Image classification under strict time constraints requires a delicate balance between feature complexity and computational overhead. This paper presents an optimized ensemble methodology developed for the NAKABA competition, focusing on identifying structural destruction. We propose a hybrid architecture that leverages two distinct Convolutional Neural Networks (EfficientNetB0 and EfficientNetB3) as base feature extractors, coupled with a Support Vector Machine (SVM) functioning as a meta-classifier. Instead of standard probability averaging or processing high-dimensional embeddings directly, the Meta-SVM acts as a learned gating mechanism to optimally combine the low-dimensional probability predictions of the base models. This ensures robust performance without the latency of heavier deep learning architectures. Empirical results demonstrate the efficacy of this approach. The model achieved a validation accuracy of 0.884 and a weighted F1-score of 0.885, with a notable F1-score of 0.839 on the challenging ’destruction’ class. On the official NAKABA leaderboard test set, the ensemble maintained strong generalization, achieving an F1-score of 0.831 and an accuracy of 0.845, which secured the 12th position overall and proved the model’s high effectiveness within the competition’s strict operational constraints.

Internet memes have become a dominant and highly accessible medium for political discourse on social media. However, their multimodal nature—combining culturally specific visual symbols with code-mixed text—presents a significant challenge for automated content analysis, particularly in low-resource languages. In this study, we describe the system submitted by team RMS for the Multi-Level Political Meme Classification shared task at DravidianLangTech @ ACL 2026, focusing exclusively on the Tamil language track. We propose a robust late-fusion multimodal architecture that leverages a pre-trained ResNet-50 network for visual feature extraction and a Transformer-based model (MuRIL) for processing code-mixed Tamil text. The modalities are aligned using bidirectional cross-modal attention and combined using a Gated Multimodal Unit, allowing the model to dynamically weight the importance of visual versus textual cues. Our system ranked 11th on the official leaderboard with a macro-averaged F1-score of 0.7382. Through detailed error analysis, we demonstrate that while our gated fusion approach excels at identifying explicit trolling stances, it struggles with complex target resolution when visual and textual cues contradict.

Cross-target Arabic stance detection presents a unique challenge due to the rich morphological complexity, regional dialects, and heavy reliance on sarcasm within social media discourse. In this paper, we describe the system submitted by Team Axiom for Track 1 (Seen Targets) of the StanceEval-2026 shared task. Our primary contribution is a resource-efficient, Two-Stage Hybrid Cascade architecture. This system utilizes a fine-tuned MARBERTv2 encoder as a high-confidence gatekeeper, dynamically routing uncertain and complex tweets to a locally hosted, 4-bit quantized Large Language Model (Llama-3.1-8B-Instruct). To prevent label bias and ground the LLM’s reasoning, the fallback stage incorporates a dynamically balanced, class-specific TF-IDF sub-word retriever. Our official submission achieved an F_avg2 score of 0.6910 (Rank 18), demonstrating a measurable improvement over our standalone encoder baseline (0.6722).

Detecting hate speech in social media memes requires disentangling complex cross-modal interactions, a challenge magnified in low-resource and morphologically rich languages like Arabic. We present Team Axiom’s submission to the ArGuard 2026 Shared Task (Task A1) on binary hate meme classification. While conventional multimodal wisdom posits that text and visual features must be jointly fused, empirical evidence across non-English domains reveals that noisy visual backbones and cross-modal space misalignment frequently degrade language signals. We validate this hypothesis by fine-tuning a unimodal MARBERTv2 sequence classifier without external multimodal fusion. Our simple, well-regularized approach achieved a Macro-F1 of 0.770 and an accuracy of 0.818 on the official gold test set, securing 6th place on the global leaderboard and outperforming both the official AraBERTv2 baseline and our own complex multimodal fusion architectures.

Sentiment analysis of regional Arabic dialects presents a formidable challenge due to severe orthographic variation, morphological complexity, and a scarcity of annotated resources. To tackle these linguistic hurdles within Subtask-1 of the DialectSentEval 2026 Shared Task, team Axiom proposes a heavily regularized, dialect-aware ensemble framework combining two domain-complementary language models: MARBERTv2 and AraBERTv2. Our pipeline incorporates preprocessed dialect tags appended directly to input sequences, Stochastic Weight Averaging (SWA) for flat-minima convergence, and simple probability averaging across model outputs. Empirical results on the hidden test set demonstrate the robustness of this approach, with our system securing 2nd place overall and achieving an official Macro F1-Score of 0.8211.

Identifying sentence boundaries in Arabic text is a challenging token classification task due to extreme dialectal variations, the absence of rigid punctuation rules, and document lengths that frequently exceed the capacity of standard transformer models. In this paper, we describe Team Axiom’s submission to the Paragraph-Aware Open track of the AraSeg Shared Task 2026. We propose a highly regularized, dual-model ensemble combining MARBERTv2 and CAMeLBERT-MSA. To address the challenge of excessively long documents, we employ a sliding-window tokenization strategy. Furthermore, we adapt our base models to the task corpus using Domain-Adaptive Pretraining (DAPT). Our final system achieved a Macro-F1 score of 90.2% (Precision: 91.9%, Recall: 89.5%), securing 7th place in the shared task.

This paper describes the "Axiom" team’s submission to the AlexandriaX-2026 Shared Task, focusing on Subtask 3: MT Error Detection and Classification. We frame the task of predicting fine-grained error spans and their corresponding LQM categories as a joint token classification problem using a BIO tagging scheme. Evaluating multiple Arabic encoders, we find that dialect-specific pretraining significantly outperforms generative extraction baselines. To address severe label imbalance across error categories, we apply a square-root scaled inverse-frequency class-weighted loss. Furthermore, we introduce a confidence-thresholded decoding strategy at inference to reduce false-positive span over-prediction. Our final system achieved highly competitive performance, ranking 5th on the official leaderboard. We highlight that while overlap-based detection is robust, exact span matching remains a persistent challenge due to the non-standardized nature of dialectal Arabic orthography.