https://arxiv.org/api/O2QJ8CSDAprpc/BAud5eFyB8YWU 2026-03-22T10:02:05Z 104601 30 15 http://arxiv.org/abs/2603.17759v2 2026-03-19T15:34:50Z

Dark humor often relies on subtle cultural nuances and implicit cues that require contextual reasoning to interpret, posing safety challenges that current static benchmarks fail to capture. To address this, we introduce a novel multimodal, multilingual benchmark for detecting and understanding harmful and offensive humor. Our manually curated dataset comprises 3,000 texts and 6,000 images in English and Arabic, alongside 1,200 videos that span English, Arabic, and language-independent (universal) contexts. Unlike standard toxicity datasets, we enforce a strict annotation guideline: distinguishing Safe jokes from Harmful ones, with the latter further classified into Explicit (overt) and Implicit (Covert) categories to probe deep reasoning. We systematically evaluate state-of-the-art (SOTA) open and closed-source models across all modalities. Our findings reveal that closed-source models significantly outperform open-source ones, with a notable difference in performance between the English and Arabic languages in both, underscoring the critical need for culturally grounded, reasoning-aware safety alignment. Warning: this paper contains example data that may be offensive, harmful, or biased.

2026-03-18T14:21:10Z Ahmed Sharshar Hosam Elgendy Saad El Dine Ahmed Yasser Rohaim Yuxia Wang http://arxiv.org/abs/2603.19017v1 2026-03-19T15:22:03Z

We present MultiTempBench, a multilingual temporal reasoning benchmark spanning three tasks, date arithmetic, time zone conversion, and temporal relation extraction across five languages (English, German, Chinese, Arabic, and Hausa) and multiple calendar conventions (Gregorian, Hijri, and Chinese Lunar). MultiTempBench contains $15,000$ examples built by translating $750$ curated English questions and expanding each into controlled date-format variants. We evaluate 20 LLMs and introduce the multilingual Date Fragmentation Ratio (mDFR), calibrated with human severity ratings, together with geometric-probing analyses of internal temporal representations. We find tokenisation quality of temporal artefacts is a resource-dependent bottleneck: in low-resource languages and rarer calendar formats, fragmentation disrupts Year/Month/Day separation and accuracy collapses, while high-resource settings are often robust to digit-level splitting. Beyond tokenisation, crossed mixed-effects regression shows that temporal linearity is the strongest predictor of temporal reasoning in high-resource languages, whereas fragmentation is the stronger predictor in low-resource languages. Code is available at: https://github.com/gagan3012/mtb

2026-03-19T15:22:03Z Gagan Bhatia Ahmad Muhammad Isa Maxime Peyrard Wei Zhao http://arxiv.org/abs/2603.16128v2 2026-03-19T15:18:31Z

As autonomous LLM-based agents increasingly populate social platforms, understanding the dynamics of AI-agent communities becomes essential for both communication research and platform governance. We present the first large-scale empirical comparison of AI-agent and human online communities, analyzing 73,899 Moltbook and 189,838 Reddit posts across five matched communities. Structurally, we find that Moltbook exhibits extreme participation inequality (Gini = 0.84 vs. 0.47) and high cross-community author overlap (33.8\% vs. 0.5\%). In terms of linguistic attributes, content generated by AI-agents is emotionally flattened, cognitively shifted toward assertion over exploration, and socially detached. These differences give rise to apparent community-level homogenization, but we show this is primarily a structural artifact of shared authorship. At the author level, individual agents are more identifiable than human users, driven by outlier stylistic profiles amplified by their extreme posting volume. As AI-mediated communication reshapes online discourse, our work offers an empirical foundation for understanding how multi-agent interaction gives rise to collective communication dynamics distinct from those of human communities.

2026-03-17T05:26:55Z Preprint: 13 pages, 4 figures, 5 tables Agam Goyal Olivia Pal Hari Sundaram Eshwar Chandrasekharan Koustuv Saha http://arxiv.org/abs/2603.19008v1 2026-03-19T15:15:58Z

Retrieval-Augmented Generation (RAG) improves Large Language Models (LLMs) by grounding generation in external, non-parametric knowledge. However, when a task requires choosing among competing options, simply grounding generation in broadly relevant context is often insufficient to drive the final decision. Existing RAG methods typically rely on a single initial query, which often favors topical relevance over decision-relevant evidence, and therefore retrieves background information that can fail to discriminate among answer options. To address this issue, here we propose Hypothesis-Conditioned Query Rewriting (HCQR), a training-free pre-retrieval framework that reorients RAG from topic-oriented retrieval to evidence-oriented retrieval. HCQR first derives a lightweight working hypothesis from the input question and candidate options, and then rewrites retrieval into three targeted queries that seek evidence to: (1) support the hypothesis, (2) distinguish it from competing alternatives, and (3) verify salient clues in the question. This approach enables context retrieval that is more directly aligned with answer selection, allowing the generator to confirm or overturn the initial hypothesis based on the retrieved evidence. Experiments on MedQA and MMLU-Med show that HCQR consistently outperforms single-query RAG and re-rank/filter baselines, improving average accuracy over Simple RAG by 5.9 and 3.6 points, respectively. Code is available at https://anonymous.4open.science/r/HCQR-1C2E.

2026-03-19T15:15:58Z Hangeol Chang Changsun Lee Seungjoon Rho Junho Yeo Jong Chul Ye http://arxiv.org/abs/2603.19002v1 2026-03-19T15:06:12Z

Simulation of surveys using LLMs is emerging as a powerful application for generating human-like responses at scale. Prior work evaluates survey simulation using metrics borrowed from other domains, which are often ad hoc, fragmented, and non-standardized, leading to results that are difficult to compare. Moreover, existing metrics focus mainly on accuracy or distributional measures, overlooking the critical dimension of ranking alignment. In practice, a simulation can achieve high accuracy while still failing to capture the option most preferred by humans - a distinction that is critical in decision-making applications. We introduce RADIUS, a comprehensive two-dimensional alignment suite for survey simulation that captures: 1) RAnking alignment and 2) DIstribUtion alignment, each complemented by statistical Significance testing. RADIUS highlights the limitations of existing metrics, enables more meaningful evaluation of survey simulation, and provides an open-source implementation for reproducible and comparable assessment.

2026-03-19T15:06:12Z Weronika Łajewska Paul Missault George Davidson Saab Mansour http://arxiv.org/abs/2510.09255v4 2026-03-19T14:52:41Z

Enhancing LLMs with the ability to actively search external knowledge is crucial for complex and real-world tasks. Current approaches either rely on prompting to elicit the model's innate agent capabilities, or suffer from performance ceilings and collapse when applying RL to complex interactive tasks, leaving their true agentic potential untapped. To address this, we introduce \textbf{D}ynamic-filter \textbf{S}equence-level \textbf{P}olicy \textbf{O}ptimization (DSPO), an improved RL algorithm designed for robust agent training through sequence-level optimization and dynamic sample filtering. We train our model purely through RL to interleave multi-turn search and reasoning, obviating the need for supervised demonstration data. Across multiple QA benchmarks, our 7B model improves over a comparable previous work by \textbf{34.1\%}, and even outperforms the 14B model from previous work in complex multihop QA such as HotpotQA by nearly \textbf{9\% relative}, maintaining exceptional training stability.

2025-10-10T10:53:25Z Chenyang Gu Yewen Pu Bruce Yang Xiaofan Li Huan Gao http://arxiv.org/abs/2602.02290v2 2026-03-19T14:20:45Z

Generative AI can turn scientific articles into narratives for diverse audiences, but evaluating these stories remains challenging. Storytelling demands abstraction, simplification, and pedagogical creativity-qualities that are not often well-captured by standard summarization metrics. Meanwhile, factual hallucinations are critical in scientific contexts, yet, detectors often misclassify legitimate narrative reformulations or prove unstable when creativity is involved. In this work, we propose StoryScore, a composite metric for evaluating AI-generated scientific stories. StoryScore integrates semantic alignment, lexical grounding, narrative control, structural fidelity, redundancy avoidance, and entity-level hallucination detection into a unified framework. Our analysis also reveals why many hallucination detection methods fail to distinguish pedagogical creativity from factual errors, highlighting a key limitation: while automatic metrics can effectively assess semantic similarity with original content, they struggle to evaluate how it is narrated and controlled.

2026-02-02T16:29:32Z Proceedings of the Text2Story'26 Workshop, Delft (The Netherlands), 29-March-2026 Alex Argese Pasquale Lisena Raphaël Troncy http://arxiv.org/abs/2603.18945v1 2026-03-19T14:20:38Z

NLP+CSS work has operationalized ideology almost exclusively on a left/right partisan axis. This approach obscures the fact that people hold interpretations of many different complex and more specific ideologies on issues like race, climate, and gender. We introduce a framework that understands ideology as an attributed, multi-level socio-cognitive concept network, and explains how ideology manifests in discourse in relation to other relevant social processes like framing. We demonstrate how this framework can clarifies overlaps between existing NLP tasks (e.g. stance detection and natural language inference) and also how it reveals new research directions. Our work provides a unique and important bridge between computational methods and ideology theory, enabling richer analysis of social discourse in a way that benefits both fields.

2026-03-19T14:20:38Z Kenneth Joseph Kim Williams David Lazer http://arxiv.org/abs/2603.18940v1 2026-03-19T14:17:16Z

Chain-of-thought (CoT) reasoning improves LLM accuracy, yet detecting failures cheaply remains elusive. We study whether the shape of uncertainty dynamics across reasoning steps--captured by sampling a few answer completions per step--predicts correctness. We introduce entropy-trajectory monotonicity: a chain is monotone if its per-step answer-distribution entropy decreases at every step. On GSM8K (n=300) with Qwen2.5-7B-Instruct, monotone chains achieve 68.8% accuracy vs. 46.8% for non-monotone chains (+21.9 pp; Fisher's p=0.0005; OR=2.50). Critically, total entropy reduction is not predictive ($ρ$=-0.06, p=0.31), revealing a shape-over-magnitude dissociation: whether entropy decreases at every step matters, not how much. Violation count 0/1/2 gives 68.8%/50.8%/28.6% accuracy. Token log-probability confidence worsens in calibration with step depth (ECE: 0.186->0.312), and monotonicity achieves +5.8 pp at 73.7% coverage, outperforming scalar baselines at approx 1,500 tokens/question--1/8 the cost of 40-chain self-consistency. Results replicate on Mistral-7B (n=300): monotone chains reach 72.3% vs. 37.6% (+34.7 pp; OR=4.33). Structural properties of uncertainty trajectories are thus more informative than aggregate measures.

2026-03-19T14:17:16Z Xinghao Zhao http://arxiv.org/abs/2603.18911v1 2026-03-19T13:48:19Z

Knowledge-grounded dialogue systems aim to generate informative, contextually relevant responses by conditioning on external knowledge sources. However, most existing approaches focus exclusively on English, lack explicit citation mechanisms for verifying factual claims, and offer limited transparency into model decision-making. We present XKD-Dial, a progressive four-stage training pipeline for explainable, knowledge-grounded dialogue generation in a bilingual (English-Hindi) setting, comprising: (1) multilingual adaptation, (2) English dialogue SFT with citation grounding, (3) bilingual dialogue SFT, and (4) GRPO alignment with citation-aware rewards. We evaluate six models spanning encoder-decoder (250M-3B) and decoder-only (1B-7B) architectures at every pipeline stage. Our key contributions are: (i) three post-hoc explainability analyses - cross-attention alignment, Integrated Gradients attribution, and occlusion-based causal grounding - applied systematically across the training trajectory to reveal how citation behaviour is learned, not only whether it is learned; (ii) citation-grounded SFT reduces hallucination to 0.0% for encoder-decoder models from Stage 2 onward; (iii) the progressive pipeline prevents catastrophic forgetting while improving Hindi capabilities; (iv) smaller models match larger models on English after SFT; and (v) GRPO provides marginal improvement over well-designed SFT for structured citation tasks. We evaluate across six automatic metrics (BLEU, ROUGE, BERTScore, FactScore, Citation-F1, and hallucination rate).

2026-03-19T13:48:19Z 30 pages, 15 figures, 11 tables. Comprehensive study across 6 LLMs (250M-7B parameters) with explainability analysis. Code and data available upon request Vedant Pandya http://arxiv.org/abs/2603.17558v2 2026-03-19T13:47:01Z

Speech Large Language Models (Speech-LLMs) have emerged as a powerful approach for automatic speech recognition (ASR) by aligning speech encoders with large language models. However, adapting these systems to multilingual settings with imbalanced data distributions remains challenging. In such scenarios, a stability-plasticity dilemma often arises: fully shared Parameter-Efficient Fine-Tuning (PEFT) can cause negative inter-lingual interference for under-represented languages, while fully language-specific tuning limits the cross-lingual beneficial knowledge transfer needed for low-resource tasks. To address this, we propose Zipper-LoRA, a novel rank-level decoupling framework with three variants (Static, Hard, and Soft) that dynamically synthesizes LoRA updates from shared and language-specific subspaces. By using a lightweight language-conditioned router, Zipper-LoRA dynamically controls the contribution of each subspace at the LoRA rank level, enabling fine-grained sharing where languages are compatible and strict decoupling when conflicts occur. To further stabilize optimization under imbalanced data, we propose a two-stage training strategy with an Initial-B warm start that significantly accelerates convergence. Experiments on a 12-language mixed-resource setting show that Zipper-LoRA consistently outperforms both fully shared and independent baselines, particularly in extremely low-resource scenarios. Moreover, we demonstrate that these gains are robust across both chunked and non-chunked encoder configurations, confirming the framework's reliability for practical, large-scale multilingual ASR. Our code and data will be available at https://github.com/YuCeong-May/Zipper-LoRA for reproducibility.

2026-03-18T10:04:50Z 13 pages, 8 figures Yuxiang Mei Delai Qiu Shengping Liu Jiaen Liang Yanhua Long http://arxiv.org/abs/2603.18886v1 2026-03-19T13:27:12Z

The ability to precisely derive mathematical objects is a core requirement for downstream STEM applications, including mathematics, physics, and chemistry, where reasoning must culminate in formally structured expressions. Yet, current LM evaluations of mathematical and scientific reasoning rely heavily on simplified answer formats such as numerical values or multiple choice options due to the convenience of automated assessment. In this paper we provide three contributions for improving reasoning over mathematical objects: (i) we build and release training data and benchmarks for deriving mathematical objects, the Principia suite; (ii) we provide training recipes with strong LLM-judges and verifiers, where we show that on-policy judge training boosts performance; (iii) we show how on-policy training can also be used to scale test-time compute via aggregation. We find that strong LMs such as Qwen3-235B and o3 struggle on Principia, while our training recipes can bring significant improvements over different LLM backbones, while simultaneously improving results on existing numerical and MCQA tasks, demonstrating cross-format generalization of reasoning abilities.

2026-03-19T13:27:12Z Pranjal Aggarwal Marjan Ghazvininejad Seungone Kim Ilia Kulikov Jack Lanchantin Xian Li Tianjian Li Bo Liu Graham Neubig Anaelia Ovalle Swarnadeep Saha Sainbayar Sukhbaatar Sean Welleck Jason Weston Chenxi Whitehouse Adina Williams Jing Xu Ping Yu Weizhe Yuan Jingyu Zhang Wenting Zhao http://arxiv.org/abs/2603.18879v1 2026-03-19T13:23:34Z

Plain Language and Easy-to-Read formats in text simplification are essential for cognitive accessibility. Yet current automatic simplification and evaluation pipelines remain largely automated, metric-driven, and fail to reflect user comprehension or normative standards. This paper introduces a hybrid framework that explicitly integrates human participation into LLM-based accessible text generation. Human-in-the-Loop (HiTL) contributions guide adjustments during generation, while Human-on-the-Loop (HoTL) supervision ensures systematic post-generation review. Empirical evidence from user studies and annotated resources is operationalized into (i) checklists aligned with standards, (ii) Event-Condition-Action trigger rules for activating expert oversight, and (iii) accessibility Key Performance Indicators (KPIs). The framework shows how human-centered mechanisms can be encoded for evaluation and reused to provide structured feedback that improves model adaptation. By embedding the human role in both generation and supervision, it establishes a traceable, reproducible, and auditable process for creating and evaluating accessible texts. In doing so, it integrates explainability and ethical accountability as core design principles, contributing to more transparent and inclusive NLP systems.

2026-03-19T13:23:34Z Accepted at LREC 2026. To appear in the Proceedings of the 14th International Conference on Language Resources and Evaluation (LREC 2026) Lourdes Moreno Paloma Martínez http://arxiv.org/abs/2603.18873v1 2026-03-19T13:17:50Z

Popular language learning applications such as Duolingo use large language models (LLMs) to generate lessons for its users. Most lessons focus on general real-world scenarios such as greetings, ordering food, or asking directions, with limited support for profession-specific contexts. This gap can hinder learners from achieving professional-level fluency, which we define as the ability to communicate comfortably various work-related and domain-specific information in the target language. We surveyed five employees from a multinational company in the Philippines on their experiences with Duolingo. Results show that respondents encountered general scenarios more frequently than work-related ones, and that the former are relatable and effective in building foundational grammar, vocabulary, and cultural knowledge. The latter helps bridge the gap toward professional fluency as it contains domain-specific vocabulary. Each participant suggested lesson scenarios that diverge in contexts hen analyzed in aggregate. With this understanding, we propose that language learning applications should generate lessons that adapt to an individual's needs through personalized, domain specific lesson scenarios while maintaining foundational support through general, relatable lesson scenarios.

2026-03-19T13:17:50Z 5 pages,3 figures,presented at the 3rd HEAL Workshop at CHI 2026 Carlos Rafael Catalan Patricia Nicole Monderin Lheane Marie Dizon Gap Estrella Raymund John Sarmimento Marie Antoinette Patalagsa http://arxiv.org/abs/2603.18863v1 2026-03-19T13:10:39Z

Better cross-lingual alignment is often assumed to yield better cross-lingual transfer. However, explicit alignment techniques -- despite increasing embedding similarity -- frequently fail to improve token-level downstream performance. In this work, we show that this mismatch arises because alignment and downstream task objectives are largely orthogonal, and because the downstream benefits from alignment vary substantially across languages and task types. We analyze four XLM-R encoder models aligned on different language pairs and fine-tuned for either POS Tagging or Sentence Classification. Using representational analyses, including embedding distances, gradient similarities, and gradient magnitudes for both task and alignment losses, we find that: (1) embedding distances alone are unreliable predictors of improvements (or degradations) in task performance and (2) alignment and task gradients are often close to orthogonal, indicating that optimizing one objective may contribute little to optimizing the other. Taken together, our findings explain why ``better'' alignment often fails to translate into ``better'' cross-lingual transfer. Based on these insights, we provide practical guidelines for combining cross-lingual alignment with task-specific fine-tuning, highlighting the importance of careful loss selection.

2026-03-19T13:10:39Z Yana Veitsman Yihong Liu Hinrich Schütze