Searcharxiv⌕ Search

arXiv subjects

Mohsen Bayati

Publications and source records attributed to Mohsen Bayati.

At least 19 recordsLinked to original sources
arXiv2026

A Few Teacher Steps Go a Long Way: Cost-Efficient On-Policy Data Augmentation for Agent Post-Training

For LLM agents, supervised fine-tuning is not only about teacher labels' quality, but also about which interaction contexts those labels condition on. Pure behavioral cloning uses full teacher demonstrations, creating a mismatch between teacher-induced contexts seen in training and student-induced contexts encountered at test time. Recent work addresses this mismatch by querying a teacher at contexts reached by the student, often with increasingly elaborate filtering of the teacher's continuations. We instead frame on-policy data construction as a budget-allocation problem: under matched supervision resources, should teacher output be spent on more start-to-finish demos, longer continuations, outcome filtering, or broader coverage of learner-induced contexts? We formalize this design space through the rollout policy, switch-time distribution, continuation horizon, filtering rules, and two complementary costs: teacher inference generated before filtering and teacher supervision retained for SFT. Across HotpotQA, ALFWorld, and Terminal-Bench-Dev, bounded unfiltered teacher continuations at learner-induced contexts improve over pure behavioral cloning at matched budgets. On HotpotQA and ALFWorld, where we run the full comparison, few-step continuations match or exceed success-filtered and critical-context-filtered alternatives. Our findings suggest that a few teacher steps, placed at learner-induced contexts, can be a more cost-efficient supervision allocation than longer or more heavily curated teacher completions.

cs.LG↗
arXiv2026

Annealed Softmax Greedy in Many-Armed Bayesian Bandits

Reinforcement learning with verifiable rewards and group-based policy optimization methods update a stochastic policy by sampling multiple completions per prompt and increasing the policy's probability on those with higher reward. These updates, unline the exploration mechanism in Thompson sampling and UCB, do not include explicit mechanisms that track epistemic uncertainty. This paper studies a stylized explanation for why such uncertainty-agnostic updates can nevertheless be effective. We analyze an annealed softmax policy that selects actions according to a softmax of empirical mean rewards in a many-armed Bayesian Bernoulli bandit. Under a linear upper-tail condition on the prior, which implies an abundance of near-optimal arms, we prove that annealed softmax greedy achieves Bayes regret $\tilde{O}(m + T/m)$, and in particular $\tilde{O}(\sqrt{T})$ when the number of arms scales as $m = Θ(\sqrt{T})$. This is the near-optimal Bayes regret rate in this regime, attained also by empirical-mean greedy. Under the upper-tail condition, many arms keep empirical means near the optimum throughout learning, so the probability that softmax places away from the empirical best falls mostly on other near-optimal arms. By contrast, with a small number of arms, the same kind of softmax policy can suffer linear regret (Cesa-Bianchi et al., 2017). The result also provides a structural analogy to RLVR, where a base policy with a non-negligible probability of producing a correct completion plays the role of the tail condition. Simulations support the theory and motivate prior-anchored variants of greedy and annealed softmax that score arms by the Beta posterior mean and skip the forced initialization; with an arm-specific prior, accurate or noisy, these variants outperform baselines, including Thompson Sampling, when the number of arms is large.

cs.LG↗
arXiv2026

Calibrating Conservatism for Scalable Oversight

Agentic AI systems capable of autonomous planning and extended environmental interaction pose a fundamental control problem: how can humans maintain meaningful oversight of systems that may exceed their own capabilities? Existing approaches to scalable oversight rely on complex assumptions, remain largely heuristic, or lack practical methods for sequential settings with statistical guarantees. We introduce Calibrated Collective Oversight (CCO), which aggregates diverse auxiliary scoring functions into a penalty measuring deviation from a conservative baseline. Inspired by Attainable Utility Preservation, CCO enables collective conservatism: actions face a penalty proportional to overseer concern, so high-utility actions are still selected when overseers find them unobjectionable and overridden only when concern accumulates. CCO calibrates this conservatism online using Conformal Decision Theory, ensuring that undesirable outcomes remain below a user-specified target threshold with finite-time bounds and no distributional assumptions. On a modified version of SWE-bench, weaker overseers successfully constrain an adversarially misaligned stronger agent; on MACHIAVELLI, CCO substantially reduces ethical violations while preserving reward. In both settings, empirical violation rates closely match the specified targets, as predicted by the theory.

cs.AI↗
arXiv2026

Text-to-Distribution Prediction with Quantile Tokens and Neighbor Context

Many applications of LLM-based text regression require predicting a full conditional distribution rather than a single point value. We study distributional regression under empirical-quantile supervision, where each input is paired with multiple observed quantile outcomes, and the target distribution is represented by a dense grid of quantiles. We address two key limitations of current approaches: the lack of local grounding for distribution estimates, and the reliance on shared representations that create an indirect bottleneck between inputs and quantile outputs. In this paper, we introduce Quantile Token Regression, which, to our knowledge, is the first work to insert dedicated quantile tokens into the input sequence, enabling direct input-output pathways for each quantile through self-attention. We further augment these quantile tokens with retrieval, incorporating semantically similar neighbor instances and their empirical distributions to ground predictions with local evidence from similar instances. We also provide the first theoretical analysis of loss functions for quantile regression, clarifying which distributional objectives each optimizes. Experiments on the Inside Airbnb and StackSample benchmark datasets with LLMs ranging from 1.7B to 14B parameters show that quantile tokens with neighbors consistently outperform baselines (~4 points lower MAPE and 2x narrower prediction intervals), with especially large gains on smaller and more challenging datasets where quantile tokens produce substantially sharper and more accurate distributions.

cs.CL↗
arXiv2026

Scalable Stewardship of an LLM-Assisted Clinical Benchmark with Physician Oversight

Reference labels for machine-learning benchmarks are increasingly synthesized with LLM assistance, but their reliability remains underexamined. We audit MedCalc-Bench, a clinical benchmark for medical score computation whose labels were partly derived with LLM assistance, and develop a scalable physician-in-the-loop stewardship pipeline to reassess them. At least 27% of test labels are likely erroneous or incomputable. On a 50-instance subset validated by physicians, our recomputed labels agree with physician ground truth 74% of the time (95% CI, 60-84%) versus 20% for the originals (95% CI, 11-33%). Using original labels to evaluate frontier LLMs underestimates accuracy by 16-23 percentage points. In a controlled reinforcement-learning experiment, a model trained on recomputed labels outperforms one trained on originals by 13.5 percentage points (95% CI, 10.6-16.6%) on physician-labeled instances, and this advantage extends to related medical tasks. LLM-assisted benchmarks can propagate systematic errors into both evaluation and post-training unless actively stewarded.

cs.AI↗
arXiv2026

Causal Effects with Unobserved Unit Types in Interacting Human-AI Systems

We study experiments on interacting populations of humans and AI agents, where both unit types and the interaction network remain unobserved. Although causal effects propagate throughout the system, the goal is to estimate effects on humans. Examples include online platforms where human users interact alongside AI-driven accounts. We assume a human-AI prior that gives each unit a probability of being human. While humans cannot be distinguished at the unit level, the prior allows us to compute the average human composition within large subpopulations. We then model outcome dynamics through a causal message passing (CMP) framework and analyze sample-mean outcomes across subpopulations. We show that by constructing subpopulations that vary in expected human composition and treatment exposure, one can consistently recover human-specific causal effects. Our results characterize when distributional knowledge of population composition (without observing unit types or the interaction network) is sufficient for identification. We validate the approach on a simulated human-AI platform driven by behaviorally differentiated LLM agents. Together, these results provide a theoretical and practical framework for experimentation in emerging human-AI systems.

stat.ML↗
arXiv2026

The Oversight Game: Learning to Cooperatively Balance an AI Agent's Safety and Autonomy

As increasingly capable agents are deployed, a central safety challenge is how to retain meaningful human control without modifying the underlying system. We study a minimal control interface in which an agent chooses whether to act autonomously (play) or defer (ask), while a human simultaneously chooses whether to be permissive (trust) or engage in oversight (oversee), and model this interaction as a two-player Markov game. When this game forms a Markov Potential Game, we prove an alignment guarantee: any increase in the agent's utility from acting more autonomously cannot decrease the human's value. This establishes a form of intrinsic alignment where the agent's incentive to seek autonomy is structurally coupled to the human's welfare. Practically, the framework induces a transparent control layer that encourages the agent to defer when risky and act when safe. While we use gridworld simulations to illustrate the emergence of this collaboration, our primary validation involves an agentic tool-use task in which two 30B parameter language models are fine-tuned via independent policy gradient. We demonstrate that even as the agents learn to coordinate on the fly, this framework effectively reduces safety violations in realistic, open-ended environments.

cs.AI↗
arXiv2026

Validating Causal Message Passing Against Network-Aware Methods on Real Experiments

Estimating total treatment effects in the presence of network interference typically requires knowledge of the underlying interaction structure. However, in many practical settings, network data is either unavailable, incomplete, or measured with substantial error. We demonstrate that causal message passing, a methodology that leverages temporal structure in outcome data rather than network topology, can recover total treatment effects comparable to network-aware approaches. We apply causal message passing to two large-scale field experiments where a recently developed bipartite graph methodology, which requires network knowledge, serves as a benchmark. Despite having no access to the interaction network, causal message passing produces effect estimates that match the network-aware approach in direction across all metrics and in statistical significance for the primary decision metric. Our findings validate the premise of causal message passing: that temporal variation in outcomes can serve as an effective substitute for network observation when estimating spillover effects. This has important practical implications: practitioners facing settings where network data is costly to collect, proprietary, or unreliable can instead exploit the temporal dynamics of their experimental data.

stat.ME↗
arXiv2025

Scaling Clinician-Grade Feature Generation from Clinical Notes with Multi-Agent Language Models

Developing accurate clinical prediction models is often bottlenecked by the difficulty of deriving meaningful structured features from unstructured EHR notes, a process that traditionally requires manual, unscalable clinical abstraction. In this study, we first established a rigorous patient-level Clinician Feature Generation (CFG) protocol, in which domain experts manually reviewed notes to define and extract nuanced features for a cohort of 147 patients with prostate cancer. As a high-fidelity ground truth, this labor-intensive process provided the blueprint for SNOW (Scalable Note-to-Outcome Workflow), a transparent multi-agent large language model (LLM) system designed to autonomously mimic the iterative reasoning and validation workflow of clinical experts. On 5-year cancer recurrence prediction, SNOW (AUC-ROC 0.767) achieved performance comparable to manual CFG (0.762) and outperformed structured baselines, clinician-guided LLM extraction, and six representational feature generation (RFG) approaches. Once configured, SNOW produced the full patient-level feature table in 12 hours with 5 hours of clinician oversight, reducing human expert effort by approximately 48-fold versus manual CFG. To test scalability where manual CFG is infeasible, we deployed SNOW on an external heart failure with preserved ejection fraction (HFpEF) cohort from MIMIC-IV (n=2,084); without task-specific tuning, SNOW generated prognostic features that outperformed baseline and RFG methods for 30-day (SNOW: 0.851) and 1-year (SNOW: 0.763) mortality prediction. These results demonstrate that a modular LLM agent-based system can scale expert-level feature generation from clinical notes, while enabling interpretable use of unstructured EHR text in outcome prediction and preserving generalizability across a variety of settings and conditions.

cs.AI↗
arXiv2025

On Evolution-Based Models for Experimentation Under Interference

Causal effect estimation in networked systems is central to data-driven decision making. In such settings, interventions on one unit can spill over to others, and in complex physical or social systems, the interaction pathways driving these interference structures remain largely unobserved. We argue that for identifying population-level causal effects, it is not necessary to recover the exact network structure; instead, it suffices to characterize how those interactions contribute to the evolution of outcomes. Building on this principle, we study an evolution-based approach that investigates how outcomes change across observation rounds in response to interventions, hence compensating for missing network information. Using an exposure-mapping perspective, we give an axiomatic characterization of when the empirical distribution of outcomes follows a low-dimensional recursive equation, and identify minimal structural conditions under which such evolution mappings exist. We frame this as a distributional counterpart to difference-in-differences. Rather than assuming parallel paths for individual units, it exploits parallel evolution patterns across treatment scenarios to estimate counterfactual trajectories. A key insight is that treatment randomization plays a role beyond eliminating latent confounding; it induces an implicit sampling from hidden interference channels, enabling consistent learning about heterogeneous spillover effects. We highlight causal message passing as an instantiation of this method in dense networks while extending to more general interference structures, including influencer networks where a small set of units drives most spillovers. Finally, we discuss the limits of this approach, showing that strong temporal trends or endogenous interference can undermine identification.

stat.ML↗
arXiv2025

Estimating Total Effects in Bipartite Experiments with Spillovers and Partial Eligibility

We study randomized experiments in bipartite systems where only a subset of treatment-side units are eligible for assignment while all units continue to interact, generating interference. We formalize eligibility-constrained bipartite experiments and define estimands aligned with full deployment: the Primary Total Treatment Effect (PTTE) on eligible units and the Secondary Total Treatment Effect (STTE) on ineligible units. Under randomization within the eligible set, we give identification conditions and develop interference-aware ensemble estimators that combine exposure mappings, generalized propensity scores, and flexible machine learning. We further introduce a projection that links treatment- and outcome-level estimands; this mapping is exact under a Linear Additive Edges condition and enables estimation on the (typically much smaller) treatment side with deterministic aggregation to outcomes. In simulations with known ground truth across realistic exposure regimes, the proposed estimators recover PTTE and STTE with low bias and variance and reduce the bias that could arise when interference is ignored. Two field experiments illustrate practical relevance: our method corrects the direction of expected interference bias for a pre-specified metric in both studies and reverses the sign and significance of the primary decision metric in one case.

stat.ME↗
arXiv2025

Simulating and Experimenting with Social Media Mobilization Using LLM Agents

Online social networks have transformed the ways in which political mobilization messages are disseminated, raising new questions about how peer influence operates at scale. Building on the landmark 61-million-person Facebook experiment \citep{bond201261}, we develop an agent-based simulation framework that integrates real U.S. Census demographic distributions, authentic Twitter network topology, and heterogeneous large language model (LLM) agents to examine the effect of mobilization messages on voter turnout. Each simulated agent is assigned demographic attributes, a personal political stance, and an LLM variant (\texttt{GPT-4.1}, \texttt{GPT-4.1-Mini}, or \texttt{GPT-4.1-Nano}) reflecting its political sophistication. Agents interact over realistic social network structures, receiving personalized feeds and dynamically updating their engagement behaviors and voting intentions. Experimental conditions replicate the informational and social mobilization treatments of the original Facebook study. Across scenarios, the simulator reproduces qualitative patterns observed in field experiments, including stronger mobilization effects under social message treatments and measurable peer spillovers. Our framework provides a controlled, reproducible environment for testing counterfactual designs and sensitivity analyses in political mobilization research, offering a bridge between high-validity field experiments and flexible computational modeling.\footnote{Code and data available at https://github.com/CausalMP/LLM-SocioPol}

cs.SI↗
arXiv2025

Can We Validate Counterfactual Estimations in the Presence of General Network Interference?

Randomized experiments have become a cornerstone of evidence-based decision-making in contexts ranging from online platforms to public health. However, in experimental settings with network interference, a unit's treatment can influence outcomes of other units, challenging both causal effect estimation and its validation. Classic validation approaches fail as outcomes are only observable under a single treatment scenario and exhibit complex correlation patterns due to interference. To address these challenges, we introduce a framework that facilitates the use of machine learning tools for both estimation and validation in causal inference. Central to our approach is the new distribution-preserving network bootstrap, a theoretically-grounded technique that generates multiple statistically-valid subpopulations from a single experiment's data. This amplification of experimental samples enables our second contribution: a counterfactual cross-validation procedure. This procedure adapts the principles of model validation to the unique constraints of causal settings, providing a rigorous, data-driven method for selecting and evaluating estimators. We extend recent causal message-passing developments by incorporating heterogeneous unit-level characteristics and varying local interactions, ensuring reliable finite-sample performance through non-asymptotic analysis. Additionally, we develop and publicly release a comprehensive benchmark toolbox featuring diverse experimental environments, from networks of interacting AI agents to ride-sharing applications. These environments provide known ground truth values while maintaining realistic complexities, enabling systematic evaluation of causal inference methods. Extensive testing across these environments demonstrates our method's robustness to diverse forms of network interference.

cs.LG↗
arXiv2025

On Aligning Prediction Models with Clinical Experiential Learning: A Prostate Cancer Case Study

Over the past decade, the use of machine learning (ML) models in healthcare applications has rapidly increased. Despite high performance, modern ML models do not always capture patterns the end user requires. For example, a model may predict a non-monotonically decreasing relationship between cancer stage and survival, keeping all other features fixed. In this paper, we present a reproducible framework for investigating this misalignment between model behavior and clinical experiential learning, focusing on the effects of underspecification of modern ML pipelines. In a prostate cancer outcome prediction case study, we first identify and address these inconsistencies by incorporating clinical knowledge, collected by a survey, via constraints into the ML model, and subsequently analyze the impact on model performance and behavior across degrees of underspecification. The approach shows that aligning the ML model with clinical experiential learning is possible without compromising performance. Motivated by recent literature in generative AI, we further examine the feasibility of a feedback-driven alignment approach in non-generative AI clinical risk prediction models through a randomized experiment with clinicians. Our findings illustrate that, by eliciting clinicians' model preferences using our proposed methodology, the larger the difference in how the constrained and unconstrained models make predictions for a patient, the more apparent the difference is in clinical interpretation.

cs.LG↗
arXiv2025

Analysis of Thompson Sampling for Controlling Unknown Linear Diffusion Processes

Linear diffusion processes serve as canonical continuous-time models for dynamic decision-making under uncertainty. These systems evolve according to drift matrices that specify the instantaneous rates of change in the expected system state, while also experiencing continuous random disturbances modeled by Brownian noise. For instance, in medical applications such as artificial pancreas systems, the drift matrices represent the internal dynamics of glucose concentrations. Classical results in stochastic control provide optimal policies under perfect knowledge of the drift matrices. However, practical decision-making scenarios typically feature uncertainty about the drift; in medical contexts, such parameters are patient-specific and unknown, requiring adaptive policies for efficiently learning the drift matrices while ensuring system stability and optimal performance. We study the Thompson sampling (TS) algorithm for decision-making in linear diffusion processes with unknown drift matrices. For this algorithm that designs control policies as if samples from a posterior belief about the parameters fully coincide with the unknown truth, we establish efficiency. That is, Thompson sampling learns optimal control actions fast, incurring only a square-root of time regret, and also learns to stabilize the system in a short time period. To our knowledge, this is the first such result for TS in a diffusion process control problem. Moreover, our empirical simulations in three settings that involve blood-glucose and flight control demonstrate that TS significantly improves regret, compared to the state-of-the-art algorithms, suggesting it explores in a more guarded fashion. Our theoretical analysis includes characterization of a certain optimality manifold that relates the geometry of the drift matrices to the optimal control of the diffusion process, among others.

cs.LG↗
arXiv2025

Quantile Regression with Large Language Models for Price Prediction

Large Language Models (LLMs) have shown promise in structured prediction tasks, including regression, but existing approaches primarily focus on point estimates and lack systematic comparison across different methods. We investigate probabilistic regression using LLMs for unstructured inputs, addressing challenging text-to-distribution prediction tasks such as price estimation where both nuanced text understanding and uncertainty quantification are critical. We propose a novel quantile regression approach that enables LLMs to produce full predictive distributions, improving upon traditional point estimates. Through extensive experiments across three diverse price prediction datasets, we demonstrate that a Mistral-7B model fine-tuned with quantile heads significantly outperforms traditional approaches for both point and distributional estimations, as measured by three established metrics each for prediction accuracy and distributional calibration. Our systematic comparison of LLM approaches, model architectures, training approaches, and data scaling reveals that Mistral-7B consistently outperforms encoder architectures, embedding-based methods, and few-shot learning methods. Our experiments also reveal the effectiveness of LLM-assisted label correction in achieving human-level accuracy without systematic bias. Our curated datasets are made available at https://github.com/vnik18/llm-price-quantile-reg/ to support future research.

cs.CL↗
arXiv2025

Conformal Arbitrage: Risk-Controlled Balancing of Competing Objectives in Language Models

Modern language model deployments must often balance competing objectives, for example, helpfulness versus harmlessness, cost versus accuracy, and reward versus safety. We introduce Conformal Arbitrage, a post hoc framework that learns a data driven threshold to mediate between a Primary model optimized for a primary objective and a more conservative Guardian which could be another model or a human domain expert aligned with a guardrail objective. The threshold is calibrated with conformal risk control, yielding finite sample, distribution free guarantees that the long run frequency of undesirable events, such as factual errors or safety violations, does not exceed a user specified quota. Because Conformal Arbitrage operates wholly at the API level, without requiring access to model logits or updating model weights, it complements weight based alignment techniques and integrates seamlessly with existing cost aware cascades. Empirically, Conformal Arbitrage traces an efficient frontier, allowing users to define an acceptable performance level for one objective while maximizing utility in another. We observe that our method outperforms, in terms of accuracy, cost matched random routing between models. These properties make Conformal Arbitrage a practical, theoretically grounded tool for trustworthy and economical deployment of large language models across a broad range of potentially competing objectives.

cs.AI↗
arXiv2025

Geometry-Aware Approaches for Balancing Performance and Theoretical Guarantees in Linear Bandits

This paper is motivated by recent research in the $d$-dimensional stochastic linear bandit literature, which has revealed an unsettling discrepancy: algorithms like Thompson sampling and Greedy demonstrate promising empirical performance, yet this contrasts with their pessimistic theoretical regret bounds. The challenge arises from the fact that while these algorithms may perform poorly in certain problem instances, they generally excel in typical instances. To address this, we propose a new data-driven technique that tracks the geometric properties of the uncertainty ellipsoid around the main problem parameter. This methodology enables us to formulate a data-driven frequentist regret bound, which incorporates the geometric information, for a broad class of base algorithms, including Greedy, OFUL, and Thompson sampling. This result allows us to identify and ``course-correct" problem instances in which the base algorithms perform poorly. The course-corrected algorithms achieve the minimax optimal regret of order $\tilde{\mathcal{O}}(d\sqrt{T})$ for a $T$-period decision-making scenario, effectively maintaining the desirable attributes of the base algorithms, including their empirical efficacy. We present simulation results to validate our findings using synthetic and real data.

cs.LG↗