Search arXiv⌕ Search

arXiv subjects

Jian Chen

Publications and source records attributed to Jian Chen.

At least 37 records · Page 2Linked to original sources
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

When Evidence Changes the Subject: Subject-Typed Claim Licensing for Learned Routing

Modern learned systems increasingly combine learned components with search, repair, or external solvers. Benchmarks often measure the resulting end-to-end system, while scientific claims may concern only one component, creating an attribution problem: evidence can fail to support the requested component-level claim while still supporting a positive conclusion about the larger system. Existing evidence-to-claim methods primarily calibrate claim strength. We argue that composite systems require a second dimension: scientific subject. We address this problem with subject-typed claim licensing, which separates weaker conclusions about the requested subject from positive but non-substitutive credit about another subject. We instantiate this idea in SCOPE-Routing for preference-conditioned multigraph routing. Non-authors reproducibly apply the declared semantics; held-out review yields fewer reference-relative upward deviations than unstructured review, while the difference from a strong evidence checklist remains unresolved; and a controlled routing study shows that score-optimal and claim-eligible methods can differ while valid hybrid-system credit is preserved. These results motivate treating claim strength and scientific subject as distinct dimensions of evidence-based evaluation.

cs.AI↗
arXiv2026

Photonics-GCCE: group collaborative-competitive evolution multi-agent framework for universal and autonomous optical design

Large language model (LLM)-empowered photonic agents connect natural-language intents to executable solvers, showing significant advantages over conventional optical design approaches. However, current multi-agent frameworks operate within a collaborative paradigm without extrinsic selective pressure, which could inherit shared blind spots, converge prematurely, and fail to accumulate transferable experience for intricate tasks. Here, we introduce a group photonics collaboration-compete evolution (GCCE) framework and its LLM instantiation, termed Photonics-GCCE. Two independent agent groups pursue the same design target and undergo structured competitive evaluation across refractive-index fidelity, fabrication sensitivity, algorithmic adequacy, and physical consistency. Each group comprises a leader and three specialist agents dedicated to materials, optimization, and code validation. Agents refine their skills through competitive evaluation across design rounds. Benchmarking across six device categories against single-agent and multi-agent baselines shows that Photonics-GCCE elevates composite scores into the high 90s, improves fabrication robustness by 15 to 17 points, and reduces solver iterations to roughly 40 rounds. A representative quasi?BIC demonstration achieves a practically fabricable design with a quality factor of 13120. Our results demonstrate Photonics-GCCE as a general-purpose and closed-loop framework for autonomous optical design, capable of producing high-performance, fabrication-ready devices across diverse nanophotonic tasks.

physics.optics↗
arXiv2026

Rethinking Backdoor Repair Evaluation: Distinguishing Aggregate Clean Utility from Benign Performance Preservation

Backdoor repair aims to suppress malicious behavior in compromised models while preserving benign task performance. Existing studies typically evaluate these objectives using Attack Success Rate (ASR) and Overall Clean Accuracy, but aggregate clean accuracy can obscure substantial degradation concentrated in a small portion of the label space. We revisit benign-performance evaluation from a preservation perspective by distinguishing aggregate clean utility from the preservation of previously available class-wise performance. We define class-wise preservation loss by comparing clean performance before and after repair and show that aggregation can hide localized degradation through localized-loss dilution and cross-class compensation. To complement Overall Clean Accuracy, we characterize localized preservation loss using Worst-Class Preservation Loss and Tail Preservation Loss. We conduct a systematic empirical study across representative backdoor attacks, repair methods, datasets, attack targets, and model architectures, with additional validation under clean-label attacks. Results show that effective attack suppression and favorable aggregate clean performance do not necessarily imply uniform preservation of previously available benign performance across classes. Substantial localized preservation losses can remain, and their severity and class-wise structure vary across repair conditions. These findings motivate preservation-oriented class-wise evaluation alongside ASR and Overall Clean Accuracy.

cs.CR↗
arXiv2026

Serial-batch scheduling to minimise the total weighted late work

We study the problem of scheduling jobs on a serial-batch machine with the aim of minimising the total weighted late work. In a serial-batch setting, jobs within a batch are processed sequentially, and none are removed from the machine until the last job in the batch completes its processing. The processing time of a batch is the sum of the processing times of the jobs within it, and the completion time for each job in the batch is equal to the makespan of the jobs in the batch. When a new batch begins, a constant setup time is required for the machine. We show that minimising the total weighted late work in this environment is $NP$-hard even if all jobs have a common due date and unit weight. For the general problem, we present a pseudo-polynomial time dynamic programming algorithm. Additionally, we explore two special cases, i.e., one with a common due date and another with an agreeable condition among due dates, processing times and weights. For both special cases, we develop specialised pseudo-polynomial time dynamic programming algorithms. The proposed approaches are equipped with specialised acceleration techniques to enhance their computational performance. The extended experiments demonstrate that the dynamic programming algorithms outperform Gurobi in time efficiency.

cs.DS↗
arXiv2026

Magnetic Skyrmion Interacting with Optical Skyrmion

Magnetic skyrmions (MSks) and optical skyrmions (OSks) embody topology in matter and in light, respectively. Here we investigate the interaction between a single MSk and an OSk beam. Three distinct nonlinear dynamical modes are identified: rotation, skipping, and trochoidal motion. By decomposing the optical driving force into gradient, orbital-angular-momentum, and spin-angular-momentum contributions, we clarify their respective roles of radial confinement, azimuthal drift, and precessional modulation. The skipping motion arises from the azimuthal asymmetry of the OSk beam and exhibits spatial selectivity originating from the magnetization-polarization coupling between the MSk and OSk. In three dimensions, the coupling acquires a propagation-dependent phase dominated by the differential Gouy phase, which yields $z$-asymmetric skipping trajectories. These results bridge topological particles and topological fields within a unified framework, offering helicity-selective and phase-programmable routes to optomagnonic control.

cond-mat.mes-hall↗
arXiv2026

Keep Evaluation Fair: Detecting Data Leakage in Code Generation Benchmarks via Membership Inference Attacks

Code generation benchmarks are widely used to evaluate Large Language Models (LLMs), but benchmark data leakage into training sets can inflate performance and undermine evaluation validity. DetectLeak, a method specifically designed for code generation benchmark leakage detection, relies on perplexity scores to identify likely leaked samples. However, perplexity mainly reflects general familiarity with code patterns and may perform poorly on complex or rare samples. It also overlooks other useful signals, such as code similarity, functional correctness, and semantic representations. To address these limitations, we propose CGMIA (Code-Generation-specific Membership Inference Attack), a method for detecting leakage in code generation benchmarks. CGMIA fine-tunes a shadow model on a subset of benchmark samples to construct labeled member and non-member data. For each sample, it collects the input prompt, generated code, and reference solution, and extracts expert features, including CodeBLEU, edit distance, test pass rate, and perplexity, together with semantic features from CodeBERT embeddings. An integrated learning module combines these features to capture both surface-level memorization signals and deeper behavioral patterns, enabling a classifier to predict whether a sample was included in the target model's training set. Experiments on eight code generation benchmarks show that CGMIA outperforms eight existing membership inference methods in most cases. It also effectively detects known leaked APPS samples in StarCoder-7B's training data.

cs.SE↗
arXiv2026

Presenting mirabolic quantum Schur algebras $\mathcal{MS}_v(2,d)$

Let $v$ be an indeterminate. We give a presentation of the mirabolic quantum Schur algebra $\mathcal{MS}_v(2,d)$ in terms of generators and defining relations. More precisely, we determine the kernel of the natural epimorphism from Rosso's mirabolic quantum $\mathfrak{sl}_2$, $\mathbf U^{\mathrm{mir}}_v(2)=\langle e,f,k^{\pm 1},\ell\rangle$, onto $\mathcal{MS}_v(2,d)$. Setting $P_d(k)=\prod_{r=0}^{d}(k-v^{2r-d})$ and $Q_d(k)=\prod_{r=1}^{d}(k-v^{2r-d})$, we prove $\mathcal{MS}_v(2,d)\cong \mathbf U^{\mathrm{mir}}_v(2)/ \left\langle P_d(k),(\ell-1)Q_d(k)\right\rangle$. We further determine the split Wedderburn decomposition of this quotient and provide an equivalent presentation in terms of weight idempotents.

math.RT↗
arXiv2026

DualPathOcc: Dual-Resolution BEV Encoder for 3D Occupancy Prediction

Predicting 3D occupancy from multi-view images requires preserving geometric detail during 2D-to-3D lifting while reasoning over sparse, volumetric scene representations. We present DualPathOcc, a camera-based framework that combines a Spatial Enhancer for high-resolution feature aggregation before BEV compression, a SENet-augmented dual-path BEV encoder for local-global context modeling, and height-aware weighted cross-entropy for near-ground occupancy. The final model is optimized with occupancy supervision and no explicit depth loss. On single-frame Occ3D-nuScenes, DualPathOcc achieves 37.37 mIoU. We further analyze how surface-centered depth targets interact with volumetric occupancy learning.

cs.CV↗
arXiv2026

REVISE: Validity-Guided Recovery for Online Revisions in Agent Workflows

Agent revisions expose a fundamental correctness--efficiency trade-off during concurrent execution. Discarding ongoing work preserves latest-version correctness but wastes progress that may remain valid, whereas reusing prior work preserves efficiency but risks propagating stale state into outputs and tool effects. Existing recovery strategies resolve this trade-off in an imbalanced way with coarse-grained policies: they either favor efficiency by allowing potentially stale work to continue, or favor correctness by restarting the workflow or recomputing a linear suffix from the earliest conflict, thereby discarding unaffected progress. We present \textsc{Revise}, a validity-guided runtime for fine-grained recovery in structured agent workflows. When a revision arrives, \textsc{Revise} first intersects its delta with recorded data and control dependencies and propagates the resulting impact through the partially executed DAG to identify affected work. It then stops invalid work, preserves validity-established progress beyond the earliest conflict, and recomputes only the affected region. Incomplete provenance conservatively expands recovery, while reused results are revalidated before commit. Analysis of real coding-agent traces show online recovery opportunities: 118 sessions retain observable work before a queued later message is delivered; across 167 overlapping assistant responses, enqueue-to-completion overlap reaches 56.55~s at p95. Across 300 challenging revision/commit executions, \textsc{Revise} matches a latest-version oracle with no stale outputs or effects. On unmodified LangGraph and LLMCompiler applications using Qwen3-14B, it reduces model calls by 40.6--56.0\% relative to full restart and by 31.3--43.6\% relative to suffix recomputation. Under serving pressure, it further reduces revision-to-correct-completion tokens by 13.26\% and improves SLO goodput by 3.07--5.43\%.

cs.AI↗
arXiv2026

Beauville--Bogomolov--Fujiki positivity and Moishezonness of complex symplectic manifolds in Fujiki class $\mathscr C$

Motivated by the principle that positivity of the Beauville--Bogomolov--Fujiki (BBF) form controls the geometry of hyperkähler manifolds, we study BBF positivity from the perspective of bimeromorphic geometry, focusing on its relation to the Moishezonness of hyperfujiki manifolds (natural bimeromorphic analogues of hyperkähler manifolds). We prove that any Moishezon hyperfujiki manifold carries a big line bundle with positive BBF square. We also prove that the existence of a BBF-positive integral $(1,1)$-class implies Moishezonness for hyperfujiki $4$-folds and, more generally, for hyperfujiki $2n$-folds under a weak Kähler minimal-model condition. The proof mainly uses the theory of primitive symplectic varieties developed by B. Bakker--C. Lehn, a construction of small bimeromorphic models for certain $K$-trivial manifolds by I. Biswas--J. Cao--S. Dumitrescu--H. Guenancia, and the theory of pull-backs of reflexive differential forms established by S. Kebekus--C. Schnell. As an application, we give a Hodge-theoretic description of the Moishezon locus in certain smooth families by combining this criterion with the period theory developed by B. Anthes--A. Cattaneo--S. Rollenske--A. Tomassini.

math.AG↗
arXiv2026

ZOTTA: Test-Time Adaptation with Gradient-Free Zeroth-Order Optimization

Test-time adaptation (TTA) aims to improve model robustness under distribution shifts by adapting to unlabeled test data, but most existing methods rely on backpropagation (BP), which is computationally costly and incompatible with non-differentiable models such as quantized models, limiting practical deployment on numerous edge devices. Recent BP-free approaches alleviate overhead but remain either architecture-specific or limited in optimization capacity to handle high-dimensional models. We propose ZOTTA, a fully BP-free TTA framework that performs efficient adaptation using only forward passes via Zeroth-Order Optimization (ZOO). While ZOO is theoretically appealing, naive application leads to slow convergence under high-dimensional parameter spaces and unstable optimization due to the lack of labels. ZOTTA overcomes these challenges through 1) Distribution-Robust Layer Selection, which automatically identifies and freezes layers that already extract distribution-invariant features, updating only domain-sensitive layers to reduce the optimization dimensionality and accelerate convergence; 2) Spatial Feature Aggregation Alignment, which stabilizes ZOO by aligning globally aggregated spatial features between source and target to reduce gradient variance. Together, these components enable architecture-agnostic and stable BP-free adaptation. Extensive experiments on ImageNet-C/R/Sketch/A show that ZOTTA outperforms or matches BP-based methods, e.g., it reduces memory usage by 84% and improves accuracy by 3.9% over SAR on ImageNet-C. Code:https://github.com/Zhang-Ronghao/zotta.

cs.CV↗
arXiv2026

Stationary electron vortex states in a plasma bubble field

Plasma wakefield accelerators (PWFAs) offer accelerating gradients of 10-100~GV/m and relativistically propagating plasma bubbles capable of confining charged particles. We study the stationary states of a vortex electron at the bubble center by solving the corresponding quasi-relativistic Schrödinger equation. Analytical solutions are obtained with Laguerre-Gaussian transverse modes and Hermite-Gaussian longitudinal envelopes. Comparing the resulting beam parameters with experimentally accessible vortex-electron bundles, we find that the transverse beam waist supported by the plasma bubble is comparable to that achieved by current electron-optical techniques. The longitudinal confinement further provides a favorable parameter regime for stable injection. Our results indicate the feasibility of maintaining localized vortex-electron states in a plasma-bubble wakefield and provide an analytical starting point for investigating their subsequent acceleration and stability.

physics.acc-ph↗
arXiv2026

ESCRAG-R1: Retrieval-Augmented Reinforcement Learning for Emotional Support Conversation

Emotional Support Conversation (ESC) systems aim to provide holistic support by balancing professional therapeutic competence with natural empathy. However, existing methods struggle to simultaneously achieve structured, stage-aware reasoning and seamless empathy-expertise alignment, often resulting in an artificial splicing of clinical strategies and generic reassurance. To overcome these limitations, we propose ESCRAG-R1, a unified framework that integrates retrieval-based psychological guidance into Group Relative Policy Optimization (GRPO). By incorporating retrieval into the reinforcement learning loop, ESCRAG-R1 transforms external knowledge into a robust learning signal that stimulates explicit internal reasoning prior to generation and fundamentally reshapes the model's internal policy. To provide the reliable supervision required for this optimization, we construct ESC-Preference, a high-quality dataset based on a Client--Counselor--Judge evaluation framework that delivers precise, empathy-aware reward signals. Extensive experiments demonstrate that ESCRAG-R1 significantly outperforms existing baselines by mitigating superficial splicing and realizing a natural integration of professional guidance and empathetic expression. Code and datasets are released at https://github.com/Matcha-Liu/ESCRAG-R1.

cs.AI↗
arXiv2026

When Is Shallow Enough? Adaptive Split Federated Learning with Client-Specific Sufficiency Estimation

\textit{Split Federated Learning} (SFL) enables distributed model training by splitting networks between the server and clients. However, under client heterogeneity, the conventional static split strategy may be suboptimal because clients can differ in data distributions, adaptation dynamics, and representation learning progress, making a single split point insufficient to accommodate client-specific training states. In this paper, we propose \textsc{FedSGA}, a \textbf{S}ufficiency-\textbf{G}uided \textbf{A}daptive split \textbf{Fed}erated learning framework that addresses this question through client-specific shallow sufficiency estimation. First, we introduce a client-specific adaptation channel based on private prompt tokens, which tracks local adaptation dynamics separately from the shared backbone and provides a lightweight signal for detecting whether client adaptation remains active. To further avoid repeated online probing over multiple candidate depths, we design a shallow sufficiency estimator that combines cross-client semantic alignment, temporal interface stability, and prompt-state variation to estimate whether the shallowest split is already sufficient. Finally, we introduce a split-compatible interface harmonization module that projects activations from different split depths into a shared semantic space, improving the comparability of heterogeneous client interfaces before server-side prediction. Extensive experiments on multiple heterogeneous benchmarks demonstrate the effectiveness of \textsc{FedSGA} in improving model performance compared with state-of-the-art methods while reducing unnecessary client-side computation.

cs.DC↗
arXiv2026

Learning What to Activate: Combinatorial Capability Allocation for Long-Horizon Multimodal Agents

Long-horizon multimodal agents rely on specialized capabilities for perception, retrieval, reasoning, verification, and execution. Existing designs typically activate a fixed capability set or invoke a predefined workflow, incurring substantial computational overhead while failing to accommodate stage-dependent capability demands. In this paper, we study the \textit{combinatorial capability allocation} problem for long-horizon multimodal agent systems, where the system selects a cost-sensitive subset of specialized capabilities at each interaction stage, which is nontrivial since capability values depend on the selected subset, while previous allocations alter the states encountered by subsequent decisions. We introduce \textsc{CoCA}, an on-policy learning framework that recovers a deployable capability-subset policy from sparse conditional comparisons. On states visited by the student policy, the stronger teacher compares the marginal net values of candidate capabilities, conditioned on the currently selected subset. Then, we adopt a conditional utility model to transform such comparisons into an autoregressive capability-subset policy, avoiding explicit enumeration. We further introduce dual-level on-policy distillation to address distribution mismatch both across environment states and within the partial subsets encountered during set construction. Finally, trajectory-level reinforcement learning refines the distilled policy toward task success, activation cost, and allocation stability. At inference time, allocation is performed solely by the lightweight student policy without teacher queries or online updates. Experiments on long-horizon multimodal environments and controlled capability-demand shifts demonstrate the superiority of our method over the state-of-the-art baseline methods.

cs.AI↗
arXiv2026

FlashDrive: Flash Vision-Language-Action Inference for Autonomous Driving

Vision-Language-Action (VLA) models promise to bring end-to-end reasoning to autonomous driving, but their computational cost remains far too high for real-time control. The core challenge is structural: VLA inference is not a single bottleneck but a cascade of four. Visual encoding wastes compute on overlapping video frames; language-model prefill recomputes context that could be carried over from the previous timestep; reasoning tokens are generated serially despite low entropy; and flow-matching denoising applies uniform compute to a non-uniform velocity field. Addressing any one stage in isolation leaves the others untouched. We propose FlashDrive, an algorithm-system co-design framework that targets all four stages simultaneously. Our key insight is that each bottleneck admits a distinct, lightweight algorithmic shortcut: temporal overlap enables streaming KV-cache reuse across frames; the low per-token entropy and strong intra-block correlations of driving-domain reasoning make a non-autoregressive diffusion drafter highly effective for speculative decoding; and the velocity field's structure---sharp at the endpoints, flat in the middle---permits adaptive step caching that concentrates compute where it matters. Layered on system-level CUDA Graph compilation and kernel fusion, these techniques compound. Applied to Alpamayo 1.5-10B with W4A8 quantization, FlashDrive reduces end-to-end latency from 717ms to 151ms (4.7x) while leaving accuracy essentially unchanged: minADE6@6.4s shifts by only 0.08m, minADE1 improves, and closed-loop collision and off-road rates improve in simulation. By raising a 10B-parameter reasoning VLA from 1.4~Hz to 6.6~Hz on a single GPU, FlashDrive moves end-to-end autonomous driving substantially closer to real-time deployment.

cs.AI↗
arXiv2026

Neural Tree Collaborative Filtering: Rethinking Graph Collaborative Filtering as Tree Collaborative Filtering with Curvature-Aware Propagation Depth

Graph Collaborative Filtering (GCF) has become the dominant paradigm in modern recommender systems by modeling user-item interactions as a bipartite graph and propagating embeddings through a fixed number of message-passing layers. However, applying a uniform propagation depth to every node ignores a fundamental property of real interaction graphs: nodes differ substantially in their local connectivity, so peripheral nodes quickly suffer from over-smoothing while hub-like nodes remain under-explored beyond their immediate neighborhood. In this paper, we revisit GCF from a tree-structured perspective and propose Neural Tree Collaborative Filtering (NTCF), a framework that re-interprets each node's local neighborhood as a rooted tree and assigns a node-specific propagation depth based on a closed-form local-degree-imbalance score that serves as a discrete Ricci-curvature proxy. We provide a theoretical analysis showing that (i) NTCF strictly generalizes NGCF, degenerating to NGCF when all curvature-induced depth adjustments vanish (a lower bound on its representation power), and (ii) the curvature-aware schedule retains strictly more discriminative information at deep layers on positively-curved (peripheral) nodes than uniform-depth propagation. NTCF can achieve higher performance than most widely used GCF backbone models and can be integrated into existing advanced self-supervised models as a backbone, replacing their original backbone to achieve enhanced performance. Extensive experiments on three public datasets demonstrate the superiority of NTCF.

cs.IR↗