arXiv subjects

Yuxin Jin

Publications and source records attributed to Yuxin Jin.

12 recordsLinked to original sources

arXiv2026

Just Let Linear States Forget the Distant Past: Prefix Caching via Suffix Replay for Hybrid LLMs

Hybrid LLMs interleave full-attention layers with linear-attention layers to reduce long-context inference cost, but this structure complicates prefix caching. Full-attention KV caches are token-addressable, whereas linear-attention layers maintain recurrent states that cannot be rolled back to arbitrary prefix boundaries. Existing systems materialize recurrent-state checkpoints, restricting prefix reuse to checkpoint-aligned positions. We present SuffixReplay, the first prefix caching system that lets hybrid LLMs reuse cached prefixes at every cache-supported page boundary without materializing recurrent-state checkpoints. Our key insight is to just let linear states forget the distant past. Modern linear-attention mechanisms use recurrent decay and gating to attenuate the influence of old inputs. Therefore, instead of checkpointing every prefix boundary, SuffixReplay approximates the state at a matched boundary by replaying only a recent suffix of the layer's input hidden states, which we retain as anchors. At the algorithmic level, SuffixReplay combines layer-wise and token-wise anchor sparsity with a bounded replay budget to control storage, computation, and quality. At the system level, it uses an independently managed anchor sidecar and a pipelined replay path to overlap anchor movement and state reconstruction with the native serving pipeline. We evaluate SuffixReplay on three hybrid LLMs: OLMo-Hybrid-7B, Qwen3.5-4B, and Qwen3.6-27B-FP8. Across these models, SuffixReplay retains 91.4-100% of full-prefill quality on average across LongBench and RULER, while using only 0.36-0.51x the amortized per-token storage of SGLang's default 8192-token checkpoint cache. Integrated into SGLang, SuffixReplay reduces median TTFT by 15-70% on branching workloads, sustains 2.3-4.3x SGLang's throughput when the working set exceeds HBM, and matches SGLang on high-hit continuation traffic.

cs.AI↗

arXiv2026

RelaxKV: Recomputation Guided by the Query with Sparse Context Attention for Efficient KV Cache Reuse

Cross-request KV caching reduces the prefill cost of Retrieval-Augmented Generation (RAG), but conventional prefix caching severely limits cache reuse across requests. Position-Independent Caching (PIC) removes this constraint by reusing independent chunks, but their KV states miss cross-chunk interactions. Existing methods selectively recompute token states to recover these missing interactions, but primarily allocate the recomputation budget to selecting which states to recompute, while fixing the recomputation context to the full causal prefix. We introduce RelaxKV, which formulates selective cache repair as a joint allocation problem over repair targets and recomputation context. Guided by the user query, RelaxKV identifies layer-specific repair targets and restricts their recomputation to a query-relevant context, reducing attention computation. Across four decoder models, RelaxKV at a 15% anchor ratio improves aggregate LongBench performance over ProphetKV on all models. On Qwen3-14B, RelaxKV provides a stronger quality-TTFT trade-off than ProphetKV across a 5%-30% anchor-ratio sweep, and achieves the best selective results on RULER-MV and LV-Eval at 16K and 32K context lengths. Controlled ablations further demonstrate the importance of recomputation context selection.

cs.AI↗

arXiv2026

Asynchronous Replanning in Two Population Linear Quadratic Mean Field Games: Information Requirements and Stability

We study asynchronous replanning in a two population linear quadratic mean field game in which the populations may begin from different beliefs and hence use different plans. Each population observes its own aggregate trajectory and a public record of implemented revisions, while its continuation best response depends on the opponent's current plan. We identify the information required for replanning as the aggregate state at the end of the initial observation interval together with the opponent's active continuation plan. For linear observations, recoverability of this state-plan pair is characterized by a kernel inclusion, and a bounded factorization quantifies sensitivity to observation error. In particular, the required pair may be recoverable even when the full hidden belief is not. Once initialized, the public event record and the common best-response map recursively determine subsequent opponent plans, and the resulting local algorithm reproduces an ideal benchmark on every finite opportunity prefix; implemented revisions alternate as a consequence of best-response persistence. For finite populations, we derive an eventwise linear recursion for sampling errors, obtain finite prefix error bounds, and prove record matching for an autonomous deadband rule under a positive decision margin. Finally, we separate unique solvability of mutual continuation responses from stability of alternating responses, and show that at a pre-terminal Zeno accumulation, spectral stability together with a moving-boundary estimate yields convergence of the continuation plans to the equilibrium restarted from the actual limiting state.

math.OC↗

arXiv2026

Linear Quadratic Mean Field Games under Heterogeneous Erroneous Initial Information

We study a finite-horizon linear--quadratic mean field game with heterogeneous observations of the initial mean field. These observations may be erroneous and are used by agents to construct their feedback laws. At the population level, the heterogeneous error profile is generally infinite-dimensional. We derive exact linear sensitivity formulas showing that its closed-loop effects nevertheless admit a finite-dimensional closure: for each agent, the deviations are governed by only two n-dimensional error channels, its private error E_i and the population-average error \bar E. The population-average error determines the displacement of the actual mean field, whereas centered private errors determine agent-specific deviations and the additional cross-sectional covariance. The same representation yields quadratic cost identities and a uniform $O(N^{-1})$ mean-square approximation of the empirical aggregate. In the deterministic model, the private state history induces a linear observation problem for the initial errors. Nonsingularity of the associated observability Gramian is necessary and sufficient for exact recovery. The recovered errors reconstruct the actual mean-field state at the revision time and initialize a revised correct-information continuation from that state. We obtain an exact continuation-cost comparison; when the population-average error vanishes, revision does not increase the continuation cost. For stochastic dynamics, we take a two-component revision signal as given and study the resulting one-shot feedback relative to an oracle continuation initialized with the actual current mean field. The actual mean-field deviation is driven only by the population-average signal error, while individual deviations also retain the private signal error. Exact covariance and quadratic continuation-cost identities quantify these effects.

math.OC↗

arXiv2026

LinearKV: One Cached State Suffices for Position-Independent Caching in Hybrid LLMs

LLM serving is increasingly accelerated by position-independent caching (PIC). Existing PIC methods, however, are built for full-attention models, where a token-indexed KV cache underlies its core operations: matching reusable token chunks, concatenating their KV entries, and selectively recomputing a few tokens to restore cross-chunk context. Hybrid LLMs break these primitives---they replace most attention layers with linear recurrences that expose only a fixed-size state, leaving no token-indexed KV to concatenate or to locally repair. This raises a natural question: can PIC benefit hybrid models, and what would it take? We present LinearKV, a training-free hybrid-PIC framework. Its key insight is a \emph{decoupled initialization}: each linear layer maps its $K$ matched local states to a single initial state, while full-attention layers concatenate their KV as before. LinearKV is therefore compatible with existing PIC methods, reusing their token selection and recomputation as-is. Under this framework, we find that a \emph{single cached state} suffices as the linear layer's initializer. The algebraically principled alternative---composing all $K$ cached states into the exact full-prefix state, as concurrent work HYPIC does---is unnecessary and, on some architectures, even harmful. We compare the two across three hybrid models and three PIC selectors. On the two GDN models the two tie, both recovering most of full quality (up to $92\%$); on the Mamba-2 model, exact composition instead collapses under every selector---under EPIC, for instance, it recovers only $46.6\%$ of full quality, versus $86.8\%$ for a single cached block initializer. A single state initializer is also cheaper, cutting time-to-first-token to $0.46\times$ full prefill versus a further $5$--$17\%$ overhead for exact composition; results hold across LongBench QA and RULER at 8K--32K.

cs.AI↗

arXiv2026

Major-Minor LQ Mean Field Games with Erroneous Initial Information: Distributed Error Estimation and Strategy Modification

This paper studies major-minor linear-quadratic mean field games (MMLQMFGs) with erroneous initial information under a constrained observation structure. Each minor agent observes only its own state and the major agent's state, while the major agent observes its own state and the states of a subset of minor agents; neither side observes the mean field state directly. We show that the initial-information errors propagate linearly through the game dynamics and lead to explicit deviations in the major state, the actual mean field, and the agents' internally updated mean field states. Based on this structure, we formulate distributed error identification as a parameter-estimation problem from discrete-time local observations and construct maximum-likelihood estimators for unknown initial errors. We then propose an estimate-based strategy modification at an intermediate time by reconstructing the current mean field from the estimated errors and switching to the corresponding control law. We also characterize the resulting estimation errors and show that, in the present symmetric setting, the major agent's estimation precision depends on the number of observed minor agents but not on their identities. Numerical results illustrate the proposed method.

math.OC↗

arXiv2026

How Far Has AI Come in Liver Fibrosis Staging? A Large-Scale Real-World Dataset and Benchmark

Despite years of methodological progress, how far AI has come in liver fibrosis staging has never been systematically evaluated under the heterogeneous, multi-center conditions that define clinical practice. To address this gap, we introduce LiFS, a large-scale dataset and benchmark derived from the MICCAI 2025 CARE-Liver challenge, comprising 610 patients across multiple centers and scanners with multi-sequence MRI. To the best of our knowledge, LiFS is the first benchmark providing complete gadoxetic acid-enhanced sequences with histopathology-confirmed annotations from diverse real-world scanners. Through systematic evaluation of 9 independently developed methods selected from 96 registered teams against in-cohort radiologist reference results, our findings address how far current AI has progressed toward clinical-level liver fibrosis staging from three complementary perspectives. First, against radiologists, the best AI methods were broadly comparable to the senior radiologist and significantly exceeded the junior radiologist in selected settings, while median AI performance generally approached junior-radiologist levels. Second, from a data perspective, cross-center heterogeneity, label imbalance, and contrast-enhanced sequence variability emerge as the dominant challenges for AI methods. Third, from a technical perspective, methodological design choices, including spatial registration, input dimensionality, multi-modal fusion strategy, and backbone architecture, appear to modulate cross-center robustness, although no single choice alone closes the gap. Overall, LiFS provides a rigorous real-world benchmark for positioning the current state of AI in liver fibrosis staging and for enabling future research on the key challenges that limit clinically reliable deployment.

cs.CV↗

arXiv2026

EXG: Self-Evolving Agents with Experience Graphs

Large language model (LLM)-based agents have demonstrated strong capabilities in complex reasoning and problem solving through multi-step interactions, yet most deployed agents remain behaviorally static, with knowledge acquired during execution rarely translating into systematic improvement over time. In response, a growing line of work on self-evolving agents explores how agents can improve through experience during deployment, but most existing approaches either rely on ad hoc reflection limited to single-task correction or adopt unstructured memory that accumulates fragmented experience with delayed usability. To address this limitation, we introduce EXG, an experience graph framework for self-evolving agents that explicitly organizes accumulated successes and failures into a structured, relational representation. EXG is the first experience graph designed for self-evolving agents, supporting both online, real-time graph growth during execution for immediate cross-task experience reuse, and offline reuse of a consolidated experience graph as an external memory module. This design also enables EXG to serve as a plug-and-play component for existing self-evolving agents, organizing prior experience into a unified experience graph and improving both solution quality and resource efficiency as deployment progresses. Extensive experiments across code generation and reasoning benchmarks show that EXG attains more favorable performance-efficiency trade-offs than reflection- and memory-based baselines in both online and offline evaluations. Our results suggest that structuring experience as a graph provides a principled foundation for scalable and transferable self-evolving agent behavior.

cs.AI↗

arXiv2026

Conflict-free chromatic index of bipartite graphs

An edge coloring of a graph $G$ is called conflict-free if, for every edge, its closed neighborhood contains a color that appears exactly once. The least number of colors required for such a coloring is the conflict-free chromatic index of $G$, denoted by $χ'_{CF}(G)$. Kamyczura, Meszka, and Przybyło conjectured that $χ'_{CF}(G)\le 3$ for any bipartite graph $G$ without isolated vertices. In this paper, we confirm this conjecture.

math.CO↗

arXiv2026

Zero-Knowledge Federated Learning: A New Trustworthy and Privacy-Preserving Distributed Learning Paradigm

Federated Learning (FL) has emerged as a promising paradigm in distributed machine learning, enabling collaborative model training while preserving data privacy. However, despite its many advantages, FL still contends with significant challenges -- most notably regarding security and trust. Zero-Knowledge Proofs (ZKPs) offer a potential solution by establishing trust and enhancing system integrity throughout the FL process. Although several studies have explored ZKP-based FL (ZK-FL), a systematic framework and comprehensive analysis are still lacking. This article makes two key contributions. First, we propose a structured ZK-FL framework that categorizes and analyzes the technical roles of ZKPs across various FL stages and tasks. Second, we introduce a novel algorithm, Verifiable Client Selection FL (Veri-CS-FL), which employs ZKPs to refine the client selection process. In Veri-CS-FL, participating clients generate verifiable proofs for the performance metrics of their local models and submit these concise proofs to the server for efficient verification. The server then selects clients with high-quality local models for uploading, subsequently aggregating the contributions from these selected clients. By integrating ZKPs, Veri-CS-FL not only ensures the accuracy of performance metrics but also fortifies trust among participants while enhancing the overall efficiency and security of FL systems.

cs.CR↗

arXiv2025

Initial Error Tolerant Distributed Mean Field Control under Partial and Discrete Information

In this paper, an initial error tolerant distributed mean field control method under partial and discrete information is introduced, where each agent only has discrete observations on its own state. First, we study agents' behavior in linear quadratic mean field games (LQMFGs) under heterogeneous erroneous information of the initial mean field state (MF-S), and formulate the relationships between initial errors and systemic deviations. Next, by capturing the initial error affection on the private trajectory of an agent, we give a distributed error estimation method based on maximum likelihood estimation (MLE), where each agent estimates information errors only based on discrete observations on its private trajectory. Furthermore, we establish an error-based segmented state estimation method, design the initial error tolerant distributed mean field control method (IET-DMFC), and demonstrate the consistent property of state estimation as observation frequency increases. Finally, simulations are performed to verify the efficiency of the algorithm and the consistent properties.

math.OC↗

arXiv2023

Depth and DOF Cues Make A Better Defocus Blur Detector

Defocus blur detection (DBD) separates in-focus and out-of-focus regions in an image. Previous approaches mistakenly mistook homogeneous areas in focus for defocus blur regions, likely due to not considering the internal factors that cause defocus blur. Inspired by the law of depth, depth of field (DOF), and defocus, we propose an approach called D-DFFNet, which incorporates depth and DOF cues in an implicit manner. This allows the model to understand the defocus phenomenon in a more natural way. Our method proposes a depth feature distillation strategy to obtain depth knowledge from a pre-trained monocular depth estimation model and uses a DOF-edge loss to understand the relationship between DOF and depth. Our approach outperforms state-of-the-art methods on public benchmarks and a newly collected large benchmark dataset, EBD. Source codes and EBD dataset are available at: https:github.com/yuxinjin-whu/D-DFFNet.

cs.CV↗