Sankarshan Damle

Sambhav Solanki, Samhita Kanaparthy, Sankarshan Damle et al.

There is a rapid increase in the cooperative learning paradigm in online learning settings, i.e., federated learning (FL). Unlike most FL settings, there are many situations where the agents are competitive. Each agent would like to learn from others, but the part of the information it shares for others to learn from could be sensitive; thus, it desires its privacy. This work investigates a group of agents working concurrently to solve similar combinatorial bandit problems while maintaining quality constraints. Can these agents collectively learn while keeping their sensitive information confidential by employing differential privacy? We observe that communicating can reduce the regret. However, differential privacy techniques for protecting sensitive information makes the data noisy and may deteriorate than help to improve regret. Hence, we note that it is essential to decide when to communicate and what shared data to learn to strike a functional balance between regret and privacy. For such a federated combinatorial MAB setting, we propose a Privacy-preserving Federated Combinatorial Bandit algorithm, P-FCB. We illustrate the efficacy of P-FCB through simulations. We further show that our algorithm provides an improvement in terms of regret while upholding quality threshold and meaningful privacy guarantees.

Sankarshan Damle, Manisha Padala, Sujit Gujar

Civic Crowdfunding (CC) uses the ``power of the crowd'' to garner contributions towards public projects. As these projects are non-excludable, agents may prefer to ``free-ride,'' resulting in the project not being funded. For single project CC, researchers propose to provide refunds to incentivize agents to contribute, thereby guaranteeing the project's funding. These funding guarantees are applicable only when agents have an unlimited budget. This work focuses on a combinatorial setting, where multiple projects are available for CC and agents have a limited budget. We study certain specific conditions where funding can be guaranteed. Further, funding the optimal social welfare subset of projects is desirable when every available project cannot be funded due to budget restrictions. We prove the impossibility of achieving optimal welfare at equilibrium for any monotone refund scheme. We then study different heuristics that the agents can use to contribute to the projects in practice. Through simulations, we demonstrate the heuristics' performance as the average-case trade-off between welfare obtained and agent utility.

Naman Gupta, Vaibhav Singh, Arun Iyer et al.

Sequential multi-agent reasoning frameworks such as Chain-of-Agents (CoA) handle long-context queries by decomposing inputs into chunks and processing them sequentially using LLM-based worker agents that read from and update a bounded shared memory. From a probabilistic perspective, CoA aims to approximate the conditional distribution corresponding to a model capable of jointly reasoning over the entire long context. CoA achieves this through a latent-state factorization in which only bounded summaries of previously processed evidence are passed between agents. The resulting bounded-memory approximation introduces a lossy information bottleneck, making the final evidence state inherently dependent on the order in which chunks are processed. In this work, we study the problem of chunk ordering for long-context reasoning. We use the well-known Chow-Liu trees to learn a dependency structure that prioritizes strongly related chunks. Empirically, we show that a breadth-first traversal of the resulting tree yields chunk orderings that reduce information loss across agents and consistently outperform both default document-chunk ordering and semantic score-based ordering in answer relevance and exact-match accuracy across three long-context benchmarks.

Divyanshu Aggarwal, Sankarshan Damle, Navin Goyal et al.

A common challenge towards the adaptability of Large Language Models (LLMs) is their ability to learn new languages over time without hampering the model's performance on languages in which the model is already proficient (usually English). Continual fine-tuning (CFT) is the process of sequentially fine-tuning an LLM to enable the model to adapt to downstream tasks with varying data distributions and time shifts. This paper focuses on the language adaptability of LLMs through CFT. We study a two-phase CFT process in which an English-only end-to-end fine-tuned LLM from Phase 1 (predominantly Task Ability) is sequentially fine-tuned on a multilingual dataset -- comprising task data in new languages -- in Phase 2 (predominantly Language Ability). We observe that the ``similarity'' of Phase 2 tasks with Phase 1 determines the LLM's adaptability. For similar phase-wise datasets, the LLM after Phase 2 does not show deterioration in task ability. In contrast, when the phase-wise datasets are not similar, the LLM's task ability deteriorates. We test our hypothesis on the open-source \mis\ and \llm\ models with multiple phase-wise dataset pairs. To address the deterioration, we analyze tailored variants of two CFT methods: layer freezing and generative replay. Our findings demonstrate their effectiveness in enhancing the language ability of LLMs while preserving task performance, in comparison to relevant baselines.

Naman Gupta, Shreeyash Gowaikar, Arun Iyer et al.

Reasoning over very long inputs remains difficult for large language models (LLMs). Common workarounds either shrink the input via retrieval (risking missed evidence), enlarge the context window (straining selectivity), or stage multiple agents to read in pieces. In staged pipelines (e.g., Chain of Agents, CoA), free-form summaries passed between agents can discard crucial details and amplify early mistakes. We introduce COSMIR (Chain Orchestrated Structured Memory for Iterative Reasoning), a chain-style framework that replaces ad hoc messages with a structured memory. A Planner agent first turns a user query into concrete, checkable sub-questions. worker agents process chunks via a fixed micro-cycle: Extract, Infer, Refine, writing all updates to the shared memory. A Manager agent then Synthesizes the final answer directly from the memory. This preserves step-wise read-then-reason benefits while changing both the communication medium (structured memory) and the worker procedure (fixed micro-cycle), yielding higher faithfulness, better long-range aggregation, and auditability. On long-context QA from the HELMET suite, COSMIR reduces propagation-stage information loss and improves accuracy over a CoA baseline.

Sankarshan Damle, Boi Faltings

Large Language Models (LLMs) are increasingly expected to handle complex decision-making tasks, yet their ability to perform structured resource allocation remains underexplored. Evaluating their reasoning is also difficult due to data contamination and the static nature of existing benchmarks. We present a dual-purpose framework leveraging Participatory Budgeting (PB) both as (i) a practical setting for LLM-based resource allocation and (ii) an adaptive benchmark for evaluating their reasoning capabilities. We task LLMs with selecting project subsets under feasibility (e.g., budget) constraints via three prompting strategies: greedy selection, direct optimization, and a hill-climbing-inspired refinement. We benchmark LLMs' allocations against a utility-maximizing oracle. Interestingly, we also test whether LLMs can infer structured preferences from natural-language voter input or metadata, without explicit votes. By comparing allocations based on inferred preferences to those from ground-truth votes, we evaluate LLMs' ability to extract preferences from open-ended input. Our results underscore the role of prompt design and show that LLMs hold promise for mechanism design with unstructured inputs.

Sankarshan Damle, Manisha Padala, Sujit Gujar

Blockchains deploy Transaction Fee Mechanisms (TFMs) to determine which user transactions to include in blocks and determine their payments (i.e., transaction fees). Increasing demand and scarce block resources have led to high user transaction fees. As these blockchains are a public resource, it may be preferable to reduce these transaction fees. To this end, we introduce Transaction Fee Redistribution Mechanisms (TFRMs) -- redistributing VCG payments collected from such TFM as rebates to minimize transaction fees. Classic redistribution mechanisms (RMs) achieve this while ensuring Allocative Efficiency (AE) and User Incentive Compatibility (UIC). Our first result shows the non-triviality of applying RM in TFMs. More concretely, we prove that it is impossible to reduce transaction fees when (i) transactions that are not confirmed do not receive rebates and (ii) the miner can strategically manipulate the mechanism. Driven by this, we propose \emph{Robust} TFRM (\textsf{R-TFRM}): a mechanism that compromises on an honest miner's individual rationality to guarantee strictly positive rebates to the users. We then introduce \emph{robust} and \emph{rational} TFRM (\textsf{R}$^2$\textsf{-TFRM}) that uses trusted on-chain randomness that additionally guarantees miner's individual rationality (in expectation) and strictly positive rebates. Our results show that TFRMs provide a promising new direction for reducing transaction fees in public blockchains.

Samhita Kanaparthy, Manisha Padala, Sankarshan Damle et al.

Fairness across different demographic groups is an essential criterion for face-related tasks, Face Attribute Classification (FAC) being a prominent example. Apart from this trend, Federated Learning (FL) is increasingly gaining traction as a scalable paradigm for distributed training. Existing FL approaches require data homogeneity to ensure fairness. However, this assumption is too restrictive in real-world settings. We propose F3, a novel FL framework for fair FAC under data heterogeneity. F3 adopts multiple heuristics to improve fairness across different demographic groups without requiring data homogeneity assumption. We demonstrate the efficacy of F3 by reporting empirically observed fairness measures and accuracy guarantees on popular face datasets. Our results suggest that F3 strikes a practical balance between accuracy and fairness for FAC.

Manisha Padala, Sankarshan Damle, Sujit Gujar

Deep learning's unprecedented success raises several ethical concerns ranging from biased predictions to data privacy. Researchers tackle these issues by introducing fairness metrics, or federated learning, or differential privacy. A first, this work presents an ethical federated learning model, incorporating all three measures simultaneously. Experiments on the Adult, Bank and Dutch datasets highlight the resulting ``empirical interplay" between accuracy, fairness, and privacy.

Sankarshan Damle, Sujit Gujar, Moin Hussain Moti

Electing democratic representatives via voting has been a common mechanism since the 17th century. However, these mechanisms raise concerns about fairness, privacy, vote concealment, fair calculations of tally, and proxies voting on their behalf for the voters. Ballot voting, and in recent times, electronic voting via electronic voting machines (EVMs) improves fairness by relying on centralized trust. Homomorphic encryption-based voting protocols also assure fairness but cannot scale to large scale elections such as presidential elections. In this paper, we leverage the blockchain technology of distributing trust to propose a smart contract-based protocol, namely, \proto. There are many existing protocols for voting using smart contracts. We observe that these either are not scalable or leak the vote tally during the voting stage, i.e., do not provide vote concealment. In contrast, we show that FASTEN preserves voter's privacy ensures vote concealment, immutability, and avoids double voting. We prove that the probability of privacy breaches is negligibly small. Further, our cost analysis of executing FASTEN over Ethereum is comparable to most of the existing cost of elections.

Sankarshan Damle, Boi Faltings, Sujit Gujar

The emergence of e-commerce and e-voting platforms has resulted in the rise in the volume of sensitive information over the Internet. This has resulted in an increased demand for secure and private means of information computation. Towards this, the Yao's Millionaires' problem, i.e., to determine the richer among two millionaires' securely, finds an application. In this work, we present a new solution to the Yao's Millionaires' problem namely, Privacy Preserving Comparison (PPC). We show that PPC achieves this comparison in constant time as well as in one execution. PPC uses semi-honest third parties for the comparison who do not learn any information about the values. Further, we show that PPC is collusion-resistance. To demonstrate the significance of PPC, we present a secure, approximate single-minded combinatorial auction, which we call TPACAS, i.e., Truthful, Privacy-preserving Approximate Combinatorial Auction for Single-minded bidders. We show that TPACAS, unlike previous works, preserves the following privacies relevant to an auction: agent privacy, the identities of the losing bidders must not be revealed to any other agent except the auctioneer (AU), bid privacy, the bid values must be hidden from the other agents as well as the AU and bid-topology privacy, the items for which the agents are bidding must be hidden from the other agents as well as the AU. We demonstrate the practicality of TPACAS through simulations. Lastly, we also look at TPACAS' implementation over a publicly distributed ledger, such as the Ethereum blockchain.