arXiv subjects

Kai Wan

Publications and source records attributed to Kai Wan.

At least 19 recordsLinked to original sources

arXiv2026

ALIGN-HOLD: Experience Alignment for Real-Time Hold Control in Large-Scale Ride-Hailing Matching at DiDi

Real-time hold control is a high-leverage mechanism in large-scale ride-hailing systems: by selectively deferring driver-order pairs, the platform can wait for better matching opportunities and improve end-to-end passenger-driver experience. Existing production systems such as EXHOLD learn bandit-based hold policies from handcrafted combinations of trip completion, cancellations, waiting time, and driver effort. However, designing such rewards becomes increasingly difficult as marketplace preferences are heterogeneous and observed passenger-driver behavior can be sparse, noisy, and affected by dynamic supply-demand conditions. We present ALIGN-HOLD, a production-scale experience alignment framework that learns hold policy from implicit marketplace preferences. ALIGN-HOLD constructs complementary preference pairs from order trajectories, driver trajectories, and contemporaneous local matching graphs, and trains an experience Reward Model (RM) using balanced multi-view sampling and model-adaptive hard preference sampling. During simulator-based policy learning, the frozen RM provides a dense, context-dependent reward and supports label-free filtering of low-identifiability interactions whose behavioral feedback is difficult to attribute to matching quality. We deploy ALIGN-HOLD on DiDi's ride-hailing platform and evaluate it in a 28-day randomized A/B experiment, covering approximately 100,000 passenger requests per day. Compared with the deployed production policy, ALIGN-HOLD achieves statistically significant improvements in trip completion rate and driver income, while significantly reducing passenger cancellations before and after driver acceptance. Complementary ablations, RM diagnostics, and behavioral analyses validate the contributions of the proposed components. ALIGN-HOLD has been fully ramped up and is currently serving DiDi's Brazil marketplace.

cs.LG↗

arXiv2026

EXHOLD: Experience-Aware Real-Time Hold Control for Large-Scale Ride-Hailing Matching at DiDi

In large-scale ride-hailing, hold control is a critical mechanism for improving passenger-driver experience. By selectively delaying certain driver-order pairs, the system waits for better opportunities, reduces cancellations, and mitigates wasted driver effort. However, existing industrial hold strategies often rely on heuristic thresholding over multiple predictive models, which can be brittle under non-stationary traffic and hard to optimize for multi-objective experience signals. We propose EXHOLD, a deployable two-stage framework decoupling experience-aware pair assessment from hold-time execution. In Stage I, we learn a decision model assigning each driver-order pair to discrete, interpretable experience tiers by optimizing a unified objective that aggregates satisfaction signals across the matching funnel. In Stage II, we solve for a monotone hold-time schedule via constrained optimization over empirical quantiles. This explicitly enforces service guardrails bounding the unnecessary holding of promising matches while maximizing overall experience improvement. We evaluate EXHOLD through randomized A/B experiments in DiDi's production system in Brazil. Results show consistent gains in marketplace efficiency and experience: EXHOLD increases trip completion and driver income, significantly reduces passenger cancellations, and improves funnel efficiency. Ablations and behavioral analyses confirm both stages are essential and that the policy makes calibrated decisions under spatiotemporal heterogeneity. EXHOLD is currently deployed, serving production traffic in Brazil.

cs.LG↗

arXiv2026

ProfiLLM: Utility-Aligned Agentic User Profiling for Industrial Ride-Hailing Dispatch

Bringing Large Language Models (LLMs) into industrial ride-hailing dispatch as semantic feature extractors over platform-scale behavioral logs is a compelling but under-explored data systems problem. Production matching pipelines remain dominated by structured numerical features, yet decisive behavioral signals (e.g., a driver's habitual aversion to certain regions) are inherently contextual and naturally expressible as LLM-generated user profiles. However, scaling such profiling to a live, millisecond-latency dispatcher faces three intertwined constraints rarely addressed together: on a platform with millions of daily orders, logs exceed any LLM's context window by orders of magnitude; most users are long-tail, with too few interactions for per-user profiling; and surface-fluent profiles do not necessarily improve downstream prediction utility. We present ProfiLLM, an agentic LLM data pipeline that operationalizes utility-aligned user profiling for production matching systems through two modules. (1) Tool-Augmented Global Knowledge Mining equips an LLM agent with 27 analytical tools to mine platform-scale data, producing reusable global knowledge, adaptive user clustering rules, and region-level supply-demand priors. (2) Utility-Aligned Profile Exploration generates multiple candidate profiles per cluster, evaluates them via a lightweight downstream utility proxy, iteratively refines the best candidates and constructs preference pairs for DPO fine-tuning. Deployed on DiDi's production dispatcher, ProfiLLM achieves up to +6.14% relative AUC improvement in outcome prediction, up to +4.35% GMV gain in dispatching simulation, and consistent improvements in a 14-day online A/B test including +0.47% GMV, +0.33% Completion Rate, and -0.82% Cancel-Before-Accept rate.

cs.AI↗

arXiv2026

Noncoherent ISAC over Block-Fading Channels: Asymptotic Performance Analysis

This paper investigates the fundamental limits and optimal signal distribution design for Integrated Sensing and Communication (ISAC) systems operating under strictly noncoherent conditions. Unlike conventional coherent frameworks that rely on perfect channel state information, we consider a block-fading MIMO channel where the channel realizations are unknown to both the transmitter and the receiver. We adopt a realization-wise perspective to characterize the noncoherent performance tradeoff across different signal-to-noise ratio (SNR) regimes. In the high-SNR regime, we derive a lower bound for the noncoherent mutual information and define a metric, termed sensing-induced rate loss, to quantify the communication penalty incurred by sensing-oriented beamforming. We then employ a projected gradient algorithm to optimize the spatial power allocation, balancing the conflict between the unitary space-time modulation-based structure for communication and the task-oriented spatial power allocation for sensing. Conversely, in the low-SNR regime, we perform a first-order asymptotic analysis of the ergodic minimum mean squared error (EMMSE). Our theoretical derivation reveals a fundamental synergy: the sensing-optimal strategy collapses to a rank-one transmission along the dominant eigenvector of the target response, which incurs no first-order communication loss in the low-SNR regime. This result demonstrates that the conflicting tradeoff observed at high SNR vanishes asymptotically at low SNR, enabling perfect alignment between sensing and communication objectives.

cs.IT↗

arXiv2026

The Capacity of Information-Theoretic Secure Aggregation in Federated Learning

Secure aggregation allows a server to aggregate users' local updates while preserving update privacy. Existing information-theoretic problems typically assume that correlated random keys are provided by a trusted third party (TTP) or generated via prescribed groupwise structures, while the communication cost for establishing such correlated keys is often ignored. Consequently, the fundamental limits under general key-distribution mechanisms remain unknown. In this paper, we study the $T$-colluding information-theoretic secure aggregation problem with $N$ users under a general two-phase framework consisting of a key distribution phase and an update aggregation phase. Unlike prior work, we model key distribution through user-to-user communication and allow arbitrary user-generated key-distribution mechanisms, eliminating TTP or prescribed structures. This enables a joint characterization of three resources: randomness for security, key-distribution communication, and aggregation communication. We completely characterize the capacity region among these three resources by constructing a novel secure aggregation scheme together with a matching information-theoretic converse. In particular, we develop an explicit deterministic capacity-achieving construction over any finite field of size at least $N$, whereas most existing schemes either rely on TTP or employ randomized or existential constructions over sufficiently large finite fields. We further show that the optimal performance can be achieved using only pairwise shared keys, enabling implementation via Diffie--Hellman key exchange. Compared with Google's seminal secure aggregation scheme, the proposed scheme requires fewer random masking keys while preserving the same aggregation communication overhead.

cs.IT↗

arXiv2026

Reliability-Constrained Blind Beam Alignment for Backscatter-MIMO mounted Target in Cluttered Multipath Channels

Practical ISAC is constrained by static clutter and NLoS multipath, which obscure target-coupled echoes and induce spurious peaks for beam alignment. Existing receiver-side methods largely model targets as passive scatterers, limiting the structural separability of target echoes from the environment. This paper establishes a structural correspondence between these limitations and target-side Backscatter-MIMO responses: reflection modulation enables waveform-domain separation from unmodulated clutter, while retro-directional passive beamforming concentrates the tagged echo toward the BS-facing direction and suppresses NLoS-induced false-peak locking. To operationalize this correspondence, dual-end spatial locking is required to overcome cascaded backscatter loss and provide beam-domain angular information. We propose a downlink-triggered blind dual-end alignment protocol that jointly selects the BS and Backscatter-MIMO codeword indices from the tagged echo observed at the BS, without pilots, CSI feedback, or target synchronization. We further derive a clutter-aware remodulation waveform robust to fractional timing offsets and construct adjustable-width BS/Backscatter-MIMO codebooks via quadratic phase spoiling. For reliability characterization, we derive closed-form expressions for the coherence-averaged end-to-end success probability. The analysis shows that beam narrowing is not universally beneficial: in NLoS-dominated regimes, enlarging the array aperture may degrade alignment reliability. The optimal beamwidth is instead governed by cross-phase competition between discovery and alignment, yielding a nontrivial feasible region with an analytically characterized boundary. Simulations validate the analysis and demonstrate improved reliability-gated locked-link performance under strong clutter, severe NLoS multipath, and finite coherence time.

cs.IT↗

arXiv2026

On the Optimality of Hierarchical Secure Aggregation with Arbitrary Heterogeneous Data Assignment

This paper studies the information theoretic secure aggregation problem in a three-layer hierarchical network with arbitrary heterogeneous data assignment, where clustered users communicate with an aggregation server through an intermediate layer of relays. We consider a more general setting with arbitrary heterogeneous data assignment across users, where `arbitrary' means that the data assignment is given in advance and `heterogeneous' means that the users may hold different numbers of datasets. Each user locally computes the partially aggregated gradients as its input based on the assigned datasets and transmits masked input to its associated relay. The relays then forward the aggregated messages to the server, which aims to recover the sum of the gradients. In this process, while some users may drop out unpredictably, the server needs to correctly recover the desired aggregation from the surviving users. Moreover, the server or any relay may collude with a subset of users. We impose the following security constraints: (i) server security, requiring the server to learn only the sum of gradients without gaining any additional information about individual inputs; and (ii) relay security, ensuring that each relay learns nothing about users' inputs. Under these constraints, we propose an aggregation scheme that guarantees information theoretic security and achieves the optimal two-layer communication loads.

cs.IT↗

arXiv2026

On Secure Gradient Coding with Uncoded Groupwise Keys

This paper considers a new secure gradient coding problem with uncoded groupwise keys, formalized as a (K, N, N_r, M, S) secure gradient coding model, where a user aims to compute the sum of the gradients from K datasets with the assistance of N distributed servers. We consider arbitrary heterogeneous data assignment, where each dataset is assigned to at least M servers. The user should recover the sum of gradients from the transmissions of any N_r servers. The security constraint guarantees that even if the user receives the transmitted messages from all servers, it cannot obtain any other information about the datasets except the sum of gradients. Compared to existing secure gradient coding works, we introduce a practical constraint on secret keys, namely uncoded groupwise keys, where the keys are mutually independent and each key is shared by precisely S servers. An achievable secure gradient coding scheme with uncoded groupwise keys is proposed, which is then proven to be optimal if S > M and to be order optimal within a factor of 2 otherwise.

cs.IT↗

arXiv2026

Information-Theoretic Secure Aggregation over Regular Graphs

Large-scale decentralized learning frameworks such as federated learning (FL), require both communication efficiency and strong data security, motivating the study of secure aggregation (SA). While information-theoretic SA is well understood in centralized and fully connected networks, its extension to decentralized networks with limited local connectivity remains largely unexplored. This paper introduces \emph{topological secure aggregation} (TSA), which studies one-shot, information-theoretically secure aggregation of neighboring users' inputs over arbitrary network topologies. We develop a unified linear design framework that characterizes TSA achievability through the spectral properties of the communication graph, specifically the kernel of a diagonally modulated adjacency matrix. For several representative classes of $d$-regular graphs including ring, prism and complete topologies, we establish the optimal communication and secret key rate region. In particular, to securely compute one symbol of the neighborhood sum, each user must (i) store at least one key symbol, (ii) broadcast at least one message symbol, and (iii) collectively, all users must hold at least $d$ i.i.d. key symbols. Notably, this total key requirement depends only on the \emph{neighborhood size} $d$, independent of the network size, revealing a fundamental limit of SA in decentralized networks with limited local connectivity.

cs.IT↗

arXiv2026

Information-Theoretic Secure Aggregation in Decentralized Networks

Motivated by the increasing demand for data security in decentralized federated learning (FL) and stochastic optimization, we formulate and investigate the problem of information-theoretic \emph{decentralized secure aggregation} (DSA). Specifically, we consider a network of $K$ interconnected users, each holding a private input, representing, for example, local model updates in FL, who aim to simultaneously compute the sum of all inputs while satisfying the security requirement that no user, even when colluding with up to $T$ others, learns anything beyond the intended sum. We characterize the optimal rate region, which specifies the minimum achievable communication and secret key rates for DSA. In particular, we show that to securely compute one bit of the desired input sum, each user must (i) transmit at least one bit to all other users, (ii) hold at least one bit of secret key, and (iii) all users must collectively hold no fewer than $K - 1$ independent key bits. Our result establishes the fundamental performance limits of DSA and offers insights into the design of provably secure and communication-efficient protocols for distributed learning systems.

cs.IT↗

arXiv2026

A Low-Complexity Framework for Multi-access Coded Caching Systems with Arbitrary User-cache Access Topology

This paper studies the multi-access coded caching (MACC) problem with arbitrary user-cache access topology, which extends existing MACC models that rely on highly structured and combinatorially designed topologies. We consider a MACC system consisting of a single server, $\Lambda$ cache-nodes, and $K$ user-nodes. The server stores $N$ equal-size files, each cache-node has a storage capacity of $M$ files, and each user-node $k\in[K]$ can access an arbitrary subset of cache-nodes $\mathcal{A}_k\subseteq[\Lambda]$ and retrieve the cached content stored in cache-nodes $\mathcal{A}_k$. The objective is to design a universal framework for the MACC delivery problem. Decoding conflicts among the requested packets are captured by a conflict graph, and the design of the delivery is reduced to a graph coloring problem, where achieving a lower transmission load corresponds to coloring the graph using fewer colors. Under this formulation, the classical DSatur algorithm achieves a transmission load close to the index-coding (IC) converse bound, thereby providing a practical benchmark. However, its computational complexity becomes prohibitive for large-scale graphs. To overcome this limitation, we develop a learning-driven approach using graph neural networks (GNNs) that efficiently constructs coded multicast transmissions with performance close to the theoretical bounds and generalizes across different user-cache access topologies and numbers of users. In addition, we extend the IC converse bound to MACC systems with arbitrary access topology and propose a low-complexity greedy approximation that closely matches the IC converse bound. Numerical results demonstrate that the proposed approach achieves performance close to the DSatur algorithm and the IC converse bound, while significantly reducing computational complexity, making it well-suited for large-scale MACC systems.

cs.IT↗

arXiv2026

Distributed Linearly Separable Computation with Arbitrary Heterogeneous Data Assignment

Distributed linearly separable computation is a fundamental problem in large-scale distributed systems, requiring the computation of linearly separable functions over different datasets across distributed workers. This paper studies a heterogeneous distributed linearly separable computation problem, including one master and N distributed workers. The linearly separable task function involves Kc linear combinations of K messages, where each message is a function of one dataset. Distinguished from the existing homogeneous settings that assume each worker holds the same number of datasets, where the data assignment is carefully designed and controlled by the data center (e.g., the cyclic assignment), we consider a more general setting with arbitrary heterogeneous data assignment across workers, where `arbitrary' means that the data assignment is given in advance and `heterogeneous' means that the workers may hold different numbers of datasets. Our objective is to characterize the fundamental tradeoff between the computable dimension of the task function and the communication cost under arbitrary heterogeneous data assignment. Under the constraint of integer communication costs, for arbitrary heterogeneous data assignment, we propose a universal computing scheme and a universal converse bound by characterizing the structure of data assignment, where they coincide under some parameter regimes. We then extend the proposed computing scheme and converse bound to the case of fractional communication costs.

cs.DC↗

arXiv2026

A New Construction Structure on MISO Coded Caching with Linear Subpacketization: Half-Sum Disjoint Packing

In the $(L,K,M,N)$ cache-aided multiple-input single-output (MISO) broadcast channel (BC) system, the server is equipped with $L$ antennas and communicates with $K$ single-antenna users through a wireless broadcast channel where the server has a library containing $N$ files, and each user is equipped with a cache of size $M$ files. Under the constraints of uncoded placement and one-shot linear delivery strategies, many schemes achieve the maximum sum Degree-of-Freedom (sum-DoF). However, for general parameters $L$, $M$, and $N$, their subpacketizations increase exponentially with the number of users. We aim to design a MISO coded caching scheme that achieves a large sum-DoF with low subpacketization $F$. An interesting combinatorial structure, called the multiple-antenna placement delivery array (MAPDA), can be used to generate MISO coded caching schemes under these two strategies; moreover, all existing schemes with these strategies can be represented by the corresponding MAPDAs. In this paper, we study the case with $F=K$ (i.e., $F$ grows linearly with $K$) by investigating MAPDAs. Specifically, based on the framework of Latin squares, we transform the design of MAPDA with $F=K$ into the construction of a combinatorial structure called the $L$-half-sum disjoint packing (HSDP). It is worth noting that a $1$-HSDP is exactly the concept of NHSDP, which is used to generate the shared-link coded caching scheme with $F=K$. By constructing $L$-HSDPs, we obtain a class of new schemes with $F=K$. Finally, theoretical and numerical analyses show that our $L$-HSDP schemes significantly reduce subpacketization compared to existing schemes with exponential subpacketization, while only slightly sacrificing sum-DoF, and achieve both a higher sum-DoF and lower subpacketization than the existing schemes with linear subpacketization.

cs.IT↗

arXiv2026

Multiaccess Coded Caching with Heterogeneous Retrieval Costs

The multiaccess coded caching (MACC) system, as formulated by Hachem {\it et al.}, consists of a central server with a library of $N$ files, connected to $K$ cache-less users via an error-free shared link, and $K$ cache nodes, each equipped with cache memory of size $M$ files. Each user can access $L$ neighboring cache nodes under a cyclic wrap-around topology. Most existing studies operate under the strong assumption that users can retrieve content from their connected cache nodes at no communication cost. In practice, each user retrieves content from its $L$ different connected cache nodes at varying costs. Additionally, the server also incurs certain costs to transmit the content to the users. In this paper, we focus on a cost-aware MACC system and aim to minimize the total system cost, which includes cache-access costs and broadcast costs. Firstly, we propose a novel coded caching framework based on superposition coding, where the MACC schemes of Cheng \textit{et al.} are layered. Then, a cost-aware optimization problem is derived that optimizes cache placement and minimizes system cost. By identifying a sparsity property of the optimal solution, we propose a structure-aware algorithm with reduced complexity. Simulation results demonstrate that our proposed scheme consistently outperforms the scheme of Cheng {\it et al.} in scenarios with heterogeneous retrieval costs.

cs.IT↗

arXiv2026

Placement Delivery Array for Cache-Aided MIMO Systems

We consider a $(G,L,K,M,N)$ cache-aided multiple-input multiple-output (MIMO) network, where a server equipped with $L$ antennas and a library of $N$ equal-size files communicates with $K$ users, each equipped with $G$ antennas and a cache of size $M$ files, over a wireless interference channel. Each user requests an arbitrary file from the library. The goal is to design coded caching schemes that simultaneously achieve the maximum sum degrees of freedom (sum-DoF) and low subpacketization. In this paper, we first introduce a unified combinatorial structure, termed the MIMO placement delivery array (MIMO-PDA), which characterizes uncoded placement and one-shot zero-forcing delivery. By analyzing the combinatorial properties of MIMO-PDAs, we derive a sum-DoF upper bound of $\min\{KG, Gt+G\lceil L/G \rceil\}$, where $t=KM/N$, which coincides with the optimal DoF characterization in prior work by Tehrani \emph{et al.}. Based on this upper bound, we present two novel constructions of MIMO-PDAs that achieve the maximum sum-DoF. The first construction achieves linear subpacketization under stringent parameter constraints, while the second achieves ordered exponential subpacketization under substantially milder constraints. Theoretical analysis and numerical comparisons demonstrate that the second construction exponentially reduces subpacketization compared to existing schemes while preserving the maximum sum-DoF.

cs.IT↗

arXiv2026

A New Construction Structure on Coded Caching with Linear Subpacketization: Non-Half-Sum Latin Rectangle

Coded caching is recognized as an effective method for alleviating network congestion during peak periods by leveraging local caching and coded multicasting gains. The key challenge in designing coded caching schemes lies in simultaneously achieving low subpacketization and low transmission load. Most existing schemes require exponential or polynomial subpacketization levels, while some linear subpacketization schemes often result in excessive transmission load. Recently, Cheng et al. proposed a construction framework for linear coded caching schemes called Non-Half-Sum Disjoint Packing (NHSDP), where the subpacketization equals the number of users $K$. This paper introduces a novel combinatorial structure, termed the Non-Half-Sum Latin Rectangle (NHSLR), which extends the framework of linear coded caching schemes from $F=K$ (i.e., the construction via NHSDP) to a broader scenario with $F=\mathcal{O}(K)$. By constructing NHSLR, we have obtained a new class of coded caching schemes that achieves linearly scalable subpacketization, while further reducing the transmission load compared with the NHSDP scheme. Theoretical and numerical analyses demonstrate that the proposed schemes not only achieves lower transmission load than existing linear subpacketization schemes but also approaches the performance of certain exponential subpacketization schemes.

cs.IT↗

arXiv2026

A New Construction Structure on Multi-access Coded Caching with Linear Subpacketization: Cyclic Multi-Access Non-Half-Sum Disjoint Packing

We consider the $(K,L,M,N)$ multi-access coded caching system introduced by Hachem et al., which consists of a central server with $N$ files and $K$ cache nodes, each of memory size $M$, where each user can access $L$ cache nodes in a cyclic wrap-around fashion. At present, several existing schemes achieve competitive transmission performance, but their subpacketization levels grow exponentially with the number of users. In contrast, schemes with linear or polynomial subpacketization always incur higher transmission loads. We aim to design a multi-access coded caching scheme with linear subpacketization $F$ while maintaining low transmission load. Recently, Cheng et al. proposed a construction framework for coded caching schemes with linear subpacketization (i.e., $F=K$) called non-half-sum disjoint packing (NHSDP). Inspired by this structure, we introduce a novel combinatorial structure named cyclic multi-access non-half-sum disjoint packing (CMA-NHSDP) by extending NHSDP to MACC system. By constructing CMA-NHSDP, we obtain a new class of multi-access coded caching schemes. Theoretical and numerical analyses show that our scheme achieves lower transmission loads than some existing schemes with linear subpacketization. Moreover, the proposed schemes achieves lower transmission load compared to existing schemes with exponential subpacketization in some case.

cs.IT↗

arXiv2026

Optimal Rate Region for Multi-server Secure Aggregation with User Collusion

Secure aggregation is a fundamental primitive in privacy-preserving distributed learning systems, where an aggregator aims to compute the sum of users' inputs without revealing individual data. In this paper, we study a multi-server secure aggregation problem in a two-hop network consisting of multiple aggregation servers and multiple users per server, under the presence of user collusion. Each user communicates only with its associated server, while the servers exchange messages to jointly recover the global sum. We adopt an information-theoretic security framework, allowing up to $T$ users to collude with any server. We characterize the complete optimal rate region in terms of user-to-server communication rate, server-to-server communication rate, individual key rate, and source key rate. Our main result shows that the minimum communication and individual key rates are all one symbol per input symbol, while the optimal source key rate is given by $\min\{U+V+T-2,\, UV-1\}$, where $U$ denotes the number of servers and $V$ the number of users per server. The achievability is established via a linear key construction that ensures correctness and security against colluding users, while the converse proof relies on tight entropy bounds derived from correctness and security constraints. The results reveal a fundamental tradeoff between security and key efficiency and demonstrate that the multi-server architecture can significantly reduce the required key randomness compared to single-server secure aggregation. Our findings provide a complete information-theoretic characterization of secure aggregation in multi-server systems with user collusion.

cs.IT↗