Ontologies are useful for automatic machine processing of domain knowledge as they represent it in a structured format. Yet, constructing ontologies requires substantial manual effort. To automate part of this process, large language models (LLMs) have been applied to solve various subtasks of ontology learning. However, this partial ontology learning does not capture the interactions between subtasks. We address this gap by introducing OLLM, a general and scalable method for building the taxonomic backbone of an ontology from scratch. Rather than focusing on subtasks, like individual relations between entities, we model entire subcomponents of the target ontology by finetuning an LLM with a custom regulariser that reduces overfitting on high-frequency concepts. We introduce a novel suite of metrics for evaluating the quality of the generated ontology by measuring its semantic and structural similarity to the ground truth. In contrast to standard metrics, our metrics use deep learning techniques to define more robust distance measures between graphs. Both our quantitative and qualitative results on Wikipedia show that OLLM outperforms subtask composition methods, producing more semantically accurate ontologies while maintaining structural integrity. We further demonstrate that our model can be effectively adapted to new domains, like arXiv, needing only a small number of training examples. Our source code and datasets are available at https://github.com/andylolu2/ollm.

The anisotropic structure of the myocardium is a key determinant of the cardiac function. To date, there is no imaging modality to assess in-vivo the cardiac fiber structure. We recently proposed Fibernet, a method for the automatic identification of the anisotropic conduction -- and thus fibers -- in the atria from local electrical recordings. Fibernet uses cardiac activation as recorded during electroanatomical mappings to infer local conduction properties using physics-informed neural networks. In this work, we extend Fibernet to cope with the uncertainty in the estimated fiber field. Specifically, we use an ensemble of neural networks to produce multiple samples, all fitting the observed data, and compute posterior statistics. We also introduce a methodology to select the best fiber orientation members and define the input of the neural networks directly on the atrial surface. With these improvements, we outperform the previous methodology in terms of fiber orientation error in 8 different atrial anatomies. Currently, our approach can estimate the fiber orientation and conduction velocities in under 7 minutes with quantified uncertainty, which opens the door to its application in clinical practice. We hope the proposed methodology will enable further personalization of cardiac digital twins for precision medicine.

Magnetoencephalography (MEG) allows the non-invasive detection of interictal epileptiform discharges (IEDs). Clinical MEG analysis in epileptic patients traditionally relies on the visual identification of IEDs, which is time consuming and partially subjective. Automatic, data-driven detection methods exist but show limited performance. Still, the rise of deep learning (DL)-with its ability to reproduce human-like abilities-could revolutionize clinical MEG practice. Here, we developed and validated STIED, a simple yet powerful supervised DL algorithm combining two convolutional neural networks with temporal (1D time-course) and spatial (2D topography) features of MEG signals inspired from current clinical guidelines. Our DL model enabled both temporal and spatial localization of IEDs in patients suffering from focal epilepsy with frequent and high amplitude spikes (FE group), with high-performance metrics-accuracy, specificity, and sensitivity all exceeding 85%-when learning from spatiotemporal features of IEDs. This performance can be attributed to our handling of input data, which mimics established clinical MEG practice. Reverse engineering further revealed that STIED encodes fine spatiotemporal features of IEDs rather than their mere amplitude. The model trained on the FE group also showed promising results when applied to a separate group of presurgical patients with different types of refractory focal epilepsy, though further work is needed to distinguish IEDs from physiological transients. This study paves the way of incorporating STIED and DL algorithms into the routine clinical MEG evaluation of epilepsy.

The first artificial quantum neuron models followed a similar path to classic models, as they work only with discrete values. Here we introduce an algorithm that generalizes the binary model manipulating the phase of complex numbers. We propose, test, and implement a neuron model that works with continuous values in a quantum computer. Through simulations, we demonstrate that our model may work in a hybrid training scheme utilizing gradient descent as a learning algorithm. This work represents another step in the direction of evaluation of the use of artificial neural networks efficiently implemented on near-term quantum devices.

The non-convex nature of trained neural networks has created significant obstacles in their incorporation into optimization models. Considering the wide array of applications that this embedding has, the optimization and deep learning communities have dedicated significant efforts to the convexification of trained neural networks. Many approaches to date have considered obtaining convex relaxations for each non-linear activation in isolation, which poses limitations in the tightness of the relaxations. Anderson et al. (2020) strengthened these relaxations and provided a framework to obtain the convex hull of the graph of a piecewise linear convex activation composed with an affine function; this effectively convexifies activations such as the ReLU together with the affine transformation that precedes it. In this article, we contribute to this line of work by developing a recursive formula that yields a tight convexification for the composition of an activation with an affine function for a wide scope of activation functions, namely, convex or ``S-shaped". Our approach can be used to efficiently compute separating hyperplanes or determine that none exists in various settings, including non-polyhedral cases. We provide computational experiments to test the empirical benefits of these convex approximations.

As large language models (LLMs) expand into performing as agents for real-world applications beyond traditional NLP tasks, evaluating their robustness becomes increasingly important. However, existing benchmarks often overlook critical dimensions like cultural and social awareness. To address these, we introduce CASA, a benchmark designed to assess LLM agents' sensitivity to cultural and social norms across two web-based tasks: online shopping and social discussion forums. Our approach evaluates LLM agents' ability to detect and appropriately respond to norm-violating user queries and observations. Furthermore, we propose a comprehensive evaluation framework that measures awareness coverage, helpfulness in managing user queries, and the violation rate when facing misleading web content. Experiments show that current LLMs perform significantly better in non-agent than in web-based agent environments, with agents achieving less than 10% awareness coverage and over 40% violation rates. To improve performance, we explore two methods: prompting and fine-tuning, and find that combining both methods can offer complementary advantages -- fine-tuning on culture-specific datasets significantly enhances the agents' ability to generalize across different regions, while prompting boosts the agents' ability to navigate complex tasks. These findings highlight the importance of constantly benchmarking LLM agents' cultural and social awareness during the development cycle.

The hallucinations of large language models (LLMs) are increasingly mitigated by allowing LLMs to search for information and to ground their answers in real sources. Observing that LLMs can learn to search for relevant facts by trying different queries and learning to up-weight queries that successfully produce relevant results, we introduce Learning to Retrieve by Trying (LeReT), a reinforcement learning framework that explores search queries and uses preference-based optimization to improve their quality. LeReT can improve the absolute retrieval accuracy by up to 29% and the downstream generator evaluations by 17%. The simplicity and flexibility of LeReT allows it to be applied to arbitrary off-the-shelf retrievers and makes it a promising technique for improving general LLM pipelines. Despite tremendous progress, large language models (LLMs) still often hallucinate, motivating significant interest in grounding LLM answers in verified sources (Guu et al., 2020; Komeili et al., 2022; PerplexityAI, 2024; Google, 2024; OpenAI, 2024). Unfortunately, simply retrieving semantically similar documents to the user question, as is prevalent in retrieval-augmented generation (RAG; Lewis et al. 2020) pipelines, tends to fail for complex information needs not answered directly by any individual document. To tackle this, multi-hop retrieval pipelines gather information incrementally over multiple steps of search. For example, if a user asks What is a good dinner place driving from the Bay Area to Lake Tahoe on Friday night to avoid traffic?, a grounded system might need to learn about towns en route Lake Tahoe from the Bay Area, followed by traffic forecast on I-80 and finally, restaurants in Auburn (and other towns). However, doing this successfully is hard as off-the-shelf LLM performance is often unsatisfactory, and producing supervision for the best search queries to generate in a sequence of "hops" is nontrivial and expensive. Recent work tackles this via prompt optimization and rejection fine-tuning given a downstream signal.

Transformer-based architectures have dominated various areas of machine learning in recent years. In this paper, we introduce a novel robust attention mechanism designed to enhance the resilience of transformer-based architectures. Crucially, this technique can be integrated into existing transformers as a plug-and-play layer, improving their robustness without the need for additional training or fine-tuning. Through comprehensive experiments and ablation studies, we demonstrate that our ProTransformer significantly enhances the robustness of transformer models across a variety of prediction tasks, attack mechanisms, backbone architectures, and data domains. Notably, without further fine-tuning, the ProTransformer consistently improves the performance of vanilla transformers by 19.5%, 28.3%, 16.1%, and 11.4% for BERT, ALBERT, DistilBERT, and RoBERTa, respectively, under the classical TextFooler attack. Furthermore, ProTransformer shows promising resilience in large language models (LLMs) against prompting-based attacks, improving the performance of T5 and LLaMA by 24.8% and 17.8%, respectively, and enhancing Vicuna by an average of 10.4% against the Jailbreaking attack. Beyond the language domain, ProTransformer also demonstrates outstanding robustness in both vision and graph domains.

We investigate how low-quality AI advisors, lacking quality disclosures, can help spread text-based lies while seeming to help people detect lies. Participants in our experiment discern truth from lies by evaluating transcripts from a game show that mimicked deceptive social media exchanges on topics with objective truths. We find that when relying on low-quality advisors without disclosures, participants' truth-detection rates fall below their own abilities, which recovered once the AI's true effectiveness was revealed. Conversely, high-quality advisor enhances truth detection, regardless of disclosure. We discover that participants' expectations about AI capabilities contribute to their undue reliance on opaque, low-quality advisors.

Lexical-semantic resources (LSRs), such as online lexicons or wordnets, are fundamental for natural language processing applications. In many languages, however, such resources suffer from quality issues: incorrect entries, incompleteness, but also, the rarely addressed issue of bias towards the English language and Anglo-Saxon culture. Such bias manifests itself in the absence of concepts specific to the language or culture at hand, the presence of foreign (Anglo-Saxon) concepts, as well as in the lack of an explicit indication of untranslatability, also known as cross-lingual \emph{lexical gaps}, when a term has no equivalent in another language. This paper proposes a novel crowdsourcing methodology for reducing bias in LSRs. Crowd workers compare lexemes from two languages, focusing on domains rich in lexical diversity, such as kinship or food. Our LingoGap crowdsourcing tool facilitates comparisons through microtasks identifying equivalent terms, language-specific terms, and lexical gaps across languages. We validated our method by applying it to two case studies focused on food-related terminology: (1) English and Arabic, and (2) Standard Indonesian and Banjarese. These experiments identified 2,140 lexical gaps in the first case study and 951 in the second. The success of these experiments confirmed the usability of our method and tool for future large-scale lexicon enrichment tasks.