Researcher profile

Zijie Zhou

Zijie Zhou contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate

Artificial Intelligence math.OC Machine Learning Computation and Language Computer Vision Distributed, Parallel, and Cluster Computing math.PR Robotics Software Engineering

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author

8works

0followers

9topics

4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

A Queueing-Theoretic Framework for Stability Analysis of LLM Inference with KV Cache Memory Constraints

The rapid adoption of large language models (LLMs) has created significant challenges for efficient inference at scale. Unlike traditional workloads, LLM inference is constrained by both computation and the memory overhead of key-value (KV) caching, which accelerates decoding but quickly exhausts GPU memory. In this paper, we introduce the first queueing-theoretic framework that explicitly incorporates both computation and GPU memory constraints into the analysis of LLM inference. Based on this framework, we derive rigorous stability and instability conditions that determine whether an LLM inference service can sustain incoming demand without unbounded queue growth. This result offers a powerful tool for system deployment, potentially addressing the core challenge of GPU provisioning. By combining an estimated request arrival rate with our derived stable service rate, operators can calculate the necessary cluster size to avoid both costly over-purchasing and performance-violating under-provisioning. We further validate our theoretical predictions through extensive experiments in real GPU production environments. Our results show that the predicted stability conditions are highly accurate, with deviations typically within 10%.

preprint2026arXiv

Automating API Documentation from Crowdsourced Knowledge

API documentation is crucial for developers to learn and use APIs. However, it is known that many official API documents are obsolete and incomplete. To address this challenge, we propose a new approach called AutoDoc that generates API documents with API knowledge extracted from online discussions on Stack Overflow (SO). AutoDoc leverages a fine-tuned dense retrieval model to identify seven types of API knowledge from SO posts. Then, it uses GPT-4o to summarize the API knowledge in these posts into concise text. Meanwhile, we designed two specific components to handle LLM hallucination and redundancy in generated content. We evaluated AutoDoc against five comparison baselines on 48 APIs of different popularity levels. Our results indicate that the API documents generated by AutoDoc are up to 77.7% more accurate, 9.5% less duplicated, and contain 34.4% knowledge uncovered by the official documents. We also measured the sensitivity of AutoDoc to the choice of different LLMs. We found that while larger LLMs produce higher-quality API documents, AutoDoc enables smaller open-source models (e.g., Mistral-7B-v0.3) to achieve comparable results. Finally, we conducted a user study to evaluate the usefulness of the API documents generated by AutoDoc. All participants found API documents generated by AutoDoc to be more comprehensive, concise, and helpful than the comparison baselines. This highlights the feasibility of utilizing LLMs for API documentation with careful design to counter LLM hallucination and information redundancy.

preprint2026arXiv

Learning Transferable Topology Priors for Multi-Agent LLM Collaboration Across Domains

Large language model (LLM)-based multi-agent systems have shown strong potential for complex reasoning by coordinating specialized agents through structured communication. However, existing topology-evolution methods typically construct or optimize a collaboration topology for each query from scratch, leading to substantial online search overhead, high inference-time token consumption, and limited scalability in multi-domain settings. We propose TopoPrior, a framework for learning transferable topology priors for multi-agent LLM collaboration across domains. Rather than repeatedly searching for effective collaboration structures online, TopoPrior learns reusable topology priors from reference collaboration graphs collected offline from multiple domains and uses them to generate query-conditioned initial collaboration graphs for downstream refinement. By shifting part of topology search from per-query online optimization to offline prior learning, TopoPrior amortizes search cost while remaining compatible with existing topology-evolution backbones. Technically, TopoPrior contains two key components. First, a transferable topology prior learning module employs a conditional variational graph framework to capture reusable structural regularities across domains in a latent space. Second, a query-conditioned latent adaptation module introduces adversarial alignment to reduce unnecessary domain discrepancy while preserving query-relevant structural variation. Experiments on multi-domain reasoning benchmarks show that TopoPrior consistently improves several heterogeneous topology-evolution backbones while reducing online inference-time token usage, with only modest additional trainable parameters. These results suggest that transferable topology initialization is an effective and lightweight mechanism for improving the efficiency of multi-agent LLM collaboration across domains.

preprint2026arXiv

Online Scheduling for LLM Inference with KV Cache Constraints

Large Language Model (LLM) inference, where a trained model generates text one word at a time in response to user prompts, is a computationally intensive process requiring efficient scheduling to optimize latency and resource utilization. A key challenge in LLM inference is the management of the Key-Value (KV) cache, which reduces redundant computations but introduces memory constraints. In this work, we model LLM inference with KV cache constraints theoretically and propose a novel batching and scheduling algorithm that minimizes inference latency while effectively managing the KV cache's memory. More specifically, we make the following contributions. First, to evaluate the performance of online algorithms for scheduling in LLM inference, we introduce a hindsight optimal benchmark, formulated as an integer program that computes the minimum total inference latency under full future information. Second, we prove that no deterministic online algorithm can achieve a constant competitive ratio when the arrival process is arbitrary. Third, motivated by the computational intractability of solving the integer program at scale, we propose a polynomial-time online scheduling algorithm and show that under certain conditions it can achieve a constant competitive ratio. We also demonstrate our algorithm's strong empirical performance by comparing it to the hindsight optimal in a synthetic dataset. Finally, we conduct empirical evaluations on a real-world public LLM inference dataset, simulating the Llama2-70B model on A100 GPUs, and show that our algorithm significantly outperforms the benchmark algorithms. Overall, our results offer a path toward more sustainable and cost-effective LLM deployment.

preprint2026arXiv

Position: LLM Serving Needs Mathematical Optimization and Algorithmic Foundations, Not Just Heuristics

This position paper argues that LLM inference serving has outgrown generic heuristics and now demands mathematical optimization and algorithmic foundations. Despite rapid advances in serving systems such as vLLM and SGLang, their algorithmic cores remain largely unchanged from classical distributed computing: request routing uses join-shortest-queue or round-robin, scheduling defaults to FIFO, and KV cache eviction follows LRU. These general-purpose policies ignore the distinctive structure of LLM inference--dynamically growing KV cache memory, prefill-decode phase asymmetry, unknown output lengths, and continuous batching constraints. We contend that the field must develop mathematical models capturing these characteristics, enabling the design of algorithms with provable performance guarantees across diverse workloads, rather than heuristics that may succeed in some scenarios but fail unpredictably in others. Emerging work at the intersection of operations research and ML systems demonstrates that principled methods can match or exceed heuristic performance while providing theoretical guarantees. We call on the community to recognize algorithmic design for LLM serving as a research frontier.

preprint2023arXiv

LCPR: A Multi-Scale Attention-Based LiDAR-Camera Fusion Network for Place Recognition

Place recognition is one of the most crucial modules for autonomous vehicles to identify places that were previously visited in GPS-invalid environments. Sensor fusion is considered an effective method to overcome the weaknesses of individual sensors. In recent years, multimodal place recognition fusing information from multiple sensors has gathered increasing attention. However, most existing multimodal place recognition methods only use limited field-of-view camera images, which leads to an imbalance between features from different modalities and limits the effectiveness of sensor fusion. In this paper, we present a novel neural network named LCPR for robust multimodal place recognition, which fuses LiDAR point clouds with multi-view RGB images to generate discriminative and yaw-rotation invariant representations of the environment. A multi-scale attention-based fusion module is proposed to fully exploit the panoramic views from different modalities of the environment and their correlations. We evaluate our method on the nuScenes dataset, and the experimental results show that our method can effectively utilize multi-view camera and LiDAR data to improve the place recognition performance while maintaining strong robustness to viewpoint changes. Our open-source code and pre-trained models are available at https://github.com/ZhouZijie77/LCPR .

preprint2022arXiv

Near-Optimal Primal-Dual Algorithms for Quantity-Based Network Revenue Management

We study the canonical quantity-based network revenue management (NRM) problem where the decision-maker must irrevocably accept or reject each arriving customer request with the goal of maximizing the total revenue given limited resources. The exact solution to the problem by dynamic programming is computationally intractable due to the well-known curse of dimensionality. Existing works in the literature make use of the solution to the deterministic linear program (DLP) to design asymptotically optimal algorithms. Those algorithms rely on repeatedly solving DLPs to achieve near-optimal regret bounds. It is, however, time-consuming to repeatedly compute the DLP solutions in real time, especially in large-scale problems that may involve hundreds of millions of demand units. In this paper, we propose innovative algorithms for the NRM problem that are easy to implement and do not require solving any DLPs. Our algorithm achieves a regret bound of $O(\log k)$, where $k$ is the system size. To the best of our knowledge, this is the first NRM algorithm that (i) has an $o(\sqrt{k})$ asymptotic regret bound, and (ii) does not require solving any DLPs.

preprint2020arXiv

Best Strategy for Each Team in The Regular Season to Win Champion in The Knockout Tournament

In J. Schwenk.(2018) ['What is the Correct Way to Seed a Knockout Tournament?' Retrieved from The American Mathematical Monthly], Schwenk identified a surprising weakness in the standard method of seeding a single elimination (or knockout) tournament. In particular, he showed that for a certain probability model for the outcomes of games it can be the case that the top seeded team would be less likely to win the tournament than the second seeded team. This raises the possibility that in certain situations it might be advantageous for a team to intentionally lose a game in an attempt to get a more optimal (though possibly lower) seed in the tournament. We examine this question in the context of a four team league which consists of a round robin "regular season" followed by a single elimination tournament with seedings determined by the results from the regular season [4]. Using the same probability model as Schwenk we show that there are situations where it is indeed optimal for a team to intentionally lose. Moreover, we show how a team can make the decision as to whether or not it should intentionally lose. We did two detailed analysis. One is for the situation where other teams always try to win every game. The other is for the situation where other teams are smart enough, namely they can also lose some games intentionally if necessary. The analysis involves computations in both probability and (multi-player) game theory.

Institution

Affiliation not imported yet

This author record came from a source that does not expose affiliation metadata. Once the author claims the profile or we enrich the record from another provider, this section will link to the concrete institution.

Topic footprint