production FL systems is not fully understood. In this work, we aim to develop a comprehensive systemization for poisoning attacks on FL by enumerating all possible threat models, variations of poisoning, and adversary capabilities. We specifically put our focus on untargeted poisoning attacks, as we argue that they are significantly relevant to production FL deployments. We present a critical analysis of untargeted poisoning attacks under practical, production FL environments by carefully characterizing the set of realistic threat models and adversarial capabilities. Our findings are rather surprising: contrary to the established belief, we show that FL is highly robust in practice even when using simple, low-cost defenses. We go even further and propose novel, state-of-the-art data and model poisoning attacks, and show via an extensive set of experiments across three benchmark datasets how (in)effective poisoning attacks are in the presence of simple defense mechanisms. We aim to correct previous misconceptions and offer concrete guidelines to conduct more accurate (and more realistic) research on this topic.
Bayesian formulations of deep learning have been shown to have compelling theoretical properties and offer practical functional benefits, such as improved predictive uncertainty quantification and model selection. The Laplace approximation (LA) is a classic, and arguably the simplest family of approximations for the intractable posteriors of deep neural networks. Yet, despite its simplicity, the LA is not as popular as alternatives like variational Bayes or deep ensembles. This may be due to assumptions that the LA is expensive due to the involved Hessian computation, that it is difficult to implement, or that it yields inferior results. In this work we show that these are misconceptions: we (i) review the range of variants of the LA including versions with minimal cost overhead; (ii) introduce "laplace", an easy-to-use software library for PyTorch offering user-friendly access to all major flavors of the LA; and (iii) demonstrate through extensive experiments that the LA is competitive with more popular alternatives in terms of performance, while excelling in terms of computational cost. We hope that this work will serve as a catalyst to a wider adoption of the LA in practical deep learning, including in domains where Bayesian approaches are not typically considered at the moment.
This manuscript revisits theoretical assumptions concerning dynamic mode decomposition (DMD) of Koopman operators, including the existence of lattices of eigenfunctions, common eigenfunctions between Koopman operators, and boundedness and compactness of Koopman operators. Counterexamples that illustrate restrictiveness of the assumptions are provided for each of the assumptions. In particular, this manuscript proves that the native reproducing kernel Hilbert space (RKHS) of the Gaussian RBF kernel function only supports bounded Koopman operators if the dynamics are affine. In addition, a new framework for DMD, that requires only densely defined Koopman operators over RKHSs is introduced, and its effectiveness is demonstrated through numerical examples.
The use of crowdworkers in NLP research is growing rapidly, in tandem with the exponential increase in research production in machine learning and AI. Ethical discussion regarding the use of crowdworkers within the NLP research community is typically confined in scope to issues related to labor conditions such as fair pay. We draw attention to the lack of ethical considerations related to the various tasks performed by workers, including labeling, evaluation, and production. We find that the Final Rule, the common ethical framework used by researchers, did not anticipate the use of online crowdsourcing platforms for data collection, resulting in gaps between the spirit and practice of human-subjects ethics in NLP research. We enumerate common scenarios where crowdworkers performing NLP tasks are at risk of harm. We thus recommend that researchers evaluate these risks by considering the three ethical principles set up by the Belmont Report. We also clarify some common misconceptions regarding the Institutional Review Board (IRB) application. We hope this paper will serve to reopen the discussion within our community regarding the ethical use of crowdworkers.
Many of the systems that appear in various signal processing applications are non-linear, for example, due to hardware impairments such as non-linear amplifiers and finite-resolution quantization. The Bussgang decomposition is a popular tool for analyzing the performance of systems that involve such non-linear components. In a nutshell, the decomposition provides an exact probabilistic relationship between the output and the input of a non-linearity: the output is equal to a scaled version of the input plus uncorrelated distortion. The decomposition can either be used to compute exact performance results or lower bounds where the uncorrelated distortion is treated as independent noise. This lecture note explains the basic theory, provides key examples, extends the theory to complex-valued vector signals, and clarifies some potential misconceptions.