arXiv subjects

Hanan Salam

Publications and source records attributed to Hanan Salam.

At least 19 recordsLinked to original sources

arXiv2026

AutoCRAT: Within-trajectory Joint Control of Stochasticity and Compute for LLM Reasoning

Large language models (LLMs) achieve strong reasoning performance, which depends critically on inference-time decisions. Yet these decisions are commonly handled by static, one-size-fits-all policies, limiting adaptation to diverse tasks and reasoning stages. Recent adaptive methods partially address this limitation, but they primarily adapt either decoding stochasticity (how the model explores) or reasoning compute (how long the model reasons) in isolation, leaving their interaction within a single reasoning trajectory unmodeled. To address this challenge, we shift toward a within-trajectory joint control view, and instantiate it in AutoCRAT, a decoder-side controller for frozen backbones. Using only signals available during decoding, AutoCRAT jointly adjusts sampling stochasticity and reasoning budget during generation. AutoCRAT operates over a discrete action space and updates control decisions only at semantic boundaries, improving stability while remaining responsive to the evolving reasoning process. Comprehensive evaluation across 6 benchmarks demonstrates that AutoCRAT (I) uses 13.8-52.7% fewer inference tokens on average than recommended static configurations, (II) surpasses recommended static and adaptive baselines by 1.5-4.5% in relative accuracy, and (III) enjoys strong cross-backbone transferability.

cs.AI↗

arXiv2026

MasDrift: Benchmarking Authorization Preservation Across Multi-Agent Architectures

Multi-agent systems (MAS) decompose long-horizon tasks across supervisors and subagents, but delegated goals do not necessarily carry their original authorization boundaries. Existing safety benchmarks mainly study adversarial compromise, while work on constraint drift lacks controlled architecture-level evaluation. We introduce MasDrift, a benchmark of 600 benign productivity tasks across eight domains. Each task pairs required work with reserved actions. MasDrift compares single-agent, centralized, and decentralized coordination while varying hierarchy depth and peer width, measuring task completion and authorization preservation. Across generic multi-agent conditions, centralized hierarchies achieve 93.9--98.6% task completion versus 85.7--87.0% for peer networks, while unauthorized actions occur in 2.7--19.8% of tasks versus 0.6--0.8%, a gap that widens with hierarchy depth. We further compare two defenses that differ in where authorization evidence resides. One re-anchors every pending call to the original user request. The other carries an attenuated policy along the delegation chain. Re-anchoring reduces unauthorized actions in every model configuration we evaluate, at a cost of 1.6 points of pooled completion. Chain propagation blocks required work instead, forfeiting up to 36.3 points. A heterogeneous case study confirms that the failure follows from coordination rather than model strength. MasDrift exposes a centralization tradeoff and makes authorization preservation a measurable property of MAS design.

cs.MA↗

arXiv2026

Four Years of GenAI: How Educators and Industry Adapted Their Assessment Strategies

GenAI's ability to solve a wide range of software engineering tasks is reshaping the software industry, raising the question of what an entry-level software engineer looks like today. In this work, we investigate whether the core skills for entry-level engineers have changed with GenAI and how the assessment of those skills has changed. Through two surveys, we consider both the educational and hiring perspectives, drawing on responses from 56 educators and 24 hiring professionals across diverse geographic regions to identify gaps and misaligned expectations. We find that educators frame their GenAI policies primarily around academic integrity, whereas GenAI use in industry is near-universal, with organizational policy governing how it should be used rather than whether to use it. Policies on GenAI use during interviews are still emerging, and respondents are split on whether they prefer candidates who demonstrate GenAI skills. Both populations are moving toward GenAI-resistant assessments built on observable real-time interactions and higher-order tasks, and both agree on which skills matter most, rating critical evaluation of AI-generated output, responsible and effective use of GenAI tools, and the ability to learn and adapt independently highly. However, they disagree on graduate readiness in these skills: hiring professionals report larger gaps, with the disagreement concentrated on foundational, non-GenAI-specific skills.

cs.SE↗

arXiv2026

AtelierEval: Agentic Evaluation of Humans & LLMs as Text-to-Image Prompters

Text-to-image (T2I) systems increasingly rely on upstream prompters, either humans or multimodal large language models (MLLMs), to translate user intent into detailed prompts. Yet current benchmarks fix the prompt and only evaluate T2I models, leaving the prompting proficiency of this upstream component entirely unmeasured. We introduce AtelierEval, the first unified benchmark that quantifies prompting proficiency across 360 expert-crafted tasks. Grounded in a cognitive view, it spans three task categories and instantiates tasks using a taxonomy of real-world challenges, with a dual interface for both humans and MLLMs. To enable scalable and reliable evaluation, we propose AtelierJudge, a skill-based, memory-augmented agentic evaluator. It produces subjective and objective scores for prompt-image pairs, achieving a Spearman correlation of 0.79 with human experts, approaching human performance. Extensive experiments benchmark 8 MLLMs against 48 human users across 4 T2I backends, validate AtelierEval as a robust diagnostic tool, and reveal the superiority of mimicry over planning, advocating for an image-augmented direction for future prompters. Our work is released to support future research.

cs.AI↗

arXiv2026

BiasIG: Benchmarking Multi-dimensional Social Biases in Text-to-Image Models

Text-to-Image (T2I) generative models have revolutionized content creation, yet they inherently risk amplifying societal biases. While sociological research provides systematic classifications of bias, existing T2I benchmarks largely conflate these nuances or focus narrowly on occupational stereotypes, leaving the multi-dimensional nature of generative bias inadequately measured. In this paper, we introduce BiasIG, a unified benchmark that quantifies social biases across a curated dataset of 47,040 prompts. Grounded in sociological and machine ethics frameworks, BiasIG disentangles biases across 4 dimensions to enable fine-grained diagnosis. To facilitate scalable and reliable evaluation, we propose a fully automated pipeline powered by a fine-tuned multi-modal large language model, achieving high alignment accuracy comparable to human experts. Extensive experiments on 8 T2I models and 3 debiasing methods not only validate BiasIG as a robust diagnostic tool, but also reveal critical insights: interventions on protected attributes often trigger unintended confounding effects on unrelated demographics, and debiasing methods exhibit a persistent tendency toward discrimination rather than mere ignorance. Our work advocates for a precise, taxonomy-driven approach to fairness in AIGC, providing a theoretical framework for using BiasIG's metrics as feedback signals in future closed-loop mitigation. The benchmark is openly available at https://github.com/Astarojth/BiasIG.

cs.CY↗

arXiv2026

GROW: A Conversational AI Coach for Goals, Reflection, Optimism, and Well-Being

College students face well-being challenges driven by academic pressure, financial strain, and social expectations. While campus counseling and student-success programs offer support, access is often limited by stigma, waitlists, and scheduling constraints. Existing digital tools focus on emotional check-ins or chatbots and may overlook structured goal setting and aligning goals with personal values. We present GROW, a goal-centered well-being coaching system that puts values-aligned goals at the center of the student experience. GROW combines the SMART framework with principles from Acceptance and Commitment Therapy in a conversational AI coach that helps students clarify aspirations, break them into concrete steps, and reflect on progress. The system links action plans with Google Calendar, sends reminders, and provides a dashboard that shows progress and engagement. We evaluated GROW through interviews with clinical psychologists, student-success staff, and faculty, followed by a one-week deployment with 30 undergraduates. Findings offer design implications for interactive systems that support engagement, accountability, and sense of purpose in higher education.

cs.HC↗

arXiv2026

PrefIx: Understand and Adapt to User Preference in Human-Agent Interaction

LLM-based agents can complete tasks correctly yet still frustrate users through poor interaction patterns, such as excessive confirmations, opaque reasoning, or misaligned pacing. Current benchmarks evaluate task accuracy but overlook how agents interact: whether they infer preferences from implicit cues, adapt dynamically, or maintain fine-grained interaction quality. We introduce Prefix, a configurable environment that evaluates both what agents accomplish and how they interact. Central to Prefix is the Interaction-as-a-Tool (IaaT) paradigm, which treats interaction behaviors as structured tool calls, unifying them with existing evaluation frameworks. We define 31 preference settings across 14 attributes and formalize user experience (UX) as a core metric alongside task accuracy. A composite LLM-as-a-Judge mechanism across seven UX dimensions achieves strong aggregate reliability (ICC > 0.79), high internal consistency (alpha = 0.943), and human correlation (rho = 0.52-0.78). Preference-aware agents show 7.6% average UX improvement and 18.5% gain in preference alignment. Our work is openly accessible.

cs.HC↗

arXiv2026

Informing Robot Wellbeing Coach Design through Longitudinal Analysis of Human-AI Dialogue

Social robots and conversational agents are being explored as supports for wellbeing, goal-setting, and everyday self-regulation. While prior work highlights their potential to motivate and guide users, much of the evidence relies on self-reported outcomes or short, researcher-mediated encounters. As a result, we know little about the interaction dynamics that unfold when people use such systems in real-world contexts, and how these dynamics should shape future robot wellbeing coaches. This paper addresses this gap through content analysis of 4352 messages exchanged longitudinally between 38 university students and an LLM-based wellbeing coach. Our results provide a fine-grained view into how users naturally shape, steer, and sometimes struggle within supportive human-AI dialogue, revealing patterns of user-led direction, guidance-seeking, and emotional expression. We discuss how these dynamics can inform the design of robot wellbeing coaches that support user autonomy, provide appropriate scaffolding, and uphold ethical boundaries in sustained wellbeing interactions.

cs.HC↗

arXiv2026

The Supportiveness-Safety Tradeoff in LLM Well-Being Agents

Large language models (LLMs) are being integrated into socially assistive robots (SARs) and other conversational agents providing mental health and well-being support. These agents are often designed to sound empathic and supportive in order to maximize user's engagement, yet it remains unclear how increasing the level of supportive framing in system prompts influences safety relevant behavior. We evaluated 6 LLMs across 3 system prompts with varying levels of supportiveness on 80 synthetic queries spanning 4 well-being domains (1440 responses). An LLM judge framework, validated against human ratings, assessed safety and care quality. Moderately supportive prompts improved empathy and constructive support while maintaining safety. In contrast, strongly validating prompts significantly degraded safety and, in some cases, care across all domains, with substantial variation across models. We discuss implications for prompt design, model selection, and domain specific safeguards in SARs deployment.

cs.HC↗

arXiv2025

Ethically-Aware Participatory Design of a Productivity Social Robot for College Students

College students often face academic and life stressors affecting productivity, especially students with Attention Deficit Hyperactivity Disorder (ADHD) who experience executive functioning challenges. Conventional productivity tools typically demand sustained self-discipline and consistent use, which many students struggle with, leading to disruptive app-switching behaviors. Socially Assistive Robots (SARs), known for their intuitive and interactive nature, offer promising potential to support productivity in academic environments, having been successfully utilized in domains like education, cognitive development, and mental health. To leverage SARs effectively in addressing student productivity, this study employed a Participatory Design (PD) approach, directly involving college students and a Student Success and Well-Being Coach in the design process. Through interviews and a collaborative workshop, we gathered detailed insights on productivity challenges and identified desirable features for a productivity-focused SAR. Importantly, ethical considerations were integrated from the onset, facilitating responsible and user-aligned design choices. Our contributions include comprehensive insights into student productivity challenges, SAR design preferences, and actionable recommendations for effective robot characteristics. Additionally, we present stakeholder-derived ethical guidelines to inform responsible future implementations of productivity-focused SARs in higher education.

cs.RO↗

arXiv2025

CentaurEval: Benchmarking Human-in-the-Loop Value in Agentic Coding

LLM-powered coding agents are reshaping the development paradigm. However, existing evaluation systems, neither traditional tests for humans nor benchmarks for LLMs, fail to capture this shift, excluding problems that require both human reasoning to guide solutions and AI efficiency for implementation. We introduce CentaurEval, a unified, ecologically valid benchmark for measuring human-in-the-loop value in coding. CentaurEval's core innovation is its "Collaboration-Necessary" problem templates, which are intractable for standalone LLMs or humans, but solvable through effective collaboration. CentaurEval dynamically instantiates tasks from 45 templates, providing a standardized IDE for humans and a reproducible 450-task toolkit for LLMs. We benchmark 45 participants against 5 LLMs under 4 levels of human intervention. Results show that while LLMs or humans alone achieve poor pass rates (0.67% and 18.89%), human-AI collaboration significantly improves to 31.11%. Our analysis reveals an emerging co-reasoning partnership, challenging the traditional human-tool hierarchy by showing that strategic breakthroughs can originate from either humans or AI.

cs.SE↗

arXiv2025

Supporting Productivity Skill Development in College Students through Social Robot Coaching: A Proof-of-Concept

College students often face academic challenges that hamper their productivity and well-being. Although self-help books and productivity apps are popular, they often fall short. Books provide generalized, non-interactive guidance, and apps are not inherently educational and can hinder the development of key organizational skills. Traditional productivity coaching offers personalized support, but is resource-intensive and difficult to scale. In this study, we present a proof-of-concept for a socially assistive robot (SAR) as an educational coach and a potential solution to the limitations of existing productivity tools and coaching approaches. The SAR delivers six different lessons on time management and task prioritization. Users interact via a chat interface, while the SAR responds through speech (with a toggle option). An integrated dashboard monitors progress, mood, engagement, confidence per lesson, and time spent per lesson. It also offers personalized productivity insights to foster reflection and self-awareness. We evaluated the system with 15 college students, achieving a System Usability Score of 79.2 and high ratings for overall experience and engagement. Our findings suggest that SAR-based productivity coaching can offer an effective and scalable solution to improve productivity among college students.

cs.RO↗

arXiv2025

T2IBias: Uncovering Societal Bias Encoded in the Latent Space of Text-to-Image Generative Models

Text-to-image (T2I) generative models are largely used in AI-powered real-world applications and value creation. However, their strategic deployment raises critical concerns for responsible AI management, particularly regarding the reproduction and amplification of race- and gender-related stereotypes that can undermine organizational ethics. In this work, we investigate whether such societal biases are systematically encoded within the pretrained latent spaces of state-of-the-art T2I models. We conduct an empirical study across the five most popular open-source models, using ten neutral, profession-related prompts to generate 100 images per profession, resulting in a dataset of 5,000 images evaluated by diverse human assessors representing different races and genders. We demonstrate that all five models encode and amplify pronounced societal skew: caregiving and nursing roles are consistently feminized, while high-status professions such as corporate CEO, politician, doctor, and lawyer are overwhelmingly represented by males and mostly White individuals. We further identify model-specific patterns, such as QWEN-Image's near-exclusive focus on East Asian outputs, Kandinsky's dominance of White individuals, and SDXL's comparatively broader but still biased distributions. These results provide critical insights for AI project managers and practitioners, enabling them to select equitable AI models and customized prompts that generate images in alignment with the principles of responsible AI. We conclude by discussing the risks of these biases and proposing actionable strategies for bias mitigation in building responsible GenAI systems. The code and Data Repository: https://github.com/Sufianlab/T2IBias

cs.LG↗

arXiv2025

CCD-Bench: Probing Cultural Conflict in Large Language Model Decision-Making

Although large language models (LLMs) are increasingly implicated in interpersonal and societal decision-making, their ability to navigate explicit conflicts between legitimately different cultural value systems remains largely unexamined. Existing benchmarks predominantly target cultural knowledge (CulturalBench), value prediction (WorldValuesBench), or single-axis bias diagnostics (CDEval); none evaluate how LLMs adjudicate when multiple culturally grounded values directly clash. We address this gap with CCD-Bench, a benchmark that assesses LLM decision-making under cross-cultural value conflict. CCD-Bench comprises 2,182 open-ended dilemmas spanning seven domains, each paired with ten anonymized response options corresponding to the ten GLOBE cultural clusters. These dilemmas are presented using a stratified Latin square to mitigate ordering effects. We evaluate 17 non-reasoning LLMs. Models disproportionately prefer Nordic Europe (mean 20.2 percent) and Germanic Europe (12.4 percent), while options for Eastern Europe and the Middle East and North Africa are underrepresented (5.6 to 5.8 percent). Although 87.9 percent of rationales reference multiple GLOBE dimensions, this pluralism is superficial: models recombine Future Orientation and Performance Orientation, and rarely ground choices in Assertiveness or Gender Egalitarianism (both under 3 percent). Ordering effects are negligible (Cramer's V less than 0.10), and symmetrized KL divergence shows clustering by developer lineage rather than geography. These patterns suggest that current alignment pipelines promote a consensus-oriented worldview that underserves scenarios demanding power negotiation, rights-based reasoning, or gender-aware analysis. CCD-Bench shifts evaluation beyond isolated bias detection toward pluralistic decision making and highlights the need for alignment strategies that substantively engage diverse worldviews.

cs.CL↗

arXiv2025

Small LLMs Do Not Learn a Generalizable Theory of Mind via Reinforcement Learning

Recent advancements in large language models (LLMs) have demonstrated emergent capabilities in complex reasoning, largely spurred by rule-based Reinforcement Learning (RL) techniques applied during the post-training. This has raised the question of whether similar methods can instill more nuanced, human-like social intelligence, such as a Theory of Mind (ToM), in LLMs. This paper investigates whether small-scale LLMs can acquire a robust and generalizable ToM capability through RL with verifiable rewards (RLVR). We conduct a systematic evaluation by training models on various combinations of prominent ToM datasets (HiToM, ExploreToM, FANToM) and testing for generalization on held-out datasets (e.g., OpenToM). Our findings indicate that small LLMs struggle to develop a generic ToM capability. While performance on in-distribution tasks improves, this capability fails to transfer to unseen ToM tasks with different characteristics. Furthermore, we demonstrate that prolonged RL training leads to models ``hacking'' the statistical patterns of the training datasets, resulting in significant performance gains on in-domain data but no change, or degradation of performance on out-of-distribution tasks. This suggests the learned behavior is a form of narrow overfitting rather than the acquisition of a true, abstract ToM capability.

cs.LG↗

arXiv2025

On the Stability of the Jacobian Matrix in Deep Neural Networks

Deep neural networks are known to suffer from exploding or vanishing gradients as depth increases, a phenomenon closely tied to the spectral behavior of the input-output Jacobian. Prior work has identified critical initialization schemes that ensure Jacobian stability, but these analyses are typically restricted to fully connected networks with i.i.d. weights. In this work, we go significantly beyond these limitations: we establish a general stability theorem for deep neural networks that accommodates sparsity (such as that introduced by pruning) and non-i.i.d., weakly correlated weights (e.g. induced by training). Our results rely on recent advances in random matrix theory, and provide rigorous guarantees for spectral stability in a much broader class of network models. This extends the theoretical foundation for initialization schemes in modern neural networks with structured and dependent randomness.

cs.LG↗

arXiv2025

AgentAuditor: Human-Level Safety and Security Evaluation for LLM Agents

Despite the rapid advancement of LLM-based agents, the reliable evaluation of their safety and security remains a significant challenge. Existing rule-based or LLM-based evaluators often miss dangers in agents' step-by-step actions, overlook subtle meanings, fail to see how small issues compound, and get confused by unclear safety or security rules. To overcome this evaluation crisis, we introduce AgentAuditor, a universal, training-free, memory-augmented reasoning framework that empowers LLM evaluators to emulate human expert evaluators. AgentAuditor constructs an experiential memory by having an LLM adaptively extract structured semantic features (e.g., scenario, risk, behavior) and generate associated chain-of-thought reasoning traces for past interactions. A multi-stage, context-aware retrieval-augmented generation process then dynamically retrieves the most relevant reasoning experiences to guide the LLM evaluator's assessment of new cases. Moreover, we developed ASSEBench, the first benchmark designed to check how well LLM-based evaluators can spot both safety risks and security threats. ASSEBench comprises 2293 meticulously annotated interaction records, covering 15 risk types across 29 application scenarios. A key feature of ASSEBench is its nuanced approach to ambiguous risk situations, employing "Strict" and "Lenient" judgment standards. Experiments demonstrate that AgentAuditor not only consistently improves the evaluation performance of LLMs across all benchmarks but also sets a new state-of-the-art in LLM-as-a-judge for agent safety and security, achieving human-level accuracy. Our work is openly accessible at https://github.com/Astarojth/AgentAuditor.

cs.AI↗

arXiv2025

Decompose-ToM: Enhancing Theory of Mind Reasoning in Large Language Models through Simulation and Task Decomposition

Theory of Mind (ToM) is the ability to understand and reflect on the mental states of others. Although this capability is crucial for human interaction, testing on Large Language Models (LLMs) reveals that they possess only a rudimentary understanding of it. Although the most capable closed-source LLMs have come close to human performance on some ToM tasks, they still perform poorly on complex variations of the task that involve more structured reasoning. In this work, we utilize the concept of "pretend-play", or ``Simulation Theory'' from cognitive psychology to propose ``Decompose-ToM'': an LLM-based inference algorithm that improves model performance on complex ToM tasks. We recursively simulate user perspectives and decompose the ToM task into a simpler set of functions: subject identification, question-reframing, world model updation, and knowledge availability. We test the algorithm on higher-order ToM tasks and a task testing for ToM capabilities in a conversational setting, demonstrating that our approach shows significant improvement across models compared to baseline methods while requiring minimal prompt tuning across tasks and no additional model training.

cs.CL↗