United States

Reasoning about time and temporal relations is an integral aspect of human cognition, essential for perceiving the world and navigating our experiences. Though large language models (LLMs) have demonstrated impressive performance in many reasoning tasks, temporal reasoning remains challenging due to its intrinsic complexity. In this work, we first study an essential task of temporal reasoning -- temporal graph generation, to unveil LLMs&#39; inherent, global reasoning capabilities. We show that this task presents great challenges even for the most powerful LLMs, such as GPT-3.5/4. We also notice a significant performance gap by small models (&lt;10B) that lag behind LLMs by 50%. Next, we study how to close this gap with a budget constraint, e.g., not using model finetuning. We propose a new prompting technique tailored for temporal reasoning, Narrative-of-Thought (NoT), that first converts the events set to a Python class, then prompts a small model to generate a temporally grounded narrative, guiding the final generation of a temporal graph. Extensive experiments showcase the efficacy of NoT in improving various metrics. Notably, NoT attains the highest F1 on the Schema-11 evaluation set, while securing an overall F1 on par with GPT-3.5. NoT also achieves the best structural similarity across the board, even compared with GPT-3.5/4. Our code is available at https://github.com/launchnlp/NoT.

EMNLP 2024

Narrative-of-Thought: Improving Temporal Reasoning of Large Language Models via Recounted Narratives

temporal graph generation

small llms

narrative-of-thought

chain-of-thought

llms

temporal reasoning

large language models

## Welcome to EMNLP 2024! 
We are excited to welcome you to one of the most prominent conferences in the field of Natural Language Processing. This year, EMNLP 2024 is being held in a hybrid format,
offering both virtual and in-person participation in beautiful Miami. Due to a record-breaking number of submissions, we've expanded the total number of accepted papers to accommodate more cutting-edge research from around the globe.
### [Conference Handbook](https://drive.google.com/file/d/1WPROgxjLAC96AJL7Ugy0tEnYm7dkrbHt/view?usp=sharing)

You are required to register for this event. **Please register [here](https://2024.emnlp.org/registration/).** The EMNLP 2024 event page on Underline will be open to public one week prior to the event.

Please register!

EMNLP 2024 will take place in Miami, Florida from Nov 12th to Nov 16th, 2024, at the Hyatt Regency Miami Hote and on Underline for remote participants.

Reasoning about time and temporal relations is an integral aspect of human cognition, essential for perceiving the world and navigating our experiences. Though large language models (LLMs) have demonstrated impressive performance in many reasoning tasks, temporal reasoning remains challenging due to its intrinsic complexity. In this work, we first study an essential task of temporal reasoning -- temporal graph generation, to unveil LLMs' inherent, global reasoning capabilities. We show that this task presents great challenges even for the most powerful LLMs, such as GPT-3.5/4. We also notice a significant performance gap by small models (<10B) that lag behind LLMs by 50%. Next, we study how to close this gap with a budget constraint, e.g., not using model finetuning. We propose a new prompting technique tailored for temporal reasoning, Narrative-of-Thought (NoT), that first converts the events set to a Python class, then prompts a small model to generate a temporally grounded narrative, guiding the final generation of a temporal graph. Extensive experiments showcase the efficacy of NoT in improving various metrics. Notably, NoT attains the highest F1 on the Schema-11 evaluation set, while securing an overall F1 on par with GPT-3.5. NoT also achieves the best structural similarity across the board, even compared with GPT-3.5/4. Our code is available at https://github.com/launchnlp/NoT.

Existing research on instruction following largely focuses on tasks with simple instructions and short responses. In this work, we explore multi-constraint instruction following for generating long-form text. We create Suri, a dataset with 20K human-written long-form texts paired with LLM-generated backtranslated instructions that contain multiple complex constraints. Because of prohibitive challenges associated with collecting human preference judgments on long-form texts, preference-tuning algorithms such as DPO are infeasible in our setting; thus, we propose Instructional ORPO (I-ORPO), an alignment method based on the ORPO algorithm. Instead of receiving negative feedback from dispreferred responses, I-ORPO obtains negative feedback from synthetically corrupted instructions generated by an LLM. Using Suri, we perform supervised and I-ORPO fine-tuning on Mistral-7b-Instruct-v0.2. The resulting models, Suri-SFT and Suri-I-ORPO, generate significantly longer texts (~5K tokens) than base models without significant quality deterioration. Our human evaluation shows that while both SFT and I-ORPO models satisfy most constraints, Suri-I-ORPO generations are generally preferred for their coherent and informative incorporation of the constraints.

Suri: Multi-constraint Instruction Following for Long-form Text Generation

While much research has explored enhancing the reasoning capabilities of large language models (LLMs) in the last few years, there is a gap in understanding the alignment of these models with social values and norms. We introduce the task of judging social acceptance. Social acceptance requires models to judge and rationalize the acceptability of people's actions in social situations. For example, is it socially acceptable for a neighbor to ask others in the community to keep their pets indoors at night? We find that LLMs' understanding of social acceptance is often misaligned with human consensus. To alleviate this, we introduce SocialGaze, a multi-step prompting framework, in which a language model verbalizes a social situation from multiple perspectives before forming a judgment. Our experiments demonstrate that the SocialGaze approach improves the alignment with human judgments by up to 11 F1 points with the GPT-3.5 model. We also identify biases and correlations in LLMs in assigning blame that is related to features such as the gender (males are significantly more likely to be judged unfairly) and age (LLMs are more aligned with humans for older narrators).

SocialGaze: Improving the Integration of Human Social Norms in Large Language Models

Visual storytelling consists in generating a natural language story given a temporally ordered sequence of images. This task is not only challenging for models, but also very difficult to evaluate with automatic metrics since there is no consensus about what makes a story 'good'. In this paper, we introduce a novel method that measures story quality in terms of human likeness regarding three key aspects highlighted in previous work: visual grounding, coherence, and repetitiveness. We then use this method to evaluate the stories generated by several models, showing that the foundation model LLaVA obtains the best result, but only slightly so compared to TAPM, a 50-times smaller visual storytelling model. Upgrading the visual and language components of TAPM results in a model that yields competitive performance with a relatively low number of parameters. Finally, we carry out a human evaluation study, whose results suggest that a 'good' story may require more than a human-like level of visual grounding, coherence, and repetition.

Not (yet) the whole story: Evaluating Visual Storytelling Requires More than Measuring Coherence, Grounding, and Repetition

Lexical Simplification (LS) increases text accessibility and aids literacy development. LS automatically replaces difficult to read words for easier alternatives and is conducted at the word-level. LS outperforms large language models (LLMs) tasked with sentence-level simplification, known as Text Simplification (TS), when dealing with low-resourced languages. Several datasets exist for LS. These LS datasets specialize on one or two sub-tasks within the LS pipeline. However, as of this moment, no single LS dataset has been developed that covers all LS sub-tasks required to train a full LS system, ranging from (1). lexical complexity prediction (LCP), (2). substitute generation, to (3). substitute ranking. We present MultiLS, the first LS framework that allows for the full training and evaluation of an LS system. Model performances are reported, ranging from transformer-based models to more recent LLMs.

MultiLS: An End-to-End Lexical Simplification Framework

In this work, we propose EASSE-multi, a framework for easier automatic sentence evaluation for languages other than English. Compared to the original EASSE framework, EASSE-multi does not focus only on English. It contains tokenizers and versions of text simplification evaluation metrics which are suitable for multiple languages. In this paper, we exemplify the usage of EASSE-multi for German TS resulting in EASSE-DE.

EASSE-DE & EASSE-multi: Easier Automatic Sentence Simplification Evaluation for German & Multiple Languages

We provide a solution for a specific morphological obstacle which often makes Hebrew literature difficult to parse for the younger generation. The morphologically-rich nature of the Hebrew language allows pronominal direct objects to be realized as bound morphemes, suffixed to the verb. Although such suffixes are often utilized in Biblical Hebrew, their use has all but disappeared in modern Hebrew. Nevertheless, authors of modern Hebrew literature, in their search for literary flair, do make use of such forms. These unusual forms are notorious for alienating young readers from Hebrew literature, especially because these rare suffixed forms are often \emph{orthographically identical} to common Hebrew words with different meanings. Upon encountering such words, readers naturally select the usual analysis of the word; yet, upon completing the sentence, they find themselves confounded. Young readers end up feeling "tricked", and this in turn contributes to their alienation from the text. In order to address this challenge, we pretrained a new BERT model specifically geared to identify such forms, so that they may be automatically simplified and/or flagged. We release this new BERT model to the public for unrestricted use.

OtoBERT: Identifying Suffixed Verbal Forms in Modern Hebrew Literature

The prediction of lexical complexity in context is assuming an increasing relevance in Natural Language Processing research, since iden- tifying complex words is often the first step of text simplification pipelines. To the best of our knowledge, though, datasets annotated with complex words are available only for English and for a limited number of Western languages. In our paper, we introduce CompLex-ZH, a dataset including words annotated with complexity scores in sentential contexts for Chinese. Our data include sentences in Mandarin and Cantonese, which were selected from a variety of sources and textual genres. We provide a first evaluation with baselines combining hand- crafted and language models-based features.

CompLex-ZH: A New Dataset for Lexical Complexity Prediction in Mandarin and Cantonese

Explanatory images play a pivotal role in accessible and easy-to-read (E2R) texts. However, the images available in online databases are not tailored toward the respective texts, and the creation of customized images is expensive. In this large-scale study, we investigated whether text-to-image generation models can close this gap by providing customizable images quickly and easily. We benchmarked seven, four open- and three closed-source, image generation models and provide an extensive evaluation of the resulting images. In addition, we performed a user study with people from the E2R target group to examine whether the images met their requirements. We find that some of the models show remarkable performance, but none of the models are ready to be used at a larger scale without human supervision. Our research is an important step toward facilitating the creation of accessible information for E2R creators and tailoring accessible images to the target group's needs.

Images Speak Volumes: User-Centric Assessment of Image Generation for Accessible Communication

The success of Large Language Models (LLMs) in other domains has raised the question of whether LLMs can reliably assess and manipulate the readability of text. We approach this question empirically. First, using a published corpus of 4,724 English text excerpts, we find that readability estimates produced “zero- shot” from GPT-4 Turbo and GPT-4o mini exhibit relatively high correlation with human judgments (r = 0.76 and r = 0.74, respec- tively), outperforming estimates derived from traditional readability formulas and various psycholinguistic indices. Then, in a pre-registered human experiment (N = 59), we ask whether Turbo can reliably make text easier or harder to read. We find evidence to support this hypothesis, though considerable variance in human judgments remains unexplained. We conclude by discussing the limitations of this approach, including limited scope, as well as the validity of the “readability” construct and its depen- dence on context, audience, and goal.

Measuring and Modifying the Readability of English Texts with GPT-4

The most reliable and up-to-date information on health questions is in the biomedical literature, but inaccessible due to the complex language full of jargon. Domain specific scientific text simplification holds the promise to make this literature accessible to a lay audience. Therefore, we create Cochrane-auto: a large corpus of pairs of aligned sentences, paragraphs, and abstracts from biomedical abstracts and lay summaries. Experiments demonstrate that a plan-guided simplification system trained on Cochrane-auto is able to outperform a strong baseline trained on unaligned abstracts and lay summaries. More generally, our freely available corpus complementing Newsela-auto and Wiki-auto facilitates text simplification research beyond the sentence-level and direct lexical and grammatical revisions.

Next from EMNLP 2024

Suri: Multi-constraint Instruction Following for Long-form Text Generation

Similar lecture

Improving Temporal Reasoning of Language Models via Recounted Narratives