We present PolaFusion, our system for SemEval-2026 Task 9, which requires detecting polarization in social media posts across 22 languages, classifying its type (Subtask 2), and identifying its rhetorical manifestation (Subtask 3). The task is characterized by severe and pervasive class imbalance across all three subtasks and all 22 languages. We address this through a combination of three strategies: a hierarchical gating architecture where a binary gatekeeper model gates two specialist classifiers trained exclusively on polarized content; an eight-model mega-ensemble combining fivefold mDeBERTa-v3-base and three-fold XLM-RoBERTa-large with soft-vote probability aggregation; and a Macro-F1-aware augmentation strategy using Qwen3-235B that generates synthetic minority-class examples only for language-label pairs that are both scarce and poorly learned. Throughout training, inverse-frequency class weighting within BCEWithLogitsLoss forces the model to attend proportionally to rare labels. Our system achieves official Macro-F1 scores of 0.800, 0.576, and 0.502 on Subtasks 1–3 respectively, outperforming the POLAR baseline by +0.040, +0.089, and +0.082 average Macro-F1 across languages. Our code is publicly available at https://github.com/Abdullah4152/PolaFuse.

Large Language Model (LLM) safety is inherently pluralistic, reflecting variations in moral norms, cultural expectations, and demographic contexts. Yet, existing alignment datasets such as Anthropic-HH and DICES rely on demographically narrow annotator pools, overlooking variation in safety perception across communities. Demo-SafetyBench addresses this gap by modeling demographic pluralism directly at the prompt level, decoupling value framing from responses. In Stage I, prompts from DICES are reclassified into 14 safety domains (adapted from BeaverTails) using Mistral-7B-Instruct-v0.3, retaining demographic metadata and expanding low-resource domains via Llama-3.1-8B-Instruct with SimHash-based deduplication, yielding 43,050 samples. In Stage II, pluralistic sensitivity is evaluated using LLMs-as-Raters—Gemma-7B, GPT-4o, and LLaMA-2-7B—under zero-shot inference. Balanced thresholds (delta = 0.5, tau = 10) achieve high reliability (ICC = 0.87) and low demographic sensitivity (DS = 0.12), confirming that pluralistic safety evaluation can be both scalable and demographically robust. Code and data available at: https://github.com/usmaann/Demo-SafetyBench

Generative AI—powered by Large Language Models (LLMs)—is increasingly deployed in industry across healthcare decision support, financial analytics, enterprise retrieval, and conversational automation, where reliability, efficiency, and cost control are critical. In such settings, models must satisfy strict constraints on energy, latency, and hardware utilization—not accuracy alone. Yet prevailing evaluation pipelines remain accuracy-centric, creating a Deployment–Evaluation Gap—the absence of operational and economic criteria in model assessment. To address this gap, we present EDGE-EVAL—a industry-oriented benchmarking framework that evaluates LLMs across their full lifecycle on legacy NVIDIA Tesla T4 GPUs. Benchmarking LLaMA and Qwen variants across three industrial tasks, we introduce five deployment metrics—Economic Break-Even (Nbreak), Intelligence-Per-Watt (IP W ), System Density (ρsys), Cold-Start Tax (Ctax), and Quantization Fidelity (Qret)—capturing profitability, energy efficiency, hardware scaling, serverless feasibility, and compression safety. Our results reveal a clear efficiency frontier—models in the < 2B parameter class dominate larger baselines across economic and ecological dimensions. LLaMA-3.2-1B (INT4) achieves ROI break-even in 14 requests (median), delivers 3× higher energy-normalized intelligence than 7B models, and exceeds 6,900 tokens/s/GB under 4-bit quantization. We further uncover an efficiency anomaly—while QLoRA reduces memory footprint, it increases adaptation energy by up to 7× for small models—challenging prevailing assumptions about quantization-aware training in edge deployment.

This study describes our submission to the CASE 2025 shared task on multimodal hate event detection, which focuses on hate detection, hate target identification, stance determination, and humour detection on text embedded images as classification challenges. Our submission contains entries in all of the subtasks. We propose FIMIF, a lightweight and efficient classification model that leverages frozen CLIP encoders. We utilise a feature interaction module that allows the model to exploit multiplicative interactions between features without any manual engineering. Our results demonstrate that the model achieves comparable or superior performance to larger models, despite having a significantly smaller parameter count