Contextualized embeddings based on large language models (LLMs) are available for various languages, but their coverage is often limited for lower resourced languages. Using LLMs for such languages is often difficult due to a high computational cost; not only during training, but also during inference. Static word embeddings are much more resource-efficient ("green"), and thus still provide value, particularly for very low-resource languages. There is, however, a notable lack of comprehensive repositories with such embeddings for diverse languages. To address this gap, we present GrEmLIn, a centralized repository of green, static baseline embeddings for 87 mid- and low-resource languages. We compute GrEmLIn embeddings with a novel method that enhances GloVe embeddings by integrating multilingual graph knowledge, which makes our static embeddings competitive with LLM representations, while being parameter-free at inference time. Our experiments demonstrate that GrEmLIn embeddings outperform state-of-the-art contextualized embeddings from E5 on the task of lexical similarity. They remain competitive in extrinsic evaluation tasks like sentiment analysis and natural language inference, with average performance gaps of just 5-10\% or less compared to state-of-the-art models, given a sufficient vocabulary overlap with the target task, and underperform only on topic classification. Our code and embeddings are publicly available at https://huggingface.co/DFKI.
Large Language Models (LLMs) are widely used in both industry and academia for various tasks, yet evaluating the consistency of generated text responses continues to be a challenge. Traditional metrics like ROUGE and BLEU show a weak correlation with human judgment. More sophisticated metrics using Natural Language Inference (NLI) have shown improved correlations but are complex to implement, require domain-specific training due to poor cross-domain generalization, and lack explainability. More recently, prompt-based metrics using LLMs as evaluators have emerged; while they are easier to implement, they still lack explainability and depend on task-specific prompts, which limits their generalizability. This work introduces Automated eXplainable Consistency Evaluation using LLMs (AXCEL), a prompt-based consistency metric which offers explanations for the consistency scores by providing detailed reasoning and pinpointing inconsistent text spans. AXCEL is also a generalizable metric which can be adopted to multiple tasks without changing the prompt. AXCEL outperforms both non-prompt and prompt-based state-of-the-art (SOTA) metrics in detecting inconsistencies across summarization by 8.7%, free text generation by 6.2%, and data-to-text conversion tasks by 29.4%. We also evaluate the influence of underlying LLMs on prompt based metric performance and recalibrate the SOTA prompt-based metrics with the latest LLMs for fair comparison. Further, we show that AXCEL demonstrates strong performance using open source LLMs.
Cutting-edge abstractive summarisers generate fluent summaries, but the factuality of the generated text is not guaranteed. Early summary factuality evaluation metrics are usually based on n-gram overlap and embedding similarity, but are reported fail to align with human annotations. Therefore, many techniques for detecting factual inconsistencies build pipelines around natural language inference (NLI) or question-answering (QA) models with additional supervised learning steps. In this paper, we revisit similarity-based metrics, showing that this failure stems from the comparison text selection and its granularity. We propose a new zero-shot factuality evaluation metric, Sentence-BERT Score (SBERTScore), which compares sentences between the summary and the source document. It outperforms widely-used word-word metrics including BERTScore and can compete with existing NLI and QA-based factuality metrics on the benchmark without needing any fine-tuning. Our experiments indicate that each technique has different strengths, with SBERTScore particularly effective in identifying correct summaries. We demonstrate how a combination of techniques is more effective in detecting various types of error.
This study examines the effectiveness of layer pruning in creating efficient Sentence BERT (SBERT) models. Our goal is to create smaller sentence embedding models that reduce complexity while maintaining strong embedding similarity. We assess BERT models like Muril and MahaBERT-v2 before and after pruning, comparing them with smaller, scratch-trained models like MahaBERT-Small and MahaBERT-Smaller. Through a two-phase SBERT fine-tuning process involving Natural Language Inference (NLI) and Semantic Textual Similarity (STS), we evaluate the impact of layer reduction on embedding quality. Our findings show that pruned models, despite fewer layers, perform competitively with fully layered versions. Moreover, pruned models consistently outperform similarly sized, scratch-trained models, establishing layer pruning as an effective strategy for creating smaller, efficient embedding models. These results highlight layer pruning as a practical approach for reducing computational demand while preserving high-quality embeddings, making SBERT models more accessible for languages with limited technological resources.
This paper explores the enhancement of small language models through strategic dataset augmentation via ChatGPT-3.5-Turbo, in the domain of Natural Language Inference (NLI). By employing knowledge distillation-based techniques and synthetic dataset augmentation, we aim to bridge the performance gap between large language models (LLMs) and small language models (SLMs) without the immense cost of human annotation. Our methods involve two forms of rationale generation--information extraction and informed reasoning--to enrich the ANLI dataset. We then fine-tune T5-Small on these augmented datasets, evaluating its performance against an established benchmark. Our findings reveal that the incorporation of synthetic rationales significantly improves the model's ability to comprehend natural language, leading to 1.3\% and 2.3\% higher classification accuracy, respectively, on the ANLI dataset, demonstrating the potential of leveraging LLMs for dataset augmentation. This approach not only enhances the performance of smaller models on complex tasks but also introduces a cost-effective method for fine-tuning smaller language models. By advancing our understanding of knowledge distillation and fine-tuning strategies, this work contributes to the ongoing effort to create more capable and efficient NLP systems.
The surge of state-of-the-art Transformer-based models has undoubtedly pushed the limits of NLP model performance, excelling in a variety of tasks. We cast the spotlight on the underexplored task of Natural Language Inference (NLI), since models trained on popular well-suited datasets are susceptible to adversarial attacks, allowing subtle input interventions to mislead the model. In this work, we validate the usage of natural language explanation as a model-agnostic defence strategy through extensive experimentation: only by fine-tuning a classifier on the explanation rather than premise-hypothesis inputs, robustness under various adversarial attacks is achieved in comparison to explanation-free baselines. Moreover, since there is no standard strategy of testing the semantic validity of the generated explanations, we research the correlation of widely used language generation metrics with human perception, in order for them to serve as a proxy towards robust NLI models. Our approach is resource-efficient and reproducible without significant computational limitations.
Large Deep Learning models are compressed and deployed for specific applications. However, current Deep Learning model compression methods do not utilize the information about the target application. As a result, the compressed models are application agnostic. Our goal is to customize the model compression process to create a compressed model that will perform better for the target application. Our method, Application Specific Compression (ASC), identifies and prunes components of the large Deep Learning model that are redundant specifically for the given target application. The intuition of our work is to prune the parts of the network that do not contribute significantly to updating the data representation for the given application. We have experimented with the BERT family of models for three applications: Extractive QA, Natural Language Inference, and Paraphrase Identification. We observe that customized compressed models created using ASC method perform better than existing model compression methods and off-the-shelf compressed models.
This paper examines the Code-Switching (CS) phenomenon where two languages intertwine within a single utterance. There exists a noticeable need for research on the CS between English and Korean. We highlight that the current Equivalence Constraint (EC) theory for CS in other languages may only partially capture English-Korean CS complexities due to the intrinsic grammatical differences between the languages. We introduce a novel Koglish dataset tailored for English-Korean CS scenarios to mitigate such challenges. First, we constructed the Koglish-GLUE dataset to demonstrate the importance and need for CS datasets in various tasks. We found the differential outcomes of various foundation multilingual language models when trained on a monolingual versus a CS dataset. Motivated by this, we hypothesized that SimCSE, which has shown strengths in monolingual sentence embedding, would have limitations in CS scenarios. We construct a novel Koglish-NLI (Natural Language Inference) dataset using a CS augmentation-based approach to verify this. From this CS-augmented dataset Koglish-NLI, we propose a unified contrastive learning and augmentation method for code-switched embeddings, ConCSE, highlighting the semantics of CS sentences. Experimental results validate the proposed ConCSE with an average performance enhancement of 1.77\% on the Koglish-STS(Semantic Textual Similarity) tasks.
Pre-training and self-training are two approaches to semi-supervised learning. The comparison between pre-training and self-training has been explored. However, the previous works led to confusing findings: self-training outperforms pre-training experienced on some tasks in computer vision, and contrarily, pre-training outperforms self-training experienced on some tasks in natural language processing, under certain conditions of incomparable settings. We propose, comparatively and exhaustively, an ensemble method to empirical study all feasible training paradigms combining pre-training, self-training, and fine-tuning within consistent foundational settings comparable to data augmentation. We conduct experiments on six datasets, four data augmentation, and imbalanced data for sentiment analysis and natural language inference tasks. Our findings confirm that the pre-training and fine-tuning paradigm yields the best overall performances. Moreover, self-training offers no additional benefits when combined with semi-supervised pre-training.
Social scientists quickly adopted large language models due to their ability to annotate documents without supervised training, an ability known as zero-shot learning. However, due to their compute demands, cost, and often proprietary nature, these models are often at odds with replication and open science standards. This paper introduces the Political DEBATE (DeBERTa Algorithm for Textual Entailment) language models for zero-shot and few-shot classification of political documents. These models are not only as good, or better than, state-of-the art large language models at zero and few-shot classification, but are orders of magnitude more efficient and completely open source. By training the models on a simple random sample of 10-25 documents, they can outperform supervised classifiers trained on hundreds or thousands of documents and state-of-the-art generative models with complex, engineered prompts. Additionally, we release the PolNLI dataset used to train these models -- a corpus of over 200,000 political documents with highly accurate labels across over 800 classification tasks.
Subjective tasks in NLP have been mostly relegated to objective standards, where the gold label is decided by taking the majority vote. This obfuscates annotator disagreement and the inherent uncertainty of the label. We argue that subjectivity should factor into model decisions and play a direct role via calibration under a selective prediction setting. Specifically, instead of calibrating confidence purely from the model's perspective, we calibrate models for subjective tasks based on crowd worker agreement. Our method, Crowd-Calibrator, models the distance between the distribution of crowd worker labels and the model's own distribution over labels to inform whether the model should abstain from a decision. On two highly subjective tasks, hate speech detection and natural language inference, our experiments show Crowd-Calibrator either outperforms or achieves competitive performance with existing selective prediction baselines. Our findings highlight the value of bringing human decision-making into model predictions.
This study demonstrates the application of instruction finetuning of pretrained Large Language Models (LLMs) to automate the generation of AI research leaderboards, extracting (Task, Dataset, Metric, Score) quadruples from articles. It aims to streamline the dissemination of advancements in AI research by transitioning from traditional, manual community curation, or otherwise taxonomy-constrained natural language inference (NLI) models, to an automated, generative LLM-based approach. Utilizing the FLAN-T5 model, this research enhances LLMs' adaptability and reliability in information extraction, offering a novel method for structured knowledge representation.
Large language models (LLMs) have demonstrated impressive performance in mathematical and commonsense reasoning tasks using chain-of-thought (CoT) prompting techniques. But can they perform emotional reasoning by concatenating `Let's think step-by-step' to the input prompt? In this paper we investigate this question along with introducing a novel approach to zero-shot emotion detection and emotional reasoning using LLMs. Existing state of the art zero-shot approaches rely on textual entailment models to choose the most appropriate emotion label for an input text. We argue that this strongly restricts the model to a fixed set of labels which may not be suitable or sufficient for many applications where emotion analysis is required. Instead, we propose framing the problem of emotion analysis as a generative question-answering (QA) task. Our approach uses a two step methodology of generating relevant context or background knowledge to answer the emotion detection question step-by-step. Our paper is the first work on using a generative approach to jointly address the tasks of emotion detection and emotional reasoning for texts. We evaluate our approach on two popular emotion detection datasets and also release the fine-grained emotion labels and explanations for further training and fine-tuning of emotional reasoning systems.
Detecting semantic arguments of a predicate word has been conventionally modeled as a sentence-level task. The typical reader, however, perfectly interprets predicate-argument relations in a much wider context than just the sentence where the predicate was evoked. In this work, we reformulate the problem of argument detection through textual entailment to capture semantic relations across sentence boundaries. We propose a method that tests whether some semantic relation can be inferred from a full passage by first encoding it into a simple and standalone proposition and then testing for entailment against the passage. Our method does not require direct supervision, which is generally absent due to dataset scarcity, but instead builds on existing NLI and sentence-level SRL resources. Such a method can potentially explicate pragmatically understood relations into a set of explicit sentences. We demonstrate it on a recent document-level benchmark, outperforming some supervised methods and contemporary language models.
This work addresses the challenge of factual consistency in text generation systems. We unify the tasks of Natural Language Inference, Summarization Evaluation, Factuality Verification and Factual Consistency Evaluation to train models capable of evaluating the factual consistency of source-target pairs across diverse domains. We rigorously evaluate these against eight baselines on a comprehensive benchmark suite comprising 22 datasets that span various tasks, domains, and document lengths. Results demonstrate that our method achieves state-of-the-art performance on this heterogeneous benchmark while addressing efficiency concerns and attaining cross-domain generalization.
The expanding applications of legged robots require their mastery of versatile motion skills. Correspondingly, researchers must address the challenge of integrating multiple diverse motion skills into controllers. While existing reinforcement learning (RL)-based approaches have achieved notable success in multi-skill integration for legged robots, these methods often require intricate reward engineering or are restricted to integrating a predefined set of motion skills constrained by specific task objectives, resulting in limited flexibility. In this work, we introduce a flexible multi-skill integration framework named Controllable Skills Integration (CSI). CSI enables the integration of a diverse set of motion skills with varying styles into a single policy without the need for complex reward tuning. Furthermore, in a hierarchical control manner, the trained low-level policy can be coupled with a high-level Natural Language Inference (NLI) module to enable preliminary language-directed skill control. Our experiments demonstrate that CSI can flexibly integrate a diverse array of motion skills more comprehensively and facilitate the transitions between different skills. Additionally, CSI exhibits good scalability as the number of motion skills to be integrated increases significantly.
This paper describes our approach to the SemEval-2024 safe biomedical Natural Language Inference for Clinical Trials (NLI4CT) task, which concerns classifying statements about Clinical Trial Reports (CTRs). We explored the capabilities of Mistral-7B, a generalist open-source Large Language Model (LLM). We developed a prompt for the NLI4CT task, and fine-tuned a quantized version of the model using an augmented version of the training dataset. The experimental results show that this approach can produce notable results in terms of the macro F1-score, while having limitations in terms of faithfulness and consistency. All the developed code is publicly available on a GitHub repository
Large language models (LLMs) have garnered significant interest in natural language processing (NLP), particularly their remarkable performance in various downstream tasks in resource-rich languages. Recent studies have highlighted the limitations of LLMs in low-resource languages, primarily focusing on binary classification tasks and giving minimal attention to South Asian languages. These limitations are primarily attributed to constraints such as dataset scarcity, computational costs, and research gaps specific to low-resource languages. To address this gap, we present datasets for sentiment and hate speech tasks by translating from English to Bangla, Hindi, and Urdu, facilitating research in low-resource language processing. Further, we comprehensively examine zero-shot learning using multiple LLMs in English and widely spoken South Asian languages. Our findings indicate that GPT-4 consistently outperforms Llama 2 and Gemini, with English consistently demonstrating superior performance across diverse tasks compared to low-resource languages. Furthermore, our analysis reveals that natural language inference (NLI) exhibits the highest performance among the evaluated tasks, with GPT-4 demonstrating superior capabilities.
Despite their impressive performance, large language models (LLMs) such as ChatGPT are known to pose important risks. One such set of risks arises from misplaced confidence, whether over-confidence or under-confidence, that the models have in their inference. While the former is well studied, the latter is not, leading to an asymmetry in understanding the comprehensive risk of the model based on misplaced confidence. In this paper, we address this asymmetry by defining two types of risk (decision and composite risk), and proposing an experimental framework consisting of a two-level inference architecture and appropriate metrics for measuring such risks in both discriminative and generative LLMs. The first level relies on a decision rule that determines whether the underlying language model should abstain from inference. The second level (which applies if the model does not abstain) is the model's inference. Detailed experiments on four natural language commonsense reasoning datasets using both an open-source ensemble-based RoBERTa model and ChatGPT, demonstrate the practical utility of the evaluation framework. For example, our results show that our framework can get an LLM to confidently respond to an extra 20.1% of low-risk inference tasks that other methods might misclassify as high-risk, and skip 19.8% of high-risk tasks, which would have been answered incorrectly.
Synthetically created Cross-Lingual Summarisation (CLS) datasets are prone to include document-summary pairs where the reference summary is unfaithful to the corresponding document as it contains content not supported by the document (i.e., hallucinated content). This low data quality misleads model learning and obscures evaluation results. Automatic ways to assess hallucinations and improve training have been proposed for monolingual summarisation, predominantly in English. For CLS, we propose to use off-the-shelf cross-lingual Natural Language Inference (X-NLI) to evaluate faithfulness of reference and model generated summaries. Then, we study training approaches that are aware of faithfulness issues in the training data and propose an approach that uses unlikelihood loss to teach a model about unfaithful summary sequences. Our results show that it is possible to train CLS models that yield more faithful summaries while maintaining comparable or better informativess.