Mourad Ouzzani

FANAR TEAM, Ummar Abbas, Mohammad Shahmeer Ahmad et al.

We present Fanar 2.0, the second generation of Qatar's Arabic-centric Generative AI platform. Sovereignty is a first-class design principle: every component, from data pipelines to deployment infrastructure, was designed and operated entirely at QCRI, Hamad Bin Khalifa University. Fanar 2.0 is a story of resource-constrained excellence: the effort ran on 256 NVIDIA H100 GPUs, with Arabic having only ~0.5% of web data despite 400 million native speakers. Fanar 2.0 adopts a disciplined strategy of data quality over quantity, targeted continual pre-training, and model merging to achieve substantial gains within these constraints. At the core is Fanar-27B, continually pre-trained from a Gemma-3-27B backbone on a curated corpus of 120 billion high-quality tokens across three data recipes. Despite using 8x fewer pre-training tokens than Fanar 1.0, it delivers substantial benchmark improvements: Arabic knowledge (+9.1 pts), language (+7.3 pts), dialects (+3.5 pts), and English capability (+7.6 pts). Beyond the core LLM, Fanar 2.0 introduces a rich stack of new capabilities. FanarGuard is a state-of-the-art 4B bilingual moderation filter for Arabic safety and cultural alignment. The speech family Aura gains a long-form ASR model for hours-long audio. Oryx vision family adds Arabic-aware image and video understanding alongside culturally grounded image generation. An agentic tool-calling framework enables multi-step workflows. Fanar-Sadiq utilizes a multi-agent architecture for Islamic content. Fanar-Diwan provides classical Arabic poetry generation. FanarShaheen delivers LLM-powered bilingual translation. A redesigned multi-layer orchestrator coordinates all components through intent-aware routing and defense-in-depth safety validation. Taken together, Fanar 2.0 demonstrates that sovereign, resource-constrained AI development can produce systems competitive with those built at far greater scale.

Frédéric Ieng, Soror Sahri, Mourad Ouzzani et al.

Knowledge Graphs (KGs) are widely used to represent structured knowledge, yet their automatic construction, especially with Large Language Models (LLMs), often results in incomplete or noisy outputs. Knowledge Graph Completion (KGC) aims to infer and add missing triples, but most existing methods either rely on structural embeddings that overlook semantics or language models that ignore the graph's structure and depend on external sources. In this work, we present OMNIA, a two-stage approach that bridges structural and semantic reasoning for KGC. It first generates candidate triples by clustering semantically related entities and relations within the KG, then validates them through lightweight embedding filtering followed by LLM-based semantic validation. OMNIA performs on the internal KG, without external sources, and specifically targets implicit semantics that are most frequent in LLM-generated graphs. Extensive experiments on multiple datasets demonstrate that OMNIA significantly improves F1-score compared to traditional embedding-based models. These results highlight OMNIA's effectiveness and efficiency, as its clustering and filtering stages reduce both search space and validation cost while maintaining high-quality completion.

Zan Ahmad Naeem, Mohammad Shahmeer Ahmad, Mohamed Eltabakh et al.

Can foundation models (such as ChatGPT) clean your data? In this proposal, we demonstrate that indeed ChatGPT can assist in data cleaning by suggesting corrections for specific cells in a data table (scenario 1). However, ChatGPT may struggle with datasets it has never encountered before (e.g., local enterprise data) or when the user requires an explanation of the source of the suggested clean values. To address these issues, we developed a retrieval-based method that complements ChatGPT's power with a user-provided data lake. The data lake is first indexed, we then retrieve the top-k relevant tuples to the user's query tuple and finally leverage ChatGPT to infer the correct value (scenario 2). Nevertheless, sharing enterprise data with ChatGPT, an externally hosted model, might not be feasible for privacy reasons. To assist with this scenario, we developed a custom RoBERTa-based foundation model that can be locally deployed. By fine-tuning it on a small number of examples, it can effectively make value inferences based on the retrieved tuples (scenario 3). Our proposed system, RetClean, seamlessly supports all three scenarios and provides a user-friendly GUI that enables the VLDB audience to explore and experiment with the system.

Fanar Team, Ummar Abbas, Mohammad Shahmeer Ahmad et al.

We present Fanar, a platform for Arabic-centric multimodal generative AI systems, that supports language, speech and image generation tasks. At the heart of Fanar are Fanar Star and Fanar Prime, two highly capable Arabic Large Language Models (LLMs) that are best in the class on well established benchmarks for similar sized models. Fanar Star is a 7B (billion) parameter model that was trained from scratch on nearly 1 trillion clean and deduplicated Arabic, English and Code tokens. Fanar Prime is a 9B parameter model continually trained on the Gemma-2 9B base model on the same 1 trillion token set. Both models are concurrently deployed and designed to address different types of prompts transparently routed through a custom-built orchestrator. The Fanar platform provides many other capabilities including a customized Islamic Retrieval Augmented Generation (RAG) system for handling religious prompts, a Recency RAG for summarizing information about current or recent events that have occurred after the pre-training data cut-off date. The platform provides additional cognitive capabilities including in-house bilingual speech recognition that supports multiple Arabic dialects, voice and image generation that is fine-tuned to better reflect regional characteristics. Finally, Fanar provides an attribution service that can be used to verify the authenticity of fact based generated content. The design, development, and implementation of Fanar was entirely undertaken at Hamad Bin Khalifa University's Qatar Computing Research Institute (QCRI) and was sponsored by Qatar's Ministry of Communications and Information Technology to enable sovereign AI technology development.

Mohammad S. Ahmad, Zan A. Naeem, Michaël Aupetit et al.

Tabular data embedded within PDF files, web pages, and other document formats are prevalent across numerous sectors such as government, engineering, science, and business. These human-centric tables (HCTs) possess a unique combination of high business value, intricate layouts, limited operational power at scale, and sometimes serve as the only data source for critical insights. However, their complexity poses significant challenges to traditional data extraction, processing, and querying methods. While current solutions focus on transforming these tables into relational formats for SQL queries, they fall short in handling the diverse and complex layouts of HCTs and hence being amenable to querying. This paper describes HCT-QA, an extensive benchmark of HCTs, natural language queries, and related answers on thousands of tables. Our dataset includes 2,188 real-world HCTs with 9,835 QA pairs and 4,679 synthetic tables with 67.5K QA pairs. While HCTs can be potentially processed by different type of query engines, in this paper, we focus on Large Language Models as potential engines and assess their ability in processing and querying such tables.

Nan Tang, Ju Fan, Fangyi Li et al.

Can AI help automate human-easy but computer-hard data preparation tasks that burden data scientists, practitioners, and crowd workers? We answer this question by presenting RPT, a denoising auto-encoder for tuple-to-X models (X could be tuple, token, label, JSON, and so on). RPT is pre-trained for a tuple-to-tuple model by corrupting the input tuple and then learning a model to reconstruct the original tuple. It adopts a Transformer-based neural translation architecture that consists of a bidirectional encoder (similar to BERT) and a left-to-right autoregressive decoder (similar to GPT), leading to a generalization of both BERT and GPT. The pre-trained RPT can already support several common data preparation tasks such as data cleaning, auto-completion and schema matching. Better still, RPT can be fine-tuned on a wide range of data preparation tasks, such as value normalization, data transformation, data annotation, etc. To complement RPT, we also discuss several appealing techniques such as collaborative training and few-shot learning for entity resolution, and few-shot learning and NLP question-answering for information extraction. In addition, we identify a series of research opportunities to advance the field of data preparation.

Saravanan Thirumuruganathan, Shameem A Puthiya Parambath, Mourad Ouzzani et al.

Entity resolution (ER) is one of the fundamental problems in data integration, where machine learning (ML) based classifiers often provide the state-of-the-art results. Considerable human effort goes into feature engineering and training data creation. In this paper, we investigate a new problem: Given a dataset D_T for ER with limited or no training data, is it possible to train a good ML classifier on D_T by reusing and adapting the training data of dataset D_S from same or related domain? Our major contributions include (1) a distributed representation based approach to encode each tuple from diverse datasets into a standard feature space; (2) identification of common scenarios where the reuse of training data can be beneficial; and (3) five algorithms for handling each of the aforementioned scenarios. We have performed comprehensive experiments on 12 datasets from 5 different domains (publications, movies, songs, restaurants, and books). Our experiments show that our algorithms provide significant benefits such as providing superior performance for a fixed training data size.

Tanay Kumar Saha, Mourad Ouzzani, Hossam M. Hammady et al.

A major task in systematic reviews is abstract screening, i.e., excluding, often hundreds or thousand of, irrelevant citations returned from a database search based on titles and abstracts. Thus, a systematic review platform that can automate the abstract screening process is of huge importance. Several methods have been proposed for this task. However, it is very hard to clearly understand the applicability of these methods in a systematic review platform because of the following challenges: (1) the use of non-overlapping metrics for the evaluation of the proposed methods, (2) usage of features that are very hard to collect, (3) using a small set of reviews for the evaluation, and (4) no solid statistical testing or equivalence grouping of the methods. In this paper, we use feature representation that can be extracted per citation. We evaluate SVM-based methods (commonly used) on a large set of reviews ($61$) and metrics ($11$) to provide equivalence grouping of methods based on a solid statistical test. Our analysis also includes a strong variability of the metrics using $500$x$2$ cross validation. While some methods shine for different metrics and for different datasets, there is no single method that dominates the pack. Furthermore, we observe that in some cases relevant (included) citations can be found after screening only 15-20% of them via a certainty based sampling. A few included citations present outlying characteristics and can only be found after a very large number of screening steps. Finally, we present an ensemble algorithm for producing a $5$-star rating of citations based on their relevance. Such algorithm combines the best methods from our evaluation and through its $5$-star rating outputs a more easy-to-consume prediction.