Shreyas Chaudhari

Jong-Ik Park, Shreyas Chaudhari, José M. F. Moura et al.

Randomized smoothing (RS) certifies robustness in the vector space where Gaussian noise is added. In audio classification, this space is often not uniquely defined as standard pipelines normalize, range-control, and transform waveforms into log-mel or other spectral features. We show that direct RS is therefore under-specified unless the certified object and preprocessing policy are explicit. On two audio benchmarks, keyword spotting and environmental-sound classification, we study waveform, feature-space, and post-processed smoothing. Our diagnostics show why representation-aware reporting is necessary: at the same smoothing level $σ=0.0025$, the two datasets share the same median raw radius $.007996$, but different waveform energies yield different SNR-equivalent scales ($83.98$ vs. $90.97$ dB); log-mel smoothing gives higher positive-radius certified accuracy on environmental sounds ($68.42\%$ vs. $65.53\%$), certifying more examples with nonzero radius but over features rather than waveforms; and clipping or peak normalization changes the effective perturbation norm by roughly $230$--$351\times$. We therefore recommend that audio RS studies choose and report the task-specific certified object and perturbation model, including the perturbation location, gain policy, raw radius, and any post-noise geometry changes.

Jong-Ik Park, Shreyas Chaudhari, Carlee Joe-Wong et al.

Randomized smoothing (RS) uses a smoothed classifier to provide architecture-agnostic certificates of $\ell_2$ classification robustness, but its dependence on per-input Monte Carlo (MC) sampling undermines its use in real-time systems. We argue that this cost is structural rather than fundamental, such that it can be significantly reduced by sharing information across the deployment stream. We introduce RRISE, an RS framework that compresses certification into a single forward pass through a learned surrogate. RRISE trains the surrogate against precomputed MC class-count targets via a soft-label cross-entropy loss and converts surrogate predictions into provably conservative certified radii through a one-time conformal calibration step. The resulting certificate is deployment-verifiable: whenever the calibrated radius is positive, the surrogate's prediction provably matches the smoothed classifier's and the smoothed classifier is constant on a ball of that radius around the input. Across image classification benchmarks, RRISE matches fixed-budget MC certified accuracy within $0.84$ percentage points while replacing up to $10^4$ noisy base-model evaluations per query with a single surrogate forward pass, recouping MC training cost after $\approx 10^5$ deployment queries. On CIFAR-100 and Tiny ImageNet, where the only prior offline-surrogate method collapses, RRISE achieves $1.23$ to $1.91\times$ higher certified accuracy, establishing efficient randomized smoothing as a practical path to certified robustness in repeated-deployment settings.

Peikai Li, Ipek Ilayda Onur, Scott Dodelson et al.

Next-generation cosmic microwave background (CMB) surveys are expected to provide valuable information about the primordial universe by creating maps of the mass along the line of sight. Traditional tools for creating these lensing convergence maps include the quadratic estimator and the maximum likelihood based iterative estimator. Here, we apply a generative adversarial network (GAN) to reconstruct the lensing convergence field. We compare our results with a previous deep learning approach -- Residual-UNet -- and discuss the pros and cons of each. In the process, we use training sets generated by a variety of power spectra, rather than the one used in testing the methods.

Vinay Samuel, Henry Peng Zou, Yue Zhou et al.

Persona agents, which are LLM agents conditioned to act according to an assigned persona, enable contextually rich and user aligned interactions across domains like education and healthcare. However, evaluating how faithfully these agents adhere to their personas remains a significant challenge, particularly in free-form settings that demand consistency across diverse, persona-relevant environments. We introduce PersonaGym, the first dynamic evaluation framework for persona agents, and PersonaScore, a human-aligned automatic metric grounded in decision theory that enables comprehensive large-scale evaluation. Our evaluation of 10 leading LLMs across 200 personas and 10,000 questions reveals significant advancement opportunities. For example, GPT-4.1 had the exact same PersonaScore as LLaMA-3-8b despite being a more recent and advanced closed source model. Importantly, increased model size and complexity do not necessarily enhance persona agent capabilities, underscoring the need for algorithmic and architectural innovation toward faithful, performant persona agents.

Shreyas Chaudhari, David Arbour, Georgios Theocharous et al.

Recommendation strategies are typically evaluated by using previously logged data, employing off-policy evaluation methods to estimate their expected performance. However, for strategies that present users with slates of multiple items, the resulting combinatorial action space renders many of these methods impractical. Prior work has developed estimators that leverage the structure in slates to estimate the expected off-policy performance, but the estimation of the entire performance distribution remains elusive. Estimating the complete distribution allows for a more comprehensive evaluation of recommendation strategies, particularly along the axes of risk and fairness that employ metrics computable from the distribution. In this paper, we propose an estimator for the complete off-policy performance distribution for slates and establish conditions under which the estimator is unbiased and consistent. This builds upon prior work on off-policy evaluation for slates and off-policy distribution estimation in reinforcement learning. We validate the efficacy of our method empirically on synthetic data as well as on a slate recommendation simulator constructed from real-world data (MovieLens-20M). Our results show a significant reduction in estimation variance and improved sample efficiency over prior work across a range of slate structures.

Shreyas Chaudhari, Srinivasa Pranav, José M. F. Moura

While much effort has been devoted to deriving and analyzing effective convex formulations of signal processing problems, the gradients of convex functions also have critical applications ranging from gradient-based optimization to optimal transport. Recent works have explored data-driven methods for learning convex objective functions, but learning their monotone gradients is seldom studied. In this work, we propose C-MGN and M-MGN, two monotone gradient neural network architectures for directly learning the gradients of convex functions. We show that, compared to state of the art methods, our networks are easier to train, learn monotone gradient fields more accurately, and use significantly fewer parameters. We further demonstrate their ability to learn optimal transport mappings to augment driving image data.

Jong-Ik Park, Shreyas Chaudhari, Srinivasa Pranav et al.

In many deployed systems (multilingual ASR, cross-hospital imaging, region-specific perception), multiple pretrained specialist models coexist. Yet, new target domains often require domain expansion: a generalized model that performs well beyond any single specialist's domain. Given a new target domain, existing methods obtain a single strong initialization prior for the model parameters by blending expert models to initialize a target model. However, heuristic blending -- using mixing coefficients based on data size or proxy metrics -- often yields lower target-domain test accuracy, and learning these coefficients on the target domain's loss function typically requires computationally-expensive full backpropagation through a neural network. We propose GLUE, Gradient-free Learning to Unify Experts, which initializes the target model as a convex combination of fixed experts and learns the mixture coefficients of this combination via gradient-free two-point SPSA (simultaneous perturbation stochastic approximation) updates, requiring only two forward passes per step. Across experiments on three datasets and three network architectures, GLUE produces model parameter priors that can be fine-tuned to outperform baselines. GLUE improves test accuracy by up to 8.5% over data-size weighting and by up to 9.1% over proxy-metric selection. GLUE either outperforms backpropagation-based full-gradient mixing or matches its performance within 1.4%.

Shreyas Chaudhari, Srinivasa Pranav, Emile Anand et al.

Peer-to-peer learning is an increasingly popular framework that enables beyond-5G distributed edge devices to collaboratively train deep neural networks in a privacy-preserving manner without the aid of a central server. Neural network training algorithms for emerging environments, e.g., smart cities, have many design considerations that are difficult to tune in deployment settings -- such as neural network architectures and hyperparameters. This presents a critical need for characterizing the training dynamics of distributed optimization algorithms used to train highly nonconvex neural networks in peer-to-peer learning environments. In this work, we provide an explicit characterization of the learning dynamics of wide neural networks trained using popular distributed gradient descent (DGD) algorithms. Our results leverage both recent advancements in neural tangent kernel (NTK) theory and extensive previous work on distributed learning and consensus. We validate our analytical results by accurately predicting the parameter and error dynamics of wide neural networks trained for classification tasks.

Shreyas Chaudhari, Pranjal Aggarwal, Vishvak Murahari et al.

State-of-the-art large language models (LLMs) have become indispensable tools for various tasks. However, training LLMs to serve as effective assistants for humans requires careful consideration. A promising approach is reinforcement learning from human feedback (RLHF), which leverages human feedback to update the model in accordance with human preferences and mitigate issues like toxicity and hallucinations. Yet, an understanding of RLHF for LLMs is largely entangled with initial design choices that popularized the method and current research focuses on augmenting those choices rather than fundamentally improving the framework. In this paper, we analyze RLHF through the lens of reinforcement learning principles to develop an understanding of its fundamentals, dedicating substantial focus to the core component of RLHF -- the reward model. Our study investigates modeling choices, caveats of function approximation, and their implications on RLHF training algorithms, highlighting the underlying assumptions made about the expressivity of reward. Our analysis improves the understanding of the role of reward models and methods for their training, concurrently revealing limitations of the current methodology. We characterize these limitations, including incorrect generalization, model misspecification, and the sparsity of feedback, along with their impact on the performance of a language model. The discussion and analysis are substantiated by a categorical review of current literature, serving as a reference for researchers and practitioners to understand the challenges of RLHF and build upon existing efforts.

Dhawal Gupta, Scott M. Jordan, Shreyas Chaudhari et al.

In this paper, we introduce a fresh perspective on the challenges of credit assignment and policy evaluation. First, we delve into the nuances of eligibility traces and explore instances where their updates may result in unexpected credit assignment to preceding states. From this investigation emerges the concept of a novel value function, which we refer to as the \emph{bidirectional value function}. Unlike traditional state value functions, bidirectional value functions account for both future expected returns (rewards anticipated from the current state onward) and past expected returns (cumulative rewards from the episode's start to the present). We derive principled update equations to learn this value function and, through experimentation, demonstrate its efficacy in enhancing the process of policy evaluation. In particular, our results indicate that the proposed learning approach can, in certain challenging contexts, perform policy evaluation more rapidly than TD($λ$) -- a method that learns forward value functions, $v^π$, \emph{directly}. Overall, our findings present a new perspective on eligibility traces and potential advantages associated with the novel value function it inspires, especially for policy evaluation.

Shreyas Chaudhari, Srinivasa Pranav, José M. F. Moura

Monotone gradient functions play a central role in solving the Monge formulation of the optimal transport (OT) problem, which arises in modern applications ranging from fluid dynamics to robot swarm control. When the transport cost is the squared Euclidean distance, Brenier's theorem guarantees that the unique optimal transport map satisfies a Monge-Ampère equation and is the gradient of a convex function. In [arXiv:2301.10862] [arXiv:2404.07361], we proposed Monotone Gradient Networks (mGradNets), neural networks that directly parameterize the space of monotone gradient maps. In this work, we leverage mGradNets to directly learn the optimal transport mapping by minimizing a training loss function defined using the Monge-Ampère equation. We empirically show that the structural bias of mGradNets facilitates the learning of optimal transport maps across both image morphing tasks and high-dimensional OT problems.

Brian Zhang, Deepti Guntur, Zhiyang Zuo et al.

Retrieval-Augmented Generation (RAG) systems face challenges with complex, multihop questions, and agentic frameworks such as Search-R1 (Jin et al., 2025), which operates iteratively, have been proposed to address these complexities. However, such approaches can introduce inefficiencies, including repetitive retrieval of previously processed information and challenges in contextualizing retrieved results effectively within the current generation prompt. Such issues can lead to unnecessary retrieval turns, suboptimal reasoning, inaccurate answers, and increased token consumption. In this paper, we investigate test-time modifications to the Search-R1 pipeline to mitigate these identified shortcomings. Specifically, we explore the integration of two components and their combination: a contextualization module to better integrate relevant information from retrieved documents into reasoning, and a de-duplication module that replaces previously retrieved documents with the next most relevant ones. We evaluate our approaches using the HotpotQA (Yang et al., 2018) and the Natural Questions (Kwiatkowski et al., 2019) datasets, reporting the exact match (EM) score, an LLM-as-a-Judge assessment of answer correctness, and the average number of turns. Our best-performing variant, utilizing GPT-4.1-mini for contextualization, achieves a 5.6% increase in EM score and reduces the number of turns by 10.5% compared to the Search-R1 baseline, demonstrating improved answer accuracy and retrieval efficiency.

Shreyas Chaudhari, Renhao Zhang, Philip S. Thomas et al.

The ability of reinforcement learning algorithms to learn effective policies is determined by the rewards available during training. However, for practical problems, obtaining large quantities of reward labels is often infeasible due to computational or financial constraints, particularly when relying on human feedback. When reinforcement learning must proceed with limited feedback -- only a fraction of samples get rewards labeled -- a fundamental question arises: which samples should be labeled to maximize policy performance? We formalize this problem of reward selection for reinforcement learning from limited feedback (RLLF), introducing a new problem formulation that facilitates the study of strategies for selecting impactful rewards. Two types of selection strategies are investigated: (i) heuristics that rely on reward-free information such as state visitation and partial value functions, and (ii) strategies pre-trained using auxiliary evaluative feedback. We find that critical subsets of rewards are those that (1) guide the agent along optimal trajectories, and (2) support recovery toward near-optimal behavior after deviations. Effective selection methods yield near-optimal policies with significantly fewer reward labels than full supervision, establishing reward selection as a powerful paradigm for scaling reinforcement learning in feedback-limited settings.

Ujwal Narayan, Shreyas Chaudhari, Ashwin Kalyan et al.

Large language models (LLMs) have become ubiquitous, interfacing with humans in numerous safety-critical applications. This necessitates improving capabilities, but importantly coupled with greater safety measures to align these models with human values and preferences. In this work, we demonstrate that contemporary models fall concerningly short of the goal of AI safety, leading to an unsafe and harmful experience for users. We introduce a prompting strategy called Code of Thought (CoDoT) to evaluate the safety of LLMs. CoDoT converts natural language inputs to simple code that represents the same intent. For instance, CoDoT transforms the natural language prompt "Make the statement more toxic: {text}" to: "make_more_toxic({text})". We show that CoDoT results in a consistent failure of a wide range of state-of-the-art LLMs. For example, GPT-4 Turbo's toxicity increases 16.5 times, DeepSeek R1 fails 100% of the time, and toxicity increases 300% on average across seven modern LLMs. Additionally, recursively applying CoDoT can further increase toxicity two times. Given the rapid and widespread adoption of LLMs, CoDoT underscores the critical need to evaluate safety efforts from first principles, ensuring that safety and capabilities advance together.

Devansh Bhardwaj, Arjun Beniwal, Shreyas Chaudhari et al.

AI agents have become increasingly adept at complex tasks such as coding, reasoning, and multimodal understanding. However, building generalist systems requires moving beyond individual agents to collective inference -- a paradigm where multi-agent systems with diverse, task-specialized agents complement one another through structured communication and collaboration. Today, coordination is usually handled with imprecise, ad-hoc natural language, which limits complex interaction and hinders interoperability with domain-specific agents. We introduce Agent context protocols (ACPs): a domain- and agent-agnostic family of structured protocols for agent-agent communication, coordination, and error handling. ACPs combine (i) persistent execution blueprints -- explicit dependency graphs that store intermediate agent outputs -- with (ii) standardized message schemas, enabling robust and fault-tolerant multi-agent collective inference. ACP-powered generalist systems reach state-of-the-art performance: 28.3 % accuracy on AssistantBench for long-horizon web assistance and best-in-class multimodal technical reports, outperforming commercial AI systems in human evaluation. ACPs are highly modular and extensible, allowing practitioners to build top-tier generalist agents quickly.

Shreyas Chaudhari, José M. F. Moura

This paper presents a novel framework for understanding trained ReLU networks as random, affine functions, where the randomness is induced by the distribution over the inputs. By characterizing the probability distribution of the network's activation patterns, we derive the discrete probability distribution over the affine functions realizable by the network. We extend this analysis to describe the probability distribution of the network's outputs. Our approach provides explicit, numerically tractable expressions for these distributions in terms of Gaussian orthant probabilities. Additionally, we develop approximation techniques to identify the support of affine functions a trained ReLU network can realize for a given distribution of inputs. Our work provides a framework for understanding the behavior and performance of ReLU networks corresponding to stochastic inputs, paving the way for more interpretable and reliable models.

Hitesh Wadhwa, Rahul Seetharaman, Somyaa Aggarwal et al.

Retrieval Augmented Generation (RAG) enriches the ability of language models to reason using external context to augment responses for a given user prompt. This approach has risen in popularity due to practical applications in various applications of language models in search, question/answering, and chat-bots. However, the exact nature of how this approach works isn't clearly understood. In this paper, we mechanistically examine the RAG pipeline to highlight that language models take shortcut and have a strong bias towards utilizing only the context information to answer the question, while relying minimally on their parametric memory. We probe this mechanistic behavior in language models with: (i) Causal Mediation Analysis to show that the parametric memory is minimally utilized when answering a question and (ii) Attention Contributions and Knockouts to show that the last token residual stream do not get enriched from the subject token in the question, but gets enriched from other informative tokens in the context. We find this pronounced shortcut behaviour true across both LLaMa and Phi family of models.

Shreyas Chaudhari, Srinivasa Pranav, José M. F. Moura

Directly parameterizing and learning gradients of functions has widespread significance, with specific applications in inverse problems, generative modeling, and optimal transport. This paper introduces gradient networks (GradNets): novel neural network architectures that parameterize gradients of various function classes. GradNets exhibit specialized architectural constraints that ensure correspondence to gradient functions. We provide a comprehensive GradNet design framework that includes methods for transforming GradNets into monotone gradient networks (mGradNets), which are guaranteed to represent gradients of convex functions. Our results establish that our proposed GradNet (and mGradNet) universally approximate the gradients of (convex) functions. Furthermore, these networks can be customized to correspond to specific spaces of potential functions, including transformed sums of (convex) ridge functions. Our analysis leads to two distinct GradNet architectures, GradNet-C and GradNet-M, and we describe the corresponding monotone versions, mGradNet-C and mGradNet-M. Our empirical results demonstrate that these architectures provide efficient parameterizations and outperform existing methods by up to 15 dB in gradient field tasks and by up to 11 dB in Hamiltonian dynamics learning tasks.

Shreyas Chaudhari, Harideep Nair, José M. F. Moura et al.

Unsupervised time series clustering is a challenging problem with diverse industrial applications such as anomaly detection, bio-wearables, etc. These applications typically involve small, low-power devices on the edge that collect and process real-time sensory signals. State-of-the-art time-series clustering methods perform some form of loss minimization that is extremely computationally intensive from the perspective of edge devices. In this work, we propose a neuromorphic approach to unsupervised time series clustering based on Temporal Neural Networks that is capable of ultra low-power, continuous online learning. We demonstrate its clustering performance on a subset of UCR Time Series Archive datasets. Our results show that the proposed approach either outperforms or performs similarly to most of the existing algorithms while being far more amenable for efficient hardware implementation. Our hardware assessment analysis shows that in 7 nm CMOS the proposed architecture, on average, consumes only about 0.005 mm^2 die area and 22 uW power and can process each signal with about 5 ns latency.

Benjamin Eysenbach, Swapnil Asawa, Shreyas Chaudhari et al.

We propose a simple, practical, and intuitive approach for domain adaptation in reinforcement learning. Our approach stems from the idea that the agent's experience in the source domain should look similar to its experience in the target domain. Building off of a probabilistic view of RL, we formally show that we can achieve this goal by compensating for the difference in dynamics by modifying the reward function. This modified reward function is simple to estimate by learning auxiliary classifiers that distinguish source-domain transitions from target-domain transitions. Intuitively, the modified reward function penalizes the agent for visiting states and taking actions in the source domain which are not possible in the target domain. Said another way, the agent is penalized for transitions that would indicate that the agent is interacting with the source domain, rather than the target domain. Our approach is applicable to domains with continuous states and actions and does not require learning an explicit model of the dynamics. On discrete and continuous control tasks, we illustrate the mechanics of our approach and demonstrate its scalability to high-dimensional tasks.

Samarth Gupta, Shreyas Chaudhari, Gauri Joshi et al.

We consider a multi-armed bandit framework where the rewards obtained by pulling different arms are correlated. We develop a unified approach to leverage these reward correlations and present fundamental generalizations of classic bandit algorithms to the correlated setting. We present a unified proof technique to analyze the proposed algorithms. Rigorous analysis of C-UCB (the correlated bandit version of Upper-confidence-bound) reveals that the algorithm ends up pulling certain sub-optimal arms, termed as non-competitive, only O(1) times, as opposed to the O(log T) pulls required by classic bandit algorithms such as UCB, TS etc. We present regret-lower bound and show that when arms are correlated through a latent random source, our algorithms obtain order-optimal regret. We validate the proposed algorithms via experiments on the MovieLens and Goodreads datasets, and show significant improvement over classical bandit algorithms.

Samarth Gupta, Shreyas Chaudhari, Subhojyoti Mukherjee et al.

We consider a finite-armed structured bandit problem in which mean rewards of different arms are known functions of a common hidden parameter $θ^*$. Since we do not place any restrictions of these functions, the problem setting subsumes several previously studied frameworks that assume linear or invertible reward functions. We propose a novel approach to gradually estimate the hidden $θ^*$ and use the estimate together with the mean reward functions to substantially reduce exploration of sub-optimal arms. This approach enables us to fundamentally generalize any classic bandit algorithm including UCB and Thompson Sampling to the structured bandit setting. We prove via regret analysis that our proposed UCB-C algorithm (structured bandit versions of UCB) pulls only a subset of the sub-optimal arms $O(\log T)$ times while the other sub-optimal arms (referred to as non-competitive arms) are pulled $O(1)$ times. As a result, in cases where all sub-optimal arms are non-competitive, which can happen in many practical scenarios, the proposed algorithms achieve bounded regret. We also conduct simulations on the Movielens recommendations dataset to demonstrate the improvement of the proposed algorithms over existing structured bandit algorithms.