We describe a comprehensive methodology for developing user-voice personalized automatic speech recognition (ASR) models by effectively training models on mobile phones, allowing user data and models to be stored and used locally. To achieve this, we propose a resource-aware sub-model-based training approach that considers the RAM, and battery capabilities of mobile phones. By considering the evaluation metric and resource constraints of the mobile phones, we are able to perform efficient training and halt the process accordingly. To simulate real users, we use speakers with various accents. The entire on-device training and evaluation framework was then tested on various mobile phones across brands. We show that fine-tuning the models and selecting the right hyperparameter values is a trade-off between the lowest achievable performance metric, on-device training time, and memory consumption. Overall, our methodology offers a comprehensive solution for developing personalized ASR models while leveraging the capabilities of mobile phones, and balancing the need for accuracy with resource constraints.

Despite the existence of various benchmarks for evaluating natural language processing models, we argue that human exams are a more suitable means of evaluating general intelligence for large language models (LLMs), as they inherently demand a much wider range of abilities such as language understanding, domain knowledge, and problem-solving skills. To this end, we introduce M3Exam, a novel benchmark sourced from real and official human exam questions for evaluating LLMs in a multilingual, multimodal, and multilevel context. M3Exam exhibits three unique characteristics: (1) multilingualism, encompassing questions from multiple countries that require strong multilingual proficiency and cultural knowledge; (2) multimodality, accounting for the multimodal nature of many exam questions to test the model's multimodal understanding capability; and (3) multilevel structure, featuring exams from three critical educational periods to comprehensively assess a model's proficiency at different levels. In total, M3Exam contains 12,317 questions in 9 diverse languages with three educational levels, where about 23\% of the questions require processing images for successful solving. We assess the performance of top-performing LLMs on M3Exam and find that current models, including GPT-4, still struggle with multilingual text, particularly in low-resource and non-Latin script languages. Multimodal LLMs also perform poorly with complex multimodal questions. We believe that M3Exam can be a valuable resource for comprehensively evaluating LLMs by examining their multilingual and multimodal abilities and tracking their development. Data and evaluation code is available at \url{https://github.com/DAMO-NLP-SG/M3Exam}.

Gender bias is a significant issue in machine translation, leading to ongoing research efforts in developing bias mitigation techniques. However, most works focus on debiasing bilingual models without much consideration for multilingual systems. In this paper, we specifically target the gender bias issue of multilingual machine translation models for unambiguous cases where there is a single correct translation, and propose a bias mitigation method based on a novel approach. Specifically, we propose Gender-Aware Contrastive Learning, GACL, which encodes contextual gender information into the representations of non-explicit gender words. Our method is target language-agnostic and is applicable to pre-trained multilingual machine translation models via fine-tuning. Through multilingual evaluation, we show that our approach improves gender accuracy by a wide margin without hampering translation performance. We also observe that incorporated gender information transfers and benefits other target languages regarding gender accuracy. Finally, we demonstrate that our method is applicable and beneficial to models of various sizes.

As the capabilities of language models continue to advance, it is conceivable that "one-size-fits-all" model will remain as the main paradigm. For instance, given the vast number of languages worldwide, many of which are low-resource, the prevalent practice is to pretrain a single model on multiple languages. In this paper, we add to the growing body of evidence that challenges this practice, demonstrating that monolingual pretraining on the target language significantly improves models already extensively trained on diverse corpora. More specifically, we further pretrain GPT-J and LLaMA models on Portuguese texts using 3% or less of their original pretraining budget. Few-shot evaluations on Poeta, a suite of 14 Portuguese datasets, reveal that our models outperform English-centric and multilingual counterparts by a significant margin. Our best model, Sabi\'a-65B, performs on par with GPT-3.5-turbo. By evaluating on datasets originally conceived in the target language as well as translated ones, we study the contributions of language-specific pretraining in terms of 1) capturing linguistic nuances and structures inherent to the target language, and 2) enriching the model's knowledge about a domain or culture. Our results indicate that the majority of the benefits stem from the domain-specific knowledge acquired through monolingual pretraining.

In human-computer conversations, extracting entities such as names, street addresses and email addresses from speech is a challenging task. In this paper, we study the impact of fine-tuning pre-trained speech encoders on extracting spoken entities in human-readable form directly from speech without the need for text transcription. We illustrate that such a direct approach optimizes the encoder to transcribe only the entity relevant portions of speech ignoring the superfluous portions such as carrier phrases, or spell name entities. In the context of dialog from an enterprise virtual agent, we demonstrate that the 1-step approach outperforms the typical 2-step approach which first generates lexical transcriptions followed by text-based entity extraction for identifying spoken entities.

Federated Learning is expected to provide strong privacy guarantees, as only gradients or model parameters but no plain text training data is ever exchanged either between the clients or between the clients and the central server. In this paper, we challenge this claim by introducing a simple but still very effective membership inference attack algorithm, which relies only on a single training step. In contrast to the popular honest-but-curious model, we investigate a framework with a dishonest central server. Our strategy is applicable to models with ReLU activations and uses the properties of this activation function to achieve perfect accuracy. Empirical evaluation on visual classification tasks with MNIST, CIFAR10, CIFAR100 and CelebA datasets show that our method provides perfect accuracy in identifying one sample in a training set with thousands of samples. Occasional failures of our method lead us to discover duplicate images in the CIFAR100 and CelebA datasets.

Google's Search Generative Experience (SGE), which currently provides generative AI summaries at the top of the search results page for select users, is about to be much more available. Just six months after its debut at I/O 2023, the company announced Wednesday that SGE is expanding to Search Labs users in 120 countries and territories, gaining support for four additional languages and receiving a handful of helpful new features. Unlike its frenetic rollout of the Bard chatbot in March, Google has taken a slightly more measured tone in distributing its AI search assistant. The company began with English language searches in the US in May, expanded to English-language users in India and Japan in August and on to teen users in September. As of Wednesday, users from Brazil to Bhutan can give the feature a try.

Existing datasets for automated fact-checking have substantial limitations, such as relying on artificial claims, lacking annotations for evidence and intermediate reasoning, or including evidence published after the claim. In this paper we introduce AVeriTeC, a new dataset of 4,568 real-world claims covering fact-checks by 50 different organizations. Each claim is annotated with question-answer pairs supported by evidence available online, as well as textual justifications explaining how the evidence combines to produce a verdict. Through a multi-round annotation process, we avoid common pitfalls including context dependence, evidence insufficiency, and temporal leakage, and reach a substantial inter-annotator agreement of $\kappa=0.619$ on verdicts. We develop a baseline as well as an evaluation scheme for verifying claims through several question-answering steps against the open web.

The limited availability of annotations in small molecule datasets presents a challenge to machine learning models. To address this, one common strategy is to collaborate with additional auxiliary datasets. However, having more data does not always guarantee improvements. Negative transfer can occur when the knowledge in the target dataset differs or contradicts that of the auxiliary molecule datasets. In light of this, identifying the auxiliary molecule datasets that can benefit the target dataset when jointly trained remains a critical and unresolved problem. Through an empirical analysis, we observe that combining graph structure similarity and task similarity can serve as a more reliable indicator for identifying high-affinity auxiliary datasets. Motivated by this insight, we propose MolGroup, which separates the dataset affinity into task and structure affinity to predict the potential benefits of each auxiliary molecule dataset. MolGroup achieves this by utilizing a routing mechanism optimized through a bi-level optimization framework. Empowered by the meta gradient, the routing mechanism is optimized toward maximizing the target dataset's performance and quantifies the affinity as the gating score. As a result, MolGroup is capable of predicting the optimal combination of auxiliary datasets for each target dataset. Our extensive experiments demonstrate the efficiency and effectiveness of MolGroup, showing an average improvement of 4.41%/3.47% for GIN/Graphormer trained with the group of molecule datasets selected by MolGroup on 11 target molecule datasets.

In this paper, we delineate the strategy employed by our team, DeepLearningBrasil, which secured us the first place in the shared task DepSign-LT-EDI@RANLP-2023, achieving a 47.0% Macro F1-Score and a notable 2.4% advantage. The task was to classify social media texts into three distinct levels of depression - "not depressed," "moderately depressed," and "severely depressed." Leveraging the power of the RoBERTa and DeBERTa models, we further pre-trained them on a collected Reddit dataset, specifically curated from mental health-related Reddit's communities (Subreddits), leading to an enhanced understanding of nuanced mental health discourse. To address lengthy textual data, we used truncation techniques that retained the essence of the content by focusing on its beginnings and endings. Our model was robust against unbalanced data by incorporating sample weights into the loss. Cross-validation and ensemble techniques were then employed to combine our k-fold trained models, delivering an optimal solution. The accompanying code is made available for transparency and further development.