Shresth Verma

Arshika Lalan, Shresth Verma, Kumar Madhu Sudan et al.

Mobile health programs are becoming an increasingly popular medium for dissemination of health information among beneficiaries in less privileged communities. Kilkari is one of the world's largest mobile health programs which delivers time sensitive audio-messages to pregnant women and new mothers. We have been collaborating with ARMMAN, a non-profit in India which operates the Kilkari program, to identify bottlenecks to improve the efficiency of the program. In particular, we provide an initial analysis of the trajectories of beneficiaries' interaction with the mHealth program and examine elements of the program that can be potentially enhanced to boost its success. We cluster the cohort into different buckets based on listenership so as to analyze listenership patterns for each group that could help boost program success. We also demonstrate preliminary results on using historical data in a time-series prediction to identify beneficiary dropouts and enable NGOs in devising timely interventions to strengthen beneficiary retention.

Shresth Verma, Niclas Boehmer, Lingkai Kong et al.

LLMs are increasingly used to design reward functions based on human preferences in Reinforcement Learning (RL). We focus on LLM-designed rewards for Restless Multi-Armed Bandits, a framework for allocating limited resources among agents. In applications such as public health, this approach empowers grassroots health workers to tailor automated allocation decisions to community needs. In the presence of multiple agents, altering the reward function based on human preferences can impact subpopulations very differently, leading to complex tradeoffs and a multi-objective resource allocation problem. We are the first to present a principled method termed Social Choice Language Model for dealing with these tradeoffs for LLM-designed rewards for multiagent planners in general and restless bandits in particular. The novel part of our model is a transparent and configurable selection component, called an adjudicator, external to the LLM that controls complex tradeoffs via a user-selected social welfare function. Our experiments demonstrate that our model reliably selects more effective, aligned, and balanced reward functions compared to purely LLM-based approaches.

Shresth Verma, Mauricio Tec, Cheol Woo Kim et al.

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.

Paritosh Verma, Shresth Verma, Aditya Mate et al.

Restless multi-arm bandits (RMABs) is a popular decision-theoretic framework that has been used to model real-world sequential decision making problems in public health, wildlife conservation, communication systems, and beyond. Deployed RMAB systems typically operate in two stages: the first predicts the unknown parameters defining the RMAB instance, and the second employs an optimization algorithm to solve the constructed RMAB instance. In this work we provide and analyze the results from a first-of-its-kind deployment of an RMAB system in public health domain, aimed at improving maternal and child health. Our analysis is focused towards understanding the relationship between prediction accuracy and overall performance of deployed RMAB systems. This is crucial for determining the value of investing in improving predictive accuracy towards improving the final system performance, and is useful for diagnosing, monitoring deployed RMAB systems. Using real-world data from our deployed RMAB system, we demonstrate that an improvement in overall prediction accuracy may even be accompanied by a degradation in the performance of RMAB system -- a broad investment of resources to improve overall prediction accuracy may not yield expected results. Following this, we develop decision-focused evaluation metrics to evaluate the predictive component and show that it is better at explaining (both empirically and theoretically) the overall performance of a deployed RMAB system.

Shresth Verma, Arpan Dasgupta, Neha Madhiwalla et al.

Maternal and child health is a critical concern around the world. In many global health programs disseminating preventive care and health information, limited healthcare worker resources prevent continuous, personalised engagement with vulnerable beneficiaries. In such scenarios, it becomes crucial to optimally schedule limited live-service resources to maximise long-term engagement. To address this fundamental challenge, the multi-year SAHELI project (2020-2025), in collaboration with partner NGO ARMMAN, leverages AI to allocate scarce resources in a maternal and child health program in India. The SAHELI system solves this sequential resource allocation problem using a Restless Multi-Armed Bandit (RMAB) framework. A key methodological innovation is the transition from a traditional Two-Stage "predict-then-optimize" approach to Decision-Focused Learning (DFL), which directly aligns the framework's learning method with the ultimate goal of maximizing beneficiary engagement. Empirical evaluation through large-scale randomized controlled trials demonstrates that the DFL policy reduced cumulative engagement drops by 31% relative to the current standard of care, significantly outperforming the Two-Stage model. Crucially, the studies also confirmed that this increased program engagement translates directly into statistically significant improvements in real-world health behaviors, notably the continued consumption of vital iron and calcium supplements by new mothers. Ultimately, the SAHELI project provides a scalable blueprint for applying sequential decision-making AI to optimize resource allocation in health programs.

Arshika Lalan, Shresth Verma, Paula Rodriguez Diaz et al.

Harnessing the wide-spread availability of cell phones, many nonprofits have launched mobile health (mHealth) programs to deliver information via voice or text to beneficiaries in underserved communities, with maternal and infant health being a key area of such mHealth programs. Unfortunately, dwindling listenership is a major challenge, requiring targeted interventions using limited resources. This paper focuses on Kilkari, the world's largest mHealth program for maternal and child care - with over 3 million active subscribers at a time - launched by India's Ministry of Health and Family Welfare (MoHFW) and run by the non-profit ARRMAN. We present a system called CHAHAK that aims to reduce automated dropouts as well as boost engagement with the program through the strategic allocation of interventions to beneficiaries. Past work in a similar domain has focused on a much smaller scale mHealth program and used markovian restless multiarmed bandits to optimize a single limited intervention resource. However this paper demonstrates the challenges in adopting a markovian approach in Kilkari; therefore CHAHAK instead relies on non-markovian time-series restless bandits, and optimizes multiple interventions to improve listenership. We use real Kilkari data from the Odisha state in India to show CHAHAK's effectiveness in harnessing multiple interventions to boost listenership, benefiting marginalized communities. When deployed CHAHAK will assist the largest maternal mHealth program to date.

Cheol Woo Kim, Jai Moondra, Shresth Verma et al.

In many real-world applications of reinforcement learning (RL), deployed policies have varied impacts on different stakeholders, creating challenges in reaching consensus on how to effectively aggregate their preferences. Generalized $p$-means form a widely used class of social welfare functions for this purpose, with broad applications in fair resource allocation, AI alignment, and decision-making. This class includes well-known welfare functions such as Egalitarian, Nash, and Utilitarian welfare. However, selecting the appropriate social welfare function is challenging for decision-makers, as the structure and outcomes of optimal policies can be highly sensitive to the choice of $p$. To address this challenge, we study the concept of an $α$-approximate portfolio in RL, a set of policies that are approximately optimal across the family of generalized $p$-means for all $p \in [-\infty, 1]$. We propose algorithms to compute such portfolios and provide theoretical guarantees on the trade-offs among approximation factor, portfolio size, and computational efficiency. Experimental results on synthetic and real-world datasets demonstrate the effectiveness of our approach in summarizing the policy space induced by varying $p$ values, empowering decision-makers to navigate this landscape more effectively.

Deniz Sezin Ayvaz, Lorenzo Belenguer, Hankun He et al.

Mastercard, a global leader in financial services, develops and deploys machine learning models aimed at optimizing card usage and preventing attrition through advanced predictive models. These models use aggregated and anonymized card usage patterns, including cross-border transactions and industry-specific spending, to tailor bank offerings and maximize revenue opportunities. Mastercard has established an AI Governance program, based on its Data and Tech Responsibility Principles, to evaluate any built and bought AI for efficacy, fairness, and transparency. As part of this effort, Mastercard has sought expertise from the Turing Institute through a Data Study Group to better assess fairness in more complex AI/ML models. The Data Study Group challenge lies in defining, measuring, and mitigating fairness in these predictions, which can be complex due to the various interpretations of fairness, gaps in the research literature, and ML-operations challenges.

Cheol Woo Kim, Shresth Verma, Mauricio Tec et al.

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.

Cheol Woo Kim, Shresth Verma, Mauricio Tec et al.

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.

Panayiotis Danassis, Shresth Verma, Jackson A. Killian et al.

The success of many healthcare programs depends on participants' adherence. We consider the problem of scheduling interventions in low resource settings (e.g., placing timely support calls from health workers) to increase adherence and/or engagement. Past works have successfully developed several classes of Restless Multi-armed Bandit (RMAB) based solutions for this problem. Nevertheless, all past RMAB approaches assume that the participants' behaviour follows the Markov property. We demonstrate significant deviations from the Markov assumption on real-world data on a maternal health awareness program from our partner NGO, ARMMAN. Moreover, we extend RMABs to continuous state spaces, a previously understudied area. To tackle the generalised non-Markovian RMAB setting we (i) model each participant's trajectory as a time-series, (ii) leverage the power of time-series forecasting models to learn complex patterns and dynamics to predict future states, and (iii) propose the Time-series Arm Ranking Index (TARI) policy, a novel algorithm that selects the RMAB arms that will benefit the most from an intervention, given our future state predictions. We evaluate our approach on both synthetic data, and a secondary analysis on real data from ARMMAN, and demonstrate significant increase in engagement compared to the SOTA, deployed Whittle index solution. This translates to 16.3 hours of additional content listened, 90.8% more engagement drops prevented, and reaching more than twice as many high dropout-risk beneficiaries.

Kai Wang, Shresth Verma, Aditya Mate et al.

This paper studies restless multi-armed bandit (RMAB) problems with unknown arm transition dynamics but with known correlated arm features. The goal is to learn a model to predict transition dynamics given features, where the Whittle index policy solves the RMAB problems using predicted transitions. However, prior works often learn the model by maximizing the predictive accuracy instead of final RMAB solution quality, causing a mismatch between training and evaluation objectives. To address this shortcoming, we propose a novel approach for decision-focused learning in RMAB that directly trains the predictive model to maximize the Whittle index solution quality. We present three key contributions: (i) we establish differentiability of the Whittle index policy to support decision-focused learning; (ii) we significantly improve the scalability of decision-focused learning approaches in sequential problems, specifically RMAB problems; (iii) we apply our algorithm to a previously collected dataset of maternal and child health to demonstrate its performance. Indeed, our algorithm is the first for decision-focused learning in RMAB that scales to real-world problem sizes.

Aditya Mate, Lovish Madaan, Aparna Taneja et al.

The widespread availability of cell phones has enabled non-profits to deliver critical health information to their beneficiaries in a timely manner. This paper describes our work to assist non-profits that employ automated messaging programs to deliver timely preventive care information to beneficiaries (new and expecting mothers) during pregnancy and after delivery. Unfortunately, a key challenge in such information delivery programs is that a significant fraction of beneficiaries drop out of the program. Yet, non-profits often have limited health-worker resources (time) to place crucial service calls for live interaction with beneficiaries to prevent such engagement drops. To assist non-profits in optimizing this limited resource, we developed a Restless Multi-Armed Bandits (RMABs) system. One key technical contribution in this system is a novel clustering method of offline historical data to infer unknown RMAB parameters. Our second major contribution is evaluation of our RMAB system in collaboration with an NGO, via a real-world service quality improvement study. The study compared strategies for optimizing service calls to 23003 participants over a period of 7 weeks to reduce engagement drops. We show that the RMAB group provides statistically significant improvement over other comparison groups, reducing ~ 30% engagement drops. To the best of our knowledge, this is the first study demonstrating the utility of RMABs in real world public health settings. We are transitioning our RMAB system to the NGO for real-world use.

Shresth Verma, Haritha S. Nair, Gaurav Agarwal et al.

Robotics has proved to be an indispensable tool in many industrial as well as social applications, such as warehouse automation, manufacturing, disaster robotics, etc. In most of these scenarios, damage to the agent while accomplishing mission-critical tasks can result in failure. To enable robotic adaptation in such situations, the agent needs to adopt policies which are robust to a diverse set of damages and must do so with minimum computational complexity. We thus propose a damage aware control architecture which diagnoses the damage prior to gait selection while also incorporating domain randomization in the damage space for learning a robust policy. To implement damage awareness, we have used a Long Short Term Memory based supervised learning network which diagnoses the damage and predicts the type of damage. The main novelty of this approach is that only a single policy is trained to adapt against a wide variety of damages and the diagnosis is done in a single trial at the time of damage.

Raj Kuwar Gupta, Shresth Verma, KV Arya et al.

This paper addresses the challenges of face attribute detection specifically in the Indian context. While there are numerous face datasets in unconstrained environments, none of them captures emotions in different face orientations. Moreover, there is an under-representation of people of Indian ethnicity in these datasets since they have been scraped from popular search engines. As a result, the performance of state-of-the-art techniques can't be evaluated on Indian faces. In this work, we introduce a new dataset, IIITM Face, for the scientific community to address these challenges. Our dataset includes 107 participants who exhibit 6 emotions in 3 different face orientations. Each of these images is further labelled on attributes like gender, presence of moustache, beard or eyeglasses, clothes worn by the subjects and the density of their hair. Moreover, the images are captured in high resolution with specific background colors which can be easily replaced by cluttered backgrounds to simulate `in the Wild' behaviour. We demonstrate the same by constructing IIITM Face-SUE. Both IIITM Face and IIITM Face-SUE have been benchmarked across key multi-label metrics for the research community to compare their results.

Shresth Verma, Joydip Dhar

Learning to communicate is considered an essential task to develop a general AI. While recent literature in language evolution has studied emergent language through discrete or continuous message symbols, there has been little work in the emergence of writing systems in artificial agents. In this paper, we present a referential game setup with two agents, where the mode of communication is a written language system that emerges during the play. We show that the agents can learn to coordinate successfully using this mode of communication. Further, we study how the game rules affect the writing system taxonomy by proposing a consistency metric.