Author name cluster

Indranil Gupta

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

7 papers

2 author rows

NeurIPS Conference 2025 Conference Paper

Generative Caching for Structurally Similar Prompts and Responses

Sarthak Chakraborty
Suman Nath
Xuchao Zhang
Chetan Bansal
Indranil Gupta

Large Language Models (LLMs) are increasingly being used to plan, reason, and execute tasks across diverse scenarios. In use cases like repeatable workflows and agentic settings, prompts are often reused with minor variations while having a similar structure for recurring tasks. This opens up opportunities for caching. However, exact prompt matching fails on such structurally similar prompts, while semantic caching may produce incorrect responses by ignoring critical differences. To address this, we introduce GenCache, a generative cache that produces variation-aware responses for structurally similar prompts. GenCache identifies reusable response patterns across similar prompt structures and synthesizes customized outputs for new requests. We show that GenCache achieves 83\% cache hit rate, while having minimal incorrect hits on datasets without prompt repetition. In agentic workflows, it improves cache hit rate by $\sim$20\% and reduces end-to-end execution latency by $\sim$34\% compared to standard prompt matching.

PDF Details

NeurIPS Conference 2020 Conference Paper

CSER: Communication-efficient SGD with Error Reset

Cong Xie
Shuai Zheng
Sanmi Koyejo
Indranil Gupta
Mu Li
Haibin Lin

The scalability of Distributed Stochastic Gradient Descent (SGD) is today limited by communication bottlenecks. We propose a novel SGD variant: \underline{C}ommunication-efficient \underline{S}GD with \underline{E}rror \underline{R}eset, or \underline{CSER}. The key idea in CSER is first a new technique called ``error reset'' that adapts arbitrary compressors for SGD, producing bifurcated local models with periodic reset of resulting local residual errors. Second we introduce partial synchronization for both the gradients and the models, leveraging advantages from them. We prove the convergence of CSER for smooth non-convex problems. Empirical results show that when combined with highly aggressive compressors, the CSER algorithms accelerate the distributed training by nearly $10\times$ for CIFAR-100, and by $4. 5\times$ for ImageNet.

PDF Details

ICML Conference 2020 Conference Paper

Zeno++: Robust Fully Asynchronous SGD

Cong Xie
Sanmi Koyejo
Indranil Gupta

We propose Zeno++, a new robust asynchronous Stochastic Gradient Descent(SGD) procedure, intended to tolerate Byzantine failures of workers. In contrast to previous work, Zeno++ removes several unrealistic restrictions on worker-server communication, now allowing for fully asynchronous updates from anonymous workers, for arbitrarily stale worker updates, and for the possibility of an unbounded number of Byzantine workers. The key idea is to estimate the descent of the loss value after the candidate gradient is applied, where large descent values indicate that the update results in optimization progress. We prove the convergence of Zeno++ for non-convex problems under Byzantine failures. Experimental results show that Zeno++ outperforms existing Byzantine-tolerant asynchronous SGD algorithms.

Details

UAI Conference 2019 Conference Paper

Fall of Empires: Breaking Byzantine-tolerant SGD by Inner Product Manipulation

Cong Xie
Sanmi Koyejo
Indranil Gupta

Recently, new defense techniques have been developed to tolerate Byzantine failures for distributed machine learning. The Byzantine model captures workers that behave arbitrarily, including malicious and compromised workers. In this paper, we break two prevailing Byzantine-tolerant techniques. Specifically we show that two robust aggregation methods for synchronous SGD–namely, coordinate-wise median and Krum–can be broken using new attack strategies based on inner product manipulation. We prove our results theoretically, as well as show empirical validation.

Details

ICML Conference 2019 Conference Paper

Zeno: Distributed Stochastic Gradient Descent with Suspicion-based Fault-tolerance

Cong Xie
Sanmi Koyejo
Indranil Gupta

We present Zeno, a technique to make distributed machine learning, particularly Stochastic Gradient Descent (SGD), tolerant to an arbitrary number of faulty workers. Zeno generalizes previous results that assumed a majority of non-faulty nodes; we need assume only one non-faulty worker. Our key idea is to suspect workers that are potentially defective. Since this is likely to lead to false positives, we use a ranking-based preference mechanism. We prove the convergence of SGD for non-convex problems under these scenarios. Experimental results show that Zeno outperforms existing approaches.

Details

TAAS Journal 2017 Journal Article

Characterizing and Adapting the Consistency-Latency Tradeoff in Distributed Key-Value Stores

Muntasir Raihan Rahman
Lewis Tseng
Son Nguyen
Indranil Gupta
Nitin Vaidya

The CAP theorem is a fundamental result that applies to distributed storage systems. In this article, we first present and prove two CAP-like impossibility theorems. To state these theorems, we present probabilistic models to characterize the three important elements of the CAP theorem: consistency (C), availability or latency (A), and partition tolerance (P). The theorems show the un-achievable envelope, that is, which combinations of the parameters of the three models make them impossible to achieve together. Next, we present the design of a class of systems called Probabilistic CAP (PCAP) that perform close to the envelope described by our theorems. In addition, these systems allow applications running on a single data center to specify either a latency Service Level Agreement (SLA) or a consistency SLA. The PCAP systems automatically adapt, in real time and under changing network conditions, to meet the SLA while optimizing the other C/A metric. We incorporate PCAP into two popular key-value stores: Apache Cassandra and Riak. Our experiments with these two deployments, under realistic workloads, reveal that the PCAP systems satisfactorily meets SLAs and perform close to the achievable envelope. We also extend PCAP from a single data center to multiple geo-distributed data centers.

Details DOI

TAAS Journal 2008 Journal Article

A new class of nature-inspired algorithms for self-adaptive peer-to-peer computing

Steven Y. Ko
Indranil Gupta
Yookyung Jo

We present, and evaluate benefits of, a design methodology for translating natural phenomena represented as mathematical models, into novel, self-adaptive, peer-to-peer (p2p) distributed computing algorithms ( protocols ). Concretely, our first contribution is a set of techniques to translate discrete sequence equations (also known as difference equations) into new p2p protocols called sequence protocols. Sequence protocols are self-adaptive, scalable, and fault-tolerant, with applicability in p2p settings like Grids. A sequence protocol is a set of probabilistic local and message-passing actions for each process. These actions are translated from terms in a set of source sequence equations. Individual processes do not simulate the source sequence equations completely. Instead, each process executes probabilistic local and message passing actions, so that the emergent round-to-round behavior of the sequence protocol in a p2p system can be probabilistically predicted by the source sequence equations. The article's second contribution is the design and evaluation of a set of sequence protocols for detection of two global triggers in a distributed system: threshold detection and interval detection. This article's third contribution is a new self-adaptive Grid computing protocol called HoneyAdapt. HoneyAdapt is derived from sequence equations modeling adaptive bee foraging behavior in nature. HoneyAdapt is intended for Grid applications that allow Grid clients, at run-time, a choice of algorithms for executing chunks of the application's dataset. HoneyAdapt tells each Grid client how to adaptively select at run-time, for each chunk it receives, a good algorithm for computing the chunk—this selection is based on continuous feedback from other clients. Finally, we design a variant of HoneyAdapt, called HoneySort, for application to Grid parallelized sorting settings using the master-worker paradigm. Our evaluation of these contributions consists of mathematical analysis, large-scale trace-based simulation results, and experimental results from a HoneySort deployment.

Details DOI