arXiv subjects

Ziliang Lai

Publications and source records attributed to Ziliang Lai.

9 recordsLinked to original sources

arXiv2026

X-Stage: Modeling Post-Issue Backpressure in GPU Communication--Computation Fusion

Fine-grained, device-initiated communication allows fused GPU kernels to issue remote stores directly from their compute pipelines, a pattern increasingly used in expert parallelism (EP), tensor parallelism (TP), and Ulysses-style sequence parallelism (UP). Existing designs reason about where communication is issued and when remote data becomes ready, but lack a quantitative model of the sender-side interval after a remote store is accepted and before it becomes visible at the destination. This interval determines whether communication remains decoupled from computation or backpressures it. We identify X-Stage, a software-visible post-issue stage with finite decoupling. Downstream pressure can dissipate while the issuer resumes useful work, whereas sustained injection consumes X-Stage headroom and eventually stalls the compute pipeline. We characterize this behavior and build a calibrated model that predicts whether remote-store arrivals accumulate backpressure or recover during intervening computation. Guided by the model, we reshape bursty arrivals when they would exhaust X-Stage headroom and exploit natural compute windows when headroom can recover concurrently. Evaluation across representative EP, TP, and UP workloads shows up to 1.62x fused-kernel, 1.75x end-to-end, and 1.43x sender-visible speedup, respectively. Microbenchmarks further validate the model's predictions of backlog accumulation, recovery, and sender-side backpressure.

cs.DC↗

arXiv2026

Anchoring Instruction Outside Mask: Exact Reference Caching for Efficient In-Context Diffusion Transformers

Omnimodal generation is central to a wide range of content creation and editing applications. In-context conditioning is essential to this paradigm. It allows diffusion transformers to process text instructions and visual references in a shared attention sequence. However, each reference image introduces thousands of tokens. Computation therefore grows rapidly with the number of references. Existing methods reduce computation through structured sparse attention, which limits interactions between reference and target tokens. This structure also makes the reference K and V independent of the denoising target, allowing them to be computed once and reused across steps. However, it blocks visual references from attending to the text instruction. This substantially degrades instruction following and reference fidelity in multi-reference editing. To resolve this conflict, we jointly redesign the token sequence and attention mask. Our beyond-mask design uses static text anchors to connect the instruction to the reference branch. It preserves exact K and V reuse without adding parameters. However, this direct architectural conversion degrades generation quality. We recover the lost performance through teacher-forced velocity distillation, followed by a short on-policy stage in which the teacher supervises student-visited states. To our knowledge, this is the first use of on-policy distillation for architectural recovery in diffusion models. Across three image-editing benchmarks, our method matches full-attention generation quality. With five reference images, it accelerates the complete 40-step denoising process by 3.92x, while static text anchors introduce negligible runtime overhead; the speedup reaches 5.47x at ten references in our scaling study.

cs.CV↗

arXiv2026

Kwai Keye-VL-2.0 Technical Report

We introduce Kwai Keye-VL-2.0-30B-A3B, an open-source Mixture-of-Experts (MoE) multimodal foundation model designed to advance long-video understanding and agentic intelligence. To address the challenges of ultra-long contexts, information redundancy, and prohibitive computational costs inherent in hour-level videos, Keye-VL-2.0 is the first to adapt DeepSeek Sparse Attention (DSA) to GQA-based multimodal architectures, enabling lossless 256K context processing while capturing critical frames and long-range temporal dependencies. This architecture is underpinned by a highly optimized training and inference infrastructure, including scalable video I/O, heterogeneous ViT-LM parallelism, and custom DSA kernels that significantly maximize throughput and minimize computational overhead. Furthermore, to overcome the algorithmic dilemma of catastrophic forgetting during multi-task alignment, we introduce Cross-Modal Multi-Teacher On-Policy Distillation (MOPD) paired with Context-RL and Video-RL. By distilling dense token-level teacher feedback from on-policy rollouts back into the MoE backbone, which activates only 3B parameters, Keye-VL-2.0 natively empowers advanced agent collaboration across Code, Tool, and Search scenarios with multimodal self-correction. Extensive evaluations across video understanding, temporal grounding, reasoning, STEM, and agent benchmarks demonstrate that Keye-VL-2.0-30B-A3B achieves state-of-the-art performance among models of similar scale, particularly excelling in fine-grained temporal localization on TimeLens and long-video comprehension on Video-MME-v2 and LongVideoBench. We release our model checkpoints to accelerate community progress toward scalable and robust multimodal agentic applications.

cs.CV↗

arXiv2023

Knock Out 2PC with Practicality Intact: a High-performance and General Distributed Transaction Protocol (Technical Report)

Two-phase-commit (2PC) has been widely adopted for distributed transaction processing, but it also jeopardizes throughput by introducing two rounds of network communications and two durable log writes to a transaction's critical path. Despite the various proposals that eliminate 2PC such as deterministic database and access localization, 2PC remains the de facto standard since the alternatives often lack generality (e.g., requiring workloads without branches based on query results). In this paper, we present Primo, a distributed transaction protocol that supports a more general set of workloads without 2PC. Primo features write-conflict-free concurrency control that guarantees once a transaction enters the commit phase, no concurrency conflict (e.g., deadlock) would occur when installing the write-set -- hence the prepare phase is no longer needed to account for any potential conflict from any partition. In addition, Primo further optimizes the transaction path using asynchronous group commit. With that, the durability delay is also taken off the transaction's critical path. Empirical results on Primo are encouraging -- in YCSB and TPC-C, Primo attains 1.42x to 8.25x higher throughput than state-of-the-art general protocols including Sundial and COCO, while having similar latency as COCO which also employs group commit.

cs.DB↗

arXiv2022

When Private Blockchain Meets Deterministic Database

Private blockchain as a replicated transactional system shares many commonalities with distributed database. However, the intimacy between private blockchain and deterministic database has never been studied. In essence, private blockchain and deterministic database both ensure replica consistency by determinism. In this paper, we present a comprehensive analysis to uncover the connections between private blockchain and deterministic database. While private blockchains have started to pursue deterministic transaction executions recently, deterministic databases have already studied deterministic concurrency control protocols for almost a decade. This motivates us to propose Harmony, a novel deterministic concurrency control protocol designed for blockchain use. We use Harmony to build a new relational blockchain, namely HarmonyBC, which features low abort rates, hotspot resiliency, and inter-block parallelism, all of which are especially important to disk-oriented blockchain. Empirical results on Smallbank, YCSB, and TPC-C show that HarmonyBC offers 2.0x to 3.5x throughput better than the state-of-the-art private blockchains.

cs.DB↗

arXiv2022

Saguaro: An Edge Computing-Enabled Hierarchical Permissioned Blockchain

We present Saguaro, a permissioned blockchain system designed specifically for edge computing networks. Saguaro leverages the hierarchical structure of edge computing networks to reduce the overhead of wide-area communication by presenting several techniques. First, Saguaro proposes coordinator-based and optimistic protocols to process cross-domain transactions with low latency where the lowest common ancestor of the involved domains coordinates the protocol or detects inconsistency. Second, data are collected over hierarchy enabling higher-level domains to aggregate their sub-domain data. Finally, transactions initiated by mobile edge devices are processed without relying on high-level fog and cloud servers. Our experimental results across a wide range of workloads demonstrate the scalability of Saguaro in supporting a range of cross-domain and mobile transactions.

cs.DB↗

arXiv2021

Top-K Deep Video Analytics: A Probabilistic Approach

The impressive accuracy of deep neural networks (DNNs) has created great demands on practical analytics over video data. Although efficient and accurate, the latest video analytic systems have not supported analytics beyond selection and aggregation queries. In data analytics, Top-K is a very important analytical operation that enables analysts to focus on the most important entities. In this paper, we present Everest, the first system that supports efficient and accurate Top-K video analytics. Everest ranks and identifies the most interesting frames/moments from videos with probabilistic guarantees. Everest is a system built with a careful synthesis of deep computer vision models, uncertain data management, and Top-K query processing. Evaluations on real-world videos and the latest Visual Road benchmark show that Everest achieves between 14.3x to 20.6x higher efficiency than baseline approaches with high result accuracy

cs.DB↗

arXiv2020

Towards Self-Tuning Parameter Servers

Recent years, many applications have been driven advances by the use of Machine Learning (ML). Nowadays, it is common to see industrial-strength machine learning jobs that involve millions of model parameters, terabytes of training data, and weeks of training. Good efficiency, i.e., fast completion time of running a specific ML job, therefore, is a key feature of a successful ML system. While the completion time of a long-running ML job is determined by the time required to reach model convergence, practically that is also largely influenced by the values of various system settings. In this paper, we contribute techniques towards building self-tuning parameter servers. Parameter Server (PS) is a popular system architecture for large-scale machine learning systems; and by self-tuning we mean while a long-running ML job is iteratively training the expert-suggested model, the system is also iteratively learning which system setting is more efficient for that job and applies it online. While our techniques are general enough to various PS-style ML systems, we have prototyped our techniques on top of TensorFlow. Experiments show that our techniques can reduce the completion times of a variety of long-running TensorFlow jobs from 1.4x to 18x.

cs.DB↗

arXiv2018

Decentralized Search on Decentralized Web

Decentralized Web, or DWeb, is envisioned as a promising future of the Web. Being decentralized, there are no dedicated web servers in DWeb; Devices that retrieve web contents also serve their cached data to peer devices with straight privacy-preserving mechanisms. The fact that contents in DWeb are distributed, replicated, and decentralized lead to a number of key advantages over the conventional web. These include better resiliency against network partitioning and distributed-denial-of-service attacks (DDoS), and better browsing experiences in terms of shorter latency and higher throughput. Moreover, DWeb provides tamper-proof contents because each content piece is uniquely identified by a cryptographic hash. DWeb also clicks well with future Internet architectures, such as Named Data Networking (NDN).Search engines have been an inseparable element of the Web. Contemporary ("Web 2.0") search engines, however, provide centralized services. They are thus subject to DDoS attacks, insider threat, and ethical issues like search bias and censorship. As the web moves from being centralized to being decentralized, search engines ought to follow. We propose QueenBee, a decentralized search engine for DWeb. QueenBee is so named because worker bees and honeycomb are a common metaphor for distributed architectures, with the queen being the one that holds the colony together. QueenBee aims to revolutionize the search engine business model by offering incentives to both content providers and peers that participate in QueenBee's page indexing and ranking operations.

cs.IR↗