arXiv subjects

Hadi Hosseini

Publications and source records attributed to Hadi Hosseini.

At least 19 recordsLinked to original sources

arXiv2026

Inference-Time Nash Alignment

Preference-based fine-tuning methods such as RLHF and DPO require substantial compute and large preference datasets. They also need direct access to the model parameters which are not provided by many state-of-the art models. Inference-time alignment offers a cost-effective alternative without updating model parameters. However, existing inference-time methods rely on a scalar reward model derived under a Bradley-Terry assumption, which cannot represent general preferences. Following recent work on fine-tuning with generalized preferences, in this work, we initiate the study of inference-time alignment under general preferences. We formulate the problem as obtaining a Nash equilibrium of a two-player zero-sum game between policies. We propose two algorithms: Best-of-Nash (BoN) and Nash Mirror Descent (NMD). We prove that both algorithms achieve a duality gap that matches the problem lower bound. Empirically, we implement the two methods on three datasets, which shows that our methods substantially outperform the base policy, converging to the performance of the fine-tuned models. Moreover, our results show that NMD remains robust across the regularization parameter.

cs.AI↗

arXiv2026

Simultaneous Envy and Equitability Guarantees

Recent work in fair division has focused on either simultaneously satisfying closely related fairness notions or achieving a single notion across the ex-ante and ex-post worlds. We study the compatibility of two fundamentally different fairness notions: envy-freeness and equitability. For indivisible goods-only and chores-only settings, we study the existence and complexity of simultaneously satisfying their relaxations, revealing sharp contrasts between the two settings. We show that EF1+EQ1 may fail to exist even for normalized, additive valuations. Our main algorithmic result computes an EF1+EQ1 allocation for normalized binary goods with at most seven agents. In sharp contrast, binary chores admit the stronger EFX+EQX guarantee for any number of agents, even without normalization. We further initiate the study of cross-notion ex-ante--ex-post guarantees, asking whether randomized allocations can provide ex-ante guarantees for one notion while preserving ex-post guarantees for another.

cs.GT↗

arXiv2026

Preference Reasoning under Indeterminacy in Large Language Models

As large language models evolve into decision-making agents, the ability to reason over preferences becomes fundamental to alignment, coordination, and collective intelligence. Yet, unlike standard benchmarks, real-world preference reasoning is inherently indeterminate: information may be incomplete, and valid solutions may not exist. We argue that indeterminacy, rather than correctness alone, is a central challenge for AI reasoning. We formalize this challenge along two axes, (i) epistemic indeterminacy, arising from incomplete, partial, or expressive preferences, and (ii) structural indeterminacy, arising from the non-existence of solutions under standard social choice concepts. Across a hierarchy of tasks, we show that state-of-the-art language models systematically fail to distinguish between determined and undetermined instances, exhibiting miscalibrated reasoning even in verification settings.

cs.AI↗

arXiv2026

To EFX OR to MMS, That is the Question

We study the agent-wise disjunction of two central fairness notions for indivisible items, where every agent must be either envy-free up to any item (EFX) or maximin-share (MMS) satisfied. One might expect this flexibility to restore existence, especially because the existence of EFX itself resisted resolution for nearly a decade. Surprisingly, it does not. We construct counterexamples with three agents and eight submodular goods, and with three agents and seven submodular chores, significantly strengthening recent EFX impossibility results. On the positive side, we prove existence for additive mixed items with at most three valuation types when one type is a singleton. Our proof extends beyond additivity for goods and yields approximation schemes for goods and three-agent chores instances. We also identify a clean separation between the disjunction and its constituents: For additive chores with two valuation types, EFX and MMS are both known to fail, whereas an EFX$\vee$MMS allocation always exists. Finally, we show that identical additive valuations even admit the conjunction EFX$\wedge$MMS for mixed items. Overall, our results show that allowing flexibility in choosing agent-specific fairness certificates expands the frontier of fair solutions while also uncovering surprising impossibilities.

cs.GT↗

arXiv2026

The Judgment-Consequence Gap: LLM Moral Reasoning in Healthcare Decisions

As large language models (LLMs) enter high-stakes domains such as healthcare, understanding their moral reasoning becomes essential. Decisions about scarce medical resources often hinge on judgments of responsibility, particularly when patients' own actions contribute to illness. We investigate how LLMs reason about responsibility and its consequences, tracing their judgments across successive levels, from the behavior, to the resulting illness, to the denial of care. We evaluate a wide range of LLMs, spanning different model families and capability levels, on various clinical vignettes adapted from prior studies. Our results identify a judgment-consequence gap: LLMs largely agree with humans that patients bear responsibility for health-harming behaviors, yet overwhelmingly refuse to let that judgment influence how they allocate scarce resources. Specifically, LLMs default to random allocation, whereas humans consistently favor the less-culpable patient. Compared to humans, LLMs also place greater emphasis on access to information, reducing responsibility judgments when health-risk knowledge is unavailable. These findings reveal that LLMs apply a systematically different moral framework than humans when responsibility and resource scarcity intersect, surprisingly often amplifying normative disagreement with humans as reasoning capability increases.

cs.CY↗

arXiv2026

The Pedagogy of AI Mistakes: Fostering Higher-Order Thinking

As generative AI becomes increasingly integrated into higher education, its frequent errors and hallucinations, often seen as limitations, offer a unique pedagogical opportunity. By framing AI as a ``learning companion'' whose imperfect outputs prompt analysis, evaluation, and reflection, we argue that instructors can engage students in the fundamental processes of higher-order thinking. This paper presents a design-oriented study in which an AI-integrated syllabus in a \textit{database design} course deliberately leverages AI's limitations to foster critical thinking and higher-order cognitive skills aligned with Bloom's taxonomy of learning. Using a mixed-methods approach, we examine how structured interaction with AI-generated errors supports metacognitive engagement, reinforces disciplinary rigor, and relates to students' perceived AI literacy and subject-matter competency.

cs.CY↗

arXiv2026

Strategic Algorithmic Monoculture: Experimental Evidence from Coordination Games

AI agents increasingly operate in multi-agent environments where outcomes depend on coordination. We distinguish primary algorithmic monoculture -- baseline action similarity -- from strategic algorithmic monoculture, whereby agents adjust similarity in response to incentives. We implement a simple experimental design that cleanly separates these forces, and deploy it on human and large language model (LLM) subjects. LLMs exhibit high levels of baseline similarity (primary monoculture) and, like humans, they regulate it in response to coordination incentives (strategic monoculture). While LLMs coordinate extremely well on similar actions, they lag behind humans in sustaining heterogeneity when divergence is rewarded.

cs.AI↗

arXiv2025

Fair Societies: Algorithms for House Allocations

House Allocations concern with matchings involving one-sided preferences, where houses serve as a proxy encoding valuable indivisible resources (e.g. organs, course seats, subsidized public housing units) to be allocated among the agents. Every agent must receive exactly one resource. We study algorithmic approaches towards ensuring fairness in such settings. Minimizing the number of envious agents is known to be NP-complete (Kamiyama et al. 2021). We present two tractable approaches to deal with the computational hardness. When the agents are presented with an initial allocation of houses, we aim to refine this allocation by reallocating a bounded number of houses to reduce the number of envious agents. We show an efficient algorithm when the agents express preference for a bounded number of houses. Next, we consider single peaked preference domain and present a polynomial time algorithm for finding an allocation that minimize the number of envious agents. We further extend it to satisfy Pareto efficiency. Our former algorithm works for other measures of envy such as total envy, or maximum envy, with suitable modifications. Finally, we present an empirical analysis recording the fairness-welfare trade-off of our algorithms.

cs.GT↗

arXiv2025

The Landscape of Almost Equitable Allocations

Equitability is a fundamental notion in fair division which requires that all agents derive equal value from their allocated bundles. We study, for general (possibly non-monotone) valuations, a popular relaxation of equitability known as equitability up to one item (EQ1). An EQ1 allocation may fail to exist even with additive non-monotone valuations; for instance, when there are two agents, one valuing every item positively and the other negatively. This motivates a mild and natural assumption: all agents agree on the sign of their value for the grand bundle. Under this assumption, we prove the existence and provide an efficient algorithm for computing EQ1 allocations for two agents with general valuations. When there are more than two agents, we show the existence and polynomial-time computability of EQ1 allocations for valuation classes beyond additivity and monotonicity, in particular for (1) doubly monotone valuations and (2) submodular (resp. supermodular) valuations where the value for the grand bundle is nonnegative (resp. nonpositive) for all agents. Furthermore, we settle an open question of Bil`o et al. by showing that an EQ1 allocation always exists for nonnegative(resp. nonpositive) valuations, i.e., when every agent values each subset of items nonnegatively (resp. nonpositively). Finally, we complete the picture by showing that for general valuations with more than two agents, EQ1 allocations may not exist even when agents agree on the sign of the grand bundle, and that deciding the existence of an EQ1 allocation is computationally intractable.

cs.GT↗

arXiv2025

Non-Monotonicity in Fair Division of Graphs

We consider the problem of fairly allocating the vertices of a graph among $n$ agents, where the value of a bundle is determined by its cut value -- the number of edges with exactly one endpoint in the bundle. This model naturally captures applications such as team formation and network partitioning, where valuations are inherently non-monotonic: the marginal values may be positive, negative, or zero depending on the composition of the bundle. We focus on the fairness notion of envy-freeness up to one item (EF1) and explore its compatibility with several efficiency concepts such as Transfer Stability (TS) that prohibits single-item transfers that benefit one agent without making the other worse-off. For general graphs, our results uncover a non-monotonic relationship between the number of agents $n$ and the existence of allocations satisfying EF1 and transfer stability (TS): such allocations always exist for $n=2$, may fail to exist for $n=3$, but exist again for all $n\geq 4$. We further show that existence can be guaranteed for any $n$ by slightly weakening the efficiency requirement or by restricting the graph to forests. All of our positive results are achieved via efficient algorithms.

cs.GT↗

arXiv2025

SP-Rank: A Dataset for Ranked Preferences with Secondary Information

We introduce $\mathbf{SP-Rank}$, the first large-scale, publicly available dataset for benchmarking algorithms that leverage both first-order preferences and second-order predictions in ranking tasks. Each datapoint includes a personal vote (first-order signal) and a meta-prediction of how others will vote (second-order signal), allowing richer modeling than traditional datasets that capture only individual preferences. SP-Rank contains over 12,000 human-generated datapoints across three domains -- geography, movies, and paintings, and spans nine elicitation formats with varying subset sizes. This structure enables empirical analysis of preference aggregation when expert identities are unknown but presumed to exist, and individual votes represent noisy estimates of a shared ground-truth ranking. We benchmark SP-Rank by comparing traditional aggregation methods that use only first-order votes against SP-Voting, a second-order method that jointly reasons over both signals to infer ground-truth rankings. While SP-Rank also supports models that rely solely on second-order predictions, our benchmarks emphasize the gains from combining both signals. We evaluate performance across three core tasks: (1) full ground-truth rank recovery, (2) subset-level rank recovery, and (3) probabilistic modeling of voter behavior. Results show that incorporating second-order signals substantially improves accuracy over vote-only methods. Beyond social choice, SP-Rank supports downstream applications in learning-to-rank, extracting expert knowledge from noisy crowds, and training reward models in preference-based fine-tuning pipelines. We release the dataset, code, and baseline evaluations (available at https://github.com/amrit19/SP-Rank-Dataset ) to foster research in human preference modeling, aggregation theory, and human-AI alignment.

cs.IR↗

arXiv2025

Matching Markets Meet LLMs: Algorithmic Reasoning with Ranked Preferences

The rise of Large Language Models (LLMs) has driven progress in reasoning tasks -- from program synthesis to scientific hypothesis generation -- yet their ability to handle ranked preferences and structured algorithms in combinatorial domains remains underexplored. We study matching markets, a core framework behind applications like resource allocation and ride-sharing, which require reconciling individual ranked preferences to ensure stable outcomes. We evaluate several state-of-the-art models on a hierarchy of preference-based reasoning tasks -- ranging from stable-matching generation to instability detection, instability resolution, and fine-grained preference queries -- to systematically expose their logical and algorithmic limitations in handling ranked inputs. Surprisingly, even top-performing models with advanced reasoning struggle to resolve instability in large markets, often failing to identify blocking pairs or execute algorithms iteratively. We further show that parameter-efficient fine-tuning (LoRA) significantly improves performance in small markets, but fails to bring about a similar improvement on large instances, suggesting the need for more sophisticated strategies to improve LLMs' reasoning with larger-context inputs.

cs.AI↗

arXiv2025

Two-Sided Manipulation Games in Stable Matching Markets

The Deferred Acceptance (DA) algorithm is an elegant procedure for finding a stable matching in two-sided matching markets. It ensures that no pair of agents prefers each other to their matched partners. In this work, we initiate the study of two-sided manipulations in matching markets as non-cooperative games. We introduce the accomplice manipulation game, where a man misreports to help a specific woman obtain a better partner, whenever possible. We provide a polynomial time algorithm for finding a pure strategy Nash equilibrium (NE) and show that our algorithm always yields a stable matching - although not every Nash equilibrium corresponds to a stable matching. Additionally, we show how our analytical techniques for the accomplice manipulation game can be applied to other manipulation games in matching markets, such as one-for-many and the standard self-manipulation games. We complement our theoretical findings with empirical evaluations of different properties of the resulting NE, such as the welfare of the agents.

cs.GT↗

arXiv2025

Who Gets the Kidney? Human-AI Alignment, Indecision, and Moral Values

The rapid integration of Large Language Models (LLMs) in high-stakes decision-making -- such as allocating scarce resources like donor organs -- raises critical questions about their alignment with human moral values. We systematically evaluate the behavior of several prominent LLMs against human preferences in kidney allocation scenarios and show that LLMs: i) exhibit stark deviations from human values in prioritizing various attributes, and ii) in contrast to humans, LLMs rarely express indecision, opting for deterministic decisions even when alternative indecision mechanisms (e.g., coin flipping) are provided. Nonetheless, we show that low-rank supervised fine-tuning with few samples is often effective in improving both decision consistency and calibrating indecision modeling. These findings illustrate the necessity of explicit alignment strategies for LLMs in moral/ethical domains.

cs.CY↗

arXiv2025

Bridging Theory and Perception in Fair Division: A Study on Comparative and Fair Share Notions

The allocation of resources among multiple agents is a fundamental problem in both economics and computer science. In these settings, fairness plays a crucial role in ensuring social acceptability and practical implementation of resource allocation algorithms. Traditional fair division solutions have given rise to a variety of approximate fairness notions, often as a response to the challenges posed by non-existence or computational intractability of exact solutions. However, the inherent incompatibility among these notions raises a critical question: which concept of fairness is most suitable for practical applications? In this paper, we examine two broad frameworks -- threshold-based and comparison-based fairness notions -- and evaluate their perceived fairness through a comprehensive human subject study. Our findings uncover novel insights into the interplay between perception of fairness, theoretical guarantees, the role of externalities and subjective valuations, and underlying cognitive processes, shedding light on the theory and practice of fair division.

cs.GT↗

arXiv2025

AI Education in a Mirror: Challenges Faced by Academic and Industry Experts

As Artificial Intelligence (AI) technologies continue to evolve, the gap between academic AI education and real-world industry challenges remains an important area of investigation. This study provides preliminary insights into challenges AI professionals encounter in both academia and industry, based on semi-structured interviews with 14 AI experts - eight from industry and six from academia. We identify key challenges related to data quality and availability, model scalability, practical constraints, user behavior, and explainability. While both groups experience data and model adaptation difficulties, industry professionals more frequently highlight deployment constraints, resource limitations, and external dependencies, whereas academics emphasize theoretical adaptation and standardization issues. These exploratory findings suggest that AI curricula could better integrate real-world complexities, software engineering principles, and interdisciplinary learning, while recognizing the broader educational goals of building foundational and ethical reasoning skills.

cs.CY↗

arXiv2025

The Algorithmic Landscape of Fair and Efficient Distribution of Delivery Orders in the Gig Economy

Distributing services, goods, and tasks in the gig economy heavily relies upon on-demand workers (aka agents), leading to new challenges varying from logistics optimization to the ethical treatment of gig workers. We focus on fair and efficient distribution of delivery tasks -- placed on the vertices of a graph -- among a fixed set of agents. We consider the fairness notion of minimax share (MMS), which aims to minimize the maximum (submodular) cost among agents and is particularly appealing in applications without monetary transfers. We propose a novel efficiency notion -- namely non-wastefulness -- that is desirable in a wide range of scenarios and, more importantly, does not suffer from computational barriers. Specifically, given a distribution of tasks, we can, in polynomial time, i) verify whether the distribution is non-wasteful and ii) turn it into an equivalent non-wasteful distribution. Moreover, we investigate several fixed-parameter tractable and polynomial-time algorithms and paint a complete picture of the (parameterized) complexity of finding fair and efficient distributions of tasks with respect to both the structure of the topology and natural restrictions of the input. Finally, we highlight how our findings shed light on computational aspects of other well-studied fairness notions, such as envy-freeness and its relaxations.

cs.GT↗

arXiv2025

Distributive Fairness in Large Language Models: Evaluating Alignment with Human Values

The growing interest in employing large language models (LLMs) for decision-making in social and economic contexts has raised questions about their potential to function as agents in these domains. A significant number of societal problems involve the distribution of resources, where fairness, along with economic efficiency, play a critical role in the desirability of outcomes. In this paper, we examine whether LLM responses adhere to fundamental fairness concepts such as equitability, envy-freeness, and Rawlsian maximin, and investigate their alignment with human preferences. We evaluate the performance of several LLMs, providing a comparative benchmark of their ability to reflect these measures. Our results demonstrate a lack of alignment between current LLM responses and human distributional preferences. Moreover, LLMs are unable to utilize money as a transferable resource to mitigate inequality. Nonetheless, we demonstrate a stark contrast when (some) LLMs are tasked with selecting from a predefined menu of options rather than generating one. In addition, we analyze the robustness of LLM responses to variations in semantic factors (e.g., intentions or personas) or non-semantic prompting changes (e.g., templates or orderings). Finally, we highlight potential strategies aimed at enhancing the alignment of LLM behavior with well-established fairness concepts.

cs.GT↗