arXiv subjects

Ramasuri Narayanam

Publications and source records attributed to Ramasuri Narayanam.

14 recordsLinked to original sources

arXiv2026

ThinkRetrieve: Retrieval-Augmented Reasoning Traces for Test-Time Scaling

Large Reasoning Models (LRMs) improve performance by allocating additional inference-time compute to generate extended chain-of-thought reasoning. However, recent studies reveal that sequential test-time scaling often yields diminishing or even negative returns, as longer traces exhibit increased uncertainty, error compounding, and drift from the original problem. We propose ThinkRetrieve, a test-time scaling framework that augments the reasoning traces of LRMs with dynamically retrieved solved examples at each reasoning step. Given an external corpus of problems paired with step-by-step solutions, ThinkRetrieve retrieves relevant exemplars at each intermediate step and injects them directly into the thinking trace, providing the model with guidance on how to reason rather than merely what facts are relevant. Experiments across five reasoning models (1.5B--8B parameters) on GSM-8K, MATH-500, AIME 2025, and SciQ demonstrate that ThinkRetrieve consistently improves accuracy over standard test-time scaling, with relative gains of up to $60\%$ on AIME 2025.

cs.AI↗

arXiv2026

From Tokens to Steps: Verification-Aware Speculative Decoding for Efficient Multi-Step Reasoning

Speculative decoding (SD) accelerates large language model inference by allowing a lightweight draft model to propose outputs that a stronger target model verifies. However, its token-centric nature allows erroneous steps to propagate. Prior approaches mitigate this using external reward models, but incur additional latency, computational overhead, and limit generalizability. We propose SpecGuard, a verification-aware speculative decoding framework that performs step-level verification using only model-internal signals. At each step, SpecGuard samples multiple draft candidates and selects the most consistent step, which is then validated using an ensemble of two lightweight model-internal signals: (i) an attention-based grounding score that measures attribution to the input and previously accepted steps, and (ii) a log-probability-based score that captures token-level confidence. These signals jointly determine whether a step is accepted or recomputed using the target, allocating compute selectively. Experiments across a range of reasoning benchmarks show that SpecGuard improves accuracy by 3.6% while reducing latency by ~11%, outperforming both SD and reward-guided SD.

cs.CL↗

arXiv2025

Sparse Linear Bandits with Blocking Constraints

We investigate the high-dimensional sparse linear bandits problem in a data-poor regime where the time horizon is much smaller than the ambient dimension and number of arms. We study the setting under the additional blocking constraint where each unique arm can be pulled only once. The blocking constraint is motivated by practical applications in personalized content recommendation and identification of data points to improve annotation efficiency for complex learning tasks. With mild assumptions on the arms, our proposed online algorithm (BSLB) achieves a regret guarantee of $\widetilde{\mathsf{O}}((1+β_k)^2k^{\frac{2}{3}} \mathsf{T}^{\frac{2}{3}})$ where the parameter vector has an (unknown) relative tail $β_k$ -- the ratio of $\ell_1$ norm of the top-$k$ and remaining entries of the parameter vector. To this end, we show novel offline statistical guarantees of the lasso estimator for the linear model that is robust to the sparsity modeling assumption. Finally, we propose a meta-algorithm (C-BSLB) based on corralling that does not need knowledge of optimal sparsity parameter $k$ at minimal cost to regret. Our experiments on multiple real-world datasets demonstrate the validity of our algorithms and theoretical framework.

cs.LG↗

arXiv2025

Tab-Shapley: Identifying Top-k Tabular Data Quality Insights

We present an unsupervised method for aggregating anomalies in tabular datasets by identifying the top-k tabular data quality insights. Each insight consists of a set of anomalous attributes and the corresponding subsets of records that serve as evidence to the user. The process of identifying these insight blocks is challenging due to (i) the absence of labeled anomalies, (ii) the exponential size of the subset search space, and (iii) the complex dependencies among attributes, which obscure the true sources of anomalies. Simple frequency-based methods fail to capture these dependencies, leading to inaccurate results. To address this, we introduce Tab-Shapley, a cooperative game theory based framework that uses Shapley values to quantify the contribution of each attribute to the data's anomalous nature. While calculating Shapley values typically requires exponential time, we show that our game admits a closed-form solution, making the computation efficient. We validate the effectiveness of our approach through empirical analysis on real-world tabular datasets with ground-truth anomaly labels.

cs.LG↗

arXiv2024

CAFIN: Centrality Aware Fairness inducing IN-processing for Unsupervised Representation Learning on Graphs

Unsupervised Representation Learning on graphs is gaining traction due to the increasing abundance of unlabelled network data and the compactness, richness, and usefulness of the representations generated. In this context, the need to consider fairness and bias constraints while generating the representations has been well-motivated and studied to some extent in prior works. One major limitation of most of the prior works in this setting is that they do not aim to address the bias generated due to connectivity patterns in the graphs, such as varied node centrality, which leads to a disproportionate performance across nodes. In our work, we aim to address this issue of mitigating bias due to inherent graph structure in an unsupervised setting. To this end, we propose CAFIN, a centrality-aware fairness-inducing framework that leverages the structural information of graphs to tune the representations generated by existing frameworks. We deploy it on GraphSAGE (a popular framework in this domain) and showcase its efficacy on two downstream tasks - Node Classification and Link Prediction. Empirically, CAFIN consistently reduces the performance disparity across popular datasets (varying from 18 to 80% reduction in performance disparity) from various domains while incurring only a minimal cost of fairness.

cs.LG↗

arXiv2021

Game of Gradients: Mitigating Irrelevant Clients in Federated Learning

The paradigm of Federated learning (FL) deals with multiple clients participating in collaborative training of a machine learning model under the orchestration of a central server. In this setup, each client's data is private to itself and is not transferable to other clients or the server. Though FL paradigm has received significant interest recently from the research community, the problem of selecting the relevant clients w.r.t. the central server's learning objective is under-explored. We refer to these problems as Federated Relevant Client Selection (FRCS). Because the server doesn't have explicit control over the nature of data possessed by each client, the problem of selecting relevant clients is significantly complex in FL settings. In this paper, we resolve important and related FRCS problems viz., selecting clients with relevant data, detecting clients that possess data relevant to a particular target label, and rectifying corrupted data samples of individual clients. We follow a principled approach to address the above FRCS problems and develop a new federated learning method using the Shapley value concept from cooperative game theory. Towards this end, we propose a cooperative game involving the gradients shared by the clients. Using this game, we compute Shapley values of clients and then present Shapley value based Federated Averaging (S-FedAvg) algorithm that empowers the server to select relevant clients with high probability. S-FedAvg turns out to be critical in designing specific algorithms to address the FRCS problems. We finally conduct a thorough empirical analysis on image classification and speech recognition tasks to show the superior performance of S-FedAvg than the baselines in the context of supervised federated learning settings.

cs.LG↗

arXiv2019

Beyond Node Embedding: A Direct Unsupervised Edge Representation Framework for Homogeneous Networks

Network representation learning has traditionally been used to find lower dimensional vector representations of the nodes in a network. However, there are very important edge driven mining tasks of interest to the classical network analysis community, which have mostly been unexplored in the network embedding space. For applications such as link prediction in homogeneous networks, vector representation (i.e., embedding) of an edge is derived heuristically just by using simple aggregations of the embeddings of the end vertices of the edge. Clearly, this method of deriving edge embedding is suboptimal and there is a need for a dedicated unsupervised approach for embedding edges by leveraging edge properties of the network. Towards this end, we propose a novel concept of converting a network to its weighted line graph which is ideally suited to find the embedding of edges of the original network. We further derive a novel algorithm to embed the line graph, by introducing the concept of collective homophily. To the best of our knowledge, this is the first direct unsupervised approach for edge embedding in homogeneous information networks, without relying on the node embeddings. We validate the edge embeddings on three downstream edge mining tasks. Our proposed optimization framework for edge embedding also generates a set of node embeddings, which are not just the aggregation of edges. Further experimental analysis shows the connection of our framework to the concept of node centrality.

cs.SI↗

arXiv2019

Computational Aspects of Equilibria in Discrete Preference Games

We study the complexity of equilibrium computation in discrete preference games. These games were introduced by Chierichetti, Kleinberg, and Oren (EC '13, JCSS '18) to model decision-making by agents in a social network that choose a strategy from a finite, discrete set, balancing between their intrinsic preferences for the strategies and their desire to choose a strategy that is `similar' to their neighbours. There are thus two components: a social network with the agents as vertices, and a metric space of strategies. These games are potential games, and hence pure Nash equilibria exist. Since their introduction, a number of papers have studied various aspects of this model, including the social cost at equilibria, and arrival at a consensus. We show that in general, equilibrium computation in discrete preference games is PLS-complete, even in the simple case where each agent has a constant number of neighbours. If the edges in the social network are weighted, then the problem is PLS-complete even if each agent has a constant number of neighbours, the metric space has constant size, and every pair of strategies is at distance 1 or 2. Further, if the social network is directed, modelling asymmetric influence, an equilibrium may not even exist. On the positive side, we show that if the metric space is a tree metric, or is the product of path metrics, then the equilibrium can be computed in polynomial time.

cs.GT↗

arXiv2017

Cogniculture: Towards a Better Human-Machine Co-evolution

Research in Artificial Intelligence is breaking technology barriers every day. New algorithms and high performance computing are making things possible which we could only have imagined earlier. Though the enhancements in AI are making life easier for human beings day by day, there is constant fear that AI based systems will pose a threat to humanity. People in AI community have diverse set of opinions regarding the pros and cons of AI mimicking human behavior. Instead of worrying about AI advancements, we propose a novel idea of cognitive agents, including both human and machines, living together in a complex adaptive ecosystem, collaborating on human computation for producing essential social goods while promoting sustenance, survival and evolution of the agents' life cycle. We highlight several research challenges and technology barriers in achieving this goal. We propose a governance mechanism around this ecosystem to ensure ethical behaviors of all cognitive agents. Along with a novel set of use-cases of Cogniculture, we discuss the road map ahead for this journey.

cs.CY↗

arXiv2017

A Generic Axiomatic Characterization of Centrality Measures in Social Network

Centrality is an important notion in complex networks; it could be used to characterize how influential a node or an edge is in the network. It plays an important role in several other network analysis tools including community detection. Even though there are a small number of axiomatic frameworks associated with this notion, the existing formalizations are not generic in nature. In this paper we propose a generic axiomatic framework to capture all the intrinsic properties of a centrality measure (a.k.a. centrality index). We analyze popular centrality measures along with other novel measures of centrality using this framework. We observed that none of the centrality measures considered satisfies all the axioms.

cs.SI↗

arXiv2016

All Fingers are not Equal: Intensity of References in Scientific Articles

Research accomplishment is usually measured by considering all citations with equal importance, thus ignoring the wide variety of purposes an article is being cited for. Here, we posit that measuring the intensity of a reference is crucial not only to perceive better understanding of research endeavor, but also to improve the quality of citation-based applications. To this end, we collect a rich annotated dataset with references labeled by the intensity, and propose a novel graph-based semi-supervised model, GraLap to label the intensity of references. Experiments with AAN datasets show a significant improvement compared to the baselines to achieve the true labels of the references (46% better correlation). Finally, we provide four applications to demonstrate how the knowledge of reference intensity leads to design better real-world applications.

cs.CL↗

arXiv2016

HEMI: Hyperedge Majority Influence Maximization

In this work, we consider the problem of influence maximization on a hypergraph. We first extend the Independent Cascade (IC) model to hypergraphs, and prove that the traditional influence maximization problem remains submodular. We then present a variant of the influence maximization problem (HEMI) where one seeks to maximize the number of hyperedges, a majority of whose nodes are influenced. We prove that HEMI is non-submodular under the diffusion model proposed.

cs.SI↗

arXiv2011

Topologies and Price of Stability of Complex Strategic Networks with Localized Payoffs : Analytical and Simulation Studies

We analyze a network formation game in a strategic setting where payoffs of individuals depend only on their immediate neighbourhood. We call these payoffs as localized payoffs. In this game, the payoff of each individual captures (1) the gain from immediate neighbors, (2) the bridging benefits, and (3) the cost to form links. This implies that the payoff of each individual can be computed using only its single-hop neighbourhood information. Based on this simple model of network formation, our study explores the structure of networks that form, satisfying one or both of the properties, namely, pairwise stability and efficiency. We analytically prove the pairwise stability of several interesting network structures, notably, the complete bi-partite network, complete equi-k-partite network, complete network and cycle network, under various configurations of the model. We validate and extend these results through extensive simulations. We characterize topologies of efficient networks by drawing upon classical results from extremal graph theory and discover that the Turan graph (or the complete equi-bi-partite network) is the unique efficient network under many configurations of parameters. We examine the tradeoffs between topologies of pairwise stable networks and efficient networks using the notion of price of stability, which is the ratio of the sum of payoffs of the players in an optimal pairwise stable network to that of an efficient network. Interestingly, we find that price of stability is equal to 1 for almost all configurations of parameters in the proposed model; and for the rest of the configurations of the parameters, we obtain a lower bound of 0.5 on the price of stability. This leads to another key insight of this paper: under mild conditions, efficient networks will form when strategic individuals choose to add or delete links based on only localized payoffs.

cs.SI↗

arXiv2009

Design of an Optimal Bayesian Incentive Compatible Broadcast Protocol for Ad hoc Networks with Rational Nodes

Nodes in an ad hoc wireless network incur certain costs for forwarding packets since packet forwarding consumes the resources of the nodes. If the nodes are rational, free packet forwarding by the nodes cannot be taken for granted and incentive based protocols are required to stimulate cooperation among the nodes. Existing incentive based approaches are based on the VCG (Vickrey-Clarke-Groves) mechanism which leads to high levels of incentive budgets and restricted applicability to only certain topologies of networks. Moreover, the existing approaches have only focused on unicast and multicast. Motivated by this, we propose an incentive based broadcast protocol that satisfies Bayesian incentive compatibility and minimizes the incentive budgets required by the individual nodes. The proposed protocol, which we call {\em BIC-B} (Bayesian incentive compatible broadcast) protocol, also satisfies budget balance. We also derive a necessary and sufficient condition for the ex-post individual rationality of the BIC-B protocol. The {\em BIC-B} protocol exhibits superior performance in comparison to a dominant strategy incentive compatible broadcast protocol.

cs.GT↗