Large language models often succeed on one formulation of a problem while failing on an equivalent formulation. Whether these failures arise from distinct internal circuits or different activation states of a shared circuit remains unknown. Recent mechanistic interpretability studies suggest that arithmetic in LLMs emerges from a "bag of heuristics," encoded by a sparse set of MLP neurons that represent distinct arithmetic strategies. We investigate whether arithmetic heuristic neurons are form-invariant across symbolic arithmetic, natural language word problems, and Python code in three Llama-3 models. In each format, we identify arithmetic heuristic neurons using a two-stage pipeline combining attribution patching and activation patching. A compact set of neurons is shared across all three formats, and targeted interventions show this shared circuit is both necessary and sufficient for late-layer arithmetic computation. Transferring the shared neurons' activations from a successful execution in one format to a failed execution in another recovers most incorrect predictions, exceeding 97% for addition and subtraction, indicating that cross-format failures arise from activation states rather than distinct circuits. Moreover, shared neurons consistently belong to the same heuristic families across formats, demonstrating that arithmetic computation in LLMs is largely form-invariant at the neuron level
Event detection (ED) systems are typically evaluated on clean, curated text, leaving their robustness to real-world noise largely unexplored, particularly for low-resource languages such as Bangla. We introduce a generalized Bangla news event ontology and a benchmark comprising 9,979 annotated sentences across 40 event subtypes, spanning clean news text, real-world Automatic Speech Recognition (ASR) transcripts, and orthographically corrupted text. We systematically evaluate fine-tuned encoder-only models (BanglaBERT and XLM-R) alongside instruction-tuned decoder-only large language models (Llama 3 and Gemma 3). Our results reveal a clear architectural trade-off: encoder models achieve higher performance on clean text but degrade substantially under noise, whereas decoder-only LLMs are markedly more robust, particularly when event triggers are corrupted. We further show that embedding annotation guidelines during instruction tuning establishes a higher performance baseline on noisy text but yields inconsistent reductions in performance degradation across noisy conditions. Finally, model scaling consistently improves the robustness of decoder-only LLMs, while combined training on clean and noisy data serves as an effective regularization strategy that disproportionately benefits encoder architectures, significantly narrowing the robustness gap.
Large Language Models have demonstrated strong multilingual fluency, yet fluency alone does not guarantee socially appropriate language use. In high-context languages, communicative competence requires sensitivity to social hierarchy, relational roles, and interactional norms that are encoded directly in everyday language. Bangla exemplifies this challenge through its three-tiered pronominal system, kinship-based addressing, and culturally embedded social customs. We introduce BanglaSocialBench, the first benchmark designed to evaluate sociopragmatic competence in Bangla through context-dependent language use rather than factual recall. The benchmark spans three domains: Bangla Address Terms, Kinship Reasoning, and Social Customs, comprising 1,719 culturally grounded instances written and verified by native Bangla speakers. We evaluate twelve contemporary LLMs in a zero-shot setting and observe systematic patterns of cultural misalignment. Models frequently default to overly formal address forms, fail to recognize multiple socially acceptable address pronouns, and conflate kinship terminology across religious contexts. Our findings show that sociopragmatic failures are often structured and non-random; for example, inappropriate addressing choices concentrate heavily in downward-hierarchy (Elder$\rightarrow$Younger) and informal contexts. This reveals persistent limitations in how current LLMs infer and apply culturally appropriate language use in realistic Bangladeshi social interactions.
Pedestrian-vehicle crashes remain a major public health concern, yet many developing cities lack the detailed roadway inventories needed to assess how roadway environments influence crash severity. This study develops a data-driven framework that integrates visual and built-environment characteristics derived from publicly available geospatial data to predict pedestrian crash severity in resource-constrained settings. **Methods**: Police-reported pedestrian crash records from Dhaka, Bangladesh, were integrated with Google Street View imagery, Overture Maps, OpenStreetMap, and Meta sociodemographic data. Built-environment variables and visual indicators derived through semantic segmentation were extracted. Four machine learning models were developed for binary crash severity prediction. SHapley Additive exPlanations (SHAP) were used to interpret model predictions, and K-means clustering of SHAP values identified roadway environments with similar risk profiles. Findings: Extreme Gradient Boosting (XGBoost) achieved the highest predictive performance (F1 score: 79%). Network density and commercial, institutional, residential, and industrial land uses were associated with higher fatal crash risk. In contrast, recreational land use, greenery, signage, building visibility, Relative Wealth Index, and the density of schools, hospitals, and bus stops were associated with lower fatality risk. SHAP-based clustering identified three distinct roadway environments, highlighting substantial spatial heterogeneity in the effects of visual and built- environment factors on crash severity. **Novelty**: This study presents an interpretable framework that combines computer vision-derived visual indicators with open geospatial datasets to assess pedestrian crash severity without requiring detailed roadway inventories. The framework demonstrates that publicly available data can generate scalable, context-sensitive safety insights for data-scarce transportation systems. **Practical Applications**: The proposed framework enables transportation agencies to identify high-risk roadway environments and prioritize location-specific pedestrian safety interventions where conventional roadway inventory data are unavailable. The approach supports evidence-based infrastructure planning and resource allocation in rapidly urbanizing, data-scarce cities.