arXiv subjects

Mauricio Tec

Publications and source records attributed to Mauricio Tec.

At least 19 recordsLinked to original sources

arXiv2026

Bilevel Optimization of Synthetic Trajectories for Multi-Turn LLM Fine-Tuning

While LLMs excel at single-turn generation, they struggle with long-horizon, multi-turn interactions. Offline reinforcement learning (RL) offers a scalable approach, yet its performance hinges on the availability and quality of multi-turn trajectory data. A common remedy is to augment training with synthetic trajectories generated by LLMs or simulators, but synthetic data is highly heterogeneous in quality, and naively treating all trajectories as equally informative can degrade performance. We propose BOOST, a bilevel optimization framework where the inner level trains the LLM on reweighted data and the outer level trains a lightweight reweighting head on held-out real validation tasks, assigning continuous trajectory-level weights without requiring an external judge. To ground this approach, we derive a PAC-Bayesian bound revealing a three-way trade-off: synthetic data increases diversity but risks task-shift, while concentrating weight on high-quality trajectories improves empirical performance at the cost of effective sample size. Empirically, our method consistently outperforms multiple baselines. Analysis reveals it upweights synthetic trajectories that align with the real data distribution and exhibit higher qualitative merit.

cs.LG↗

arXiv2025

Guided Transfer Learning for Discrete Diffusion Models

Discrete diffusion models (DMs) have achieved strong performance in language and other discrete domains, offering a compelling alternative to autoregressive modeling. Yet this performance typically depends on large training datasets, challenging the performance of DMs in small-data regimes -- common under real-world constraints. Aimed at this challenge, recent work in continuous DMs suggests that transfer learning via classifier ratio-based guidance can adapt a pretrained DM to a related target distribution, often outperforming alternatives such as full-weight fine-tuning on the target data. By contrast, transfer learning for discrete DMs remains unexplored. We address this gap by exploring practical analogues of ratio-based transfer learning for discrete DMs. Our theoretical analysis shows that a direct extension of existing ratio-based guidance is computationally prohibitive, scaling with vocabulary size. To overcome this limitation, we introduce a scheduling mechanism that yields a practical algorithm, Guided Transfer Learning for discrete diffusion models (GTL). GTL enables sampling from a target distribution without modifying the pretrained denoiser and reduces the cost to linear scaling in vocabulary size, which in turn supports longer sequence generation. We evaluate GTL on sequential data, including synthetic Markov chains and language modeling tasks, and provide a detailed empirical analysis of its behavior. The results highlight a clear trade-off: when target datasets are large, weight fine-tuning is often preferable, whereas GTL becomes increasingly effective as target data shrinks. Finally, we experimentally demonstrate a key failure mode of GTL: when the source and target distributions overlap poorly, the ratio-based classifier required for guidance becomes unreliable, limiting transfer performance.

cs.LG↗

arXiv2025

Lightweight Robust Direct Preference Optimization

Direct Preference Optimization (DPO) has become a popular method for fine-tuning large language models (LLMs) due to its stability and simplicity. However, it is also known to be sensitive to noise in the data and prone to overfitting. Recent works have proposed using distributionally robust optimization (DRO) to address potential noise and distributional shift in the data. However, these methods often suffer from excessive conservatism and high computational cost. We propose DPO-PRO (DPO with Preference Robustness), a robust fine-tuning algorithm based on DPO which accounts for uncertainty in the preference distribution through a lightweight DRO formulation. Unlike prior DRO-based variants, DPO-PRO focuses solely on uncertainty in preferences, avoiding unnecessary conservatism and incurring negligible computational overhead. We further show that DPO-PRO is equivalent to a regularized DPO objective that penalizes model overconfidence under weak preference signals. We evaluate DPO-PRO on standard alignment benchmarks and a real-world public health task. Experimental results show that our method consistently improves robustness to noisy preference signals compared to existing DPO variants.

cs.LG↗

arXiv2025

Preference Robustness for DPO with Applications to Public Health

We study an LLM fine-tuning task for designing reward functions for sequential resource allocation problems in public health, guided by human preferences expressed in natural language. This setting presents a challenging testbed for alignment due to complex and ambiguous objectives and limited data availability. We propose DPO-PRO, a robust fine-tuning algorithm based on Direct Preference Optimization (DPO), which accounts for uncertainty in the preference distribution using a lightweight Distributionally Robust Optimization (DRO) formulation. Unlike prior DRO-based DPO methods, DPO-PRO is significantly less conservative. We evaluate DPO-PRO on a real-world maternal mobile health program operated by the non-profit organization ARMMAN, as well as on standard alignment benchmarks. Experimental results demonstrate that our method consistently improves robustness to noisy preference signals compared to existing DPO variants. Moreover, DPO-PRO achieves comparable performance to prior self-reflection-based baseline for reward function design, while requiring significantly lower inference-time cost.

cs.LG↗

arXiv2025

Rule-Bottleneck Reinforcement Learning: Joint Explanation and Decision Optimization for Resource Allocation with Language Agents

Deep Reinforcement Learning (RL) is remarkably effective in addressing sequential resource allocation problems in domains such as healthcare, public policy, and resource management. However, deep RL policies often lack transparency and adaptability, challenging their deployment alongside human decision-makers. In contrast, Language Agents, powered by large language models (LLMs), provide human-understandable reasoning but may struggle with effective decision making. To bridge this gap, we propose Rule-Bottleneck Reinforcement Learning (RBRL), a novel framework that jointly optimizes decision and explanations. At each step, RBRL generates candidate rules with an LLM, selects among them using an attention-based RL policy, and determines the environment action with an explanation via chain-of-thought reasoning. The RL rule selection is optimized using the environment rewards and an explainability metric judged by the LLM. Evaluations in real-world scenarios highlight RBRL's competitive performance with deep RL and efficiency gains over LLM fine-tuning. A survey further confirms the enhanced quality of its explanations.

cs.LG↗

arXiv2024

An Instrumental Variables Framework to Unite Spatial Confounding Methods

Studies investigating the causal effects of spatially varying exposures on outcomes often rely on observational and spatially indexed data. A prevalent challenge is unmeasured spatial confounding, where an unobserved spatially varying variable affects both exposure and outcome, leading to biased estimates and invalid confidence intervals. There is a very large literature on spatial statistics that attempts to address unmeasured spatial confounding bias; most of this literature is not framed in the context of causal inference and relies on strict assumptions. In this paper, we propose an instrumental variables (IV) framework that unifies and extends existing methods for addressing unmeasured spatial confounding bias. This framework reveals that many spatial confounding methods can be viewed as IV methods, in which small-scale spatial variation in exposure operates as the instrument, providing a common theoretical foundation for approaches that previously appeared distinct. The framework clarifies that these methods share a common set of assumptions and differ primarily in how small-scale spatial variation is defined, while offering a general strategy for constructing instruments. It also extends to identify and estimate a broad class of causal effects, including the exposure response curve, without requiring a linear outcome model. We apply our methodology in simulation and to a national data set of 33,255 zip codes to estimate the effect of enforcing air pollution exposure levels below 6-12 $\mu g/m^3$ on all-cause mortality while adjusting for unmeasured spatial confounding.

stat.ME↗

arXiv2024

ICML Topological Deep Learning Challenge 2024: Beyond the Graph Domain

This paper describes the 2nd edition of the ICML Topological Deep Learning Challenge that was hosted within the ICML 2024 ELLIS Workshop on Geometry-grounded Representation Learning and Generative Modeling (GRaM). The challenge focused on the problem of representing data in different discrete topological domains in order to bridge the gap between Topological Deep Learning (TDL) and other types of structured datasets (e.g. point clouds, graphs). Specifically, participants were asked to design and implement topological liftings, i.e. mappings between different data structures and topological domains --like hypergraphs, or simplicial/cell/combinatorial complexes. The challenge received 52 submissions satisfying all the requirements. This paper introduces the main scope of the challenge, and summarizes the main results and findings.

cs.LG↗

arXiv2024

TopoBench: A Framework for Benchmarking Topological Deep Learning

This work introduces TopoBench, an open-source library designed to standardize benchmarking and accelerate research in topological deep learning (TDL). TopoBench decomposes TDL into a sequence of independent modules for data generation, loading, transforming and processing, as well as model training, optimization and evaluation. This modular organization provides flexibility for modifications and facilitates the adaptation and optimization of various TDL pipelines. A key feature of TopoBench is its support for transformations and lifting across topological domains. Mapping the topology and features of a graph to higher-order topological domains, such as simplicial and cell complexes, enables richer data representations and more fine-grained analyses. The applicability of TopoBench is demonstrated by benchmarking several TDL architectures across diverse tasks and datasets.

cs.LG↗

arXiv2024

E(n) Equivariant Topological Neural Networks

Graph neural networks excel at modeling pairwise interactions, but they cannot flexibly accommodate higher-order interactions and features. Topological deep learning (TDL) has emerged recently as a promising tool for addressing this issue. TDL enables the principled modeling of arbitrary multi-way, hierarchical higher-order interactions by operating on combinatorial topological spaces, such as simplicial or cell complexes, instead of graphs. However, little is known about how to leverage geometric features such as positions and velocities for TDL. This paper introduces E(n)-Equivariant Topological Neural Networks (ETNNs), which are E(n)-equivariant message-passing networks operating on combinatorial complexes, formal objects unifying graphs, hypergraphs, simplicial, path, and cell complexes. ETNNs incorporate geometric node features while respecting rotation, reflection, and translation equivariance. Moreover, being TDL models, ETNNs are natively ready for settings with heterogeneous interactions. We provide a theoretical analysis to show the improved expressiveness of ETNNs over architectures for geometric graphs. We also show how E(n)-equivariant variants of TDL models can be directly derived from our framework. The broad applicability of ETNNs is demonstrated through two tasks of vastly different scales: i) molecular property prediction on the QM9 benchmark and ii) land-use regression for hyper-local estimation of air pollution with multi-resolution irregular geospatial data. The results indicate that ETNNs are an effective tool for learning from diverse types of richly structured data, as they match or surpass SotA equivariant TDL models with a significantly smaller computational burden, thus highlighting the benefits of a principled geometric inductive bias. Our implementation of ETNNs can be found at https://github.com/NSAPH-Projects/topological-equivariant-networks.

cs.LG↗

arXiv2023

Optimizing Heat Alert Issuance with Reinforcement Learning

A key strategy in societal adaptation to climate change is using alert systems to prompt preventative action and reduce the adverse health impacts of extreme heat events. This paper implements and evaluates reinforcement learning (RL) as a tool to optimize the effectiveness of such systems. Our contributions are threefold. First, we introduce a new publicly available RL environment enabling the evaluation of the effectiveness of heat alert policies to reduce heat-related hospitalizations. The rewards model is trained from a comprehensive dataset of historical weather, Medicare health records, and socioeconomic/geographic features. We use scalable Bayesian techniques tailored to the low-signal effects and spatial heterogeneity present in the data. The transition model uses real historical weather patterns enriched by a data augmentation mechanism based on climate region similarity. Second, we use this environment to evaluate standard RL algorithms in the context of heat alert issuance. Our analysis shows that policy constraints are needed to improve RL's initially poor performance. Third, a post-hoc contrastive analysis provides insight into scenarios where our modified heat alert-RL policies yield significant gains/losses over the current National Weather Service alert policy in the United States.

cs.LG↗

arXiv2023

SpaCE: The Spatial Confounding Environment

Spatial confounding poses a significant challenge in scientific studies involving spatial data, where unobserved spatial variables can influence both treatment and outcome, possibly leading to spurious associations. To address this problem, we introduce SpaCE: The Spatial Confounding Environment, the first toolkit to provide realistic benchmark datasets and tools for systematically evaluating causal inference methods designed to alleviate spatial confounding. Each dataset includes training data, true counterfactuals, a spatial graph with coordinates, and smoothness and confounding scores characterizing the effect of a missing spatial confounder. It also includes realistic semi-synthetic outcomes and counterfactuals, generated using state-of-the-art machine learning ensembles, following best practices for causal inference benchmarks. The datasets cover real treatment and covariates from diverse domains, including climate, health and social sciences. SpaCE facilitates an automated end-to-end pipeline, simplifying data loading, experimental setup, and evaluating machine learning and causal inference models. The SpaCE project provides several dozens of datasets of diverse sizes and spatial complexity. It is publicly available as a Python package, encouraging community feedback and contributions.

cs.LG↗

arXiv2023

Causal Estimation of Exposure Shifts with Neural Networks

A fundamental task in causal inference is estimating the effect of distribution shift in the treatment variable. We refer to this problem as shift-response function (SRF) estimation. Existing neural network methods for causal inference lack theoretical guarantees and practical implementations for SRF estimation. In this paper, we introduce Targeted Regularization for Exposure Shifts with Neural Networks (TRESNET), a method to estimate SRFs with robustness and efficiency guarantees. Our contributions are twofold. First, we propose a targeted regularization loss for neural networks with theoretical properties that ensure double robustness and asymptotic efficiency specific to SRF estimation. Second, we extend targeted regularization to support loss functions from the exponential family to accommodate non-continuous outcome distributions (e.g., discrete counts). We conduct benchmark experiments demonstrating TRESNET's broad applicability and competitiveness. We then apply our method to a key policy question in public health to estimate the causal effect of revising the US National Ambient Air Quality Standards (NAAQS) for PM 2.5 from 12 ${\mu}g/m^3$ to 9 ${\mu}g/m^3$. This change has been recently proposed by the US Environmental Protection Agency (EPA). Our goal is to estimate the reduction in deaths that would result from this anticipated revision using data consisting of 68 million individuals across the U.S.

cs.LG↗

arXiv2022

Targeted active learning for probabilistic models

A fundamental task in science is to design experiments that yield valuable insights about the system under study. Mathematically, these insights can be represented as a utility or risk function that shapes the value of conducting each experiment. We present PDBAL, a targeted active learning method that adaptively designs experiments to maximize scientific utility. PDBAL takes a user-specified risk function and combines it with a probabilistic model of the experimental outcomes to choose designs that rapidly converge on a high-utility model. We prove theoretical bounds on the label complexity of PDBAL and provide fast closed-form solutions for designing experiments with common exponential family likelihoods. In simulation studies, PDBAL consistently outperforms standard untargeted approaches that focus on maximizing expected information gain over the design space. Finally, we demonstrate the scientific potential of PDBAL through a study on a large cancer drug screen dataset where PDBAL quickly recovers the most efficacious drugs with a small fraction of the total number of experiments.

cs.LG↗

arXiv2022

Weather2vec: Representation Learning for Causal Inference with Non-Local Confounding in Air Pollution and Climate Studies

Estimating the causal effects of a spatially-varying intervention on a spatially-varying outcome may be subject to non-local confounding (NLC), a phenomenon that can bias estimates when the treatments and outcomes of a given unit are dictated in part by the covariates of other nearby units. In particular, NLC is a challenge for evaluating the effects of environmental policies and climate events on health-related outcomes such as air pollution exposure. This paper first formalizes NLC using the potential outcomes framework, providing a comparison with the related phenomenon of causal interference. Then, it proposes a broadly applicable framework, termed "weather2vec", that uses the theory of balancing scores to learn representations of non-local information into a scalar or vector defined for each observational unit, which is subsequently used to adjust for confounding in conjunction with causal inference methods. The framework is evaluated in a simulation study and two case studies on air pollution where the weather is an (inherently regional) known confounder.

cs.LG↗

arXiv2022

A Comparative Tutorial of Bayesian Sequential Design and Reinforcement Learning

Reinforcement Learning (RL) is a computational approach to reward-driven learning in sequential decision problems. It implements the discovery of optimal actions by learning from an agent interacting with an environment rather than from supervised data. We contrast and compare RL with traditional sequential design, focusing on simulation-based Bayesian sequential design (BSD). Recently, there has been an increasing interest in RL techniques for healthcare applications. We introduce two related applications as motivating examples. In both applications, the sequential nature of the decisions is restricted to sequential stopping. Rather than a comprehensive survey, the focus of the discussion is on solutions using standard tools for these two relatively simple sequential stopping problems. Both problems are inspired by adaptive clinical trial design. We use examples to explain the terminology and mathematical background that underlie each framework and map one to the other. The implementations and results illustrate the many similarities between RL and BSD. The results motivate the discussion of the potential strengths and limitations of each approach.

stat.ME↗

arXiv2022

Bayesian Nonparametric Adjustment of Confounding

Analysis of observational studies increasingly confronts the challenge of determining which of a possibly high-dimensional set of available covariates are required to satisfy the assumption of ignorable treatment assignment for estimation of causal effects. We propose a Bayesian nonparametric approach that simultaneously 1) prioritizes inclusion of adjustment variables in accordance with existing principles of confounder selection; 2) estimates causal effects in a manner that permits complex relationships among confounders, exposures, and outcomes; and 3) provides causal estimates that account for uncertainty in the nature of confounding. The proposal relies on specification of multiple Bayesian Additive Regression Trees models, linked together with a common prior distribution that accrues posterior selection probability to covariates on the basis of association with both the exposure and the outcome of interest. A set of extensive simulation studies demonstrates that the proposed method performs well relative to similarly-motivated methodologies in a variety of scenarios. We deploy the method to investigate the causal effect of emissions from coal-fired power plants on ambient air pollution concentrations, where the prospect of confounding due to local and regional meteorological factors introduces uncertainty around the confounding role of a high-dimensional set of measured variables. Ultimately, we show that the proposed method produces more efficient and more consistent results across adjacent years than alternative methods, lending strength to the evidence of the causal relationship between SO2 emissions and ambient particulate pollution.

stat.ME↗

arXiv2022

Estimating Importation Risk of Covid-19 in Hurricane Evacuations: A Prediction Framework Applied to Hurricane Laura in Texas

In August 2020, as Texas was coming down from a large summer COVID-19 surge, forecasts suggested that Hurricane Laura was tracking towards 6M residents along the East Texas coastline, threatening to spread COVID-19 across the state and cause pandemic resurgences. To assist local authorities facing the dual-threat, we integrated survey expectations of coastal residents and observed hurricane evacuation rates in a statistical framework that combined with local pandemic conditions predicts how COVID-19 would spread in response to a hurricane. For Hurricane Laura, we estimate that 499,500 [90% Credible Interval (CI): 347,500, 624,000] people evacuated the Texan counties, that no single county accumulated more than 2.5% of hurricane evacuees, and that there were 2,900 [90% CI: 1,700, 5,800] exportations of Covid-19 across the state. In general, reception estimates were concentrated in regions with higher population densities. Nonetheless, higher importation risk is expected in small Districts, with a maximum number of importations of 10 per 10,000 residents in our case study. Overall, we present a flexible and transferable framework that captures spatial heterogeneity and incorporates geographic components for predicting population movement in the wake of a natural disaster. As hurricanes continue to increase in both frequency and strength, our framework can be deployed in response to anticipated hurricane paths to guide disaster preparedness and planning.

stat.AP↗

arXiv2021

Adversarial Intrinsic Motivation for Reinforcement Learning

Learning with an objective to minimize the mismatch with a reference distribution has been shown to be useful for generative modeling and imitation learning. In this paper, we investigate whether one such objective, the Wasserstein-1 distance between a policy's state visitation distribution and a target distribution, can be utilized effectively for reinforcement learning (RL) tasks. Specifically, this paper focuses on goal-conditioned reinforcement learning where the idealized (unachievable) target distribution has full measure at the goal. This paper introduces a quasimetric specific to Markov Decision Processes (MDPs) and uses this quasimetric to estimate the above Wasserstein-1 distance. It further shows that the policy that minimizes this Wasserstein-1 distance is the policy that reaches the goal in as few steps as possible. Our approach, termed Adversarial Intrinsic Motivation (AIM), estimates this Wasserstein-1 distance through its dual objective and uses it to compute a supplemental reward function. Our experiments show that this reward function changes smoothly with respect to transitions in the MDP and directs the agent's exploration to find the goal efficiently. Additionally, we combine AIM with Hindsight Experience Replay (HER) and show that the resulting algorithm accelerates learning significantly on several simulated robotics tasks when compared to other rewards that encourage exploration or accelerate learning.

cs.LG↗