Hallucinations pose a significant challenge to the reliability of neural models for abstractive summarisation. While automatically generated summaries may be fluent, they often lack faithfulness to the original document. This issue becomes even more pronounced in low-resource languages, where summarisation requires cross-lingual transfer. With the existing faithful metrics focusing on English, even measuring the extent of this phenomenon in cross-lingual settings is hard. To address this, we first develop a novel metric, mFACT, evaluating the faithfulness of non-English summaries, leveraging translation-based transfer from multiple English faithfulness metrics. Through extensive experiments in multiple languages, we demonstrate that mFACT is best suited to detect hallucinations compared to alternative metrics. With mFACT, we assess a broad range of multilingual large language models, and find that they all tend to hallucinate often in languages different from English. We then propose a simple but effective method to reduce hallucinations in cross-lingual transfer, which weighs the loss of each training example by its faithfulness score. This method drastically increases both performance and faithfulness according to both automatic and human evaluation when compared to strong baselines for cross-lingual transfer such as MAD-X. Our code and dataset are available at https://github.com/yfqiu-nlp/mfact-summ.
论文检索
输入标题、作者或关键词，从 7,591 篇学术成果中精准定位
While Large Language Models (LLMs) have achieved remarkable performance in many tasks, much about their inner workings remains unclear. In this study, we present novel experimental insights into the resilience of LLMs, particularly GPT-4, when subjected to extensive character-level permutations. To investigate this, we first propose the Scrambled Bench, a suite designed to measure the capacity of LLMs to handle scrambled input, in terms of both recovering scrambled sentences and answering questions given scrambled context. The experimental results indicate that multiple advanced LLMs demonstrate the capability akin to typoglycemia, a phenomenon where humans can understand the meaning of words even when the letters within those words are scrambled, as long as the first and last letters remain in place. More surprisingly, we found that only GPT-4 nearly flawlessly processes inputs with unnatural errors, a task that poses significant challenges for other LLMs and often even for humans. Specifically, GPT-4 can almost perfectly reconstruct the original sentences from scrambled ones, decreasing the edit distance by 95%, even when all letters within each word are entirely scrambled. It is counter-intuitive that LLMs can exhibit such resilience despite severe disruption to input tokenization caused by scrambled text.
“Mistakes Help Us Grow”: Facilitating and Evaluating Growth Mindset Supportive Language in Classrooms
PDF ↗Teachers’ growth mindset supportive language (GMSL)—rhetoric emphasizing that one’s skills can be improved over time—has been shown to significantly reduce disparities in academic achievement and enhance students’ learning outcomes. Although teachers espouse growth mindset principles, most find it difficult to adopt GMSL in their practice due the lack of effective coaching in this area. We explore whether large language models (LLMs) can provide automated, personalized coaching to support teachers’ use of GMSL. We establish an effective coaching tool to reframe unsupportive utterances to GMSL by developing (i) a parallel dataset containing GMSL-trained teacher reframings of unsupportive statements with an accompanying annotation guide, (ii) a GMSL prompt framework to revise teachers’ unsupportive language, and (iii) an evaluation framework grounded in psychological theory for evaluating GMSL with the help of students and teachers. We conduct a large-scale evaluation involving 174 teachers and 1,006 students, finding that both teachers and students perceive GMSL-trained teacher and model reframings as more effective in fostering a growth mindset and promoting challenge-seeking behavior, among other benefits. We also find that model-generated reframings outperform those from the GMSL-trained teachers. These results show promise for harnessing LLMs to provide automated GMSL feedback for teachers and, more broadly, LLMs’ potentiality for supporting students’ learning in the classroom. Our findings also demonstrate the benefit of large-scale human evaluations when applying LLMs in educational domains.
Sharing knowledge between information extraction tasks has always been a challenge due to the diverse data formats and task variations. Meanwhile, this divergence leads to information waste and increases difficulties in building complex applications in real scenarios. Recent studies often formulate IE tasks as a triplet extraction problem. However, such a paradigm does not support multi-span and n-ary extraction, leading to weak versatility. To this end, we reorganize IE problems into unified multi-slot tuples and propose a universal framework for various IE tasks, namely Mirror. Specifically, we recast existing IE tasks as a multi-span cyclic graph extraction problem and devise a non-autoregressive graph decoding algorithm to extract all spans in a single step. It is worth noting that this graph structure is incredibly versatile, and it supports not only complex IE tasks, but also machine reading comprehension and classification tasks. We manually construct a corpus containing 57 datasets for model pretraining, and conduct experiments on 30 datasets across 8 downstream tasks. The experimental results demonstrate that our model has decent compatibility and outperforms or reaches competitive performance with SOTA systems under few-shot and zero-shot settings. The code, model weights, and pretraining corpus are available at https://github.com/Spico197/Mirror .
We propose Chain-of-Questions, a framework that trains a model to robustly answer multistep questions by generating and answering sub-questions. We obtain supervision for sub-questions from human-annotated question decomposition meaning representation (QDMR), but QDMR does not include annotated answers to sub-questions. To overcome this technical challenge, we treat sub-answers as latent variables and infer them with a novel dynamic mixture of Hard-EM and MAPO. Chain-of-Questions is effective and robust, greatly outperforming strong neuro-symbolic methods by 9.0 F1 on a DROP contrast set and GPT-3.5 by 24.3 F1 on a HotpotQA adversarial set.
Fine-tuning pretrained language models on helpful intermediate tasks often greatly improves the performance of target tasks. However, how to efficiently find the source tasks that can successfully transfer still remains under-explored. In this work, we propose to learn an affinity scoring function to predict transferability between tasks. Specifically, we conduct prompt tuning and regard soft prompts as task embeddings that summarize task-specific information. Then we randomly sample task pairs to train an affinity scoring function. The goal is to predict the transfer gain (i.e., affinity) between a task pair, by conditioning on their task embeddings. Once the scoring function is trained, given a novel target task, we use it to predict the most transferable source tasks, without a brute-force search for all possible source-target pairs. Experimental results across 50 tasks show that our method efficiently identifies beneficial tasks for transfer learning.
To address the data scarcity issue in Conversational question answering (ConvQA), a dialog inpainting method, which utilizes documents to generate ConvQA datasets, has been proposed. However, the original dialog inpainting model is trained solely on the dialog reconstruction task, resulting in the generation of questions with low contextual relevance due to insufficient learning of question-answer alignment. To overcome this limitation, we propose a novel framework called Dialogizer, which has the capability to automatically generate ConvQA datasets with high contextual relevance from textual sources. The framework incorporates two training tasks: question-answer matching (QAM) and topic-aware dialog generation (TDG). Moreover, re-ranking is conducted during the inference phase based on the contextual relevance of the generated questions. Using our framework, we produce four ConvQA datasets by utilizing documents from multiple domains as the primary source. Through automatic evaluation using diverse metrics, as well as human evaluation, we validate that our proposed framework exhibits the ability to generate datasets of higher quality compared to the baseline dialog inpainting model.
This work focuses on the spatial dimension of narrative understanding and presents the task of event-location tracking in narrative texts. The task intends to extract the sequence of locations where the narrative is set through its progression. We present several architectures for the task that seeks to model the global structure of the sequence, with varying levels of context awareness. We compare these methods to several baselines, including the use of strong methods applied over narrow contexts. We also develop methods for the generation of location embeddings and show that learning to predict a sequence of continuous embeddings, rather than a string of locations, is advantageous in terms of performance. We focus on the test case of Holocaust survivor testimonies. We argue for the moral and historical importance of studying this dataset in computational means and that it provides a unique case of a large set of narratives with a relatively restricted set of location trajectories. Our results show that models that are aware of the larger context of the narrative can generate more accurate location chains. We further corroborate the effectiveness of our methods by showing similar trends from experiments on an additional domain.
Evaluation Metrics in the Era of GPT-4: Reliably Evaluating Large Language Models on Sequence to Sequence Tasks
PDF ↗Large Language Models (LLMs) evaluation is a patchy and inconsistent landscape, and it is becoming clear that the quality of automatic evaluation metrics is not keeping up with the pace of development of generative models. We aim to improve the understanding of current models’ performance by providing a preliminary and hybrid evaluation on a range of open and closed-source generative LLMs on three NLP benchmarks: text summarisation, text simplification and grammatical error correction (GEC), using both automatic and human evaluation. We also explore the potential of the recently released GPT-4 to act as an evaluator. We find that ChatGPT consistently outperforms many other popular models according to human reviewers on the majority of metrics, while scoring much more poorly when using classic automatic evaluation metrics. We also find that human reviewers rate the gold reference as much worse than the best models’ outputs, indicating the poor quality of many popular benchmarks. Finally, we find that GPT-4 is capable of ranking models’ outputs in a way which aligns reasonably closely to human judgement despite task-specific variations, with a lower alignment in the GEC task.
Are Embedded Potatoes Still Vegetables? On the Limitations of WordNet Embeddings for Lexical Semantics
PDF ↗Knowledge Base Embedding (KBE) models have been widely used to encode structured information from knowledge bases, including WordNet. However, the existing literature has predominantly focused on link prediction as the evaluation task, often neglecting exploration of the models’ semantic capabilities. In this paper, we investigate the potential disconnect between the performance of KBE models of WordNet on link prediction and their ability to encode semantic information, highlighting the limitations of current evaluation protocols. Our findings reveal that some top-performing KBE models on the WN18RR benchmark exhibit subpar results on two semantic tasks and two downstream tasks. These results demonstrate the inadequacy of link prediction benchmarks for evaluating the semantic capabilities of KBE models, suggesting the need for a more targeted assessment approach.
Evidently, words can have multiple senses. For example, the word mess refers to a place to have food or to a confusing situation. How exactly multiple senses emerge is less clear. In this work, we propose and analyze a mathematical model of the evolution of lexical meaning to investigate mechanisms leading to polysemy. This model features factors that have been discussed to impact the semantic processing and transmission of words: word frequency, non-conformism, and semantic discriminability. We formally derive conditions under which a sense of a word tends to diversify itself into multiple senses that coexist stably. The model predicts that diversification is promoted by low frequency, a strong bias for non-conformist usage, and high semantic discriminability. We statistically validate these predictions with historical language data covering semantic developments of a set of English words. Multiple alternative measures are used to operationalize each variable involved, and we confirm the predicted tendencies for twelve combinations of measures.
MedEval: A Multi-Level, Multi-Task, and Multi-Domain Medical Benchmark for Language Model Evaluation
PDF ↗Curated datasets for healthcare are often limited due to the need of human annotations from experts. In this paper, we present MedEval, a multi-level, multi-task, and multi-domain medical benchmark to facilitate the development of language models for healthcare. MedEval is comprehensive and consists of data from several healthcare systems and spans 35 human body regions from 8 examination modalities. With 22,779 collected sentences and 21,228 reports, we provide expert annotations at multiple levels, offering a granular potential usage of the data and supporting a wide range of tasks. Moreover, we systematically evaluated 10 generic and domain-specific language models under zero-shot and finetuning settings, from domain-adapted baselines in healthcare to general-purposed state-of-the-art large language models (e.g., ChatGPT). Our evaluations reveal varying effectiveness of the two categories of language models across different tasks, from which we notice the importance of instruction tuning for few-shot usage of large language models. Our investigation paves the way toward benchmarking language models for healthcare and provides valuable insights into the strengths and limitations of adopting large language models in medical domains, informing their practical applications and future advancements.
The scientific innovation in Natural Language Processing (NLP) and more broadly in artificial intelligence (AI) is at its fastest pace to date. As large language models (LLMs) unleash a new era of automation, important debates emerge regarding the benefits and risks of their development, deployment and use. Currently, these debates have been dominated by often polarized narratives mainly led by the AI Safety and AI Ethics movements. This polarization, often amplified by social media, is swaying political agendas on AI regulation and governance and posing issues of regulatory capture. Capture occurs when the regulator advances the interests of the industry it is supposed to regulate, or of special interest groups rather than pursuing the general public interest. Meanwhile in NLP research, attention has been increasingly paid to the discussion of regulating risks and harms. This often happens without systematic methodologies or sufficient rooting in the disciplines that inspire an extended scope of NLP research, jeopardizing the scientific integrity of these endeavors. Regulation studies are a rich source of knowledge on how to systematically deal with risk and uncertainty, as well as with scientific evidence, to evaluate and compare regulatory options. This resource has largely remained untapped so far. In this paper, we argue how NLP research on these topics can benefit from proximity to regulatory studies and adjacent fields. We do so by discussing basic tenets of regulation, and risk and uncertainty, and by highlighting the shortcomings of current NLP discussions dealing with risk assessment. Finally, we advocate for the development of a new multidisciplinary research space on regulation and NLP (RegNLP), focused on connecting scientific knowledge to regulatory processes based on systematic methodologies.
Sentiment analysis is a well-established natural language processing task, with sentiment polarity classification being one of its most popular and representative tasks. However, despite the success of pre-trained language models in this area, they often fall short of capturing the broader complexities of sentiment analysis. To address this issue, we propose a new task called Sentiment and Opinion Understanding of Language (SOUL). SOUL aims to evaluate sentiment understanding through two subtasks: Review Comprehension (RC) and Justification Generation (JG). RC seeks to validate statements that focus on subjective information based on a review text, while JG requires models to provide explanations for their sentiment predictions. To enable comprehensive evaluation, we annotate a new dataset comprising 15,028 statements from 3,638 reviews. Experimental results indicate that SOUL is a challenging task for both small and large language models, with a performance gap of up to 27% when compared to human performance. Furthermore, evaluations conducted with both human experts and GPT-4 highlight the limitations of the small language model in generating reasoning-based justifications. These findings underscore the challenging nature of the SOUL task for existing models, emphasizing the need for further advancements in sentiment analysis to address its complexities. The new dataset and code are available at https://github.com/DAMO-NLP-SG/SOUL.
Structure-aware Knowledge Graph-to-text Generation with Planning Selection and Similarity Distinction
PDF ↗The knowledge graph-to-text (KG-to-text) generation task aims to synthesize coherent and engaging sentences that accurately convey the complex information derived from an input knowledge graph. One of the primary challenges in this task is bridging the gap between the diverse structures of the KG and the target text, while preserving the details of the input KG. To address this, we propose a novel approach that efficiently integrates graph structure-aware modules with pre-trained language models. Unlike conventional techniques, which only consider direct connections between first-order neighbors, our method delves deeper by incorporating Relative Distance Encoding as a bias within the graph structure-aware module. This enables our model to better capture the intricate topology information present in the KG. To further elevate the fidelity of the generated text, Planning Selection and Similarity Distinction are introduced. Our approach filters the most relevant linearized sequences by employing a planning scorer, while simultaneously distinguishing similar input KGs through contrastive learning techniques. Experiments on two datasets demonstrate the superiority of our model.
Not all quantifiers are equal: Probing Transformer-based language models’ understanding of generalised quantifiers
PDF ↗How do different generalised quantifiers affect the behaviour of transformer-based language models (TLMs)? The recent popularity of TLMs and the central role generalised quantifiers have traditionally played in linguistics and logic bring this question into particular focus. The current research investigating this subject has not utilised a task defined purely in a logical sense, and thus, has not captured the underlying logical significance of generalised quantifiers. Consequently, they have not answered the aforementioned question faithfully or adequately. Therefore, we investigate how different generalised quantifiers affect TLMs by employing a textual entailment problem defined in a purely logical sense, namely, model-checking with natural language. Our approach permits the automatic construction of datasets with respect to which we can assess the ability of TLMs to learn the meanings of generalised quantifiers. Our investigation reveals that TLMs generally can comprehend the logical semantics of the most common generalised quantifiers, but that distinct quantifiers influence TLMs in varying ways.
In this paper we consider the task of conversational semantic parsing over general purpose knowledge graphs (KGs) with millions of entities, and thousands of relation-types. We focus on models which are capable of interactively mapping user utterances into executable logical forms (e.g., Sparql) in the context of the conversational history. Our key idea is to represent information about an utterance and its context via a subgraph which is created dynamically, i.e., the number of nodes varies per utterance. Rather than treating the subgraph as a sequence, we exploit its underlying structure and encode it with a graph neural network which further allows us to represent a large number of (unseen) nodes. Experimental results show that dynamic context modeling is superior to static approaches, delivering performance improvements across the board (i.e., for simple and complex questions). Our results further confirm that modeling the structure of context is better at processing discourse information, (i.e., at handling ellipsis and resolving coreference) and longer interactions.
Topic Modelling is an established research area where the quality of a given topic is measured using coherence metrics. Often, we infer topics from Neural Topic Models (NTM) by interpreting their decoder weights, consisting of top-activated words projected from individual neurons. Transformer-based Language Models (TLM) similarly consist of decoder weights. However, due to its hypothesised superposition properties, the final logits originating from the residual path are considered uninterpretable. Therefore, we posit that we can interpret TLM as superposed NTM by proposing a novel weight-based, model-agnostic and corpus-agnostic approach to search and disentangle decoder-only TLM, potentially mapping individual neurons to multiple coherent topics. Our results show that it is empirically feasible to disentangle coherent topics from GPT-2 models using the Wikipedia corpus. We validate this approach for GPT-2 models using Zero-Shot Topic Modelling. Finally, we extend the proposed approach to disentangle and analyse LLaMA models.
The CoNLL-03 corpus is arguably the most well-known and utilized benchmark dataset for named entity recognition (NER). However, prior works found significant numbers of annotation errors, incompleteness, and inconsistencies in the data. This poses challenges to objectively comparing NER approaches and analyzing their errors, as current state-of-the-art models achieve F1-scores that are comparable to or even exceed the estimated noise level in CoNLL-03. To address this issue, we present a comprehensive relabeling effort assisted by automatic consistency checking that corrects 7.0% of all labels in the English CoNLL-03. Our effort adds a layer of entity linking annotation both for better explainability of NER labels and as additional safeguard of annotation quality. Our experimental evaluation finds not only that state-of-the-art approaches reach significantly higher F1-scores (97.1%) on our data, but crucially that the share of correct predictions falsely counted as errors due to annotation noise drops from 47% to 6%. This indicates that our resource is well suited to analyze the remaining errors made by state-of-the-art models, and that the theoretical upper bound even on high resource, coarse-grained NER is not yet reached. To facilitate such analysis, we make CleanCoNLL publicly available to the research community.
MT2: Towards a Multi-Task Machine Translation Model with Translation-Specific In-Context Learning
PDF ↗Sentence-level translation, document-level translation, translation memory, and terminology constrained translation play an important role in machine translation. Most of the previous work uses separate models or methods to solve these tasks, which is not conducive to knowledge transfer of different tasks and increases the complexity of system construction. In this work, we explore the potential of pre-trained language model in machine translation tasks and propose a Multi-Task Machine Translation (MT2) model to integrate these translation tasks. We design a novel translation-specific In-Context Learning (ICL) paradigm for model training, in which all of the translation tasks can be modeled as context-learning tasks that integrate contextual information for performance improvement. Specifically, we propose a retrieval and alignment method to obtain a large scale context-enhancement training data, then we train the model in an in-context learning manner. Furthermore, we adopt two context-dependent training strategies to encourage the model to better understand and utilize contextual information for translation. Extensive experiments on translation memory, terminology constrained translation, document-level translation, and few-shot domain-adaptation tasks demonstrate the superior performance of our model, verifying the effectiveness of our proposed approach.