Warning: this paper contains content that may be inappropriate or offensive.As generative models become available for public use in various applications, testing and analyzing vulnerabilities of these models has become a priority. In this work, we propose an automatic red teaming framework that evaluates a given black-box model and exposes its vulnerabilities against unsafe and inappropriate content generation. Our framework uses in-context learning in a feedback loop to red team models and trigger them into unsafe content generation. In particular, taking text-to-image models as target models, we explore different feedback mechanisms to automatically learn effective and diverse adversarial prompts. Our experiments demonstrate that even with enhanced safety features, Stable Diffusion (SD) models are vulnerable to our adversarial prompts, raising concerns on their robustness in practical uses. Furthermore, we demonstrate that the proposed framework is effective for red teaming text-to-text models.

Data are crucial element in large language model (LLM) alignment. Recent studies have explored using LLMs for efficient data collection. However, LLM-generated data often suffers from quality issues, with underrepresented or absent aspects and low-quality datapoints. To address these problems, we propose Data Advisor, an enhanced LLM-based method for generating data that takes into account the characteristics of the desired dataset. Starting from a set of pre-defined principles in hand, Data Advisor monitors the status of the generated data, identifies weaknesses in the current dataset, and advises the next iteration of data generation accordingly. Data Advisor can be easily integrated into existing data generation methods to enhance data quality and coverage. Experiments on safety alignment of three representative LLMs (i.e., Mistral, Llama2, and Falcon) demonstrate the effectiveness of Data Advisor in enhancing model safety against various fine-grained safety issues without sacrificing model utility.

Large Language Models (LLMs) are powerful tools which have been both dominant and commonplace in the field of Artificial Intelligence. Yet, LLMs have a tendency to devolve into toxic degeneration, wherein otherwise safe and unproblematic models begin generating toxic content. For the sake of social responsibility and inspired by the biological mechanisms of inhibition control, we introduce the paradigm of Education for Societal Norms (ESN). By collecting and labeling examples as acceptable and unacceptable (in this case toxic and non-toxic), and including a corresponding acceptable rewrite with every unacceptable example, we introduce a new mechanism for LLM detoxification. We annotate a dataset of 2,850 entries and use it to fine-tune a model, which we call a Model with Inhibition Control (MICo). Evaluating this model on toxicity detection capability, rewrite detoxification, meaning preservation, and overall toxicity reduction, we discover significant improvements over the baseline model. In our experiments we show that overall toxicity of this model is more than 60% reduced, with over 75% reduction in severe toxicity.

Gender-inclusive NLP research has documented the harmful limitations of gender binary-centric large language models (LLM), such as the inability to correctly use gender-diverse English neopronouns (e.g., xe, zir, fae). While data scarcity is a known culprit, the precise mechanisms through which scarcity affects this behavior remain underexplored. We discover LLM misgendering is significantly influenced by Byte-Pair Encoding (BPE) tokenization, the tokenizer powering many popular LLMs. Unlike binary pronouns, BPE overfragments neopronouns, a direct consequence of data scarcity during tokenizer training. This disparate tokenization mirrors tokenizer limitations observed in multilingual and low-resource NLP, unlocking new misgendering mitigation strategies. We propose two techniques: (1) pronoun tokenization parity, a method to enforce consistent tokenization across gendered pronouns, and (2) utilizing pre-existing LLM pronoun knowledge to improve neopronoun proficiency. Our proposed methods outperform finetuning with standard BPE, improving neopronoun accuracy from 14.1% to 58.4%. Our paper is the first to link LLM misgendering to tokenization and deficient neopronoun grammar, indicating that LLMs unable to correctly treat neopronouns as pronouns are more prone to misgender.

In a plethora of recent work, large language models (LLMs) demonstrated impressive reasoning ability, but many proposed downstream reasoning tasks only focus on performance-wise evaluation. Two fundamental questions persist: 1) how consistent is the reasoning, and 2) can models detect unreliable reasoning? In this paper, we investigate self-contradictory (Self-Contra) reasoning, where the model reasoning does not support answers. To answer 1), we define and assess the Self-Contra rate across three datasets and delve into finer-grained categories of Self-Contra reasoning. We find that LLMs often contradict themselves in reasoning tasks involving contextual information understanding or commonsense. The model may generate correct answers by taking shortcuts in reasoning or overlooking contextual evidence, leading to compromised reasoning. For 2), we task the state-of-the-art model GPT-4 with identifying Self-Contra reasoning and finer-grained fallacies. We find that finer-grained aided detection can improve GPT-4’s ability to detect Self-Contra. However, it is only able to detect Self-Contra with a 52.2% F1 score, much lower compared to 66.7% for humans. Our results indicate that current LLMs lack the robustness necessary for reliable reasoning and we emphasize the urgent need for establishing best practices in comprehensive reasoning evaluations beyond pure performance-based metrics.

We propose a constraint learning schema forfine-tuning Large Language Models (LLMs)with attribute control. Given a training corpusand control criteria formulated as a sequence-level constraint on model outputs, our methodfine-tunes the LLM on the training corpus whileenhancing constraint satisfaction with minimalimpact on its utility and generation quality.Specifically, our approach regularizes the LLMtraining by penalizing the KL divergence be-tween the desired output distribution, which sat-isfies the constraints, and the LLM’s posterior.This regularization term can be approximatedby an auxiliary model trained to decomposethe sequence-level constraints into token-levelguidance, allowing the term to be measuredby a closed-form formulation. To further im-prove efficiency, we design a parallel schemefor concurrently updating both the LLM andthe auxiliary model. We evaluate the empiricalperformance of our approach by controlling thetoxicity when training an LLM. We show thatour approach leads to an LLM that producesfewer inappropriate responses while achievingcompetitive performance on benchmarks and atoxicity detection task

Recent advancement in large language models (LLMs) has offered a strong potential for natural language systems to process informal language. A representative form of informal language is slang, used commonly in daily conversations and online social media. To date, slang has not been comprehensively evaluated in LLMs due partly to the absence of a carefully designed and publicly accessible benchmark. Using movie subtitles, we construct a dataset that supports evaluation on a diverse set of tasks pertaining to automatic processing of slang. For both evaluation and finetuning, we show the effectiveness of our dataset on two core applications: 1) slang detection, and 2) identification of regional and historical sources of slang from natural sentences. We also show how our dataset can be used to probe the output distributions of LLMs for interpretive insights. We find that while LLMs such as GPT-4 achieve good performance in a zero-shot setting, smaller BERT-like models finetuned on our dataset achieve comparable performance. Furthermore, we show that our dataset enables finetuning of LLMs such as GPT-3.5 that achieve substantially better performance than strong zero-shot baselines. Our work offers a comprehensive evaluation and a high-quality benchmark on English slang based on the OpenSubtitles corpus, serving both as a publicly accessible resource and a platform for applying tools for informal language processing.

Large language models (LLMs) are known to generate biased responses where the opinions of certain groups and populations are underrepresented. Here, we present a novel approach to achieve controllable generation of specific viewpoints using LLMs, that can be leveraged to produce multiple perspectives and to reflect the diverse opinions. Moving beyond the traditional reliance on demographics like age, gender, or party affiliation, we introduce a data-driven notion of persona grounded in collaborative filtering, which is defined as either a single individual or a cohort of individuals manifesting similar views across specific inquiries. As individuals in the same demographic group may have different personas, our data-driven persona definition allows for a more nuanced understanding of different (latent) social groups present in the population. In addition to this, we also explore an efficient method to steer LLMs toward the personas that we define. We show that our data-driven personas significantly enhance model steerability, with improvements of between 57%-77% over our best performing baselines.

Knowledge graphs (KGs) complement Large Language Models (LLMs) by providing reliable, structured, domain-specific, and up-to-date external knowledge. However, KGs and LLMs are often developed separately and must be integrated after training. We introduce Tree-of-Traversals, a novel zero-shot reasoning algorithm that enables augmentation of black-box LLMs with one or more KGs. The algorithm equips a LLM with actions for interfacing a KG and enables the LLM to perform tree search over possible thoughts and actions to find high confidence reasoning paths. Tree-of-Traversals significantly improves performance on question answering and KG question answering tasks. Code is available at https://github.com/amazon-science/tree-of-traversals

Large language models incorporate world knowledge and present breakthrough performances on zero-shot learning. However, these models capture societal bias (e.g., gender or racial bias) due to bias during the training process which raises ethical concerns or can even be potentially harmful. The issue is more pronounced in multi-modal settings, such as image captioning, as images can also add onto biases (e.g., due to historical non-equal representation of genders in different occupations). In this study, we investigate the removal of potentially problematic knowledge from multi-modal models used for image captioning. We relax the gender bias issue in captioning models by degenderizing generated captions through the use of a simple linear mask, trained via adversarial training. Our proposal makes no assumption on the architecture of the model and freezes the model weights during the procedure, which also enables the mask to be turned off. We conduct experiments on COCO caption datasets using our masking solution. The results suggest that the proposed mechanism can effectively mask the targeted biased knowledge, by replacing more than 99% gender words with neutral ones, and maintain a comparable captioning quality performance with minimal (e.g., -1.4 on BLEU4 and ROUGE) impact to accuracy metrics.

Language models, pre-trained on large amounts of unmoderated content, have been shown to contain societal biases. Mitigating such biases typically requires access to model parameters and training schemas. In this work, we address bias mitigation at inference time, such that it can be applied to any black-box model. To this end, we propose a belief generation and augmentation framework, BELIEVE, that demonstrates effective bias mitigation for natural language generation by augmenting input prompts with automatically generated instruction-based beliefs. Our framework eases the bottleneck required for manually crafting these instruction-based beliefs, by extending a recently proposed iterative in-context learning framework to automatically generate beliefs via a language model. We assess the impact of this system on fairness, and demonstrate effective bias mitigation on pretrained and instruction-tuned models for both sentiment and regard with respect to multiple protected classes including race, gender, and political ideology.

Dialect differences caused by regional, social, and economic factors cause performance discrepancies for many groups of language technology users. Inclusive and equitable language technology must critically be dialect invariant, meaning that performance remains constant over dialectal shifts. Current systems often fall short of this ideal since they are designed and tested on a single dialect: Standard American English (SAE). We introduce a suite of resources for evaluating and achieving English dialect invariance. The resource is called Multi-VALUE, a controllable rule-based translation system spanning 50 English dialects and 189 unique linguistic features. Multi-VALUE maps SAE to synthetic forms of each dialect. First, we use this system to stress tests question answering, machine translation, and semantic parsing. Stress tests reveal significant performance disparities for leading models on non-standard dialects. Second, we use this system as a data augmentation technique to improve the dialect robustness of existing systems. Finally, we partner with native speakers of Chicano and Indian English to release new gold-standard variants of the popular CoQA task. To execute the transformation code, run model checkpoints, and download both synthetic and gold-standard dialectal benchmark datasets, see http://value-nlp.org.

Natural language often contains ambiguities that can lead to misinterpretation and miscommunication. While humans can handle ambiguities effectively by asking clarifying questions and/or relying on contextual cues and common-sense knowledge, resolving ambiguities can be notoriously hard for machines. In this work, we study ambiguities that arise in text-to-image generative models. We curate the Text-to-image Ambiguity Benchmark (TAB) dataset to study different types of ambiguities in text-to-image generative models. We then propose the Text-to-ImagE Disambiguation (TIED) framework to disambiguate the prompts given to the text-to-image generative models by soliciting clarifications from the end user. Through automatic and human evaluations, we show the effectiveness of our framework in generating more faithful images aligned with end user intention in the presence of ambiguities.

Large Language Models (LLMs) are known to memorize significant portions of their training data. Parts of this memorized content have been shown to be extractable by simply querying the model, which poses a privacy risk. We present a novel approach which uses prompt-tuning to control the extraction rates of memorized content in LLMs. We present two prompt training strategies to increase and decrease extraction rates, which correspond to an attack and a defense, respectively. We demonstrate the effectiveness of our techniques by using models from the GPT-Neo family on a public benchmark. For the 1.3B parameter GPT-Neo model, our attack yields a 9.3 percentage point increase in extraction rate compared to our baseline. Our defense can be tuned to achieve different privacy-utility trade-offs by a user-specified hyperparameter. We achieve an extraction rate reduction of up to 97.7% relative to our baseline, with a perplexity increase of 16.9%.

The multi-sentential long sequence textual data unfolds several interesting research directions pertaining to natural language processing and generation. Though we observe several high-quality long-sequence datasets for English and other monolingual languages, there is no significant effort in building such resources for code-mixed languages such as Hinglish (code-mixing of Hindi-English). In this paper, we propose a novel task of identifying multi-sentential code-mixed text (MCT) from multilingual articles. As a use case, we leverage multilingual articles from two different data sources and build a first-of-its-kind multi-sentential code-mixed Hinglish dataset i.e., MUTANT. We propose a token-level language-aware pipeline and extend the existing metrics measuring the degree of code-mixing to a multi-sentential framework and automatically identify MCT in the multilingual articles. The MUTANT dataset comprises 67k articles with 85k identified Hinglish MCTs. To facilitate future research directions, we will make the dataset and the code publicly available upon publication.

Recent advancements in Large language models (LLMs) have enabled them to hold free form conversations over multiple turns, but they exhibit a tendency to make unfounded and incorrect statements, commonly known as hallucinations. In particular, LLMs hallucinate frequently when given invalid questions, i.e. ones with incorrect assumptions. The most common approach to evaluate LLMs on hallucinations is to test them on Question Answering (QA) test sets such as TruthfulQA. However, LLMs are increasingly pretrained on massive text corpora scraped from the Internet, which may inevitably expose these test sets to the model during training, leading eventually to an overestimation of model performances on these test sets. In this work, we present an alternative framework to address this risk and to foster further research towards making LLMs robust against invalid questions. We name our framework INVITE: a testbed of automatically generated INValId questions to evaluaTE large language models for hallucinations. In each instantiation, our framework is set up to create a fresh batch of invalid questions by distorting valid facts in which subjects or objects are replaced by similar entities. We evaluate several state of the art LLMs against a testset generated by our framework and highlight its capacity to trigger hallucinations in these models.

While recent studies have looked into the abilities of large language models in various benchmark tasks, including question generation, reading comprehension, multilingual and etc, there have been few studies looking into the controllability of large language models on generation tasks. We present an extensive analysis of various benchmarks including a sentence planning benchmark with different granularities. After comparing large language models against state-of-the-start finetuned smaller models, we present a spectrum showing large language models falling behind, are comparable, or exceed the ability of smaller models. We conclude that *large language models struggle at meeting fine-grained hard constraints*.

Statistical significance testing is used in natural language processing (NLP) to determine whether the results of a study or experiment are likely to be due to chance or if they reflect a genuine relationship. A key step in significance testing is the estimation of confidence interval which is a function of sample variance. Sample variance calculation is straightforward when evaluating against ground truth. However, in many cases, a metric model is often used for evaluation. For example, to compare toxicity of two large language models, a toxicity classifier is used for evaluation. Existing works usually do not consider the variance change due to metric model errors, which can lead to wrong conclusions. In this work, we establish the mathematical foundation of significance testing for model-based metrics. With experiments on public benchmark datasets and a production system, we show that considering metric model errors to calculate sample variances for model-based metrics changes the conclusions in certain experiments.

Continual Federated Learning (CFL) combines Federated Learning (FL), the decentralized learning of a central model on a number of client devices that may not communicate their data, and Continual Learning (CL), the learning of a model from a continual stream of data without keeping the entire history. In CL, the main challenge is forgetting what was learned from past data. While replay-based algorithms that keep a small pool of past training data are effective to reduce forgetting, only simple replay sample selection strategies have been applied to CFL in prior work, and no previous work has explored coordination among clients for better sample selection. To bridge this gap, we adapt a replay sample selection objective based on loss gradient diversity to CFL and propose a new relaxation-based selection of samples to optimize the objective. Next, we propose a practical algorithm to coordinate gradient-based replay sample selection across clients without communicating private data. We benchmark our coordinated and uncoordinated replay sample selection algorithms against random sampling-based baselines with language models trained on a large scale de-identified real-world text dataset. We show that gradient-based sample selection methods both boost performance and reduce forgetting compared to random sampling methods, with our coordination method showing gains early in the low replay size regime (when the budget for storing past data is small).

Training mixed-domain translation models is a complex task that demands tailored architec- tures and costly data preparation techniques. In this work, we leverage federated learning (FL) in order to tackle the problem. Our investiga- tion demonstrates that with slight modifications in the training process, neural machine trans- lation (NMT) engines can be easily adapted when an FL-based aggregation is applied to fuse different domains. Experimental results also show that engines built via FL are able to perform on par with state-of-the-art baselines that rely on centralized training techniques. We evaluate our hypothesis in the presence of five datasets with different sizes, from different domains, to translate from German into English and discuss how FL and NMT can mutually benefit from each other. In addition to provid- ing benchmarking results on the union of FL and NMT, we also propose a novel technique to dynamically control the communication band- width by selecting impactful parameters during FL updates. This is a significant achievement considering the large size of NMT engines that need to be exchanged between FL parties.

Machine Learning (ML) systems are getting increasingly popular, and drive more and more applications and services in our daily life. Thishas led to growing concerns over user privacy, since human interaction data typically needs to be transmitted to the cloud in order to trainand improve such systems. Federated learning (FL) has recently emerged as a method for training ML models on edge devices using sensitive user data and is seen as a way to mitigate concerns over data privacy. However, since ML models are most commonly trained with label supervision, we need a way to extract labels on edge to make FL viable. In this work, we propose a strategy for training FL models using positive and negative user feedback. We also design a novel framework to study different noise patterns in user feedback, and explore how well standard noise-robust objectives can help mitigate this noise when training models in a federated setting. We evaluate our proposed training setup through detailed experiments on two text classification datasets and analyze the effects of varying levels of user reliability and feedback noise on model performance. We show that our method improves substantially over a self-training baseline, achieving performance closer to models trained with full supervision.

Use of synthetic data is rapidly emerging as a realistic alternative to manually annotating live traffic for industry-scale model building. Manual data annotation is slow, expensive and not preferred for meeting customer privacy expectations. Further, commercial natural language applications are required to support continuously evolving features as well as newly added experiences. To address these requirements, we propose a targeted synthetic data generation technique by inserting tokens into a given semantic signature. The generated data are used as additional training samples in the tasks of intent classification and named entity recognition. We evaluate on a real-world voice assistant dataset, and using only 33% of the available training set, we achieve the same accuracy as training with all available data. Further, we analyze the effects of data generation across varied real-world applications and propose heuristics that improve the task performance further.

With the rapid growth in language processing applications, fairness has emerged as an important consideration in data-driven solutions. Although various fairness definitions have been explored in the recent literature, there is lack of consensus on which metrics most accurately reflect the fairness of a system. In this work, we propose a new formulation – accumulated prediction sensitivity, which measures fairness in machine learning models based on the model’s prediction sensitivity to perturbations in input features. The metric attempts to quantify the extent to which a single prediction depends on a protected attribute, where the protected attribute encodes the membership status of an individual in a protected group. We show that the metric can be theoretically linked with a specific notion of group fairness (statistical parity) and individual fairness. It also correlates well with humans’ perception of fairness. We conduct experiments on two text classification datasets – Jigsaw Toxicity, and Bias in Bios, and evaluate the correlations between metrics and manual annotations on whether the model produced a fair outcome. We observe that the proposed fairness metric based on prediction sensitivity is statistically significantly more correlated with human annotation than the existing counterfactual fairness metric.

Natural Language Understanding (NLU) models can be trained on sensitive information such as phone numbers, zip-codes etc. Recent literature has focused on Model Inversion Attacks (ModIvA) that can extract training data from model parameters. In this work, we present a version of such an attack by extracting canaries inserted in NLU training data. In the attack, an adversary with open-box access to the model reconstructs the canaries contained in the model’s training set. We evaluate our approach by performing text completion on canaries and demonstrate that by using the prefix (non-sensitive) tokens of the canary, we can generate the full canary. As an example, our attack is able to reconstruct a four digit code in the training dataset of the NLU model with a probability of 0.5 in its best configuration. As countermeasures, we identify several defense mechanisms that, when combined, effectively eliminate the risk of ModIvA in our experiments.

Multiple metrics have been introduced to measure fairness in various natural language processing tasks. These metrics can be roughly categorized into two categories: 1) extrinsic metrics for evaluating fairness in downstream applications and 2) intrinsic metrics for estimating fairness in upstream contextualized language representation models. In this paper, we conduct an extensive correlation study between intrinsic and extrinsic metrics across bias notions using 19 contextualized language models. We find that intrinsic and extrinsic metrics do not necessarily correlate in their original setting, even when correcting for metric misalignments, noise in evaluation datasets, and confounding factors such as experiment configuration for extrinsic metrics.

This paper presents an approach to identify samples from live traffic where the customer implicitly communicated satisfaction with Alexa’s responses, by leveraging interpretations of model behavior. Such customer signals are noisy and adding a large number of samples from live traffic to training set makes re-training infeasible. Our work addresses these challenges by identifying a small number of samples that grow training set by ~0.05% while producing statistically significant improvements in both offline and online tests.

Language models excel at generating coherent text, and model compression techniques such as knowledge distillation have enabled their use in resource-constrained settings. However, these models can be biased in multiple ways, including the unfounded association of male and female genders with gender-neutral professions. Therefore, knowledge distillation without any fairness constraints may preserve or exaggerate the teacher model’s biases onto the distilled model. To this end, we present a novel approach to mitigate gender disparity in text generation by learning a fair model during knowledge distillation. We propose two modifications to the base knowledge distillation based on counterfactual role reversal—modifying teacher probabilities and augmenting the training set. We evaluate gender polarity across professions in open-ended text generated from the resulting distilled and finetuned GPT–2 models and demonstrate a substantial reduction in gender disparity with only a minor compromise in utility. Finally, we observe that language models that reduce gender polarity in language generation do not improve embedding fairness or downstream classification fairness.

Increasing concerns and regulations about data privacy and sparsity necessitate the study of privacy-preserving, decentralized learning methods for natural language processing (NLP) tasks. Federated learning (FL) provides promising approaches for a large number of clients (e.g., personal devices or organizations) to collaboratively learn a shared global model to benefit all clients while allowing users to keep their data locally. Despite interest in studying FL methods for NLP tasks, a systematic comparison and analysis is lacking in the literature. Herein, we present the FedNLP, a benchmarking framework for evaluating federated learning methods on four different task formulations: text classification, sequence tagging, question answering, and seq2seq. We propose a universal interface between Transformer-based language models (e.g., BERT, BART) and FL methods (e.g., FedAvg, FedOPT, etc.) under various non-IID partitioning strategies. Our extensive experiments with FedNLP provide empirical comparisons between FL methods and help us better understand the inherent challenges of this direction. The comprehensive analysis points to intriguing and exciting future research aimed at developing FL methods for NLP tasks.

Natural language understanding (NLU) tasks are typically defined by creating an annotated dataset in which each utterance is encountered once. Such data does not resemble real-world natural language interactions in which certain utterances are encountered frequently, others rarely. For deployed NLU systems this is a vital problem, since the underlying machine learning (ML) models are often fine-tuned on typical NLU data, and then applied to real-world data with a very different distribution. Such systems need to maintain interpretation consistency for both high-frequency utterances and low-frequency utterances. We propose an alternative strategy that explicitly uses utterance frequency in training data to learn models that are more robust to unknown distributions. We present a methodology to simulate utterance usage in two public NLU corpora and create new corpora with head, body and tail segments. We evaluate several methods for joint intent classification and named entity recognition (IC-NER), and use two domain generalization approaches that we adapt to NER. The proposed approaches demonstrate upto 7.02% relative improvement in semantic accuracy over baselines on the tail data. We provide insights as to why the proposed approaches work and show that the reasons for observed improvements do not align with those reported in previous work.

Privacy is an important concern when building statistical models on data containing personal information. Differential privacy offers a strong definition of privacy and can be used to solve several privacy concerns. Multiple solutions have been proposed for the differentially-private transformation of datasets containing sensitive information. However, such transformation algorithms offer poor utility in Natural Language Processing (NLP) tasks due to noise added in the process. This paper addresses this issue by providing a utility-preserving differentially private text transformation algorithm using auto-encoders. Our algorithm transforms text to offer robustness against attacks and produces transformations with high semantic quality that perform well on downstream NLP tasks. We prove our algorithm’s theoretical privacy guarantee and assess its privacy leakage under Membership Inference Attacks (MIA) on models trained with transformed data. Our results show that the proposed model performs better against MIA attacks while offering lower to no degradation in the utility of the underlying transformation process compared to existing baselines.

There is an increasing interest in continuous learning (CL), as data privacy is becoming a priority for real-world machine learning applications. Meanwhile, there is still a lack of academic NLP benchmarks that are applicable for realistic CL settings, which is a major challenge for the advancement of the field. In this paper we discuss some of the unrealistic data characteristics of public datasets, study the challenges of realistic single-task continuous learning as well as the effectiveness of data rehearsal as a way to mitigate accuracy loss. We construct a CL NER dataset from an existing publicly available dataset and release it along with the code to the research community.

Neural Architecture Search (NAS) methods, which automatically learn entire neural model or individual neural cell architectures, have recently achieved competitive or state-of-the-art (SOTA) performance on variety of natural language processing and computer vision tasks, including language modeling, natural language inference, and image classification. In this work, we explore the applicability of a SOTA NAS algorithm, Efficient Neural Architecture Search (ENAS) (Pham et al., 2018) to two sentence pair tasks, paraphrase detection and semantic textual similarity. We use ENAS to perform a micro-level search and learn a task-optimized RNN cell architecture as a drop-in replacement for an LSTM. We explore the effectiveness of ENAS through experiments on three datasets (MRPC, SICK, STS-B), with two different models (ESIM, BiLSTM-Max), and two sets of embeddings (Glove, BERT). In contrast to prior work applying ENAS to NLP tasks, our results are mixed – we find that ENAS architectures sometimes, but not always, outperform LSTMs and perform similarly to random architecture search.