arXiv subjects

Kareem Darwish

Publications and source records attributed to Kareem Darwish.

At least 19 recordsLinked to original sources

arXiv2026

From RAG to Agentic RAG for Faithful Islamic Question Answering

Large Language Models (LLMs) are increasingly used for Islamic question answering, where ungrounded responses may carry serious religious consequences. Yet standard MCQ/MRC-style evaluations (MCQ: Multiple choice questions, MRC: Machine Reading Comprehension) do not capture key real-world failure modes, notably free-form hallucinations and the ability to abstain when evidence is insufficient. To address this gap, we introduce IslamicFaithQA, a 3,810-item bilingual (Arabic/English) generative benchmark with atomic single-gold answers, which enables direct measurement of hallucination and abstention. We additionally developed an end-to-end grounded Islamic modeling suite consisting of (i) 25K Arabic text-grounded SFT reasoning pairs, (ii) 5K bilingual preference samples for reward-guided alignment, and (iii) a verse-level Qur'an retrieval corpus of ~6k atomic verses (ayat). Building on these resources, we develop an agentic Quran-grounding framework (agentic RAG) that uses structured tool calls for iterative evidence seeking and answer revision. Experiments across Arabic-centric and multilingual LLMs show that retrieval improves correctness and that agentic RAG yields the largest gains beyond standard RAG, achieving state-of-the-art performance and stronger Arabic-English robustness even with a small model (i.e., Qwen3 4B). We made the datasets are publicly available. https://huggingface.co/datasets/QCRI/IslamicFaithQA

cs.CL↗

arXiv2026

Fanar-Sadiq: A Multi-Agent Architecture for Grounded Islamic QA

Large language models (LLMs) can answer religious knowledge queries fluently, yet they often hallucinate and misattribute sources, which is especially consequential in Islamic settings where users expect grounding in canonical texts (Qur'an and Hadith) and jurisprudential (fiqh) nuance. Retrieval-augmented generation (RAG) improves grounding, however, a single retrieve-then-generate pipeline is insufficient for diverse Islamic queries, including verbatim scripture, citation-grounded guidance, and rule-constrained computations such as zakat and inheritance. To address these challenges, we present Fanar-Sadiq, a bilingual Arabic-English Islamic QA system built on a multi-agent, tool-augmented architecture. It is a core component of the Fanar AI platform. Fanar-Sadiq routes Islamic queries to specialized modules within an agentic tool architecture. It supports intent-aware routing, retrieval-grounded fiqh answers with normalized citations and verification traces, exact verse lookup with quotation validation, and deterministic Sunni zakat and inheritance calculators with madhhab-sensitive branching. We evaluate the end-to-end system on public Islamic QA benchmarks and show strong effectiveness and efficiency. It is publicly accessible through an API and Web application and has received over 1.9M accesses in less than a year (https://api.fanar.qa/docs).

cs.CL↗

arXiv2026

NativQA Framework: Enabling LLMs and VLMs with Native, Local, and Everyday Knowledge

The rapid progress of large language models (LLMs) raises concerns about cultural bias, fairness, and performance in diverse languages and underrepresented regions. Addressing these gaps requires large-scale resources grounded in multilingual, local, and cultural contexts. We systematize and extend the earlier NativQA framework to multimodality by adding image, audio, and video support, enabling scalable construction of culturally and regionally aligned QA datasets in native languages. Given user-defined seed queries, the framework uses search engines to collect location-specific everyday information. We evaluate it across 39 locations in 24 countries and 7 languages, spanning extremely low-resource to high-resource settings, and collect over $\sim$300K text QA pairs, $\sim$312K images, and $\sim$29K videos with associated audio. The developed resources can be used for LLMs benchmarking and further fine-tuning. The framework has been made publicly available for the community (https://gitlab.com/nativqa/nativqa-framework). Demo video is available here: \href{https://shorturl.at/DAVn9}{https://shorturl.at/DAVn9}.

cs.CL↗

arXiv2026

Fanar 2.0: Arabic Generative AI Stack

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.

cs.CL↗

arXiv2026

More Data, Fewer Diacritics: Scaling Arabic TTS

Arabic Text-to-Speech (TTS) research has been hindered by the availability of both publicly available training data and accurate Arabic diacritization models. In this paper, we address the limitation by exploring Arabic TTS training on large automatically annotated data. Namely, we built a robust pipeline for collecting Arabic recordings and processing them automatically using voice activity detection, speech recognition, automatic diacritization, and noise filtering, resulting in around 4,000 hours of Arabic TTS training data. We then trained several robust TTS models with voice cloning using varying amounts of data, namely 100, 1,000, and 4,000 hours with and without diacritization. We show that though models trained on diacritized data are generally better, larger amounts of training data compensate for the lack of diacritics to a significant degree. We plan to release a public Arabic TTS model that works without the need for diacritization.

cs.CL↗

arXiv2025

Tool Calling for Arabic LLMs: Data Strategies and Instruction Tuning

Tool calling is a critical capability that allows Large Language Models (LLMs) to interact with external systems, significantly expanding their utility. However, research and resources for tool calling are predominantly English-centric, leaving a gap in our understanding of how to enable this functionality for other languages, such as Arabic. This paper investigates three key research questions: (1) the necessity of in-language (Arabic) tool-calling data versus relying on cross-lingual transfer, (2) the effect of general-purpose instruction tuning on tool-calling performance, and (3) the value of fine-tuning on specific, high-priority tools. To address these questions, we conduct extensive experiments using base and post-trained variants of an open-weight Arabic LLM. To enable this study, we bridge the resource gap by translating and adapting two open-source tool-calling datasets into Arabic. Our findings provide crucial insights into the optimal strategies for developing robust tool-augmented agents for Arabic.

cs.CL↗

arXiv2025

Sacred or Synthetic? Evaluating LLM Reliability and Abstention for Religious Questions

Despite the increasing usage of Large Language Models (LLMs) in answering questions in a variety of domains, their reliability and accuracy remain unexamined for a plethora of domains including the religious domains. In this paper, we introduce a novel benchmark FiqhQA focused on the LLM generated Islamic rulings explicitly categorized by the four major Sunni schools of thought, in both Arabic and English. Unlike prior work, which either overlooks the distinctions between religious school of thought or fails to evaluate abstention behavior, we assess LLMs not only on their accuracy but also on their ability to recognize when not to answer. Our zero-shot and abstention experiments reveal significant variation across LLMs, languages, and legal schools of thought. While GPT-4o outperforms all other models in accuracy, Gemini and Fanar demonstrate superior abstention behavior critical for minimizing confident incorrect answers. Notably, all models exhibit a performance drop in Arabic, highlighting the limitations in religious reasoning for languages other than English. To the best of our knowledge, this is the first study to benchmark the efficacy of LLMs for fine-grained Islamic school of thought specific ruling generation and to evaluate abstention for Islamic jurisprudence queries. Our findings underscore the need for task-specific evaluation and cautious deployment of LLMs in religious applications.

cs.CL↗

arXiv2025

BALSAM: A Platform for Benchmarking Arabic Large Language Models

The impressive advancement of Large Language Models (LLMs) in English has not been matched across all languages. In particular, LLM performance in Arabic lags behind, due to data scarcity, linguistic diversity of Arabic and its dialects, morphological complexity, etc. Progress is further hindered by the quality of Arabic benchmarks, which typically rely on static, publicly available data, lack comprehensive task coverage, or do not provide dedicated platforms with blind test sets. This makes it challenging to measure actual progress and to mitigate data contamination. Here, we aim to bridge these gaps. In particular, we introduce BALSAM, a comprehensive, community-driven benchmark aimed at advancing Arabic LLM development and evaluation. It includes 78 NLP tasks from 14 broad categories, with 52K examples divided into 37K test and 15K development, and a centralized, transparent platform for blind evaluation. We envision BALSAM as a unifying platform that sets standards and promotes collaborative research to advance Arabic LLM capabilities.

cs.CL↗

arXiv2025

Fanar: An Arabic-Centric Multimodal Generative AI Platform

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.

cs.CL↗

arXiv2024

Creating Arabic LLM Prompts at Scale

The debut of chatGPT and BARD has popularized instruction following text generation using LLMs, where a user can interrogate an LLM using natural language requests and obtain natural language answers that matches their requests. Training LLMs to respond in this manner requires a large number of worked out examples of user requests (aka prompts) with corresponding gold responses. In this paper, we introduce two methods for creating such prompts for Arabic cheaply and quickly. The first methods entails automatically translating existing prompt datasets from English, such as PromptSource and Super-NaturalInstructions, and then using machine translation quality estimation to retain high quality translations only. The second method involves creating natural language prompts on top of existing Arabic NLP datasets. Using these two methods we were able to create more than 67.4 million Arabic prompts that cover a variety of tasks including summarization, headline generation, grammar checking, open/closed question answering, creative writing, etc. We show that fine tuning an open 7 billion parameter large language model, namely base Qwen2 7B, enables it to outperform a state-of-the-art 70 billion parameter instruction tuned model, namely Llama3 70B, in handling Arabic prompts.

cs.CL↗

arXiv2024

An Automated End-to-End Open-Source Software for High-Quality Text-to-Speech Dataset Generation

Data availability is crucial for advancing artificial intelligence applications, including voice-based technologies. As content creation, particularly in social media, experiences increasing demand, translation and text-to-speech (TTS) technologies have become essential tools. Notably, the performance of these TTS technologies is highly dependent on the quality of the training data, emphasizing the mutual dependence of data availability and technological progress. This paper introduces an end-to-end tool to generate high-quality datasets for text-to-speech (TTS) models to address this critical need for high-quality data. The contributions of this work are manifold and include: the integration of language-specific phoneme distribution into sample selection, automation of the recording process, automated and human-in-the-loop quality assurance of recordings, and processing of recordings to meet specified formats. The proposed application aims to streamline the dataset creation process for TTS models through these features, thereby facilitating advancements in voice-based technologies.

eess.AS↗

arXiv2022

NatiQ: An End-to-end Text-to-Speech System for Arabic

NatiQ is end-to-end text-to-speech system for Arabic. Our speech synthesizer uses an encoder-decoder architecture with attention. We used both tacotron-based models (tacotron-1 and tacotron-2) and the faster transformer model for generating mel-spectrograms from characters. We concatenated Tacotron1 with the WaveRNN vocoder, Tacotron2 with the WaveGlow vocoder and ESPnet transformer with the parallel wavegan vocoder to synthesize waveforms from the spectrograms. We used in-house speech data for two voices: 1) neutral male "Hamza"- narrating general content and news, and 2) expressive female "Amina"- narrating children story books to train our models. Our best systems achieve an average Mean Opinion Score (MOS) of 4.21 and 4.40 for Amina and Hamza respectively. The objective evaluation of the systems using word and character error rate (WER and CER) as well as the response time measured by real-time factor favored the end-to-end architecture ESPnet. NatiQ demo is available on-line at https://tts.qcri.org

cs.CL↗

arXiv2022

Cross-lingual Emotion Detection

Emotion detection can provide us with a window into understanding human behavior. Due to the complex dynamics of human emotions, however, constructing annotated datasets to train automated models can be expensive. Thus, we explore the efficacy of cross-lingual approaches that would use data from a source language to build models for emotion detection in a target language. We compare three approaches, namely: i) using inherently multilingual models; ii) translating training data into the target language; and iii) using an automatically tagged parallel corpus. In our study, we consider English as the source language with Arabic and Spanish as target languages. We study the effectiveness of different classification models such as BERT and SVMs trained with different features. Our BERT-based monolingual models that are trained on target language data surpass state-of-the-art (SOTA) by 4% and 5% absolute Jaccard score for Arabic and Spanish respectively. Next, we show that using cross-lingual approaches with English data alone, we can achieve more than 90% and 80% relative effectiveness of the Arabic and Spanish BERT models respectively. Lastly, we use LIME to analyze the challenges of training cross-lingual models for different language pairs

cs.CL↗

arXiv2022

Embeddings-Based Clustering for Target Specific Stances: The Case of a Polarized Turkey

On June 24, 2018, Turkey conducted a highly consequential election in which the Turkish people elected their president and parliament in the first election under a new presidential system. During the election period, the Turkish people extensively shared their political opinions on Twitter. One aspect of polarization among the electorate was support for or opposition to the reelection of Recep Tayyip Erdoğan. In this paper, we present an unsupervised method for target-specific stance detection in a polarized setting, specifically Turkish politics, achieving 90% precision in identifying user stances, while maintaining more than 80% recall. The method involves representing users in an embedding space using Google's Convolutional Neural Network (CNN) based multilingual universal sentence encoder. The representations are then projected onto a lower dimensional space in a manner that reflects similarities and are consequently clustered. We show the effectiveness of our method in properly clustering users of divergent groups across multiple targets that include political figures, different groups, and parties. We perform our analysis on a large dataset of 108M Turkish election-related tweets along with the timeline tweets of 168k Turkish users, who authored 213M tweets. Given the resultant user stances, we are able to observe correlations between topics and compute topic polarization.

cs.SI↗

arXiv2021

Automatic Expansion and Retargeting of Arabic Offensive Language Training

Rampant use of offensive language on social media led to recent efforts on automatic identification of such language. Though offensive language has general characteristics, attacks on specific entities may exhibit distinct phenomena such as malicious alterations in the spelling of names. In this paper, we present a method for identifying entity specific offensive language. We employ two key insights, namely that replies on Twitter often imply opposition and some accounts are persistent in their offensiveness towards specific targets. Using our methodology, we are able to collect thousands of targeted offensive tweets. We show the efficacy of the approach on Arabic tweets with 13% and 79% relative F1-measure improvement in entity specific offensive language detection when using deep-learning based and support vector machine based classifiers respectively. Further, expanding the training set with automatically identified offensive tweets directed at multiple entities can improve F1-measure by 48%.

cs.CL↗

arXiv2021

A Panoramic Survey of Natural Language Processing in the Arab World

The term natural language refers to any system of symbolic communication (spoken, signed or written) without intentional human planning and design. This distinguishes natural languages such as Arabic and Japanese from artificially constructed languages such as Esperanto or Python. Natural language processing (NLP) is the sub-field of artificial intelligence (AI) focused on modeling natural languages to build applications such as speech recognition and synthesis, machine translation, optical character recognition (OCR), sentiment analysis (SA), question answering, dialogue systems, etc. NLP is a highly interdisciplinary field with connections to computer science, linguistics, cognitive science, psychology, mathematics and others. Some of the earliest AI applications were in NLP (e.g., machine translation); and the last decade (2010-2020) in particular has witnessed an incredible increase in quality, matched with a rise in public awareness, use, and expectations of what may have seemed like science fiction in the past. NLP researchers pride themselves on developing language independent models and tools that can be applied to all human languages, e.g. machine translation systems can be built for a variety of languages using the same basic mechanisms and models. However, the reality is that some languages do get more attention (e.g., English and Chinese) than others (e.g., Hindi and Swahili). Arabic, the primary language of the Arab world and the religious language of millions of non-Arab Muslims is somewhere in the middle of this continuum. Though Arabic NLP has many challenges, it has seen many successes and developments. Next we discuss Arabic's main challenges as a necessary background, and we present a brief history of Arabic NLP. We then survey a number of its research areas, and close with a critical discussion of the future of Arabic NLP.

cs.CL↗

arXiv2021

News Consumption in Time of Conflict: 2021 Palestinian-Israel War as an Example

This paper examines news consumption in response to a major polarizing event, and we use the May 2021 Israeli-Palestinian conflict as an example. We conduct a detailed analysis of the news consumption of more than eight thousand Twitter users who are either pro-Palestinian or pro-Israeli and authored more than 29 million tweets between January 1 and August 17, 2021. We identified the stance of users using unsupervised stance detection. We observe that users may consume more topically-related content from foreign and less popular sources, because, unlike popular sources, they may reaffirm their views, offer more extreme, hyper-partisan, or sensational content, or provide more in depth coverage of the event. The sudden popularity of such sources may not translate to longer-term or general popularity on other topics.

cs.SI↗

arXiv2021

Fighting the COVID-19 Infodemic: Modeling the Perspective of Journalists, Fact-Checkers, Social Media Platforms, Policy Makers, and the Society

With the emergence of the COVID-19 pandemic, the political and the medical aspects of disinformation merged as the problem got elevated to a whole new level to become the first global infodemic. Fighting this infodemic has been declared one of the most important focus areas of the World Health Organization, with dangers ranging from promoting fake cures, rumors, and conspiracy theories to spreading xenophobia and panic. Addressing the issue requires solving a number of challenging problems such as identifying messages containing claims, determining their check-worthiness and factuality, and their potential to do harm as well as the nature of that harm, to mention just a few. To address this gap, we release a large dataset of 16K manually annotated tweets for fine-grained disinformation analysis that (i) focuses on COVID-19, (ii) combines the perspectives and the interests of journalists, fact-checkers, social media platforms, policy makers, and society, and (iii) covers Arabic, Bulgarian, Dutch, and English. Finally, we show strong evaluation results using pretrained Transformers, thus confirming the practical utility of the dataset in monolingual vs. multilingual, and single task vs. multitask settings.

cs.CL↗