Transformer-based pre-trained language models (PLMs) have achieved remarkable performance in various natural language processing (NLP) tasks. However, pre-training such models can take considerable resources that are almost only available to high-resource languages. On the contrary, static word embeddings are easier to train in terms of computing resources and the amount of data required. In this paper, we introduce MoSECroT Model Stitching with Static Word Embeddings for Crosslingual Zero-shot Transfer), a novel and challenging task that is especially relevant to low-resource languages for which static word embeddings are available. To tackle the task, we present the first framework that leverages relative representations to construct a common space for the embeddings of a source language PLM and the static word embeddings of a target language. In this way, we can train the PLM on source-language training data and perform zero-shot transfer to the target language by simply swapping the embedding layer. However, through extensive experiments on two classification datasets, we show that although our proposed framework is competitive with weak baselines when addressing MoSECroT, it fails to achieve competitive results compared with some strong baselines. In this paper, we attempt to explain this negative result and provide several thoughts on possible improvement.

We investigate large language model performance across five orders of magnitude of compute scaling in eleven recent model architectures. We show that average benchmark performance, aggregating over many individual tasks and evaluations as in the commonly-used BIG-Bench dataset, is decently predictable as a function of training compute scale. Specifically, when extrapolating BIG-Bench Hard performance across one order of magnitude in compute, we observe average absolute errors of 6 percentage points (pp). By contrast, extrapolation for individual BIG-Bench tasks across an order of magnitude in compute yields higher average errors of 18pp. Nonetheless, individual task performance remains significantly more predictable than chance. Overall, our work suggests compute scaling provides a promising basis to forecast AI capabilities in diverse benchmarks, though predicting performance in specific tasks poses challenges.

Stakeholders constantly make assumptions in the development of deep learning (DL) frameworks. These assumptions are related to various types of software artifacts (e.g., requirements, design decisions, and technical debt) and can turn out to be invalid, leading to system failures. Existing approaches and tools for assumption management usually depend on manual identification of assumptions. However, assumptions are scattered in various sources (e.g., code comments, commits, pull requests, and issues) of DL framework development, and manually identifying assumptions has high costs (e.g., time and resources). To overcome the issues of manually identifying assumptions in DL framework development, we constructed a new and largest dataset (i.e., AssuEval) of assumptions collected from the TensorFlow and Keras repositories on GitHub; explored the performance of seven traditional machine learning models (e.g., Support Vector Machine, Classification and Regression Trees), a popular DL model (i.e., ALBERT), and a large language model (i.e., ChatGPT) of identifying assumptions on the AssuEval dataset. The experiment results show that: ALBERT achieves the best performance (f1-score: 0.9584) of identifying assumptions on the AssuEval dataset, which is much better than the other models (the 2nd best f1-score is 0.6211, achieved by ChatGPT). Though ChatGPT is the most popular large language model, we do not recommend using it to identify assumptions in DL framework development because of its low performance on the task. Fine-tuning ChatGPT specifically for assumption identification could improve the performance. This study provides researchers with the largest dataset of assumptions for further research (e.g., assumption classification, evaluation, and reasoning) and helps practitioners better understand assumptions and how to manage them in their projects.

Effective disaster response is critical for affected communities. Responders and decision-makers would benefit from reliable, timely measures of the issues impacting their communities during a disaster, and social media offers a potentially rich data source. Social media can reflect public concerns and demands during a disaster, offering valuable insights for decision-makers to understand evolving situations and optimize resource allocation. We used Bidirectional Encoder Representations from Transformers (BERT) topic modeling to cluster topics from Twitter data. Then, we conducted a temporal-spatial analysis to examine the distribution of these topics across different regions during the 2020 western U.S. wildfire season. Our results show that Twitter users mainly focused on three topics:"health impact," "damage," and "evacuation." We used the Susceptible-Infected-Recovered (SIR) theory to explore the magnitude and velocity of topic diffusion on Twitter. The results displayed a clear relationship between topic trends and wildfire propagation patterns. The estimated parameters obtained from the SIR model in selected cities revealed that residents exhibited a high level of several concerns during the wildfire. Our study details how the SIR model and topic modeling using social media data can provide decision-makers with a quantitative approach to measure disaster response and support their decision-making processes.

In this paper, we explore effective prompting techniques to enhance zero- and few-shot Visual Question Answering (VQA) performance in contemporary Vision-Language Models (VLMs). Central to our investigation is the role of question templates in guiding VLMs to generate accurate answers. We identify that specific templates significantly influence VQA outcomes, underscoring the need for strategic template selection. Another pivotal aspect of our study is augmenting VLMs with image captions, providing them with additional visual cues alongside direct image features in VQA tasks. Surprisingly, this augmentation significantly improves the VLMs' performance in many cases, even though VLMs "see" the image directly! We explore chain-of-thought (CoT) reasoning and find that while standard CoT reasoning causes drops in performance, advanced methods like self-consistency can help recover it. Furthermore, we find that text-only few-shot examples enhance VLMs' alignment with the task format, particularly benefiting models prone to verbose zero-shot answers. Lastly, to mitigate the challenges associated with evaluating free-form open-ended VQA responses using string-matching based VQA metrics, we introduce a straightforward LLM-guided pre-processing technique to adapt the model responses to the expected ground-truth answer distribution. In summary, our research sheds light on the intricacies of prompting strategies in VLMs for VQA, emphasizing the synergistic use of captions, templates, and pre-processing to enhance model efficacy.

Conventional task- and modality-specific artificial intelligence (AI) models are inflexible in real-world deployment and maintenance for biomedicine. At the same time, the growing availability of biomedical data, coupled with the advancements in modern multi-modal multi-task AI techniques, has paved the way for the emergence of generalist biomedical AI solutions. These solutions hold the potential to interpret different medical modalities and produce expressive outputs such as free-text reports or disease diagnosis. Here, we propose BiomedGPT, the first open-source and generalist visual language AI for diverse biomedical tasks. BiomedGPT achieved 16 state-of-the-art results across five clinically significant tasks on 26 datasets. Notably, it outperformed OpenAI's GPT-4 with vision (GPT-4V) in radiology human evaluation and surpassed Google's Med-PaLM M (12B) in breast cancer diagnosis and medical visual question answering. Moreover, BiomedGPT facilitates zero-shot transfer learning, greatly enhancing its utility as a biomedical assistant, similar to ChatGPT. Our method demonstrates effective training with diverse datasets can lead to more practical biomedical AI.

Effective representation of molecules is a crucial factor affecting the performance of artificial intelligence models. This study introduces a flexible, fragment-based, multiscale molecular representation framework called t-SMILES (tree-based SMILES) with three code algorithms: TSSA (t-SMILES with Shared Atom), TSDY (t-SMILES with Dummy Atom) and TSID (t-SMILES with ID). It describes molecules using SMILES-type strings obtained by performing a breadth-first search on a full binary tree formed from a fragmented molecular graph. Systematic evaluations using JTVAE, BRICS, MMPA, and Scaffold show the feasibility to construct a multi-code molecular description system, where various descriptions complement each other, enhancing the overall performance. Additionally, it exhibits impressive performance on low-resource datasets, whether the model is original, data augmented, or pre-training fine-tuned. It significantly outperforms classical SMILES, DeepSMILES, SELFIES and baseline models in goal-directed tasks. Furthermore, it surpasses start-of-the-art fragment, graph and SMILES based approaches on ChEMBL, Zinc, and QM9.

Factual questions typically can be answered correctly at different levels of granularity. For example, both ``August 4, 1961'' and ``1961'' are correct answers to the question ``When was Barack Obama born?''. Standard question answering (QA) evaluation protocols, however, do not explicitly take this into account and compare a predicted answer against answers of a single granularity level. In this work, we propose GRANOLA QA, a novel evaluation setting where a predicted answer is evaluated in terms of accuracy and informativeness against a set of multi-granularity answers. We present a simple methodology for enriching existing datasets with multi-granularity answers, and create GRANOLA-EQ, a multi-granularity version of the EntityQuestions dataset. We evaluate a range of decoding methods on GRANOLA-EQ, including a new algorithm, called Decoding with Response Aggregation (DRAG), that is geared towards aligning the response granularity with the model's uncertainty. Our experiments show that large language models with standard decoding tend to generate specific answers, which are often incorrect. In contrast, when evaluated on multi-granularity answers, DRAG yields a nearly 20 point increase in accuracy on average, which further increases for rare entities. Overall, this reveals that standard evaluation and decoding schemes may significantly underestimate the knowledge encapsulated in LMs.

Effective prioritization of issue reports is crucial in software engineering to optimize resource allocation and address critical problems promptly. However, the manual classification of issue reports for prioritization is laborious and lacks scalability. Alternatively, many open source software (OSS) projects employ automated processes for this task, albeit relying on substantial datasets for adequate training. This research seeks to devise an automated approach that ensures reliability in issue prioritization, even when trained on smaller datasets. Our proposed methodology harnesses the power of Generative Pre-trained Transformers (GPT), recognizing their potential to efficiently handle this task. By leveraging the capabilities of such models, we aim to develop a robust system for prioritizing issue reports accurately, mitigating the necessity for extensive training data while maintaining reliability. In our research, we have developed a reliable GPT-based approach to accurately label and prioritize issue reports with a reduced training dataset. By reducing reliance on massive data requirements and focusing on few-shot fine-tuning, our methodology offers a more accessible and efficient solution for issue prioritization in software engineering. Our model predicted issue types in individual projects up to 93.2% in precision, 95% in recall, and 89.3% in F1-score.

In the contemporary digital era, the Internet functions as an unparalleled catalyst, dismantling geographical and linguistic barriers particularly evident in texting. This evolution facilitates global communication, transcending physical distances and fostering dynamic cultural exchange. A notable trend is the widespread use of transliteration, where the English alphabet is employed to convey messages in native languages, posing a unique challenge for language technology in accurately detecting the source language. This paper addresses this challenge through a dataset of phone text messages in Hindi and Russian transliterated into English utilizing BERT for language classification and Google Translate API for transliteration conversion. The research pioneers innovative approaches to identify and convert transliterated text, navigating challenges in the diverse linguistic landscape of digital communication. Emphasizing the pivotal role of comprehensive datasets for training Large Language Models LLMs like BERT, our model showcases exceptional proficiency in accurately identifying and classifying languages from transliterated text. With a validation accuracy of 99% our models robust performance underscores its reliability. The comprehensive exploration of transliteration dynamics supported by innovative approaches and cutting edge technologies like BERT, positions our research at the forefront of addressing unique challenges in the linguistic landscape of digital communication. Beyond contributing to language identification and transliteration capabilities this work holds promise for applications in content moderation, analytics and fostering a globally connected community engaged in meaningful dialogue.