arXiv subjects

Suyash Gupta

Publications and source records attributed to Suyash Gupta.

At least 19 recordsLinked to original sources

arXiv2026

Cassandra: Consensus with Partial Progress via Robust Partitionable View Synchronization

Replicated databases and permissioned blockchain systems rely on Byzantine Fault-Tolerant (BFT) consensus to maintain a globally consistent order of transactions across distributed replicas. These protocols preserve safety even under asynchrony, as they commit a transaction only after agreement among a strong quorum of replicas. During network partitions, however, when no strong quorum is reachable, they lose liveness and cannot make useful progress. In this paper, we present Cassandra, a consensus protocol that enables partial progress without sacrificing safety. Cassandra achieves this through a two-tier certification framework that decouples availability from commitment, allowing each partition to extend its own chain and reconcile these chains once the network is restored. To support this, Cassandra introduces a pacemaker that advances views without requiring a strong quorum and calibrates each replica's timeout off the critical path. Our evaluation results show that Cassandra remains competitive with state-of-the-art BFT protocols under stable conditions, sustaining 900K TPS at 16 replicas and 480K TPS at 104 replicas, with latency ranging from 0.31s at 16 replicas to 0.75s at 104 replicas. Under severe partitions, Cassandra maintains non-zero speculative throughput through PoA-backed progress, preserving work that can be reconciled once connectivity is restored.

cs.DC↗

arXiv2026

Sampling for Quality: Training-Free Reward-Guided LLM Decoding via Sequential Monte Carlo

We introduce a principled probabilistic framework for reward-guided decoding in large language models, addressing the limitations of standard decoding methods that optimize token-level likelihood rather than sequence-level quality. Our method defines a reward-augmented target distribution over complete sequences by combining model transition probabilities with prefix-dependent reward potentials. Importantly, the approach is training-free: it leaves model weights unchanged and instead modifies the inference distribution via reward potentials, with all gains arising purely from inference-time sampling. To sample from this distribution, we develop Sequential Monte Carlo algorithms, including a computationally efficient prefix-only variant and a lookahead variant whose intermediate targets match the exact marginals of the full sequence distribution. The framework also integrates resample-move updates with Metropolis-Hastings rejuvenation and supports block-wise generation, subsuming common decoding strategies such as temperature sampling and power-tempered objectives. Empirical results across three 7B models show significant gains. On code generation (HumanEval), our method improves base performance by up to 54.9% and surpasses the strongest sampling baselines by 9.1%-15.3%. On mathematical reasoning (MATH500), it achieves gains of up to 8.8%. Notably, it reaches 87.8% on HumanEval and 78.4% on MATH500 with Qwen2.5-7B, consistently outperforming the reinforcement learning method GRPO.

cs.LG↗

arXiv2026

Support Tokens, Stability Margins, and a New Foundation for Robust LLMs

Self-attention is usually described as a flexible, content-adaptive way to mix a token with information from its past. We reinterpret causal self-attention transformers, the backbone of modern foundation models, within a probabilistic framework, much as classical PCA is extended to probabilistic PCA. This reformulation reveals a key structural consequence of the underlying change of variables: a barrier constraint emerges on the parameters of self-attention. The resulting geometry exposes a degeneracy boundary where the attention-induced mapping becomes locally ill-conditioned, yielding a stability-margin interpretation analogous to the margin in support vector machines. This, in turn, naturally gives rise to the concept of support tokens. We further show that causal transformers define a consistent stochastic process over infinite token sequences, providing a rigorous probabilistic foundation for sequence modeling. Building on this view, we derive a Bayesian MAP training objective that requires only a minimal modification to standard LLM training: adding a smooth log-barrier penalty to the usual cross-entropy loss. Empirically, the resulting training objective improves robustness to input perturbations and sharpens the margin geometry of the learned representations without sacrificing out-of-sample accuracy.

cs.LG↗

arXiv2026

Heterogeneity-Aware Client Selection Methodology For Efficient Federated Learning

Federated Learning (FL) enables a distributed client-server architecture where multiple clients collaboratively train a global Machine Learning (ML) model without sharing sensitive local data. However, FL often results in lower accuracy than traditional ML algorithms due to statistical heterogeneity across clients. Prior works attempt to address this by using model updates, such as loss and bias, from client models to select participants that can improve the global model's accuracy. However, these updates neither accurately represent a client's heterogeneity nor are their selection methods deterministic. We mitigate these limitations by introducing Terraform, a novel client selection methodology that uses gradient updates and a deterministic selection algorithm to select heterogeneous clients for retraining. This bi-pronged approach allows Terraform to achieve up to 47 percent higher accuracy over prior works. We further demonstrate its efficiency through comprehensive ablation studies and training time analyses, providing strong justification for the robustness of Terraform.

cs.DC↗

arXiv2025

Carry the Tail in Consensus Protocols

We present Carry-the-Tail, the first deterministic atomic broadcast protocol in partial synchrony that, after GST, guarantees a constant fraction of commits by non-faulty leaders against tail-forking attacks, and maintains optimal, worst-case quadratic communication under a cascade of faulty leaders. The solution also guarantees linear amortized communication, i.e., the steady-state is linear. Prior atomic broadcast solutions achieve quadratic word communication complexity in the worst case. However, they face a significant degradation in throughput under tail-forking attack. Existing solutions to tail-forking attacks require either quadratic communication steps or computationally-prohibitive SNARK generation. The key technical contribution is Carry, a practical drop-in mechanism for streamlined protocols in the HotStuff family. Carry guarantees good performance against tail-forking and removes most leader-induced stalls, while retaining linear traffic and protocol simplicity.

cs.DB↗

arXiv2024

Did we miss P In CAP? Partial Progress Conjecture under Asynchrony

Each application developer desires to provide its users with consistent results and an always-available system despite failures. Boldly, the CALM theorem disagrees. It states that it is hard to design a system that is both consistent and available under network partitions; select at most two out of these three properties. One possible solution is to design coordination-free monotonic applications. However, a majority of real-world applications require coordination. We resolve this dilemma by conjecturing that partial progress is possible under network partitions. This partial progress ensures the system appears responsive to a subset of clients and achieves non-zero throughput during failures. To this extent, we present the design of our CASSANDRA consensus protocol that allows partitioned replicas to order client requests.

cs.DC↗

arXiv2024

HotStuff-1: Linear Consensus with One-Phase Speculation

This paper introduces HotStuff-1, a BFT consensus protocol that improves the latency of HotStuff-2 by two network hops while maintaining linear communication complexity against faults. Furthermore, HotStuff-1 incorporates an incentive-compatible leader rotation design that motivates leaders to propose transactions promptly. HotStuff-1 achieves a reduction of two network hops by speculatively sending clients early confirmations, after one phase of the protocol. Introducing speculation into streamlined protocols is challenging because, unlike stable-leader protocols, these protocols cannot stop the consensus and recover from failures. Thus, we identify prefix speculation dilemma in the context of streamlined protocols; HotStuff-1 is the first streamlined protocol to resolve it. HotStuff-1 embodies an additional mechanism, slotting, that thwarts delays caused by (1) rationally-incentivized leaders and (2) malicious leaders inclined to sabotage other's progress. The slotting mechanism allows leaders to dynamically drive as many decisions as allowed by network transmission delays before view timers expire, thus mitigating both threats.

cs.DB↗

arXiv2024

Spatial Transfer Learning for Estimating PM2.5 in Data-poor Regions

Air pollution, especially particulate matter 2.5 (PM2.5), is a pressing concern for public health and is difficult to estimate in developing countries (data-poor regions) due to a lack of ground sensors. Transfer learning models can be leveraged to solve this problem, as they use alternate data sources to gain knowledge (i.e., data from data-rich regions). However, current transfer learning methodologies do not account for dependencies between the source and the target domains. We recognize this transfer problem as spatial transfer learning and propose a new feature named Latent Dependency Factor (LDF) that captures spatial and semantic dependencies of both domains and is subsequently added to the feature spaces of the domains. We generate LDF using a novel two-stage autoencoder model that learns from clusters of similar source and target domain data. Our experiments show that transfer learning models using LDF have a 19.34% improvement over the baselines. We additionally support our experiments with qualitative findings.

cs.LG↗

arXiv2024

Predictive Inference in Multi-environment Scenarios

We address the challenge of constructing valid confidence intervals and sets in problems of prediction across multiple environments. We investigate two types of coverage suitable for these problems, extending the jackknife and split-conformal methods to show how to obtain distribution-free coverage in such non-traditional, potentially hierarchical data-generating scenarios. We demonstrate a novel resizing method to adapt to problem difficulty, which applies both to existing approaches for predictive inference and the methods we develop; this reduces prediction set sizes using limited information from the test environment, a key to the methods' practical performance, which we evaluate through neurochemical sensing and species classification datasets. Our contributions also include extensions for settings with non-real-valued responses, a theory of consistency for predictive inference in these general problems, and insights on the limits of conditional coverage.

stat.ML↗

arXiv2023

Picsou: Enabling Replicated State Machines to Communicate Efficiently

Replicated state machines (RSMs) cannot communicate effectively today as there is no formal framework or efficient protocol to do so. To address this issue, we introduce a new primitive, Cross-Cluster Consistent Broadcast (C3B) and present PICSOU, a practical implementation of the C3B primitive. PICSOU draws inspiration from networking and TCP to allow two RSMs to communicate with constant metadata overhead in the failure-free case and a minimal number of message resends in the case of failures. PICSOU is flexible and allows both crash fault tolerant and Byzantine fault tolerant consensus protocols to communicate. At the heart of PICSOU's good performance and generality is the concept of QUACKs (quorum acknowledgments). QUACKs allow nodes in each RSM to precisely determine when messages have definitely been received, or likely lost. Our results are promising: we obtain up to 24x better performance than prior solutions on microbenchmarks and applications, ranging from disaster recovery to data reconciliation.

cs.DC↗

arXiv2023

Securing Consensus from Long-Range Attacks through Collaboration

Decentralized systems built around blockchain technology promise clients an immutable ledger. They add a transaction to the ledger after it undergoes consensus among the replicas that run a Proof-of-Stake (PoS) or Byzantine Fault-Tolerant (BFT) consensus protocol. Unfortunately, these protocols face a long-range attack where an adversary having access to the private keys of the replicas can rewrite the ledger. One solution is forcing each committed block from these protocols to undergo another consensus, Proof-of-Work(PoW) consensus; PoW protocol leads to wastage of computational resources as miners compete to solve complex puzzles. In this paper, we present the design of our Power-of-Collaboration (PoC) protocol, which guards existing PoS/BFT blockchains against long-range attacks and requires miners to collaborate rather than compete. PoC guarantees fairness and accountability and only marginally degrades the throughput of the underlying system.

cs.CR↗

arXiv2022

Dissecting BFT Consensus: In Trusted Components we Trust!

The growing interest in reliable multi-party applications has fostered widespread adoption of Byzantine Fault-Tolerant (BFT) consensus protocols. Existing BFT protocols need f more replicas than Paxos-style protocols to prevent equivocation attacks. Trust-BFT protocols instead seek to minimize this cost by making use of trusted components at replicas. This paper makes two contributions. First, we analyze the design of existing Trust-BFT protocols and uncover three fundamental limitations that preclude most practical deployments. Some of these limitations are fundamental, while others are linked to the state of trusted components today. Second, we introduce a novel suite of consensus protocols, FlexiTrust, that attempts to sidestep these issues. We show that our FlexiTrust protocols achieve up to 185% more throughput than their Trust-BFT counterparts.

cs.DB↗

arXiv2022

Reliable Transactions in Serverless-Edge Architecture

Modern edge applications demand novel solutions where edge applications do not have to rely on a single cloud provider (which cannot be in the vicinity of every edge device) or dedicated edge servers (which cannot scale as clouds) for processing compute-intensive tasks. A recent computing philosophy, Sky computing, proposes giving each user ability to select between available cloud providers. In this paper, we present our serverless-edge co-design, which extends the Sky computing vision. In our serverless-edge co-design, we expect edge devices to collaborate and spawn required number of serverless functions. This raises several key challenges: (1) how will this collaboration take place, (2) what if some edge devices are compromised, and (3) what if a selected cloud provider is malicious. Hence, we design ServerlessBFT, the first protocol to guarantee Byzantine fault-tolerant (BFT) transactional flow between edge devices and serverless functions. We present an exhaustive list of attacks and their solutions on our serverless-edge co-design. Further, we extensively benchmark our architecture on a variety of parameters.

cs.DB↗

arXiv2022

On the Correctness of Speculative Consensus

The introduction of Bitcoin fueled the development of blockchain-based resilient data management systems that are resilient against failures, enable federated data management, and can support data provenance. The key factor determining the performance of such resilient data management systems is the consensus protocol used by the system to replicate client transactions among all participants. Unfortunately, existing high-throughput consensus protocols are costly and impose significant latencies on transaction processing, which rules out their usage in responsive high-performance data management systems. In this work, we improve on this situation by introducing the Proof-of-Execution consensus protocol (PoE), a consensus protocol designed for high-performance low-latency resilient data management. PoE introduces speculative execution, which minimizes latencies by starting execution before consensus is reached, and PoE introduces proof-of-executions to guarantee successful execution to clients. Furthermore, PoE introduces a single-round check-commit protocol to reduce the overall communication costs of consensus. Hence, we believe that PoE is a promising step towards flexible general-purpose low-latency resilient data management systems.

cs.DB↗

arXiv2022

RingBFT: Resilient Consensus over Sharded Ring Topology

The recent surge in federated data management applications has brought forth concerns about the security of underlying data and the consistency of replicas in the presence of malicious attacks. A prominent solution in this direction is to employ a permissioned blockchain framework that is modeled around traditional Byzantine Fault-Tolerant (BFT) consensus protocols. Any federated application expects its data to be globally scattered to achieve faster access. But, prior works have shown that traditional BFT protocols are slow. This has led to the rise of sharded-replicated blockchains. Existing BFT protocols for these sharded blockchains are efficient if client transactions require access to a single-shard, but face performance degradation if there is a cross-shard transaction that requires access to multiple shards. As cross-shard transactions are common, to resolve this dilemma, we present RingBFT, a novel meta-BFT protocol for sharded blockchains. RingBFT requires shards to adhere to the ring order, and follow the principle of process, forward, and re-transmit while ensuring the communication between shards is linear. Our evaluation of RingBFT against state-of-the-art sharding BFT protocols illustrates that RingBFT achieves up to 18x higher throughput, gracefully scales to nearly 500 globally distributed nodes, and achieves a peak throughput of 1.2 million transactions per second.

cs.DB↗

arXiv2022

Predictive Inference with Weak Supervision

The expense of acquiring labels in large-scale statistical machine learning makes partially and weakly-labeled data attractive, though it is not always apparent how to leverage such data for model fitting or validation. We present a methodology to bridge the gap between partial supervision and validation, developing a conformal prediction framework to provide valid predictive confidence sets -- sets that cover a true label with a prescribed probability, independent of the underlying distribution -- using weakly labeled data. To do so, we introduce a (necessary) new notion of coverage and predictive validity, then develop several application scenarios, providing efficient algorithms for classification and several large-scale structured prediction problems. We corroborate the hypothesis that the new coverage definition allows for tighter and more informative (but valid) confidence sets through several experiments.

stat.ML↗

arXiv2021

Blockchain Transaction Processing

A blockchain is an append-only linked-list of blocks, which is maintained at each participating node. Each block records a set of transactions and their associated metadata. Blockchain transactions act on the identical ledger data stored at each node. Blockchain was first perceived by Satoshi Nakamoto as a peer-to-peer digital-commodity (also known as crypto-currency) exchange system. Blockchains received traction due to their inherent property of immutability-once a block is accepted, it cannot be reverted.

cs.DB↗

arXiv2021

The $s$-value: evaluating stability with respect to distributional shifts

Common statistical measures of uncertainty such as $p$-values and confidence intervals quantify the uncertainty due to sampling, that is, the uncertainty due to not observing the full population. However, sampling is not the only source of uncertainty. In practice, distributions change between locations and across time. This makes it difficult to gather knowledge that transfers across data sets. We propose a measure of instability that quantifies the distributional instability of a statistical parameter with respect to Kullback-Leibler divergence, that is, the sensitivity of the parameter under general distributional perturbations within a Kullback-Leibler divergence ball. In addition, we quantify the instability of parameters with respect to directional or variable-specific shifts. Measuring instability with respect to directional shifts can be used to detect the type of shifts a parameter is sensitive to. We discuss how such knowledge can inform data collection for improved estimation of statistical parameters under shifted distributions. We evaluate the performance of the proposed measure on real data and show that it can elucidate the distributional instability of a parameter with respect to certain shifts and can be used to improve estimation accuracy under shifted distributions.

stat.ME↗