Sebastian Gehrmann

• PhD Student at Harvard University

Datasets are foundational to many breakthroughs in modern artificial intelligence. Many recent achievements in the space of natural language processing (NLP) can be attributed to the finetuning of pre-trained models on a diverse set of tasks that enables a large language model (LLM) to respond to instructions. Instruction fine-tuning (IFT) requires...

We investigate a formalism for the conditions of a successful explanation of AI. We consider “success” to depend not only on what information the explanation contains, but also on what information the human explainee understands from it. Theory of mind literature discusses the folk concepts that humans use to understand and generalize behavior. We...

Pre-trained large language models (PLMs) underlie most new developments in natural language processing. They have shifted the field from application-specific model pipelines to a single model that is adapted to a wide range of tasks. Autoregressive PLMs like GPT-3 or PaLM, alongside techniques like few-shot learning, have additionally shifted the o...

Evaluation practices in natural language generation (NLG) have many known flaws, but improved evaluation approaches are rarely widely adopted. This issue has become more urgent, since neural generation models have improved to the point where their outputs can often no longer be distinguished based on the surface-level features that older metrics re...

Reliable automatic evaluation of summarization systems is challenging due to the multifaceted and subjective nature of the task. This is especially the case for languages other than English, where human evaluations are scarce. In this work, we introduce SEAHORSE, a dataset for multilingual, multifaceted summarization evaluation. SEAHORSE consists o...

We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a mixture of objectives. Through extensive evaluations on English and multilingual language, and reasoning tasks, we demonstrate...

The acquisition of high-quality human annotations through crowdsourcing platforms like Amazon Mechanical Turk (MTurk) is more challenging than expected. The annotation quality might be affected by various aspects like annotation instructions, Human Intelligence Task (HIT) design, and wages paid to annotators, etc. To avoid potentially low-quality a...

The paper presents an approach to semantic grounding of language models (LMs) that conceptualizes the LM as a conditional model generating text given a desired semantic message formalized as a set of entity-relationship triples. It embeds the LM in an auto-encoder by feeding its output to a semantic parser whose output is in the same representation...

Evaluation metrics that are not robust to dialect variation make it impossible to tell how well systems perform for many groups of users, and can even penalize systems for producing text in lower-resource dialects. However, currently, there exists no way to quantify how metrics respond to change in the dialect of a generated utterance. We thus form...

Existing data-to-text generation datasets are mostly limited to English. To address this lack of data, we create Table-to-Text in African languages (TaTa), the first large multilingual table-to-text dataset with a focus on African languages. We created TaTa by transcribing figures and accompanying text in bilingual reports by the Demographic and He...

BIG-Bench (Srivastava et al., 2022) is a diverse evaluation suite that focuses on tasks believed to be beyond the capabilities of current language models. Language models have already made good progress on this benchmark, with the best model in the BIG-Bench paper outperforming average reported human-rater results on 65% of the BIG-Bench tasks via...

Background Care coordination is challenging but crucial for children with medical complexity (CMC). Technology-based solutions are increasingly prevalent but little is known about how to successfully deploy them in the care of CMC. Objective The aim of this study was to assess the feasibility and acceptability of GoalKeeper (GK), an internet-based...

Evaluation in machine learning is usually informed by past choices, for example which datasets or metrics to use. This standardization enables the comparison on equal footing using leaderboards, but the evaluation choices become sub-optimal as better alternatives arise. This problem is especially pertinent in natural language generation which requi...

Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we traine...

Evaluation practices in natural language generation (NLG) have many known flaws, but improved evaluation approaches are rarely widely adopted. This issue has become more urgent, since neural NLG models have improved to the point where they can often no longer be distinguished based on the surface-level features that older metrics rely on. This pape...

When explaining AI behavior to humans, how is the communicated information being comprehended by the human explainee, and does it match what the explanation attempted to communicate? When can we say that an explanation is explaining something? We aim to provide an answer by leveraging theory of mind literature about the folk concepts that humans us...

Data augmentation is an important component in the robustness evaluation of models in natural language processing (NLP) and in enhancing the diversity of the data they are trained on. In this paper, we present NL-Augmenter, a new participatory Python-based natural language augmentation framework which supports the creation of both transformations (...

NLP researchers need more, higher-quality text datasets. Human-labeled datasets are expensive to collect, while datasets collected via automatic retrieval from the web such as WikiBio are noisy and can include undesired biases. Moreover, data sourced from the web is often included in datasets used to pretrain models, leading to inadvertent cross-co...

While different language models are ubiquitous in NLP, it is hard to contrast their outputs and identify which contexts one can handle better than the other. To address this question, we introduce LMdiff, a tool that visually compares probability distributions of two models that differ, e.g., through finetuning, distillation, or simply training wit...

Recent developments in machine translation and multilingual text generation have led researchers to adopt trained metrics such as COMET or BLEURT, which treat evaluation as a regression problem and use representations from multilingual pre-trained models such as XLM-RoBERTa or mBERT. Yet studies on related tasks suggest that these models are most e...

Developing documentation guidelines and easy-to-use templates for datasets and models is a challenging task, especially given the variety of backgrounds, skills, and incentives of the people involved in the building of natural language processing (NLP) tools. Nevertheless, the adoption of standard documentation practices across the field of NLP pro...

Machine learning approaches applied to NLP are often evaluated by summarizing their performance in a single number, for example accuracy. Since most test sets are constructed as an i.i.d. sample from the overall data, this approach overly simplifies the complexity of language and encourages overfitting to the head of the data distribution. As such,...

Targeted syntactic evaluations have demonstrated the ability of language models to perform subject-verb agreement given difficult contexts. To elucidate the mechanisms by which the models accomplish this behavior, this study applies causal mediation analysis to pre-trained neural language models. We investigate the magnitude of models' preferences...

The de novo design of antimicrobial therapeutics involves the exploration of a vast chemical repertoire to find compounds with broad-spectrum potency and low toxicity. Here, we report an efficient computational method for the generation of antimicrobials with desired attributes. The method leverages guidance from classifiers trained on an informati...

We introduce GEM, a living benchmark for natural language Generation (NLG), its Evaluation, and Metrics. Measuring progress in NLG relies on a constantly evolving ecosystem of automated metrics, datasets, and human evaluation standards. However, due to this moving target, new models often still evaluate on divergent anglo-centric corpora with well-...

The quality of machine translation systems has dramatically improved over the last decade, and as a result, evaluation has become an increasingly challenging problem. This paper describes our contribution to the WMT 2020 Metrics Shared Task, the main benchmark for automatic evaluation of translation. Our submission is based on BLEURT, a previously...

We present the Language Interpretability Tool (LIT), an open-source platform for visualization and understanding of NLP models. We focus on core questions about model behavior: Why did my model make this prediction? When does it perform poorly? What happens under a controlled change in the input? LIT integrates local explanations, aggregate analysi...

Objective Shared decision-making (SDM) may improve outcomes for children with medical complexity (CMC). CMC have lower rates of SDM than other children, but little is known about how to improve SDM for CMC. The objective of this study is to describe parent perspectives of SDM for CMC and identify opportunities to improve elements of SDM specific to...

De novo therapeutic design is challenged by a vast chemical repertoire and multiple constraints such as high broad-spectrum potency and low toxicity. We propose CLaSS (Controlled Latent attribute Space Sampling) - a novel and efficient computational method for attribute-controlled generation of molecules, which leverages guidance from classifiers t...

We present ToTTo, an open-domain English table-to-text dataset with over 120,000 training examples that proposes a controlled generation task: given a Wikipedia table and a set of highlighted table cells, produce a one-sentence description. To obtain generated targets that are natural but also faithful to the source table, we introduce a dataset co...

Common methods for interpreting neural models in natural language processing typically examine either their structure or their behavior, but not both. We propose a methodology grounded in the theory of causal mediation analysis for interpreting which parts of a model are causally implicated in its behavior. It enables us to analyze the mechanisms b...

A crucial step within secondary analysis of electronic health records (EHRs) is to identify the patient cohort under investigation. While EHRs contain medical billing codes that aim to represent the conditions and treatments patients may have, much of the information is only present in the patient notes. Therefore, it is critical to develop robust...

We introduce three memory-augmented Recurrent Neural Networks (MARNNs) and explore their capabilities on a series of simple language modeling tasks whose solutions require stack-based mechanisms. We provide the first demonstration of neural networks recognizing the generalized Dyck languages, which express the core of what it means to be a language...

Large language models can produce powerful contextual representations that lead to improvements across many NLP tasks. Since these models are typically guided by a sequence of learned self attention mechanisms and may comprise undesired inductive biases, it is paramount to be able to explore what the attention has learned. While static analyses of...

Large pretrained language models have changed the way researchers approach discriminative natural language understanding tasks, leading to the dominance of approaches that adapt a pretrained model for arbitrary downstream tasks. However it is an open-question how to use similar techniques for language generation. Early results in the encoder-agnost...

Automation of tasks can have critical consequences when humans lose agency over decision processes. Deep learning models are particularly susceptible since current black-box approaches lack explainable reasoning. We argue that both the visual interface and model structure of deep learning systems need to take into account interaction design. We pro...

Automation of tasks can have critical consequences when humans lose agency over decision processes. Deep learning models are particularly susceptible since current black-box approaches lack explainable reasoning. We argue that both the visual interface and model structure of deep learning systems need to take into account interaction design. We pro...

The rapid improvement of language models has raised the specter of abuse of text generation systems. This progress motivates the development of simple methods for detecting generated text that can be used by and explained to non-experts. We develop GLTR, a tool to support humans in detecting whether a text was generated by a model. GLTR applies a s...

In this paper, we systematically assess the ability of standard recurrent networks to perform dynamic counting and to encode hierarchical representations. All the neural models in our experiments are designed to be small-sized networks both to prevent them from memorizing the training sets and to visualize and interpret their behaviour at test time...

Titles of short sections within long documents support readers by guiding their focus towards relevant passages and by providing anchor-points that help to understand the progression of the document. The positive effects of section titles are even more pronounced when measured on readers with less developed reading abilities, for example in communi...

Neural sequence-to-sequence models have proven to be accurate and robust for many sequence prediction tasks, and have become the standard approach for automatic translation of text. The models work with a five-stage blackbox pipeline that begins with encoding a source sequence to a vector space and then decoding out to a new target sequence. This p...

Learning to generate fluent natural language from structured data with neural networks has become an common approach for NLG. This problem can be challenging when the form of the structured data varies between examples. This paper presents a survey of several extensions to sequence-to-sequence models to account for the latent content selection proc...

Neural network-based methods for abstractive summarization produce outputs that are more fluent than other techniques, but which can be poor at content selection. This work proposes a simple technique for addressing this issue: use a data-efficient content selector to over-determine phrases in a source document that should be part of the summary. W...

Neural Sequence-to-Sequence models have proven to be accurate and robust for many sequence prediction tasks, and have become the standard approach for automatic translation of text. The models work in a five stage blackbox process that involves encoding a source sequence to a vector space and then decoding out to a new target sequence. This process...

In secondary analysis of electronic health records, a crucial task consists in correctly identifying the patient cohort under investigation. In many cases, the most valuable and relevant information for an accurate classification of medical conditions exist only in clinical narratives. Therefore, it is necessary to use natural language processing (...

Overview of CNN results with different convolution widths. Each column name shows the minimum and maximum width of the convolution. (PDF)

Overview of cTAKES results with different models with all input features and the filtered lists of inputs. While in most cases, the clinician-defined phrase dictionary improves the model performance, the full input performs almost as well and outperforms the filtered model in some. (PDF)

Additional training information for the models. (PDF)

Results of different n-gram based models. Each column name shows the minimum and maximum length of phrase that has been considered. We observe that in most cases, a simple bag of words (phrase length 1) outperforms all other models. (PDF)

Advancement of Artificial Intelligence (AI) capabilities in medicine can help address many pressing problems in healthcare. However, AI research endeavors in healthcare may not be clinically relevant, may have unrealistic expectations, or may not be explicit enough about their limitations. A diverse and well-functioning multidisciplinary team (MDT)...

Objective: We investigate whether deep learning techniques for natural language processing (NLP) can be used efficiently for patient phenotyping. Patient phenotyping is a classification task for determining whether a patient has a medical condition, and is a crucial part of secondary analysis of healthcare data. We assess the performance of deep le...

Recurrent neural networks, and in particular long short-term memory networks (LSTMs), are a remarkably effective tool for sequence modeling that learn a dense black-box hidden representation of their sequential input. Researchers interested in better understanding these models have studied the changes in hidden state representations over time and n...

Recurrent neural networks, and in particular long short-term memory (LSTM) networks, are a remarkably effective tool for sequence modeling that learn a dense black-box hidden representation of their sequential input. Researchers interested in better understanding these models have studied the changes in hidden state representations over time and no...

Many documents (e.g., academic papers, government reports) are typically written by multiple authors. While existing tools facilitate and support such collaborative efforts (e.g., Dropbox, Google Docs), these tools lack intelligent information sharing mechanisms. Capabilities such as " track changes " and " diff " visualize changes to authors, but...