Search arXiv⌕ Search

arXiv subjects

Usman Naseem

Publications and source records attributed to Usman Naseem.

At least 19 recordsLinked to original sources
arXiv2026

They Know Where It Is, But Not What It Says: A Bilingual Hallucination Benchmark of Hadith Recall in Open LLMs

Large language models are increasingly consulted for religious knowledge, including the recall of specific Hadith (recorded sayings and actions of the Prophet Muhammad, peace be upon him). Unlike open-ended generation, Hadith recall is a high-stakes, verifiable task: a reference (collection, chapter, and Hadith number) has exactly one correct text, and any deviation from it constitutes a fabrication that can be mistaken for authentic scripture. This study introduces a bilingual (English/Arabic) benchmark that probes five open-weight, locally-deployable language models (Mistral-7B, Gemma3-4B, Jais-7B, ALLaM-7B, and Phi4-mini-3.8B) across all six canonical Sunni Hadith collections (Sahih al-Bukhari, Sahih Muslim, Sunan Abi Da'ud, Jami' al-Tirmidhi, Sunan an-Nasa'i, Sunan Ibn Majah), sampling 500 Hadith per collection (3,000 per model per language, 30,000 responses total). Each model is given only the bibliographic reference and asked to reproduce the exact Hadith text, with no retrieval augmentation, and is scored on lexical fidelity (BLEU) and semantic fidelity (multilingual cosine similarity), with refusals flagged separately. Across all five models and both languages, 0% of 30,000 responses achieve both high lexical and high semantic fidelity, i.e., an exact, verifiable match to the source text. Arabic responses fare uniformly worse than English. These findings suggest that current open-weight LLMs, used without retrieval grounding, are unreliable and potentially dangerous sources for exact religious-text recall, and that deployment for Islamic scholarship should mandate retrieval-augmented generation over a verified corpus rather than parametric memory. The complete code, prompts, evaluation pipeline, and scored outputs are publicly available at https://github.com/dayam8696/Hadith-Hallucination-Benchmark.

cs.CL↗
arXiv2026

VISPA: Pluralistic Alignment via Automatic Value Selection and Activation

As large language models are increasingly used in high-stakes domains, it is essential that their outputs reflect not average} human preference, rather range of varying perspectives. Achieving such pluralism, however, remains challenging. Existing approaches consider limited values or rely on prompt-level interventions, lacking value control and representation. To address this, we introduce VISPA, a training-free pluralistic alignment framework, that enables direct control over value expression by dynamic selection and internal model activation steering. Across extensive empirical studies spanning multiple models and evaluation settings, we show VISPA is performant across all pluralistic alignment modes in healthcare and beyond. Further analysis reveals VISPA is adaptable with different steering initiations, model, and/or values. These results suggest that pluralistic alignment can be achieved through internal activation mechanisms, offering a scalable path toward language models that serves all.

cs.CL↗
arXiv2026

HUMAN-TCI: Hierarchical Multi-Stream Motion-Aware Network with Torso-Centered Interaction for Text-to-Motion Retrieval

Accurate retrieval of human motions is a crucial first step in text-guided human motion modeling and synthesis, as it selects semantically relevant sequences from large datasets and provides grounded references for downstream tasks. Retrieving motions from natural language descriptions remains challenging because sentences can describe multiple actions, overlapping movements, and intricate dependencies between body parts. Existing methods often focus on simple, single-action descriptions and typically process body parts independently or by merely concatenating features, without explicitly modeling how torso movements influence other parts. In addition, their processing pipelines often rely on computationally heavy models, introducing considerable overhead, particularly when modeling longer or more complex motion sequences. This limits learning discriminative motion-pattern representations, reducing retrieval accuracy, interpretability, and efficiency in practical applications. To address these limitations, we propose HUMAN-TCI, a Hierarchical Multi-Stream Motion-Aware Network for text-guided human motion retrieval. HUMAN-TCI employs a three-stream architecture that separately models upper-body, lower-body, and torso motions while explicitly capturing their interactions, allowing torso-related movements to influence the positioning and dynamics of other body parts. By incorporating tailored torso attention, our model effectively recognizes complex human motion patterns, captures fine-grained motion relationships and handles complex multi-action descriptions. Our framework supports retrieval for both simple, single-action sentences and long, compositional descriptions containing sequential or overlapping actions without relying on complex models.

cs.CV↗
arXiv2026

Beyond Retrieval: Joint Supervision and Multimodal Document Ranking for Textbook Question Answering

Textbook question answering (TQA) is a complex task, requiring the interpretation of complex multimodal context. Although recent advances have improved overall performance, they often encounter difficulties in educational settings where accurate semantic alignment and task-specific document retrieval are essential. In this paper, we propose a novel approach to multimodal textbook question answering by introducing a mechanism for enhancing semantic representations through multi-objective joint training. Our model, Joint Embedding Training With Ranking Supervision for Textbook Question Answering (JETRTQA), is a multimodal learning framework built on a retriever--generator architecture that uses a retrieval-augmented generation setup, in which a multimodal large language model generates answers. JETRTQA is designed to improve the relevance of retrieved documents in complex educational contexts. Unlike traditional direct scoring approaches, JETRTQA learns to refine the semantic representations of questions and documents through a supervised signal that combines pairwise ranking and implicit supervision derived from answers. We evaluate our method on the CK12-QA dataset and demonstrate that it significantly improves the discrimination between informative and irrelevant documents, even when they are long, complex, and multimodal. JETRTQA outperforms the previous state of the art, achieving a 2.4\% gain in accuracy on the validation set and 11.1\% on the test set.

cs.IR↗
arXiv2026

Beyond Shallow Alignment: How Post-Training Methods Determine Refusal Circuits And Steering Robustness

How do the methods used to train language models to refuse harmful requests shape how that refusal actually works inside the model? We compare three post-training methods - supervised fine-tuning, reasoning-augmented fine-tuning (training on reasoning chains that justify a safety decision), and preference optimization (ORPO) - across three architecturally distinct models (Llama-3.1-8B, Gemma-2-9B, Qwen3-8B). We find that training method, not just data, reshapes how refusal is computed internally: reasoning-augmented training consistently produces a distinct kind of refusal computation, visible across all three models, while architecture independently shapes internal structure and how reliably refusal can be steered. Most importantly, no method we study achieves all three properties we would want from safe alignment at once: refusal that isn't concentrated in a few fragile components, safety gains that don't cost general capability, and safety behavior correctable through small, targeted edits. We caution against treating current post-training methods as a solved, reliable defense, especially for security-critical use. Code and models are available in https://github.com/hoangcuongnguyen2001/Beyond-Shallow-Alignment.

cs.CL↗
arXiv2026

Modeling and Optimizing User Preferences in AI Copilots: A Comprehensive Survey and Taxonomy

AI copilots represent a new generation of AI-powered systems designed to assist users, particularly knowledge workers and developers, in complex, context-rich tasks. As these systems become more embedded in daily workflows, personalization has emerged as a critical factor for improving usability, effectiveness, and user satisfaction. Central to this personalization is preference optimization: the system's ability to detect, interpret, and align with individual user preferences. While prior work in intelligent assistants and optimization algorithms is extensive, their intersection within AI copilots remains underexplored. This survey addresses that gap by examining how user preferences are operationalized in AI copilots. We investigate how preference signals are sourced, modeled across different interaction stages, and refined through feedback loops. Building on a comprehensive literature review, we define the concept of an AI copilot and introduce a taxonomy of preference optimization techniques across pre-, mid-, and post-interaction phases. Each technique is evaluated in terms of advantages, limitations, and design implications. By consolidating fragmented efforts across AI personalization, human-AI interaction, and language model adaptation, this work offers both a unified conceptual foundation and a practical design perspective for building user-aligned, persona-aware AI copilots that support end-to-end adaptability and deployment.

cs.AI↗
arXiv2026

Over-Refusal and Representation Subspaces: A Mechanistic Analysis of Task-Conditioned Refusal in Aligned LLMs

Aligned language models that are trained to refuse harmful requests also exhibit over-refusal: they decline safe instructions that seemingly resemble harmful instructions. A natural approach is to ablate the global refusal direction, steering the hidden-state vectors away or towards the harmful-refusal examples, but this corrects over-refusal only incidentally while disrupting the broader refusal mechanism. In this work, we analyse the representational geometry of both refusal types to understand why this happens. We show that harmful-refusal directions are task-agnostic and can be captured by a single global vector, whereas over-refusal directions are task-dependent: they reside within the benign task-representation clusters, vary across tasks, and span a higher-dimensional subspace. Linear probing suggests that the two refusal types are representationally distinct from the early transformer layers. These findings provide a mechanistic explanation of why global direction ablation alone cannot address over-refusal, and establish that task-specific geometric interventions are necessary.

cs.CL↗
arXiv2026

Pak3H: Evaluating the Cost of Cultural Mismatch in LLM Alignment with a Human-Contextualized Urdu Benchmark

Large language models (LLMs) demonstrate strong Helpfulness, Harmlessness, and Honesty (3H) alignment in English-centric settings, but these gains transfer poorly to low-resource languages due to cultural mismatches. Existing multilingual 3H benchmarks rely predominantly on automated translation or LLM based synthesis, propagating source-language biases while sacrificing local relevance. To address this gap, we introduce Pak3H1, the first human-validated, culturally contextualized Urdu benchmark suite for 3H alignment, comprising PakAlpaca (helpfulness), PakBeaverTails (harmlessness), and PakTruthfulQA (honesty). Our multi-stage pipeline integrates manual cultural adaptation and dictionary-guided post editing to prioritize native speaker judgment, ensuring both semantic fidelity and contextual authenticity. Zero-shot evaluations across multiple open and proprietary LLM architectures reveal systematic cross-lingual alignment gaps: helpfulness win rates decline under localized contexts, harmlessness guardrails break down against regional safety risks, and composite honesty metrics degrade substantially due to localized factual constraints. These findings expose structural limitations in current alignment approaches, underscoring the necessity of human-guided localization for equitable multilingual evaluation.

cs.CL↗
arXiv2026

Divergent Paths to Depolarization: Dialogue Design Shapes the Intergroup Attitudinal Effects of AI-Assisted Political Argumentation

Structured argumentative dialogues where interlocutors deliberate on opposing political ideas are known to promote perspective-taking and reduce political polarization, but finding willing partners is difficult as Americans increasingly shun political discussions. AI dialogue partners offer a scalable framework for such open-mindedness exercises, but how the format of human-AI dialogues shapes their benefits remains unclear. This study seeks to fill the gap with a preregistered two-session online experiment with 527 US participants. As the primary experimental manipulation, participants were assigned to argue either for or against their pre-existing attitude on a contested political issue, engaging either with an AI chatbot or a solitary essay task. The AI conditions further varied in the chatbot's interaction style (adversarial or collaborative) and the presence of an additional financial incentive. The results show that attitude-congruent dialogues more strongly reduced polarization than attitude-incongruent dialogues immediately after the exchange. By contrast, an exploratory analysis suggested a delayed increase in cognitive empathy following attitude-incongruent dialogues, a pattern consistent with the account of sleeper effects. While the conversation style had little influence on the effects of attitude-congruent dialogues, a collaborative discussion tended to make attitude-incongruent dialogues more effective, narrowing the immediate effect gap. Additional financial incentives did not alter outcomes. Given the heterogeneity, the AI conditions were not universally more effective forming favorable intergroup attitudes in pooled comparisons between AI and non-AI conditions. The findings caution against a simplistic view of AI dialogues as a silver bullet for depolarization and highlight dialogue design as a key determinant of effective AI-mediated attitudinal interventions.

cs.CY↗
arXiv2026

Can Argus Judge Them All? Comparing VLMs Across Domains

Vision-Language Models (VLMs) are increasingly used in industry VLM applications such as retrieval systems, content generation platforms, and decision-support workflows, where model selection is commonly guided by benchmark rankings. These rankings are largely determined by retrieval, captioning, and reasoning downstream tasks; however, models with similar task performance often show substantially different behavior across datasets. This creates a Capability-Reliability Gap between benchmark performance and observed model stability. We present ARGUS-EVAL, a capability-reliability-oriented evaluation framework for VLMs that characterizes model behavior through Benchmark Capability P(M), Cross-Dataset Consistency CDC(M), Robustness Retention RR(M), and Efficiency E(M). We evaluate CLIP, BLIP, LXMERT, Gemma-3-4B, and Qwen-2.5VL-3B-Instruct across retrieval, captioning, and reasoning downstream tasks. The results reveal notable differences between capability-oriented and reliability-oriented rankings. Qwen-2.5VL-3BInstruct achieves the strongest overall capability (R@1 = 82.7%, BLEU-4 = 47.2%, CIDEr = 141.6, CDC = 0.91), whereas CLIP records the lowest latency (31 ms) and memory footprint (0.9 GB).

cs.IR↗
arXiv2026

SemEval-2026 Task 9: Detecting Multilingual, Multicultural and Multievent Online Polarization

We present SemEval-2026 Task 9, a shared task on online polarization detection, covering 22 languages and comprising over 110K annotated instances. Each data instance is multi-labeled with the presence of polarization, polarization type, and polarization manifestation. Participants were asked to predict labels in three sub-tasks: (1) detecting the presence of polarization, (2) identifying the type of polarization, and (3) recognizing the polarization manifestation. The three tasks attracted over 1,000 participants worldwide and more than 10k submission on Codabench. We received final submissions from 67 teams and 73 system description papers. We report the baseline results and analyze the performance of the best-performing systems, highlighting the most common approaches and the most effective methods across different subtasks and languages. The dataset of this task is publicly available.

cs.CL↗
arXiv2026

Health-ORSC-Bench: A Benchmark for Measuring Over-Refusal and Safety Completion in Health Context

Safety alignment in Large Language Models is critical for healthcare; however, reliance on binary refusal boundaries often results in over-refusal of benign queries or unsafe compliance with harmful ones. While existing benchmarks measure these extremes, they fail to evaluate Safe Completion: the model's ability to maximise helpfulness on dual-use or borderline queries by providing safe, high-level guidance without crossing into actionable harm. We introduce Health-ORSC-Bench, the first large-scale benchmark designed to systematically measure Over-Refusal and Safe Completion quality in healthcare. Comprising 31,920 benign boundary prompts across seven health categories (e.g., self-harm, medical misinformation), our framework uses an automated pipeline with human validation to test models at varying levels of intent ambiguity. We evaluate 30 state-of-the-art LLMs, including GPT-5 and Claude-4, revealing a significant tension: safety-optimised models frequently refuse up to 80% of "Hard" benign prompts, while domain-specific models often sacrifice safety for utility. Our findings demonstrate that model family and size significantly influence calibration: larger frontier models (e.g., GPT-5, Llama-4) exhibit "safety-pessimism" and higher over-refusal than smaller or MoE-based counterparts (e.g., Qwen-3-Next), highlighting that current LLMs struggle to balance refusal and compliance. Health-ORSC-Bench provides a rigorous standard for calibrating the next generation of medical AI assistants toward nuanced, safe, and helpful completions. Furthermore, our benchmark facilitates reproducible evaluation, encourages safety calibration, and supports development of clinically reliable, context-aware, human-aligned medical AI systems. Our code and data are available at: https://github.com/ZhihaoZhang97/Health-ORSC-Bench. Warning: Some contents may include toxic or undesired contents.

cs.AI↗
arXiv2026

Med-R2: Perception and Reflection-driven Complex Reasoning for Medical Report Generation

Automated medical report generation (MRG) is increasingly used to reduce the burden of manual reporting and for decision support. Large vision-language models (LVLMs) hold great promise for automated MRG due to their fine-grained image-text alignment and advanced text-generation capabilities. Currently, state-of-the-art MRGs primarily focus on adapting pre-trained LVLMs with direct supervised fine-tuning (SFT), a fine-tuning strategy with medical image-report pairs. However, several factors limit the performance of these LVLMs. Firstly, direct SFT enables LVLMs to generate medical reports directly without an intermediate thinking process of pathological feature perception and diagnostic reasoning. This causes a potential failure to perceive pathological features and thus leads to misdiagnosis. Secondly, direct SFT lacks the incorporation of radiology-specific knowledge guidance, causing LVLMs to misinterpret perceived pathological features and make incorrect diagnoses. To address these gaps, we propose a novel fine-tuning strategy named Med-R2. We introduce a perception-driven long reasoning process that precedes report generation and incorporates radiology-specific knowledge as guidance. Additionally, to alleviate potential perceptual errors in complex reasoning, a reflection mechanism is introduced to refine the perception of pathological features and the generated report. Our experiments demonstrate that Med-R2 effectively enhances the capability of pathological features perception and diagnosis accuracy for MRG via fine-tuned LVLMs.

cs.CL↗
arXiv2026

Beyond Uniform Forgetting: A Study of Sequential Direct Preference Optimization Across Preference Settings

Aligning language models with human preferences often requires optimising multiple behavioural objectives. A practical approach is to apply these objectives sequentially using preference optimisation methods such as Direct Preference Optimisation (DPO), but it remains unclear whether later training uniformly degrades preferences learned earlier or whether the effect depends on the relationship between objectives. We study sequential DPO across four preference settings covering distributional conflict, multi-attribute interaction, strong safety signal, and compatible response-quality objectives. Using Llama-3.1-8B-Instruct with LoRA adapters, we evaluate all objectives after every stage with a fixed base-model reference. We find that sequential DPO does not produce a single forgetting pattern; preference change ranges from partial degradation to stability, pair-level redistribution, or positive transfer depending on objective relationship, signal strength, and training order. Pair-level analysis using length-normalised policy margins shows that aggregate metrics can mask heterogeneous changes across preference pairs, whereas quartile decomposition reveals that high-confidence pairs can either degrade or improve depending on the setting. Mechanistic diagnostics show that Stage~2 gradients and adapter updates are near-orthogonal to the previous objective across all settings, providing little evidence that direct gradient opposition is the primary driver. These findings suggest that future sequential alignment pipelines should account for objective compatibility and signal strength, rather than assuming that later objectives affect earlier preferences uniformly.

cs.CL↗
arXiv2026

ChildGuard: A Specialized Dataset for Combatting Child-Targeted Hate Speech

Mental health industry faces growing concerns regarding hate speech directed at children's on social media, as exposure to such content can contribute to adverse psychological outcomes during critical stages of development. Current hate speech datasets and detection systems provide limited support for child-focused applications because they are primarily designed for adults and lack dedicated representations of age-specific characteristics associated with hate speech directed at children's. To address this gap, we introduce ChildGuard, a large-scale English dataset for child-targeted hate speech containing 351,877 annotated instances collected from X (formerly Twitter), Reddit, and YouTube. The dataset covers three age groups such as younger children's (under 11), pre-teens (11-12), and teens (13-17). ChildGuard contains two subsets such as a contextual subset (157K) and a lexical subset (194K). Evaluation using recent transformer-based models and LLMs achieves a best Macro-F1 of 82.07%, decreasing to 79.41%, 79.24%, 76.04%, and 74.88% on younger children's, contextual, implicit hate, and cross-subset settings, respectively.

cs.CL↗
arXiv2026

Benchmarking Large Language Models for Cryptanalysis and Side-Channel Vulnerabilities

Recent advancements in large language models (LLMs) have transformed natural language understanding and generation, leading to extensive benchmarking across diverse tasks. However, cryptanalysis - a critical area for data security and its connection to LLMs' generalization abilities - remains underexplored in LLM evaluations. To address this gap, we evaluate the cryptanalytic potential of state-of-the-art LLMs on ciphertexts produced by a range of cryptographic algorithms. We introduce a benchmark dataset of diverse plaintexts, spanning multiple domains, lengths, writing styles, and topics, paired with their encrypted versions. Using zero-shot and few-shot settings along with chain-of-thought prompting, we assess LLMs' decryption success rate and discuss their comprehension abilities. Our findings reveal key insights into LLMs' strengths and limitations in side-channel scenarios and raise concerns about their susceptibility to under-generalization-related attacks. This research highlights the dual-use nature of LLMs in security contexts and contributes to the ongoing discussion on AI safety and security.

cs.CL↗
arXiv2026

Truth, Trust, and Trouble: Medical AI on the Edge

Large Language Models (LLMs) hold significant promise for transforming digital health by enabling automated medical question answering. However, ensuring these models meet critical industry standards for factual accuracy, usefulness, and safety remains a challenge, especially for open-source solutions. We present a rigorous benchmarking framework using a dataset of over 1,000 health questions. We assess model performance across honesty, helpfulness, and harmlessness. Our results highlight trade-offs between factual reliability and safety among evaluated models -- Mistral-7B, BioMistral-7B-DARE, and AlpaCare-13B. AlpaCare-13B achieves the highest accuracy (91.7%) and harmlessness (0.92), while domain-specific tuning in BioMistral-7B-DARE boosts safety (0.90) despite its smaller scale. Few-shot prompting improves accuracy from 78% to 85%, and all models show reduced helpfulness on complex queries, highlighting ongoing challenges in clinical QA.

cs.CL↗
arXiv2026

CBT-Audio: Evaluating Audio Language Models for Patient-Side Distress Intensity Estimation in CBT Session Recordings

Cognitive behavioural therapy is widely used to help patients understand and manage psychological distress. It is often delivered through spoken conversation, where therapists attend not only to what patients say, but also to how they say it, because these cues can help therapists decide how to respond and adapt treatment. Progress in building AI systems for CBT remains largely limited to text, partly because most available datasets are text based and shareable spoken CBT data are scarce under ethical and privacy constraints. This creates a blind spot because text based models and evaluations cannot capture the mismatch between the transcript and the patient's voice, even though therapists often rely on this mismatch to understand patient distress. We introduce CBT-Audio, a dataset for evaluating patient distress estimation from spoken CBT sessions with audio language models. CBT-Audio contains 1,802 patient turns from 96 publicly available CBT recordings, with turn-level distress labels validated on an experts-annotated subset. We evaluate 10 open source audio language models under three input conditions, where models receive only patient audio, only the transcript, or both audio and transcript. Our results show that audio can provide useful information beyond text, especially when combined with transcripts. Adding audio to transcript input improves distress estimation over using the transcript alone in 8 of 10 model families, with significant gains in 4, and case studies show the clearest benefit when verbal content and vocal delivery diverge. CBT-Audio makes spoken patient behaviour measurable for AI evaluation in CBT-related tasks and supports future work on audio language models for mental health interaction.

cs.AI↗