Large language models (LLMs) have significantly advanced natural language processing, but their progress has yet to be equal across languages. While most LLMs are trained in high-resource languages like English, multilingual models generally underperform monolingual ones. Additionally, aspects of their multilingual foundation sometimes restrict the byproducts they produce, like computational demands and licensing regimes. In this study, we document the development of open-foundation models tailored for use in low-resource settings, their limitations, and their benefits. This is the TeenyTinyLlama pair: two compact models for Brazilian Portuguese text generation. We release them under the permissive Apache 2.0 license on GitHub and Hugging Face for community use and further development.

Fine-tuning large pre-trained language models (LLMs) on particular datasets is a commonly employed strategy in Natural Language Processing (NLP) classification tasks. However, this approach usually results in a loss of models generalizability. In this paper, we present a framework that allows for maintaining generalizability, and enhances the performance on the downstream task by utilizing task-specific context attribution. We show that a linear transformation of the text representation from any transformer model using the task-specific concept operator results in a projection onto the latent concept space, referred to as context attribution in this paper. The specific concept operator is optimized during the supervised learning stage via novel loss functions. The proposed framework demonstrates that context attribution of the text representation for each task objective can improve the capacity of the discriminator function and thus achieve better performance for the classification task. Experimental results on three datasets, namely HateXplain, IMDB reviews, and Social Media Attributions, illustrate that the proposed model attains superior accuracy and generalizability. Specifically, for the non-fine-tuned BERT on the HateXplain dataset, we observe 8% improvement in accuracy and 10% improvement in F1-score. Whereas for the IMDB dataset, fine-tuned state-of-the-art XLNet is outperformed by 1% for both accuracy and F1-score. Furthermore, in an out-of-domain cross-dataset test, DistilBERT fine-tuned on the IMDB dataset in conjunction with the proposed model improves the F1-score on the HateXplain dataset by 7%. For the Social Media Attributions dataset of YouTube comments, we observe 5.2% increase in F1-metric. The proposed framework is implemented with PyTorch and provided open-source on GitHub.

Translation into severely low-resource languages has both the cultural goal of saving and reviving those languages and the humanitarian goal of assisting the everyday needs of local communities that are accelerated by the recent COVID-19 pandemic. In many humanitarian efforts, translation into severely low-resource languages often does not require a universal translation engine, but a dedicated text-specific translation engine. For example, healthcare records, hygienic procedures, government communication, emergency procedures and religious texts are all limited texts. While generic translation engines for all languages do not exist, translation of multilingually known limited texts into new, low-resource languages may be possible and reduce human translation effort. We attempt to leverage translation resources from rich-resource languages to efficiently produce best possible translation quality for well known texts, which are available in multiple languages, in a new, low-resource language. To reach this goal, we argue that in translating a closed text into low-resource languages, generalization to out-of-domain texts is not necessary, but generalization to new languages is. Performance gain comes from massive source parallelism by careful choice of close-by language families, style-consistent corpus-level paraphrases within the same language and strategic adaptation of existing large pretrained multilingual models to the domain first and then to the language. Such performance gain makes it possible for machine translation systems to collaborate with human translators to expedite the translation process into new, low-resource languages.

The intricate relationship between human decision-making and emotions, particularly guilt and regret, has significant implications on behavior and well-being. Yet, these emotions subtle distinctions and interplay are often overlooked in computational models. This paper introduces a dataset tailored to dissect the relationship between guilt and regret and their unique textual markers, filling a notable gap in affective computing research. Our approach treats guilt and regret recognition as a binary classification task and employs three machine learning and six transformer-based deep learning techniques to benchmark the newly created dataset. The study further implements innovative reasoning methods like chain-of-thought and tree-of-thought to assess the models interpretive logic. The results indicate a clear performance edge for transformer-based models, achieving a 90.4% macro F1 score compared to the 85.3% scored by the best machine learning classifier, demonstrating their superior capability in distinguishing complex emotional states.

A multiagent system can be viewed as a society of autonomous agents, whose interactions can be effectively regulated via social norms. In general, the norms of a society are not hardcoded but emerge from the agents' interactions. Specifically, how the agents in a society react to each other's behavior and respond to the reactions of others determines which norms emerge in the society. We think of these reactions by an agent to the satisfactory or unsatisfactory behaviors of another agent as communications from the first agent to the second agent. Understanding these communications is a kind of social intelligence: these communications provide natural drivers for norm emergence by pushing agents toward certain behaviors, which can become established as norms. Whereas it is well-known that sanctioning can lead to the emergence of norms, we posit that a broader kind of social intelligence can prove more effective in promoting cooperation in a multiagent system. Accordingly, we develop Nest, a framework that models social intelligence in the form of a wider variety of communications and understanding of them than in previous work. To evaluate Nest, we develop a simulated pandemic environment and conduct simulation experiments to compare Nest with baselines considering a combination of three kinds of social communication: sanction, tell, and hint. We find that societies formed of Nest agents achieve norms faster; moreover, Nest agents effectively avoid undesirable consequences, which are negative sanctions and deviation from goals, and yield higher satisfaction for themselves than baseline agents despite requiring only an equivalent amount of information.

Peer-to-peer (P2P) lending has emerged as a distinctive financing mechanism, linking borrowers with lenders through online platforms. However, P2P lending faces the challenge of information asymmetry, as lenders often lack sufficient data to assess the creditworthiness of borrowers. This paper proposes a novel approach to address this issue by leveraging the textual descriptions provided by borrowers during the loan application process. Our methodology involves processing these textual descriptions using a Large Language Model (LLM), a powerful tool capable of discerning patterns and semantics within the text. Transfer learning is applied to adapt the LLM to the specific task at hand. Our results derived from the analysis of the Lending Club dataset show that the risk score generated by BERT, a widely used LLM, significantly improves the performance of credit risk classifiers. However, the inherent opacity of LLM-based systems, coupled with uncertainties about potential biases, underscores critical considerations for regulatory frameworks and engenders trust-related concerns among end-users, opening new avenues for future research in the dynamic landscape of P2P lending and artificial intelligence.

The goal of inductive logic programming (ILP) is to search for a logic program that generalises training examples and background knowledge. We introduce an ILP approach that identifies minimal unsatisfiable subprograms (MUSPs). We show that finding MUSPs allows us to efficiently and soundly prune the search space. Our experiments on multiple domains, including program synthesis and game playing, show that our approach can reduce learning times by 99%.

Large language models are trained on massive scrapes of the web, which are often unstructured, noisy, and poorly phrased. Current scaling laws show that learning from such data requires an abundance of both compute and data, which grows with the size of the model being trained. This is infeasible both because of the large compute costs and duration associated with pre-training, and the impending scarcity of high-quality data on the web. In this work, we propose Web Rephrase Augmented Pre-training ($\textbf{WRAP}$) that uses an off-the-shelf instruction-tuned model prompted to paraphrase documents on the web in specific styles such as "like Wikipedia" or in "question-answer format" to jointly pre-train LLMs on real and synthetic rephrases. First, we show that using WRAP on the C4 dataset, which is naturally noisy, speeds up pre-training by $\sim3x$. At the same pre-training compute budget, it improves perplexity by more than 10% on average across different subsets of the Pile, and improves zero-shot question answer accuracy across 13 tasks by more than 2%. Second, we investigate the impact of the re-phrasing style on the performance of the model, offering insights into how the composition of the training data can impact the performance of LLMs in OOD settings. Our gains are attributed to the fact that re-phrased synthetic data has higher utility than just real data because it (i) incorporates style diversity that closely reflects downstream evaluation style, and (ii) has higher 'quality' than web-scraped data.

Cross-modal retrieval is the task of retrieving samples of a given modality by using queries of a different one. Due to the wide range of practical applications, the problem has been mainly focused on the vision and language case, e.g. text to image retrieval, where models like CLIP have proven effective in solving such tasks. The dominant approach to learning such coordinated representations consists of projecting them onto a common space where matching views stay close and those from non-matching pairs are pushed away from each other. Although this cross-modal coordination has been applied also to other pairwise combinations, extending it to an arbitrary number of diverse modalities is a problem that has not been fully explored in the literature. In this paper, we propose two different approaches to the problem. The first is based on an extension of the CLIP contrastive objective to an arbitrary number of input modalities, while the second departs from the contrastive formulation and tackles the coordination problem by regressing the cross-modal similarities towards a target that reflects two simple and intuitive constraints of the cross-modal retrieval task. We run experiments on two different datasets, over different combinations of input modalities and show that the approach is not only simple and effective but also allows for tackling the retrieval problem in novel ways. Besides capturing a more diverse set of pair-wise interactions, we show that we can use the learned representations to improve retrieval performance by combining the embeddings from two or more such modalities.

Geometry problem solving presents a formidable challenge within the NLP community. Existing approaches often rely on models designed for solving math word problems, neglecting the unique characteristics of geometry math problems. Additionally, the current research predominantly focuses on geometry calculation problems, while overlooking other essential aspects like proving. In this study, we address these limitations by proposing the Geometry-Aware Problem Solver (GAPS) model. GAPS is specifically designed to generate solution programs for geometry math problems of various types with the help of its unique problem-type classifier. To achieve this, GAPS treats the solution program as a composition of operators and operands, segregating their generation processes. Furthermore, we introduce the geometry elements enhancement method, which enhances the ability of GAPS to recognize geometry elements accurately. By leveraging these improvements, GAPS showcases remarkable performance in resolving geometry math problems. Our experiments conducted on the UniGeo dataset demonstrate the superiority of GAPS over the state-of-the-art model, Geoformer. Specifically, GAPS achieves an accuracy improvement of more than 5.3% for calculation tasks and an impressive 41.1% for proving tasks. Notably, GAPS achieves an impressive accuracy of 97.5% on proving problems, representing a significant advancement in solving geometry proving tasks.