This paper investigates a critical design decision in the practice of massively multilingual continual pre-training -- the inclusion of parallel data. Specifically, we study the impact of bilingual translation data for massively multilingual language adaptation of the Llama3 family of models to 500 languages. To this end, we construct the MaLA bilingual translation corpus, containing data from more than 2,500 language pairs. Subsequently, we develop the EMMA-500 Llama 3 suite of four massively multilingual models -- continually pre-trained from the Llama 3 family of base models extensively on diverse data mixes up to 671B tokens -- and explore the effect of continual pre-training with or without bilingual translation data. Comprehensive evaluation across 7 tasks and 12 benchmarks demonstrates that bilingual data tends to enhance language transfer and performance, particularly for low-resource languages. We open-source the MaLA corpus, EMMA-500 Llama 3 suite artefacts, code, and model generations.

In real practice, questions are typically complex and knowledge-intensive, requiring Large Language Models (LLMs) to recognize the multifaceted nature of the question and reason across multiple information sources. Iterative and adaptive retrieval, where LLMs decide when and what to retrieve based on their reasoning, has been shown to be a promising approach to resolve complex, knowledge-intensive questions. However, the performance of such retrieval frameworks is limited by the accumulation of reasoning errors and misaligned retrieval results. To overcome these limitations, we propose TreeRare (Syntax Tree-Guided Retrieval and Reasoning), a framework that utilizes syntax trees to guide information retrieval and reasoning for question answering. Following the principle of compositionality, TreeRare traverses the syntax tree in a bottom-up fashion, and in each node, it generates subcomponent-based queries and retrieves relevant passages to resolve localized uncertainty. A subcomponent question answering module then synthesizes these passages into concise, context-aware evidence. Finally, TreeRare aggregates the evidence across the tree to form a final answer. Experiments across five question answering datasets involving ambiguous or multi-hop reasoning demonstrate that TreeRare achieves substantial improvements over existing state-of-the-art methods.

The size and compute characteristics of modern large language models have led to an increased interest in developing specialized kernels tailored for particular training and inference workloads. Existing kernels primarily optimize for compute utilization, targeting the large-batch training and inference settings. However, low-batch inference, where memory bandwidth and kernel launch overheads are significant factors, remains important for many applications of interest such as in edge deployment and latency-sensitive applications. This paper describes FlashFormer, which fuses the entire transformer forward pass into a single kernel for accelerating low-batch inference of large language models. Across various model sizes and quantizations settings, FlashFormer achieves nontrivial speedups compared to existing inference kernels.

With the rise of artificial intelligence (A.I.) and large language models like ChatGPT, a new race for achieving artificial general intelligence (A.G.I) has started. While many speculate how and when A.I. will achieve A.G.I., there is no clear agreement on how A.G.I. can be detected in A.I. models, even when popular tools like the Turing test (and its modern variations) are used to measure their intelligence. In this work, we discuss why traditional methods like the Turing test do not suffice for measuring or detecting A.G.I. and provide a new, practical method that can be used to decide if a system (computer or any other) has reached or surpassed A.G.I. To achieve this, we make two new contributions. First, we present a clear definition for general intelligence (G.I.) and set a G.I. Threshold (G.I.T.) that can be used to distinguish between systems that achieve A.G.I. and systems that do not. Second, we present a new framework on how to construct tests that can detect if a system has achieved G.I. in a simple, comprehensive, and clear-cut fail/pass way. We call this novel framework the Turing test 2.0. We then demonstrate real-life examples of applying tests that follow our Turing test 2.0 framework on modern A.I. models.

The deployment of pre-trained perception models in novel environments often leads to performance degradation due to distributional shifts. Although recent artificial intelligence approaches for metacognition use logical rules to characterize and filter model errors, improving precision often comes at the cost of reduced recall. This paper addresses the hypothesis that leveraging multiple pre-trained models can mitigate this recall reduction. We formulate the challenge of identifying and managing conflicting predictions from various models as a consistency-based abduction problem, building on the idea of abductive learning (ABL) but applying it to test-time instead of training. The input predictions and the learned error detection rules derived from each model are encoded in a logic program. We then seek an abductive explanation--a subset of model predictions--that maximizes prediction coverage while ensuring the rate of logical inconsistencies (derived from domain constraints) remains below a specified threshold. We propose two algorithms for this knowledge representation task: an exact method based on Integer Programming (IP) and an efficient Heuristic Search (HS). Through extensive experiments on a simulated aerial imagery dataset featuring controlled, complex distributional shifts, we demonstrate that our abduction-based framework outperforms individual models and standard ensemble baselines, achieving, for instance, average relative improvements of approximately 13.6\% in F1-score and 16.6\% in accuracy across 15 diverse test datasets when compared to the best individual model. Our results validate the use of consistency-based abduction as an effective mechanism to robustly integrate knowledge from multiple imperfect models in challenging, novel scenarios.

Large Language Models (LLMs) are typically trained on data mixtures: most data come from web scrapes, while a small portion is curated from high-quality sources with dense domain-specific knowledge. In this paper, we show that when training LLMs on such data mixtures, knowledge acquisition from knowledge-dense datasets, unlike training exclusively on knowledge-dense data (arXiv:2404.05405), does not always follow a smooth scaling law but can exhibit phase transitions with respect to the mixing ratio and model size. Through controlled experiments on a synthetic biography dataset mixed with web-scraped data, we demonstrate that: (1) as we increase the model size to a critical value, the model suddenly transitions from memorizing very few to most of the biographies; (2) below a critical mixing ratio, the model memorizes almost nothing even with extensive training, but beyond this threshold, it rapidly memorizes more biographies. We attribute these phase transitions to a capacity allocation phenomenon: a model with bounded capacity must act like a knapsack problem solver to minimize the overall test loss, and the optimal allocation across datasets can change discontinuously as the model size or mixing ratio varies. We formalize this intuition in an information-theoretic framework and reveal that these phase transitions are predictable, with the critical mixing ratio following a power-law relationship with the model size. Our findings highlight a concrete case where a good mixing recipe for large models may not be optimal for small models, and vice versa.

Detecting vulnerabilities in source code remains a critical yet challenging task, especially when benign and vulnerable functions share significant similarities. In this work, we introduce VulTrial, a courtroom-inspired multi-agent framework designed to identify vulnerable code and to provide explanations. It employs four role-specific agents, which are security researcher, code author, moderator, and review board. Using GPT-4o as the base LLM, VulTrial almost doubles the efficacy of prior best-performing baselines. Additionally, we show that role-specific instruction tuning with small quantities of data significantly further boosts VulTrial's efficacy. Our extensive experiments demonstrate the efficacy of VulTrial across different LLMs, including an open-source, in-house-deployable model (LLaMA-3.1-8B), as well as the high quality of its generated explanations and its ability to uncover multiple confirmed zero-day vulnerabilities in the wild.

The visually impaired population faces significant challenges in daily activities. While prior works employ vision language models for assistance, most focus on static content and cannot address real-time perception needs in complex environments. Recent VideoLLMs enable real-time vision and speech interaction, offering promising potential for assistive tasks. In this work, we conduct the first study evaluating their effectiveness in supporting daily life for visually impaired individuals. We first conducted a user survey with visually impaired participants to design the benchmark VisAssistDaily for daily life evaluation. Using VisAssistDaily, we evaluate popular VideoLLMs and find GPT-4o achieves the highest task success rate. We further conduct a user study to reveal concerns about hazard perception. To address this, we propose SafeVid, an environment-awareness dataset, and fine-tune VITA-1.5, improving risk recognition accuracy from 25.00% to 76.00%.We hope this work provides valuable insights and inspiration for future research in this field.

Harmful algal blooms are a growing threat to inland water quality and public health worldwide, creating an urgent need for e fficient, accurate, and cost-e ff ective detection methods. This research introduces a high-performing methodology that integrates multiple open-source remote sensing data with advanced artificial intelligence models. Key data sources include Copernicus Sentinel-2 optical imagery, the Copernicus Digital Elevation Model (DEM), and NOAA's High-Resolution Rapid Refresh (HRRR) climate data, all e ffi ciently retrieved using platforms like Google Earth Engine (GEE) and Microsoft Planetary Computer (MPC). The NIR and two SWIR bands from Sentinel-2, the altitude from the elevation model, the temperature and wind from NOAA as well as the longitude and latitude were the most important features. The approach combines two types of machine learning models--tree-based models and a neural network--into an ensemble for classifying algal bloom severity. While the tree models performed strongly on their own, incorporating a neural network added robustness and demonstrated how deep learning models can e ff ectively use diverse remote sensing inputs. The method leverages high-resolution satellite imagery and AI-driven analysis to monitor algal blooms dynamically, and although initially developed for a NASA competition in the U.S., it shows potential for global application. Keywords: Machine learning; Inland Water; Algal Bloom; Remote Sensing; Data Fusion; Water Quality 1. Introduction Algal blooms are becoming the greatest inland water quality threat to public health and aquatic ecosystems that can degrade water quality to a greater extent than many chemicals (Brooks et al., 2016). Human nutrient loading and climate change (warming, altered rainfall) synergistically enhance cyanobacterial blooms in aquatic ecosystems (Paerl and Paul, 2012). Excessive nutrient loads in many cases comes from agricultural, industrial and other sources (Novotny, 2011). Phenology and trends of chlorophyll-a and cyanobacterial blooms are established (Matthews, 2014).

--Anomaly Detection (AD) defines the task of identifying observations or events that deviate from typical - or normal - patterns, a critical capability in IT security for recognizing incidents such as system misconfigurations, malware infections, or cyberattacks. In enterprise environments like SAP HANA Cloud systems, this task often involves monitoring high-dimensional, multivariate time series (MTS) derived from telemetry and log data. One approach is the Quantum Autoencoder (QAE), an emerging and promising method with potential for application in both data compression and AD. However, prior applications of QAEs to time series AD have been restricted to univariate data, limiting their relevance for real-world enterprise systems. In this work, we introduce a novel QAE-based framework designed specifically for MTS AD towards enterprise scale. We theoretically develop and experimentally validate the architecture, demonstrating that our QAE achieves performance competitive with neural-network-based autoencoders while requiring fewer trainable parameters. We evaluate our model on datasets that closely reflect SAP system telemetry and show that the proposed QAE is a viable and efficient alternative for semisupervised AD in real-world enterprise settings. Anomaly Detection (AD) refers to the process of identifying patterns or events that deviate from typical - or normal - behavior [1]. It plays a critical role in IT security and many other domains, as anomalies often correspond to potential security breaches, frauds, or system failures [2], [3]. Modern enterprise infrastructure, such as SAP HANA Cloud and other large scale cloud native applications, rely on continuous monitoring to ensure optimal performance, availability, and reliability. With increasing system complexity and scale, observability platforms generate large volumes of telemetry data, including structured multivariate time series (MTS) and unstructured log streams.