Search

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.

Pre-trained large-scale language models (LLMs) excel at producing coherent articles, yet their outputs may be untruthful, toxic, or fail to align with user expectations. Current approaches focus on using reinforcement learning with human feedback (RLHF) to improve model alignment, which works by transforming coarse human preferences of LLM outputs into a feedback signal that guides the model learning process. However, because this approach operates on sequence-level feedback, it lacks the precision to identify the exact parts of the output affecting user preferences. To address this gap, we propose a method to enhance LLM alignment through fine-grained token-level supervision. Specifically, we ask annotators to minimally edit less preferred responses within the standard reward modeling dataset to make them more favorable, ensuring changes are made only where necessary while retaining most of the original content. The refined dataset is used to train a token-level reward model, which is then used for training our fine-grained Proximal Policy Optimization (PPO) model. Our experiment results demonstrate that this approach can achieve up to an absolute improvement of $5.1\%$ in LLM performance, in terms of win rate against the reference model, compared with the traditional PPO model.

Large language models are trained on vast amounts of internet data, prompting concerns and speculation that they have memorized public benchmarks. Going from speculation to proof of contamination is challenging, as the pretraining data used by proprietary models are often not publicly accessible. We show that it is possible to provide provable guarantees of test set contamination in language models without access to pretraining data or model weights. Our approach leverages the fact that when there is no data contamination, all orderings of an exchangeable benchmark should be equally likely. In contrast, the tendency for language models to memorize example order means that a contaminated language model will find certain canonical orderings to be much more likely than others. Our test flags potential contamination whenever the likelihood of a canonically ordered benchmark dataset is significantly higher than the likelihood after shuffling the examples. We demonstrate that our procedure is sensitive enough to reliably prove test set contamination in challenging situations, including models as small as 1.4 billion parameters, on small test sets of only 1000 examples, and datasets that appear only a few times in the pretraining corpus. Using our test, we audit five popular publicly accessible language models for test set contamination and find little evidence for pervasive contamination.

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)., Comment: preprint

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., Comment: ACL 2023

Language models (LMs) are increasingly being used in open-ended contexts, where the opinions reflected by LMs in response to subjective queries can have a profound impact, both on user satisfaction, as well as shaping the views of society at large. In this work, we put forth a quantitative framework to investigate the opinions reflected by LMs -- by leveraging high-quality public opinion polls and their associated human responses. Using this framework, we create OpinionsQA, a new dataset for evaluating the alignment of LM opinions with those of 60 US demographic groups over topics ranging from abortion to automation. Across topics, we find substantial misalignment between the views reflected by current LMs and those of US demographic groups: on par with the Democrat-Republican divide on climate change. Notably, this misalignment persists even after explicitly steering the LMs towards particular demographic groups. Our analysis not only confirms prior observations about the left-leaning tendencies of some human feedback-tuned LMs, but also surfaces groups whose opinions are poorly reflected by current LMs (e.g., 65+ and widowed individuals). Our code and data are available at https://github.com/tatsu-lab/opinions_qa.

Large language models (LLMs) have shown promise for automatic summarization but the reasons behind their successes are poorly understood. By conducting a human evaluation on ten LLMs across different pretraining methods, prompts, and model scales, we make two important observations. First, we find instruction tuning, and not model size, is the key to the LLM's zero-shot summarization capability. Second, existing studies have been limited by low-quality references, leading to underestimates of human performance and lower few-shot and finetuning performance. To better evaluate LLMs, we perform human evaluation over high-quality summaries we collect from freelance writers. Despite major stylistic differences such as the amount of paraphrasing, we find that LMM summaries are judged to be on par with human written summaries.

Recent work has identified noisy and misannotated data as a core cause of hallucinations and unfaithful outputs in Natural Language Generation (NLG) tasks. Consequently, identifying and removing these examples is a key open challenge in creating reliable NLG systems. In this work, we introduce a framework to identify and remove low-quality training instances that lead to undesirable outputs, such as faithfulness errors in text summarization. We show that existing approaches for error tracing, such as gradient-based influence measures, do not perform reliably for detecting faithfulness errors in NLG datasets. We overcome the drawbacks of existing error tracing methods through a new, contrast-based estimate that compares undesired generations to human-corrected outputs. Our proposed method can achieve a mean average precision of 0.93 at detecting known data errors across synthetic tasks with known ground truth, substantially outperforming existing approaches. Using this approach and re-training models on cleaned data leads to a 70% reduction in entity hallucinations on the NYT dataset and a 55% reduction in semantic errors on the E2E dataset., Comment: ACL 2023

Many real-world applications of language models (LMs), such as writing assistance and code autocomplete, involve human-LM interaction. However, most benchmarks are non-interactive in that a model produces output without human involvement. To evaluate human-LM interaction, we develop a new framework, Human-AI Language-based Interaction Evaluation (HALIE), that defines the components of interactive systems and dimensions to consider when designing evaluation metrics. Compared to standard, non-interactive evaluation, HALIE captures (i) the interactive process, not only the final output; (ii) the first-person subjective experience, not just a third-party assessment; and (iii) notions of preference beyond quality (e.g., enjoyment and ownership). We then design five tasks to cover different forms of interaction: social dialogue, question answering, crossword puzzles, summarization, and metaphor generation. With four state-of-the-art LMs (three variants of OpenAI's GPT-3 and AI21 Labs' Jurassic-1), we find that better non-interactive performance does not always translate to better human-LM interaction. In particular, we highlight three cases where the results from non-interactive and interactive metrics diverge and underscore the importance of human-LM interaction for LM evaluation., Comment: Authored by the Center for Research on Foundation Models (CRFM) at the Stanford Institute for Human-Centered Artificial Intelligence (HAI)

Language models (LMs) are becoming the foundation for almost all major language technologies, but their capabilities, limitations, and risks are not well understood. We present Holistic Evaluation of Language Models (HELM) to improve the transparency of language models. First, we taxonomize the vast space of potential scenarios (i.e. use cases) and metrics (i.e. desiderata) that are of interest for LMs. Then we select a broad subset based on coverage and feasibility, noting what's missing or underrepresented (e.g. question answering for neglected English dialects, metrics for trustworthiness). Second, we adopt a multi-metric approach: We measure 7 metrics (accuracy, calibration, robustness, fairness, bias, toxicity, and efficiency) for each of 16 core scenarios when possible (87.5% of the time). This ensures metrics beyond accuracy don't fall to the wayside, and that trade-offs are clearly exposed. We also perform 7 targeted evaluations, based on 26 targeted scenarios, to analyze specific aspects (e.g. reasoning, disinformation). Third, we conduct a large-scale evaluation of 30 prominent language models (spanning open, limited-access, and closed models) on all 42 scenarios, 21 of which were not previously used in mainstream LM evaluation. Prior to HELM, models on average were evaluated on just 17.9% of the core HELM scenarios, with some prominent models not sharing a single scenario in common. We improve this to 96.0%: now all 30 models have been densely benchmarked on the same core scenarios and metrics under standardized conditions. Our evaluation surfaces 25 top-level findings. For full transparency, we release all raw model prompts and completions publicly for further analysis, as well as a general modular toolkit. We intend for HELM to be a living benchmark for the community, continuously updated with new scenarios, metrics, and models., Comment: Authored by the Center for Research on Foundation Models (CRFM) at the Stanford Institute for Human-Centered Artificial Intelligence (HAI). Project page: https://crfm.stanford.edu/helm/v1.0

This paper introduces the shared task of summarizing documents in several creative domains, namely literary texts, movie scripts, and television scripts. Summarizing these creative documents requires making complex literary interpretations, as well as understanding non-trivial temporal dependencies in texts containing varied styles of plot development and narrative structure. This poses unique challenges and is yet underexplored for text summarization systems. In this shared task, we introduce four sub-tasks and their corresponding datasets, focusing on summarizing books, movie scripts, primetime television scripts, and daytime soap opera scripts. We detail the process of curating these datasets for the task, as well as the metrics used for the evaluation of the submissions. As part of the CREATIVESUMM workshop at COLING 2022, the shared task attracted 18 submissions in total. We discuss the submissions and the baselines for each sub-task in this paper, along with directions for facilitating future work in the field., Comment: 4 pages + 3 for references and appendix

Summarizing novel chapters is a difficult task due to the input length and the fact that sentences that appear in the desired summaries draw content from multiple places throughout the chapter. We present a pipelined extractive-abstractive approach where the extractive step filters the content that is passed to the abstractive component. Extremely lengthy input also results in a highly skewed dataset towards negative instances for extractive summarization; we thus adopt a margin ranking loss for extraction to encourage separation between positive and negative examples. Our extraction component operates at the constituent level; our approach to this problem enriches the text with spinal tree information which provides syntactic context (in the form of constituents) to the extraction model. We show an improvement of 3.71 Rouge-1 points over best results reported in prior work on an existing novel chapter dataset.

Machine learning models that convert user-written text descriptions into images are now widely available online and used by millions of users to generate millions of images a day. We investigate the potential for these models to amplify dangerous and complex stereotypes. We find a broad range of ordinary prompts produce stereotypes, including prompts simply mentioning traits, descriptors, occupations, or objects. For example, we find cases of prompting for basic traits or social roles resulting in images reinforcing whiteness as ideal, prompting for occupations resulting in amplification of racial and gender disparities, and prompting for objects resulting in reification of American norms. Stereotypes are present regardless of whether prompts explicitly mention identity and demographic language or avoid such language. Moreover, stereotypes persist despite mitigation strategies; neither user attempts to counter stereotypes by requesting images with specific counter-stereotypes nor institutional attempts to add system ``guardrails'' have prevented the perpetuation of stereotypes. Our analysis justifies concerns regarding the impacts of today's models, presenting striking exemplars, and connecting these findings with deep insights into harms drawn from social scientific and humanist disciplines. This work contributes to the effort to shed light on the uniquely complex biases in language-vision models and demonstrates the ways that the mass deployment of text-to-image generation models results in mass dissemination of stereotypes and resulting harms., Comment: FAccT 2023 paper. The published version is available at 10.1145/3593013.3594095

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 requires ever-improving suites of datasets, metrics, and human evaluation to make definitive claims. To make following best model evaluation practices easier, we introduce GEMv2. The new version of the Generation, Evaluation, and Metrics Benchmark introduces a modular infrastructure for dataset, model, and metric developers to benefit from each others work. GEMv2 supports 40 documented datasets in 51 languages. Models for all datasets can be evaluated online and our interactive data card creation and rendering tools make it easier to add new datasets to the living benchmark.

Hate speech detection is complex; it relies on commonsense reasoning, knowledge of stereotypes, and an understanding of social nuance that differs from one culture to the next. It is also difficult to collect a large-scale hate speech annotated dataset. In this work, we frame this problem as a few-shot learning task, and show significant gains with decomposing the task into its "constituent" parts. In addition, we see that infusing knowledge from reasoning datasets (e.g. Atomic2020) improves the performance even further. Moreover, we observe that the trained models generalize to out-of-distribution datasets, showing the superiority of task decomposition and knowledge infusion compared to previously used methods. Concretely, our method outperforms the baseline by 17.83% absolute gain in the 16-shot case., Comment: Accepted at EMNLP 2022

Model-based, reference-free evaluation metrics have been proposed as a fast and cost-effective approach to evaluate Natural Language Generation (NLG) systems. Despite promising recent results, we find evidence that reference-free evaluation metrics of summarization and dialog generation may be relying on spurious correlations with measures such as word overlap, perplexity, and length. We further observe that for text summarization, these metrics have high error rates when ranking current state-of-the-art abstractive summarization systems. We demonstrate that these errors can be mitigated by explicitly designing evaluation metrics to avoid spurious features in reference-free evaluation., Comment: Published in ACL 2022 main conference

Despite recent progress in abstractive summarization, systems still suffer from faithfulness errors. While prior work has proposed models that improve faithfulness, it is unclear whether the improvement comes from an increased level of extractiveness of the model outputs as one naive way to improve faithfulness is to make summarization models more extractive. In this work, we present a framework for evaluating the effective faithfulness of summarization systems, by generating a faithfulnessabstractiveness trade-off curve that serves as a control at different operating points on the abstractiveness spectrum. We then show that the Maximum Likelihood Estimation (MLE) baseline as well as a recently proposed method for improving faithfulness, are both worse than the control at the same level of abstractiveness. Finally, we learn a selector to identify the most faithful and abstractive summary for a given document, and show that this system can attain higher faithfulness scores in human evaluations while being more abstractive than the baseline system on two datasets. Moreover, we show that our system is able to achieve a better faithfulness-abstractiveness trade-off than the control at the same level of abstractiveness., Comment: Published in ACL 2022 main conference

AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature., Comment: Authored by the Center for Research on Foundation Models (CRFM) at the Stanford Institute for Human-Centered Artificial Intelligence (HAI). Report page with citation guidelines: https://crfm.stanford.edu/report.html

Typical ASR systems segment the input audio into utterances using purely acoustic information, which may not resemble the sentence-like units that are expected by conventional machine translation (MT) systems for Spoken Language Translation. In this work, we propose a model for correcting the acoustic segmentation of ASR models for low-resource languages to improve performance on downstream tasks. We propose the use of subtitles as a proxy dataset for correcting ASR acoustic segmentation, creating synthetic acoustic utterances by modeling common error modes. We train a neural tagging model for correcting ASR acoustic segmentation and show that it improves downstream performance on MT and audio-document cross-language information retrieval (CLIR).

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. Due to this moving target, new models often still evaluate on divergent anglo-centric corpora with well-established, but flawed, metrics. This disconnect makes it challenging to identify the limitations of current models and opportunities for progress. Addressing this limitation, GEM provides an environment in which models can easily be applied to a wide set of tasks and in which evaluation strategies can be tested. Regular updates to the benchmark will help NLG research become more multilingual and evolve the challenge alongside models. This paper serves as the description of the data for which we are organizing a shared task at our ACL 2021 Workshop and to which we invite the entire NLG community to participate.

Leveraging large amounts of unlabeled data using Transformer-like architectures, like BERT, has gained popularity in recent times owing to their effectiveness in learning general representations that can then be further fine-tuned for downstream tasks to much success. However, training these models can be costly both from an economic and environmental standpoint. In this work, we investigate how to effectively use unlabeled data: by exploring the task-specific semi-supervised approach, Cross-View Training (CVT) and comparing it with task-agnostic BERT in multiple settings that include domain and task relevant English data. CVT uses a much lighter model architecture and we show that it achieves similar performance to BERT on a set of sequence tagging tasks, with lesser financial and environmental impact., Comment: Accepted in the Proceedings of 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP 2020)(https://2020.emnlp.org/papers/main)

Users of machine translation (MT) may want to ensure the use of specific lexical terminologies. While there exist techniques for incorporating terminology constraints during inference for MT, current APE approaches cannot ensure that they will appear in the final translation. In this paper, we present both autoregressive and non-autoregressive models for lexically constrained APE, demonstrating that our approach enables preservation of 95% of the terminologies and also improves translation quality on English-German benchmarks. Even when applied to lexically constrained MT output, our approach is able to improve preservation of the terminologies. However, we show that our models do not learn to copy constraints systematically and suggest a simple data augmentation technique that leads to improved performance and robustness., Comment: To appear in WMT, 2020

We introduce WikiLingua, a large-scale, multilingual dataset for the evaluation of crosslingual abstractive summarization systems. We extract article and summary pairs in 18 languages from WikiHow, a high quality, collaborative resource of how-to guides on a diverse set of topics written by human authors. We create gold-standard article-summary alignments across languages by aligning the images that are used to describe each how-to step in an article. As a set of baselines for further studies, we evaluate the performance of existing cross-lingual abstractive summarization methods on our dataset. We further propose a method for direct crosslingual summarization (i.e., without requiring translation at inference time) by leveraging synthetic data and Neural Machine Translation as a pre-training step. Our method significantly outperforms the baseline approaches, while being more cost efficient during inference., Comment: Findings of EMNLP 2020

We present a new summarization task, generating summaries of novel chapters using summary/chapter pairs from online study guides. This is a harder task than the news summarization task, given the chapter length as well as the extreme paraphrasing and generalization found in the summaries. We focus on extractive summarization, which requires the creation of a gold-standard set of extractive summaries. We present a new metric for aligning reference summary sentences with chapter sentences to create gold extracts and also experiment with different alignment methods. Our experiments demonstrate significant improvement over prior alignment approaches for our task as shown through automatic metrics and a crowd-sourced pyramid analysis. We make our data collection scripts available at https://github.com/manestay/novel-chapter-dataset ., Comment: ACL 2020

Research in the social sciences and psychology has shown that the persuasiveness of an argument depends not only the language employed, but also on attributes of the source/communicator, the audience, and the appropriateness and strength of the argument's claims given the pragmatic and discourse context of the argument. Among these characteristics of persuasive arguments, prior work in NLP does not explicitly investigate the effect of the pragmatic and discourse context when determining argument quality. This paper presents a new dataset to initiate the study of this aspect of argumentation: it consists of a diverse collection of arguments covering 741 controversial topics and comprising over 47,000 claims. We further propose predictive models that incorporate the pragmatic and discourse context of argumentative claims and show that they outperform models that rely only on claim-specific linguistic features for predicting the perceived impact of individual claims within a particular line of argument., Comment: EMNLP 2019

Systems for automatic argument generation and debate require the ability to (1) determine the stance of any claims employed in the argument and (2) assess the specificity of each claim relative to the argument context. Existing work on understanding claim specificity and stance, however, has been limited to the study of argumentative structures that are relatively shallow, most often consisting of a single claim that directly supports or opposes the argument thesis. In this paper, we tackle these tasks in the context of complex arguments on a diverse set of topics. In particular, our dataset consists of manually curated argument trees for 741 controversial topics covering 95,312 unique claims; lines of argument are generally of depth 2 to 6. We find that as the distance between a pair of claims increases along the argument path, determining the relative specificity of a pair of claims becomes easier and determining their relative stance becomes harder.

We incorporate an explicit neural interlingua into a multilingual encoder-decoder neural machine translation (NMT) architecture. We demonstrate that our model learns a language-independent representation by performing direct zero-shot translation (without using pivot translation), and by using the source sentence embeddings to create an English Yelp review classifier that, through the mediation of the neural interlingua, can also classify French and German reviews. Furthermore, we show that, despite using a smaller number of parameters than a pairwise collection of bilingual NMT models, our approach produces comparable BLEU scores for each language pair in WMT15., Comment: Accepted in WMT 18