Griffin Adams

Griffin Adams, Alexander Fabbri, Faisal Ladhak et al.

Selecting the ``right'' amount of information to include in a summary is a difficult task. A good summary should be detailed and entity-centric without being overly dense and hard to follow. To better understand this tradeoff, we solicit increasingly dense GPT-4 summaries with what we refer to as a ``Chain of Density'' (CoD) prompt. Specifically, GPT-4 generates an initial entity-sparse summary before iteratively incorporating missing salient entities without increasing the length. Summaries generated by CoD are more abstractive, exhibit more fusion, and have less of a lead bias than GPT-4 summaries generated by a vanilla prompt. We conduct a human preference study on 100 CNN DailyMail articles and find that that humans prefer GPT-4 summaries that are more dense than those generated by a vanilla prompt and almost as dense as human written summaries. Qualitative analysis supports the notion that there exists a tradeoff between informativeness and readability. 500 annotated CoD summaries, as well as an extra 5,000 unannotated summaries, are freely available on HuggingFace (https://huggingface.co/datasets/griffin/chain_of_density).

Griffin Adams, Han-Chin Shing, Qing Sun et al.

In real-world scenarios with naturally occurring datasets, reference summaries are noisy and may contain information that cannot be inferred from the source text. On large news corpora, removing low quality samples has been shown to reduce model hallucinations. Yet, for smaller, and/or noisier corpora, filtering is detrimental to performance. To improve reference quality while retaining all data, we propose a new approach: to selectively re-write unsupported reference sentences to better reflect source data. We automatically generate a synthetic dataset of positive and negative revisions by corrupting supported sentences and learn to revise reference sentences with contrastive learning. The intensity of revisions is treated as a controllable attribute so that, at inference, diverse candidates can be over-generated-then-rescored to balance faithfulness and abstraction. To test our methods, we extract noisy references from publicly available MIMIC-III discharge summaries for the task of hospital-course summarization, and vary the data on which models are trained. According to metrics and human evaluation, models trained on revised clinical references are much more faithful, informative, and fluent than models trained on original or filtered data.

Griffin Adams, Jason Zucker, Noémie Elhadad

Long-form clinical summarization of hospital admissions has real-world significance because of its potential to help both clinicians and patients. The faithfulness of summaries is critical to their safe usage in clinical settings. To better understand the limitations of abstractive systems, as well as the suitability of existing evaluation metrics, we benchmark faithfulness metrics against fine-grained human annotations for model-generated summaries of a patient's Brief Hospital Course. We create a corpus of patient hospital admissions and summaries for a cohort of HIV patients, each with complex medical histories. Annotators are presented with summaries and source notes, and asked to categorize manually highlighted summary elements (clinical entities like conditions and medications as well as actions like "following up") into one of three categories: ``Incorrect,'' ``Missing,'' and ``Not in Notes.'' We meta-evaluate a broad set of proposed faithfulness metrics and, across metrics, explore the importance of domain adaptation (e.g. the impact of in-domain pre-training and metric fine-tuning), the use of source-summary alignments, and the effects of distilling a single metric from an ensemble of pre-existing metrics. Off-the-shelf metrics with no exposure to clinical text correlate well yet overly rely on summary extractiveness. As a practical guide to long-form clinical narrative summarization, we find that most metrics correlate best to human judgments when provided with one summary sentence at a time and a minimal set of relevant source context.

Melanie Subbiah, Faisal Ladhak, Akankshya Mishra et al.

Human evaluation has been the gold standard for checking faithfulness in abstractive summarization. However, with a challenging source domain like narrative, multiple annotators can agree a summary is faithful, while missing details that are obvious errors only once pointed out. We therefore introduce a new dataset, STORYSUMM, comprising LLM summaries of short stories with localized faithfulness labels and error explanations. This benchmark is for evaluation methods, testing whether a given method can detect challenging inconsistencies. Using this dataset, we first show that any one human annotation protocol is likely to miss inconsistencies, and we advocate for pursuing a range of methods when establishing ground truth for a summarization dataset. We finally test recent automatic metrics and find that none of them achieve more than 70% balanced accuracy on this task, demonstrating that it is a challenging benchmark for future work in faithfulness evaluation.

Benjamin Clavié, Antoine Chaffin, Griffin Adams

Over the last few years, multi-vector retrieval methods, spearheaded by ColBERT, have become an increasingly popular approach to Neural IR. By storing representations at the token level rather than at the document level, these methods have demonstrated very strong retrieval performance, especially in out-of-domain settings. However, the storage and memory requirements necessary to store the large number of associated vectors remain an important drawback, hindering practical adoption. In this paper, we introduce a simple clustering-based token pooling approach to aggressively reduce the number of vectors that need to be stored. This method can reduce the space & memory footprint of ColBERT indexes by 50% with virtually no retrieval performance degradation. This method also allows for further reductions, reducing the vector count by 66%-to-75% , with degradation remaining below 5% on a vast majority of datasets. Importantly, this approach requires no architectural change nor query-time processing, and can be used as a simple drop-in during indexation with any ColBERT-like model.

Griffin Adams, Jason Zucker, Noémie Elhadad

Clinician must write a lengthy summary each time a patient is discharged from the hospital. This task is time-consuming due to the sheer number of unique clinical concepts covered in the admission. Identifying and covering salient entities is vital for the summary to be clinically useful. We fine-tune open-source LLMs (Mistral-7B-Instruct and Zephyr-7B-beta) on the task and find that they generate incomplete and unfaithful summaries. To increase entity coverage, we train a smaller, encoder-only model to predict salient entities, which are treated as content-plans to guide the LLM. To encourage the LLM to focus on specific mentions in the source notes, we propose SPEER: Sentence-level Planning via Embedded Entity Retrieval. Specifically, we mark each salient entity span with special "{ }" boundary tags and instruct the LLM to retrieve marked spans before generating each sentence. Sentence-level planning acts as a form of state tracking in that the model is explicitly recording the entities it uses. We fine-tune Mistral and Zephyr variants on a large-scale, diverse dataset of ~167k in-patient hospital admissions and evaluate on 3 datasets. SPEER shows gains in both coverage and faithfulness metrics over non-guided and guided baselines.

Benjamin Warner, Antoine Chaffin, Benjamin Clavié et al.

Encoder-only transformer models such as BERT offer a great performance-size tradeoff for retrieval and classification tasks with respect to larger decoder-only models. Despite being the workhorse of numerous production pipelines, there have been limited Pareto improvements to BERT since its release. In this paper, we introduce ModernBERT, bringing modern model optimizations to encoder-only models and representing a major Pareto improvement over older encoders. Trained on 2 trillion tokens with a native 8192 sequence length, ModernBERT models exhibit state-of-the-art results on a large pool of evaluations encompassing diverse classification tasks and both single and multi-vector retrieval on different domains (including code). In addition to strong downstream performance, ModernBERT is also the most speed and memory efficient encoder and is designed for inference on common GPUs.

Griffin Adams, Alexander R. Fabbri, Faisal Ladhak et al.

Two-step approaches, in which summary candidates are generated-then-reranked to return a single summary, can improve ROUGE scores over the standard single-step approach. Yet, standard decoding methods (i.e., beam search, nucleus sampling, and diverse beam search) produce candidates with redundant, and often low quality, content. In this paper, we design a novel method to generate candidates for re-ranking that addresses these issues. We ground each candidate abstract on its own unique content plan and generate distinct plan-guided abstracts using a model's top beam. More concretely, a standard language model (a BART LM) auto-regressively generates elemental discourse unit (EDU) content plans with an extractive copy mechanism. The top K beams from the content plan generator are then used to guide a separate LM, which produces a single abstractive candidate for each distinct plan. We apply an existing re-ranker (BRIO) to abstractive candidates generated from our method, as well as baseline decoding methods. We show large relevance improvements over previously published methods on widely used single document news article corpora, with ROUGE-2 F1 gains of 0.88, 2.01, and 0.38 on CNN / Dailymail, NYT, and Xsum, respectively. A human evaluation on CNN / DM validates these results. Similarly, on 1k samples from CNN / DM, we show that prompting GPT-3 to follow EDU plans outperforms sampling-based methods by 1.05 ROUGE-2 F1 points. Code to generate and realize plans is available at https://github.com/griff4692/edu-sum.

Griffin Adams, Bichlien H Nguyen, Jake Smith et al.

Summarization models often generate text that is poorly calibrated to quality metrics because they are trained to maximize the likelihood of a single reference (MLE). To address this, recent work has added a calibration step, which exposes a model to its own ranked outputs to improve relevance or, in a separate line of work, contrasts positive and negative sets to improve faithfulness. While effective, much of this work has focused on how to generate and optimize these sets. Less is known about why one setup is more effective than another. In this work, we uncover the underlying characteristics of effective sets. For each training instance, we form a large, diverse pool of candidates and systematically vary the subsets used for calibration fine-tuning. Each selection strategy targets distinct aspects of the sets, such as lexical diversity or the size of the gap between positive and negatives. On three diverse scientific long-form summarization datasets (spanning biomedical, clinical, and chemical domains), we find, among others, that faithfulness calibration is optimal when the negative sets are extractive and more likely to be generated, whereas for relevance calibration, the metric margin between candidates should be maximized and surprise--the disagreement between model and metric defined candidate rankings--minimized. Code to create, select, and optimize calibration sets is available at https://github.com/griff4692/calibrating-summaries

Griffin Adams

The rapid adoption of Electronic Health Records (EHRs) has been instrumental in streamlining administrative tasks, increasing transparency, and enabling continuity of care across providers. An unintended consequence of the increased documentation burden, however, has been reduced face-time with patients and, concomitantly, a dramatic rise in clinician burnout. In this thesis, we pinpoint a particularly time-intensive, yet critical, documentation task: generating a summary of a patient's hospital admissions, and propose and evaluate automated solutions. In Chapter 2, we construct a dataset based on 109,000 hospitalizations (2M source notes) and perform exploratory analyses to motivate future work on modeling and evaluation [NAACL 2021]. In Chapter 3, we address faithfulness from a modeling perspective by revising noisy references [EMNLP 2022] and, to reduce the reliance on references, directly calibrating model outputs to metrics [ACL 2023]. These works relied heavily on automatic metrics as human annotations were limited. To fill this gap, in Chapter 4, we conduct a fine-grained expert annotation of system errors in order to meta-evaluate existing metrics and better understand task-specific issues of domain adaptation and source-summary alignments. To learn a metric less correlated to extractiveness (copy-and-paste), we derive noisy faithfulness labels from an ensemble of existing metrics and train a faithfulness classifier on these pseudo labels [MLHC 2023]. Finally, in Chapter 5, we demonstrate that fine-tuned LLMs (Mistral and Zephyr) are highly prone to entity hallucinations and cover fewer salient entities. We improve both coverage and faithfulness by performing sentence-level entity planning based on a set of pre-computed salient entities from the source text, which extends our work on entity-guided news summarization [ACL, 2023], [EMNLP, 2023].

Griffin Adams, Emily Alsentzer, Mert Ketenci et al.

Summarization of clinical narratives is a long-standing research problem. Here, we introduce the task of hospital-course summarization. Given the documentation authored throughout a patient's hospitalization, generate a paragraph that tells the story of the patient admission. We construct an English, text-to-text dataset of 109,000 hospitalizations (2M source notes) and their corresponding summary proxy: the clinician-authored "Brief Hospital Course" paragraph written as part of a discharge note. Exploratory analyses reveal that the BHC paragraphs are highly abstractive with some long extracted fragments; are concise yet comprehensive; differ in style and content organization from the source notes; exhibit minimal lexical cohesion; and represent silver-standard references. Our analysis identifies multiple implications for modeling this complex, multi-document summarization task.

Griffin Adams, Sarguna Janani Padmanabhan, Shivang Shekhar

We address two major challenges of implicit coordination in multi-agent deep reinforcement learning: non-stationarity and exponential growth of state-action space, by combining Deep-Q Networks for policy learning with Nash equilibrium for action selection. Q-values proxy as payoffs in Nash settings, and mutual best responses define joint action selection. Coordination is implicit because multiple/no Nash equilibria are resolved deterministically. We demonstrate that knowledge of game type leads to an assumption of mirrored best responses and faster convergence than Nash-Q. Specifically, the Friend-or-Foe algorithm demonstrates signs of convergence to a Set Controller which jointly chooses actions for two agents. This encouraging given the highly unstable nature of decentralized coordination over joint actions. Inspired by the dueling network architecture, which decouples the Q-function into state and advantage streams, as well as residual networks, we learn both a single and joint agent representation, and merge them via element-wise addition. This simplifies coordination by recasting it is as learning a residual function. We also draw high level comparative insights on key MADRL and game theoretic variables: competitive vs. cooperative, asynchronous vs. parallel learning, greedy versus socially optimal Nash equilibria tie breaking, and strategies for the no Nash equilibrium case. We evaluate on 3 custom environments written in Python using OpenAI Gym: a Predator Prey environment, an alternating Warehouse environment, and a Synchronization environment. Each environment requires successively more coordination to achieve positive rewards.

Griffin Adams, Mert Ketenci, Shreyas Bhave et al.

We introduce Latent Meaning Cells, a deep latent variable model which learns contextualized representations of words by combining local lexical context and metadata. Metadata can refer to granular context, such as section type, or to more global context, such as unique document ids. Reliance on metadata for contextualized representation learning is apropos in the clinical domain where text is semi-structured and expresses high variation in topics. We evaluate the LMC model on the task of zero-shot clinical acronym expansion across three datasets. The LMC significantly outperforms a diverse set of baselines at a fraction of the pre-training cost and learns clinically coherent representations. We demonstrate that not only is metadata itself very helpful for the task, but that the LMC inference algorithm provides an additional large benefit.

Monica Agrawal, Griffin Adams, Nathan Nussbaum et al.

Oral drugs are becoming increasingly common in oncology care. In contrast to intravenous chemotherapy, which is administered in the clinic and carefully tracked via structure electronic health records (EHRs), oral drug treatment is self-administered and therefore not tracked as well. Often, the details of oral cancer treatment occur only in unstructured clinic notes. Extracting this information is critical to understanding a patient's treatment history. Yet, this a challenging task because treatment intervals must be inferred longitudinally from both explicit mentions in the text as well as from document timestamps. In this work, we present TIFTI (Temporally Integrated Framework for Treatment Intervals), a robust framework for extracting oral drug treatment intervals from a patient's unstructured notes. TIFTI leverages distinct sources of temporal information by breaking the problem down into two separate subtasks: document-level sequence labeling and date extraction. On a labeled dataset of metastatic renal-cell carcinoma (RCC) patients, it exactly matched the labeled start date in 46% of the examples (86% of the examples within 30 days), and it exactly matched the labeled end date in 52% of the examples (78% of the examples within 30 days). Without retraining, the model achieved a similar level of performance on a labeled dataset of advanced non-small-cell lung cancer (NSCLC) patients.