arXiv subjects

Gengrui Zhang

Publications and source records attributed to Gengrui Zhang.

17 recordsLinked to original sources

arXiv2026

REMOP: REmote-Memory-aware OPerator Optimization

Remote and disaggregated memory tiers expand the effective memory capacity of analytical database engines, but they also reshape the cost structure of out-of-memory query processing. When an operator spills beyond local DRAM, moving pages to remote memory incurs both data-transfer time and a fixed round-trip latency per transfer. Classical operator analyses and buffer-allocation heuristics primarily target disk spilling by minimizing total I/O volume. Under remote memory, these strategies can be suboptimal because they may trigger excessive transfer rounds. We present REMOP, a remote-memory-aware operator optimization framework that uses transfer-round-aware intra-operator memory policies to improve out-of-memory execution under tight memory budgets. REMOP introduces the number of transfer rounds into the latency cost model and derives operator-specific buffer-partitioning strategies, instantiating the approach for blocked nested-loop join, external merge sort, and external hash join in DuckDB. Our evaluation on a two-node compute-memory testbed shows that REMOP reduces transfer rounds by up to 97% and operator runtime by up to 48% on spill-heavy microbenchmarks, and lowers the average runtime of spilling TPC-H and TPC-DS queries by 22.7% and 26.4% end-to-end.

cs.DB↗

arXiv2026

Epoch-based Optimistic Concurrency Control in Geo-replicated Databases

Geo-distribution is essential for modern online applications to ensure service reliability and high availability. However, supporting high-performance serializable transactions in geo-replicated databases remains a significant challenge. This difficulty stems from the extensive over-coordination inherent in distributed atomic commitment, concurrency control, and fault-tolerance replication protocols under high network latency. To address these challenges, we introduce Minerva, a unified distributed concurrency control designed for highly scalable multi-leader replication. Minerva employs a novel epoch-based asynchronous replication protocol that decouples data propagation from the commitment process, enabling continuous transaction replication. Optimistic concurrency control is used to allow any replicas to execute transactions concurrently and commit without coordination. In stead of aborting transactions when conflicts are detected, Minerva uses deterministic re-execution to resolve conflicts, ensuring serializability without sacrificing performance. To further enhance concurrency, we construct a conflict graph and use a maximum weight independent set algorithm to select the optimal subset of transactions for commitment, minimizing the number of re-executed transactions. Our evaluation demonstrates that Minerva significantly outperforms state-of-the-art replicated databases, achieving over $3\times$ higher throughput in scalability experiments and $2.8\times$ higher throughput during a high network latency simulation with the TPC-C benchmark.

cs.DB↗

arXiv2025

WOC: Dual-Path Weighted Object Consensus Made Efficient

Modern distributed systems face a critical challenge: existing consensus protocols optimize for either node heterogeneity or workload independence, but not both. For example, Cabinet leverages weighted quorums to handle node heterogeneity but serializes all operations through a global leader, limiting parallelism. EPaxos enables parallel execution for independent operations but treats all nodes uniformly, ignoring performance differences. To tackle this problem, we present WOC, a dual-path consensus protocol that dynamically routes operations into two paths based on their access patterns. Independent operations execute through a fast path that uses object-specific weighted quorums and completes in one network round-trip. Conflicting or shared objects route through a leader-coordinated slow path employing node-weighted consensus. Our evaluation demonstrates that WOC achieves up to 4X higher throughput than Cabinet for workloads with >70% independent objects, while maintaining equivalent performance under high contention.

cs.DC↗

arXiv2025

The Pervasive Blind Spot: Benchmarking VLM Inference Risks on Everyday Personal Videos

The proliferation of Vision-Language Models (VLMs) introduces profound privacy risks from personal videos. This paper addresses the critical yet unexplored inferential privacy threat, the risk of inferring sensitive personal attributes over the data. To address this gap, we crowdsourced a dataset of 508 everyday personal videos from 58 individuals. We then conducted a benchmark study evaluating VLM inference capabilities against human performance. Our findings reveal three critical insights: (1) VLMs possess superhuman inferential capabilities, significantly outperforming human evaluators, leveraging a shift from object recognition to behavioral inference from temporal streams. (2) Inferential risk is strongly correlated with factors such as video characteristics and prompting strategies. (3) VLM-driven explanation towards the inference is unreliable, as we revealed a disconnect between the model-generated explanations and evidential impact, identifying ubiquitous objects as misleading confounders.

cs.HC↗

arXiv2025

GPoS: Geospatially-aware Proof of Stake

Geospatial decentralization is essential for blockchains, ensuring regulatory resilience, robustness, and fairness. We empirically analyze five major Proof of Stake (PoS) blockchains: Aptos, Avalanche, Ethereum, Solana, and Sui, revealing that a few geographic regions dominate consensus voting power, resulting in limited geospatial decentralization. To address this, we propose Geospatially aware Proof of Stake (GPoS), which integrates geospatial diversity with stake-based voting power. Experimental evaluation demonstrates an average 45% improvement in geospatial decentralization, as measured by the Gini coefficient of Eigenvector centrality, while incurring minimal performance overhead in BFT protocols, including HotStuff and CometBFT. These results demonstrate that GPoS can improve geospatial decentralization {while, in our experiments, incurring minimal overhead} to consensus performance.

cs.DC↗

arXiv2025

Learning Compact Representations of LLM Abilities via Item Response Theory

Recent years have witnessed a surge in the number of large language models (LLMs), yet efficiently managing and utilizing these vast resources remains a significant challenge. In this work, we explore how to learn compact representations of LLM abilities that can facilitate downstream tasks, such as model routing and performance prediction on new benchmarks. We frame this problem as estimating the probability that a given model will correctly answer a specific query. Inspired by the item response theory (IRT) in psychometrics, we model this probability as a function of three key factors: (i) the model's multi-skill ability vector, (2) the query's discrimination vector that separates models of differing skills, and (3) the query's difficulty scalar. To learn these parameters jointly, we introduce a Mixture-of-Experts (MoE) network that couples model- and query-level embeddings. Extensive experiments demonstrate that our approach leads to state-of-the-art performance in both model routing and benchmark accuracy prediction. Moreover, analysis validates that the learned parameters encode meaningful, interpretable information about model capabilities and query characteristics.

cs.AI↗

arXiv2025

Truly Assessing Fluid Intelligence of Large Language Models through Dynamic Reasoning Evaluation

Recent advances in large language models (LLMs) have demonstrated impressive reasoning capacities that mirror human-like thinking. However, whether LLMs possess genuine fluid intelligence (i.e., the ability to reason abstractly and generalize rules in novel situations) remains an open question. Existing reasoning benchmarks either focus on domain-specific knowledge (crystallized intelligence) or lack interpretability. To address these limitations, we propose DRE-Bench, a dynamic reasoning evaluation benchmark grounded in a hierarchical cognitive framework. DRE-Bench consists of 36 abstract reasoning tasks organized across four cognitive levels, with each task featuring multiple dynamic variants that test the same underlying latent rule. This design enables fine-grained, interpretable, and reliable assessments of fluid intelligence. We evaluate a range of state-of-the-art LLMs, including both general LLMs (GPT-4o, Claude 3.7) and reasoning LLMs (o1, DeepSeek-R1, QwQ, Skywork-OR1). Experimental results reveal that although most LLMs achieve competent and robust performance in low-level cognition, they struggle with high-level cognition and exhibit limited generalization as task complexity grows. Our findings highlight the gap between current LLMs and true human-like fluid intelligence and offer a new path for systematically tracking reasoning progress in LLMs.

cs.AI↗

arXiv2025

Through Their Eyes: User Perceptions on Sensitive Attribute Inference of Social Media Videos by Visual Language Models

The rapid advancement of Visual Language Models (VLMs) has enabled sophisticated analysis of visual content, leading to concerns about the inference of sensitive user attributes and subsequent privacy risks. While technical capabilities of VLMs are increasingly studied, users' understanding, perceptions, and reactions to these inferences remain less explored, especially concerning videos uploaded on the social media. This paper addresses this gap through a semi-structured interview (N=17), investigating user perspectives on VLM-driven sensitive attribute inference from their visual data. Findings reveal that users perceive VLMs as capable of inferring a range of attributes, including location, demographics, and socioeconomic indicators, often with unsettling accuracy. Key concerns include unauthorized identification, misuse of personal information, pervasive surveillance, and harm from inaccurate inferences. Participants reported employing various mitigation strategies, though with skepticism about their ultimate effectiveness against advanced AI. Users also articulate clear expectations for platforms and regulators, emphasizing the need for enhanced transparency, user control, and proactive privacy safeguards. These insights are crucial for guiding the development of responsible AI systems, effective privacy-enhancing technologies, and informed policymaking that aligns with user expectations and societal values.

cs.HC↗

arXiv2025

Large Language Models are Near-Optimal Decision-Makers with a Non-Human Learning Behavior

Human decision-making belongs to the foundation of our society and civilization, but we are on the verge of a future where much of it will be delegated to artificial intelligence. The arrival of Large Language Models (LLMs) has transformed the nature and scope of AI-supported decision-making; however, the process by which they learn to make decisions, compared to humans, remains poorly understood. In this study, we examined the decision-making behavior of five leading LLMs across three core dimensions of real-world decision-making: uncertainty, risk, and set-shifting. Using three well-established experimental psychology tasks designed to probe these dimensions, we benchmarked LLMs against 360 newly recruited human participants. Across all tasks, LLMs often outperformed humans, approaching near-optimal performance. Moreover, the processes underlying their decisions diverged fundamentally from those of humans. On the one hand, our finding demonstrates the ability of LLMs to manage uncertainty, calibrate risk, and adapt to changes. On the other hand, this disparity highlights the risks of relying on them as substitutes for human judgment, calling for further inquiry.

cs.AI↗

arXiv2025

Cabinet: Dynamically Weighted Consensus Made Fast

Conventional consensus algorithms, such as Paxos and Raft, encounter inefficiencies when applied to large-scale distributed systems due to the requirement of waiting for replies from a majority of nodes. To address these challenges, we propose Cabinet, a novel consensus algorithm that introduces dynamically weighted consensus, allocating distinct weights to nodes based on any given failure thresholds. Cabinet dynamically adjusts nodes' weights according to their responsiveness, assigning higher weights to faster nodes. The dynamic weight assignment maintains an optimal system performance, especially in large-scale and heterogeneous systems where node responsiveness varies. We evaluate Cabinet against Raft with distributed MongoDB and PostgreSQL databases using YCSB and TPC-C workloads. The evaluation results show that Cabinet outperforms Raft in throughput and latency under increasing system scales, complex networks, and failures in both homogeneous and heterogeneous clusters, offering a promising high-performance consensus solution.

cs.DC↗

arXiv2024

Cache-Aware Reinforcement Learning in Large-Scale Recommender Systems

Modern large-scale recommender systems are built upon computation-intensive infrastructure and usually suffer from a huge difference in traffic between peak and off-peak periods. In peak periods, it is challenging to perform real-time computation for each request due to the limited budget of computational resources. The recommendation with a cache is a solution to this problem, where a user-wise result cache is used to provide recommendations when the recommender system cannot afford a real-time computation. However, the cached recommendations are usually suboptimal compared to real-time computation, and it is challenging to determine the items in the cache for each user. In this paper, we provide a cache-aware reinforcement learning (CARL) method to jointly optimize the recommendation by real-time computation and by the cache. We formulate the problem as a Markov decision process with user states and a cache state, where the cache state represents whether the recommender system performs recommendations by real-time computation or by the cache. The computational load of the recommender system determines the cache state. We perform reinforcement learning based on such a model to improve user engagement over multiple requests. Moreover, we show that the cache will introduce a challenge called critic dependency, which deteriorates the performance of reinforcement learning. To tackle this challenge, we propose an eigenfunction learning (EL) method to learn independent critics for CARL. Experiments show that CARL can significantly improve the users' engagement when considering the result cache. CARL has been fully launched in Kwai app, serving over 100 million users.

cs.LG↗

arXiv2024

UNEX-RL: Reinforcing Long-Term Rewards in Multi-Stage Recommender Systems with UNidirectional EXecution

In recent years, there has been a growing interest in utilizing reinforcement learning (RL) to optimize long-term rewards in recommender systems. Since industrial recommender systems are typically designed as multi-stage systems, RL methods with a single agent face challenges when optimizing multiple stages simultaneously. The reason is that different stages have different observation spaces, and thus cannot be modeled by a single agent. To address this issue, we propose a novel UNidirectional-EXecution-based multi-agent Reinforcement Learning (UNEX-RL) framework to reinforce the long-term rewards in multi-stage recommender systems. We show that the unidirectional execution is a key feature of multi-stage recommender systems, bringing new challenges to the applications of multi-agent reinforcement learning (MARL), namely the observation dependency and the cascading effect. To tackle these challenges, we provide a cascading information chain (CIC) method to separate the independent observations from action-dependent observations and use CIC to train UNEX-RL effectively. We also discuss practical variance reduction techniques for UNEX-RL. Finally, we show the effectiveness of UNEX-RL on both public datasets and an online recommender system with over 100 million users. Specifically, UNEX-RL reveals a 0.558% increase in users' usage time compared with single-agent RL algorithms in online A/B experiments, highlighting the effectiveness of UNEX-RL in industrial recommender systems.

cs.IR↗

arXiv2023

Reaching Consensus in the Byzantine Empire: A Comprehensive Review of BFT Consensus Algorithms

Byzantine fault-tolerant (BFT) consensus algorithms are at the core of providing safety and liveness guarantees for distributed systems that must operate in the presence of arbitrary failures. Recently, numerous new BFT algorithms have been proposed, not least due to the traction blockchain technologies have garnered in the search for consensus solutions that offer high throughput, low latency, and robust system designs. In this paper, we conduct a systematic survey of selected and distinguished BFT algorithms that have received extensive attention in academia and industry alike. We perform a qualitative comparison among all algorithms we review considering message and time complexities. Furthermore, we decompose each consensus algorithm into its constituent subprotocols for replication and view change backed by intuitive figures to illustrate the message-passing pattern. We also elaborate on the strengths and weaknesses of each algorithm as compared to the state-of-the-art approaches.

cs.DC↗

arXiv2023

Lifting the Fog of Uncertainties: Dynamic Resource Orchestration for the Containerized Cloud

The advances in virtualization technologies have sparked a growing transition from virtual machine (VM)-based to container-based infrastructure for cloud computing. From the resource orchestration perspective, containers' lightweight and highly configurable nature not only enables opportunities for more optimized strategies, but also poses greater challenges due to additional uncertainties and a larger configuration parameter search space. Towards this end, we propose Drone, a resource orchestration framework that adaptively configures resource parameters to improve application performance and reduce operational cost in the presence of cloud uncertainties. Built on Contextual Bandit techniques, Drone is able to achieve a balance between performance and resource cost on public clouds, and optimize performance on private clouds where a hard resource constraint is present. We show that our algorithms can achieve sub-linear growth in cumulative regret, a theoretically sound convergence guarantee, and our extensive experiments show that Drone achieves an up to 45% performance improvement and a 20% resource footprint reduction across batch processing jobs and microservice workloads.

cs.DC↗

arXiv2023

PrestigeBFT: Revolutionizing View Changes in BFT Consensus Algorithms with Reputation Mechanisms

This paper proposes PrestigeBFT, a novel leader-based BFT consensus algorithm that addresses the weaknesses of passive view-change protocols. Passive protocols blindly rotate leadership among servers on a predefined schedule, potentially selecting unavailable or slow servers as leaders. PrestigeBFT proposes an active view-change protocol using reputation mechanisms that calculate a server's potential correctness based on historic behavior. The active protocol enables servers to campaign for leadership by performing reputation-associated work. As such, up-to-date and correct servers with good reputations are more likely to be elected as leaders as they perform less work, whereas faulty servers with bad reputations are suppressed from becoming leaders by being required to perform more work. Under normal operation, PrestigeBFT achieves 5X higher throughput than the baseline that uses passive view-change protocols. In addition, PrestigeBFT remains unaffected under benign faults and experiences only a 24% drop in throughput under a variety of Byzantine faults, while the baseline throughput drops by 62% and 69%, respectively.

cs.DC↗

arXiv2023

V-Guard: An Efficient Permissioned Blockchain for Achieving Consensus under Dynamic Memberships in V2X

This paper presents V-Guard, a new permissioned blockchain that achieves consensus for vehicular data under changing memberships, targeting the problem in V2X networks where vehicles are often intermittently connected on the roads. To achieve this goal, V-Guard integrates membership management into the consensus process for agreeing on data entries. It binds a data entry with a membership configuration profile that describes responsible vehicles for achieving consensus for the data entry. As such, V-Guard produces chained consensus results of both data entries and their residing membership profiles, which enables consensus to be achieved seamlessly under changing memberships. In addition, V-Guard separates the ordering of transactions from consensus, allowing concurrent ordering instances and periodic consensus instances to order and commit data entries. These features make V-Guard efficient for achieving consensus under dynamic memberships with high throughput and latency performance.

cs.DC↗

arXiv2022

ESCAPE to Precaution against Leader Failures

Leader-based consensus protocols must undergo a view-change phase to elect a new leader when the current leader fails. The new leader is often decided upon a candidate server that collects votes from a quorum of servers. However, voting-based election mechanisms intrinsically cause competition in leadership candidacy when each candidate collects only partial votes. This split-vote scenario can result in no leadership winner and prolong the undesired view-change period. In this paper, we investigate a case study of Raft's leader election mechanism and propose a new leader election protocol, called ESCAPE, that fundamentally solves split votes by prioritizing servers based on their log responsiveness. ESCAPE dynamically assigns servers with a configuration that offers different priorities through Raft's periodic heartbeat. In each assignment, ESCAPE keeps track of server log responsiveness and assigns configurations that are inclined to win an election to more up-to-date servers, thereby preparing a pool of prioritized candidates. Consequently, when the next election takes place, the candidate with the highest priority will defeat its counterparts and becomes the next leader without competition. The evaluation results show that ESCAPE progressively reduces the leader election time when the cluster scales up, and the improvement becomes more significant under message loss.

cs.DC↗