Large language models (LLMs) can generate fluent Arabic answers, yet factual errors remain difficult to detect, localize, explain, and verify. Existing hallucination benchmarks often provide response-level labels, with limited support for identifying the exact erroneous content, explaining why it is incorrect, or selecting the correct factual answer. We introduce HalluTruthQA, a fine-grained benchmark for hallucination evaluation in Arabic question answering. The benchmark contains 2,400 expert-curated examples across four knowledge-intensive domains: Islamic knowledge, history, science, and geography. Each example pairs an Arabic question and a model-generated answer with a verified reference answer, a binary hallucination label, and six candidate answers for factual verification. Hallucinated answers additionally include character-level erroneous spans, human-written explanations, and macro- and micro-level hallucination types. We evaluate four open-source LLMs, ALLaM-7B, Falcon-H1R-7B, Qwen3-32B, and SILMA, in a zero-shot setting across hallucination detection, span-level localization, factual verification, and explanation evaluation. Results show that these tasks capture different abilities: no single model performs best across all tasks. The best scores are 0.880 Macro-F1 for detection, 0.516 F1-Sp for localization, 0.852 LO-Score for factual verification, and 0.644 for explanation evaluation. These findings show that hallucination evaluation should move beyond response-level detection toward the localization, verification, and explanation of factual errors.
Large language models can generate fluent Arabic answers while introducing factual errors that are difficult to identify and verify. Existing Arabic hallucination resources often assign a binary label to an entire response, indicating whether it is hallucinated or non-hallucinated, but provide limited information about the exact erroneous content, the reason for the error, or the correct factual answer. We present HalluTruthQA-4K, an expanded version of the HalluTruthQA resource containing 4,000 expert-curated Arabic question-answering instances across four knowledge-intensive domains: Islamic knowledge, history, science, and geography. Serving as the official dataset for Track 2 of the HalluScoring 2026 shared task, HalluTruthQA-4K extends our original corpus to 4,000 instances. Each instance pairs an Arabic question with a model-generated response, a verified reference answer, and five plausible distractors. Hallucinated responses are additionally annotated with character-level erroneous spans, human-written explanations, and hierarchical hallucination types. The corpus contains 1,643 hallucinated and 2,357 non-hallucinated responses, with 1,843 annotated erroneous spans. We describe the resource construction and annotation methodology, including question selection, controlled answer generation, candidate construction, expert annotation, independent verification, adjudication, and quality control. We also document the annotation guidelines, taxonomy, data format, inter-annotator agreement, and corpus statistics. HalluTruthQA-4K provides a reusable resource for hallucination detection, span-level error localization, explanation generation, factual verification, and the broader evaluation of factual reliability in Arabic language models.
S. E. Bekhouche, Abdessalam Bouchekif, H. Telli et al.· 0 citations
Large language models can generate fluent responses to Islamic questions while introducing factual errors that are difficult to identify. This paper presents our system for \textsc{HalluScoring 2026} Task 2.1, \textit{Islamic Hallucination Detection and Find the Truth}. The task requires a unified two-step prediction: determining whether an Arabic answer generated by an LLM is hallucinated and selecting the verified answer from six closely related candidate options. We use the Islamic knowledge dataset provided by the shared task, which contains 600 question--answer instances, including 341 hallucinated and 259 non-hallucinated answers. Our system is based on the fine-tuned \texttt{google/gemma-4-12B-it} model and uses deterministic decoding during inference. The generated outputs are normalized to extract the hallucination label and the selected option. The system achieves a Macro-F1 score of 0.928 and a label accuracy of 0.935 for hallucination detection, together with an option accuracy of 0.895 for answer selection. These results yield a combined score of 0.912, demonstrating strong performance across both stages of the task. The lower option-selection accuracy indicates that distinguishing the verified answer from plausible alternatives remains more challenging than detecting hallucinated responses.
HallDetect, a lightweight, reference-free, and black-box framework for hallucination detection, is presented, a lightweight, reference-free, and black-box framework for hallucination detection that is evaluated not only on summarization but across a broader range of source-grounded generation settings.
Achir Oukelmoun, N. Semmar, Gäel de Chalendar· 0 citations
Fine-grained hallucination diagnosis for MLLMs is proposed, a new unified task that jointly performs hallucination detection, classification, and interpretable explanation generation and feedback experiments show that the fine-grained diagnostic explanations produced by the model effectively guide target models to correct their hallucinations.
Weilin Jin, Mingyu Wang, Wenbo Li et al.· 0 citations
Hallucination remains a critical challenge for developing trustworthy Multimodal Large Language Models (MLLMs). While existing benchmarks mainly focus on entity, attribute, and relation hallucinations, knowledge-related failures are often investigated separately, lacking a unified evaluation framework across different hallucination dimensions. To overcome this, we propose \textbf{KnowHal}, a benchmark that explicitly incorporates knowledge hallucination into multimodal hallucination evaluation spanning four dimensions: entity, attribute, relation, and knowledge. KnowHal constructs paired positive and negative questions over shared images and entities, enabling controlled comparisons among perceptual errors, knowledge-related errors, and false-premise acceptance. The benchmark contains 1,800 samples across 10 domains and 50 categories, constructed through a semi-automated pipeline combining LLM assistance, CLIP-based filtering, and human verification. We evaluate 14 representative MLLMs on KnowHal and conduct extensive analyses. Results show that the knowledge dimension consistently presents the greatest challenge for nearly all evaluated models, while most models exhibit substantial performance degradation on negative questions, revealing limited robustness to false premises. By unifying four hallucination dimensions with paired question design, KnowHal addresses an important gap in existing evaluation frameworks and enables a more comprehensive assessment of hallucinations in MLLMs.
Ruihan Li, Jiyang Tan, Kailin Jiang et al.· 0 citations
Large language models can produce fluent responses that are unsupported by the provided evidence or inconsistent with reference answers. Adapter-based fine-tuning offers a practical way to modify model behavior without updating all model parameters, but its effect on hallucination mitigation depends strongly on how improvement is measured. This paper presents TruthShield, a metric-aware trigger-guided QLoRA adapter training and evaluation pipeline for hallucination-aware language model adaptation. We construct hallucination-related failure cases from baseline evaluations, assign rule-based failure labels, generate counterexample triggers, and train a Mistral-7B adapter to respond to these trigger patterns. We then compare the baseline model and the adapter across TruthfulQA generation, TruthfulQA multiple choice, and HaluEval using two evaluation views: strict matching and an LLM judge. The adapter substantially improves strict scores across all tasks, increasing strict accuracy by 1.0000 on both TruthfulQA settings and by 0.6957 on HaluEval. However, these gains do not translate into judge-based improvement: judge scores remain nearly unchanged for TruthfulQA generation and decrease for TruthfulQA multiple choice and HaluEval. The results suggest that trigger-guided adapter training may learn surface-level response patterns without clear evidence of semantic hallucination mitigation under our current single-judge setting. This finding highlights the risk of overestimating reliability gains when strict metrics are used without complementary semantic evaluation.