Significant advances have been made in natural language processing in recent years. However, our current deep learning approach to language modeling requires substantial resources in terms of data and computation. One of the side effects of this data-hungry paradigm is the current schism between languages, separating those considered high-resource, where most of the development happens and resources are available, and the low-resource ones, which struggle to attain the same level of performance and autonomy. This study aims to introduce a new set of resources to stimulate the future development of neural text generation in Portuguese. In this work, we document the development of GigaVerbo, a concatenation of deduplicated Portuguese text corpora amounting to 200 billion tokens. Via this corpus, we trained a series of decoder-transformers named Tucano. Our models perform equal or superior to other Portuguese and multilingual language models of similar size in several Portuguese benchmarks. The evaluation of our models also reveals that model performance on many currently available benchmarks used by the Portuguese NLP community has little to no correlation with the scaling of token ingestion during training, highlighting the limitations of such evaluations when it comes to the assessment of Portuguese generative language models. All derivatives of our study are openly released on GitHub and Hugging Face. See https://nkluge-correa.github.io/Tucano/

In an effort to mitigate the harms of large language models (LLMs), learning from human feedback (LHF) has been used to steer LLMs towards outputs that are intended to be both less harmful and more helpful. Despite the widespread adoption of LHF in practice, the quality of this feedback and its effectiveness as a safety mitigation technique remain unclear. This study addresses these issues by auditing the widely-used Helpful and Harmless (HH) dataset by Anthropic. Our work includes: (1) a thorough investigation of the dataset's content through both manual and automated evaluation; (2) experiments demonstrating the dataset's impact on models' safety; and (3) an analysis of the 100 most influential papers citing this dataset. Through our audit, we showcase how conceptualization failures and quality issues identified in the HH dataset can create additional harms by leading to disparate safety behaviors across demographic groups. Our findings highlight the need for more nuanced, context-sensitive approaches to safety mitigation in LLMs.

Podcasts provide highly diverse content to a massive listener base through a unique on-demand modality. However, limited data has prevented large-scale computational analysis of the podcast ecosystem. To fill this gap, we introduce a massive dataset of over 1.1M podcast transcripts that is largely comprehensive of all English language podcasts available through public RSS feeds from May and June of 2020. This data is not limited to text, but rather includes audio features and speaker turns for a subset of 370K episodes, and speaker role inferences and other metadata for all 1.1M episodes. Using this data, we also conduct a foundational investigation into the content, structure, and responsiveness of this ecosystem. Together, our data and analyses open the door to continued computational research of this popular and impactful medium.

Architecture performance evaluation is the most time-consuming part of neural architecture search (NAS). Zero-Shot NAS accelerates the evaluation by utilizing zero-cost proxies instead of training. Though effective, existing zero-cost proxies require invoking backpropagations or running networks on input data, making it difficult to further accelerate the computation of proxies. To alleviate this issue, architecture topologies are used to evaluate the performance of networks in this study. We prove that particular architectural topologies decrease the local entropy of feature maps, which degrades specific features to a bias, thereby reducing network performance. Based on this proof, architectural topologies are utilized to quantify the suppression of local entropy decrease (SED) as a data-free and running-free proxy. Experimental results show that SED outperforms most state-of-the-art proxies in terms of architecture selection on five benchmarks, with computation time reduced by three orders of magnitude. We further compare the SED-based NAS with state-of-the-art proxies. SED-based NAS selects the architecture with higher accuracy and fewer parameters in only one second. The theoretical analyses of local entropy and experimental results demonstrate that the suppression of local entropy decrease facilitates selecting optimal architectures in Zero-Shot NAS.

Africa is still early in the process of adopting AI technologies. But researchers say the continent is uniquely hospitable to it for several reasons, including a relatively young and increasingly well-educated population, a rapidly growing ecosystem of AI startups, and lots of potential consumers. However, ambitious efforts to develop AI tools that answer the needs of Africans face numerous hurdles. The biggest are inadequate funding and poor infrastructure. Limited internet access and a scarcity of domestic data centers also mean that developers might not be able to deploy cutting-edge AI capabilities.

Okinga-Koumu pulled a phone from the pocket of her blue jeans and opened a prototype web app she's built. Using VR and AI features, the app allows students to simulate using the necessary lab equipment--exploring 3D models of the tools in a real-world setting, like a classroom or lab. "Students could have detailed VR of lab equipment, making their hands-on experience more effective," she said. Established in 2017, the Deep Learning Indaba now has chapters in 47 of the 55 African nations and aims to boost AI development across the continent by providing training and resources to African AI researchers like Okinga-Koumu. Africa is still early in the process of adopting AI technologies, but organizers say the continent is uniquely hospitable to it for several reasons, including a relatively young and increasingly well-educated population, a rapidly growing ecosystem of AI startups, and lots of potential consumers.

This study investigates the use of generative AI and multi-agent systems to provide automatic feedback in educational contexts, particularly for student constructed responses in science assessments. The research addresses a key gap in the field by exploring how multi-agent systems, called AutoFeedback, can improve the quality of GenAI-generated feedback, overcoming known issues such as over-praise and over-inference that are common in single-agent large language models (LLMs). The study developed a multi-agent system consisting of two AI agents: one for generating feedback and another for validating and refining it. The system was tested on a dataset of 240 student responses, and its performance was compared to that of a single-agent LLM. Results showed that AutoFeedback significantly reduced the occurrence of over-praise and over-inference errors, providing more accurate and pedagogically sound feedback. The findings suggest that multi-agent systems can offer a more reliable solution for generating automated feedback in educational settings, highlighting their potential for scalable and personalized learning support. These results have important implications for educators and researchers seeking to leverage AI in formative assessments, offering a pathway to more effective feedback mechanisms that enhance student learning outcomes.

Traffic accidents pose a significant threat to public safety, resulting in numerous fatalities, injuries, and a substantial economic burden each year. The development of predictive models capable of real-time forecasting of post-accident impact using readily available data can play a crucial role in preventing adverse outcomes and enhancing overall safety. However, existing accident predictive models encounter two main challenges: first, reliance on either costly or non-real-time data, and second the absence of a comprehensive metric to measure post-accident impact accurately. To address these limitations, this study proposes a deep neural network model known as the cascade model. It leverages readily available real-world data from Los Angeles County to predict post-accident impacts. The model consists of two components: Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN). The LSTM model captures temporal patterns, while the CNN extracts patterns from the sparse accident dataset. Furthermore, an external traffic congestion dataset is incorporated to derive a new feature called the "accident impact" factor, which quantifies the influence of an accident on surrounding traffic flow. Extensive experiments were conducted to demonstrate the effectiveness of the proposed hybrid machine learning method in predicting the post-accident impact compared to state-of-the-art baselines. The results reveal a higher precision in predicting minimal impacts (i.e., cases with no reported accidents) and a higher recall in predicting more significant impacts (i.e., cases with reported accidents).

Utilizing fault diagnosis methods is crucial for nuclear power professionals to achieve efficient and accurate fault diagnosis for nuclear power plants (NPPs). The performance of traditional methods is limited by their dependence on complex feature extraction and skilled expert knowledge, which can be time-consuming and subjective. This paper proposes a novel intelligent fault diagnosis method for NPPs that combines enhanced temporal convolutional network (ETCN) with sparrow search algorithm (SSA). ETCN utilizes temporal convolutional network (TCN), self-attention (SA) mechanism and residual block for enhancing performance. ETCN excels at extracting local features and capturing time series information, while SSA adaptively optimizes its hyperparameters for superior performance. The proposed method's performance is experimentally verified on a CPR1000 simulation dataset. Compared to other advanced intelligent fault diagnosis methods, the proposed one demonstrates superior performance across all evaluation metrics. This makes it a promising tool for NPP intelligent fault diagnosis, ultimately enhancing operational reliability.

Characterizing the computational power of neural network architectures in terms of formal language theory remains a crucial line of research, as it describes lower and upper bounds on the reasoning capabilities of modern AI. However, when empirically testing these bounds, existing work often leaves a discrepancy between experiments and the formal claims they are meant to support. The problem is that formal language theory pertains specifically to recognizers: machines that receive a string as input and classify whether it belongs to a language. On the other hand, it is common to instead use proxy tasks that are similar in only an informal sense, such as language modeling or sequence-to-sequence transduction. We correct this mismatch by training and evaluating neural networks directly as binary classifiers of strings, using a general method that can be applied to a wide variety of languages. As part of this, we extend an algorithm recently proposed by Sn{\ae}bjarnarson et al. (2024) to do length-controlled sampling of strings from regular languages, with much better asymptotic time complexity than previous methods. We provide results on a variety of languages across the Chomsky hierarchy for three neural architectures: a simple RNN, an LSTM, and a causally-masked transformer. We find that the RNN and LSTM often outperform the transformer, and that auxiliary training objectives such as language modeling can help, although no single objective uniformly improves performance across languages and architectures. Our contributions will facilitate theoretically sound empirical testing of language recognition claims in future work. We have released our datasets as a benchmark called FLaRe (Formal Language Recognition), along with our code.