arXiv subjects

Ying Xiong

Publications and source records attributed to Ying Xiong.

At least 19 recordsLinked to original sources

arXiv2026

VRL-Bench: Benchmarking agents on computer control tasks under finite trial budgets

Learning from trial and error is a promising way to improve language agents on complex tasks such as computer control. Reflexion introduced verbal reinforcement learning, which turns failed trials into text that guides later attempts without updating model parameters. We introduce VRL-Bench, a harness for fair evaluation of trial-and-error learning under finite trial budgets. Across three models on MiniWoB and WebShop, we evaluate updates from several prominent verbal-memory methods spanning Reflexion and later work: each improves observed success over memory-free retry in some settings but reduces it in others. Replay experiments show that using reflection can reduce success rates, revealing a trade-off between exploiting experience and continued exploration. We propose VEX$^2$, a verbal exploration--exploitation scheduler that uses a language model to jointly select policies and allocate the remaining trial budget. VEX$^2$ is the only evaluated update to achieve positive observed success-rate gains over retry in all six settings.

cs.AI↗

arXiv2026

Beyond Semantic Similarity: Reducing Unnecessary API Calls via Behavior-Aligned Retriever

Tool-augmented LLMs invoke external functions to extend their capabilities, but errors in the invocation decision, such as calling a tool when none is needed or omitting a needed call, can produce unreliable outputs and unnecessary cost. A lightweight remedy is to prepend retrieved examples so LLMs decide tool use in context. However, existing retrievers rank examples by semantic similarity alone. Lexically close or semantically close queries can require opposite behavior, so the retrieved examples may be behaviorally inconsistent and silently mislead the model. We propose Behavior Aligned Retrieval (BAR), a backbone-agnostic training recipe that teaches a dense retriever a behavior-aware similarity, keeping semantically related candidates close only when their tool-use behavior is compatible. BAR does not predict invocation labels; instead, it ranks demonstrations while leaving the final tool-use decision to the LLM. Applied to multiple retrieval backbones, including BERT, Contriever, and Qwen-based representation backbone, BAR consistently improves invocation reliability and reduces unnecessary API calls across 14 LLMs and 3 benchmarks.

cs.CL↗

arXiv2026

Splashing-regime transitions and secondary-droplet scaling in oblique drop impacts on a deep pool

Oblique drop impact onto a deep liquid pool produces asymmetric crowns, directional jetting, and splashing transitions that cannot be characterized by the total impact inertia alone. We numerically investigate water drops impacting a quiescent deep pool over $41\leq We\leq1790$ and $10^\circ\leqθ\leq90^\circ$. The simulations reproduce the principal features observed experimentally and identify five post-impact regimes in the $We$--$θ$ plane: deposition, front splashing, side splashing, side-front splashing, and crown splashing. The deposition--front-splashing transition is described by the tangential-inertial parameter $K_s=We\cosθ$, with $K_s^c\approx120$. This criterion follows from the competition between downstream crown-rim inertia and capillary retraction at the Taylor--Culick velocity. The transition from front to side-front splashing is instead governed primarily by normal impact inertia, with a critical normal Weber number $We_N^c\approx318$. Beyond these regime transitions, the secondary-droplet statistics reveal fragmentation behavior common to the different splashing regimes. The droplet-size distributions are positively skewed, and the median diameter follows $d_{s,\mathrm{med}}/D\sim We^{-3/5}$. Second-order velocity structure functions support a scale-dependent capillary--inertial description of rim and ligament breakup. Combined with mass conservation, this scaling gives $N_s\sim We^{9/5}$, providing a numerical explanation for the secondary-droplet-number scaling observed experimentally. Thus, directional impact inertia governs the macroscopic selection of splashing regimes, whereas the secondary-droplet populations across these regimes exhibit a common capillary--inertial fragmentation scaling.

physics.flu-dyn↗

arXiv2026

Ultralow p-type contact resistance for ultra-nanoscaled 2D-materials transistors

High contact resistance is one of the main bottlenecks for practical two-dimensional (2D) materials transistors, especially for p-type transistors and future 2D ultra-nanoscaled (sub-10 nm) FETs (PMOS + CMOS). We develop self-consistent contact resistance models for metal-2D semiconductor-metal devices to capture the essential interface physics for both vertical and edge configurations. Our calculations have been verified with various recent experiments of p-type and n-type contacts. For a given set of materials, the model determines the scaling of contact resistance over a wide range of device parameters including channel length (100s nm down to sub-10 nm), doping and mobility of the 2D materials, contact length of the electrodes, and applied voltages. These results identify the key factors in order to reduce the contact resistance for p-type 2D semiconductor WSe$_2$ towards the sub-10 nm channel length scale that are readily to be realized by future experiments. It is found that the effect of source-limited current saturation is the key challenge for down scaling 2D FET to sub-10 nm channel length. Two topological semi-metals as potential electrodes are proposed for 2D p-type semiconducting WSe$_2$ with our predicted contact resistance $R_c<$ 100 $Ω\; {\rm μm}$ approaching the quantum limit. Our model is also verified with the computational expensive full quantum atomistic model that is currently limited to a few nm scale.

cond-mat.mes-hall↗

arXiv2026

MASPRM: Multi-Agent System Process Reward Model

Inference-time search over multi-agent systems (MAS) wastes compute when it cannot identify which agent's intermediate message advanced progress. We present the Multi-Agent System Process Reward Model (MASPRM), which scores routed transcripts (ordered sequences of messages between agents) and acts as an inference controller for step-level beam search (SBS) and Monte Carlo Tree Search (MCTS). MASPRM is trained from multi-agent MCTS rollouts labeled only with terminal outcome rewards, without human step-level annotations. We evaluate on GSM8K, MATH, MMLU, and LogiQA. Under matched scorer size and comparable MCTS budget, MASPRM exceeds a size-matched ORM by $+2.0$ to $+3.0$ points at 1.5B and $+4.1$ to $+14.5$ at 7B across all four benchmarks, with additional scorer-scaling gains over policy likelihood at 7B (avg $+13.4$ under MCTS). MASPRM also improves ranking quality, reducing Hit@1 to Hit@5 gaps by up to $10.3$ points, with the largest gains under stepwise search that uses intermediate decisions. Code: https://github.com/milad1378yz/MASPRM

cs.MA↗

arXiv2026

Retrieval-Augmented Generation for Natural Language Processing: A Survey

Large language models (LLMs) have achieved strong empirical performance in various fields, benefiting from their huge amount of parameters that store knowledge. However, LLMs still suffer from several key issues, such as hallucination problems, knowledge update issues, and lacking domain-specific expertise. The appearance of retrieval-augmented generation (RAG), which leverages an external knowledge base to augment LLMs, mitigates these limitations. This paper presents a systematic review of RAG techniques for natural language processing (NLP), with a focus on retrievers and retrieval fusions. We introduce a novel taxonomy of retrieval fusions, such as query-based, logits-based, latent, and parametric fusion, and provide structured comparisons across accessibility, efficiency, and use cases. The paper further examines RAG applications across diverse NLP tasks, discusses evaluation methodologies and benchmark limitations, and analyzes training paradigms with and without knowledge base updates. Finally, we explore industrial deployment considerations and identify emerging challenges and future directions, including security, efficiency, and graph-based retrieval.

cs.CL↗

arXiv2026

HFX: Joint Design of Algorithms and Systems for Multi-SLO Serving and Fast Scaling

Large language model (LLM) serving faces the dual challenge of meeting strict user-specific service-level objectives (SLOs) while minimizing computational cost under dynamic, multi-task workloads. Existing approaches either rely on static scheduling policies or focus on single-task settings, limiting their applicability in real-world deployments with heterogeneous requests, variable prompt lengths, and elastic scaling requirements. We present HFX, a production LLM serving system that jointly optimizes request scheduling and elastic scaling across model replicas to satisfy diverse SLOs. HFX introduces a \textbf{scheduler} that performs proactive budget estimation and prioritization to ensure SLO compliance for both new and in-flight requests. HFX also integrates a \textbf{scaler} that supports fast device-to-device (D2D) weight transfer, reducing cold-start latency. Additionally, the system supports both colocated and disaggregated prefill/decode deployments, enabling adaptation to diverse workload patterns and cloud environments. Through extensive experiments on multi-task workloads, we demonstrate consistently higher SLO attainment, lower end-to-end latency, and lower NPU usage cost by up to 4.44$\times$, 65.82\%, and 49.81\%, respectively, compared to state-of-the-art systems. Our results highlight the effectiveness of SLO-aware scheduling and scaling in practical LLM serving, providing a robust framework for cost-efficient and SLO-compliant deployments.

cs.DC↗

arXiv2026

Geometric Direction Finding on Dynamic Manifolds: Unambiguous DOA Estimation for Spatially Undersampled UWB Arrays

Traditional Direction of Arrival (DOA) estimation methods struggle to simultaneously address three physical constraints in Ultra-Wideband (UWB) electromagnetic sensing: spatial undersampling, asynchronous array phase, and beam squint. Existing solutions treat these issues in isolation, leading to limited performance in complex scenarios. This paper proposes a novel dynamic manifold perspective, which models UWB signal observations as a continuous manifold curve in a high-dimensional space driven by temporal evolution and array topology. We theoretically demonstrate that the DOA can be uniquely determined solely by the geometric shape of the manifold, rather than the absolute arrival phase. Based on this perspective, we construct a geometric parameter system comprising extrinsic and intrinsic parameters, along with a corresponding DOA estimation framework. Extrinsic vector parameters serve as a dynamic extension of traditional array processing, effectively expanding the degrees of freedom to suppress grating lobes. Intrinsic scalar invariants provide a new geometric perspective independent of traditional phase models, offering intrinsic robustness against array channel phase errors. Simulation results show that the derived analytical expressions for geometric parameters are highly consistent with numerical truths. The proposed framework not only completely eliminates spatial ambiguity in sparse arrays but also achieves high-precision direction finding under conditions with calibration-free phase errors.

eess.SP↗

arXiv2026

RAEE: A Robust Retrieval-Augmented Early Exit Framework for Efficient Inference

Deploying large language model inference remains challenging due to their high computational overhead. Early exit optimizes model inference by adaptively reducing the number of inference layers. Current methods typically train internal classifiers or use heuristic methods to determine the exit layer. However, those methods either introduce significant training overheads or lead to performance degradation. To address these limitations, this paper proposes RAEE, a robust Retrieval-Augmented Early Exit framework that not only enables early exit but also enhances model performance through corrective exit information at intermediate layers. This paper first demonstrates that the early exit problem can be effectively modeled as a distribution prediction problem, in which the distribution can be further approximated through the exit information of similar data. Subsequently, this paper introduces the process of collecting exit information of correct predictions and the steps to construct the retrieval database. Finally, leveraging the pre-constructed retrieval database, RAEE utilizes the exit information from retrieved similar data to guide the backbone model's exit. Experimental results demonstrate that RAEE can not only accelerate inference while achieving robust zero-shot performance across eight downstream tasks.

cs.CL↗

arXiv2026

Huawei Cloud Model-as-a-Service on the CloudMatrix384 SuperPod

Scaled-out MoE LLMs and scaled-up SuperPods create new systems challenges for production Model-as-a-Service (MaaS), requiring disaggregation, low-latency communication, and decentralized serving. This report presents xDeepServe, the production serving system behind Huawei Cloud's MaaS offering on CloudMatrix384, a 48-server SuperPod with 384 Ascend 910C chips connected by a high-bandwidth UB fabric and global shared memory. It serves models including DeepSeek, Kimi, GLM, Qwen, and MiniMax, among others. xDeepServe is built around Transformerless, a disaggregated execution architecture that decomposes transformer inference into modular units -- attention, feedforward, and MoE -- and supports disaggregated Prefill-Decode and MoE-Attention deployments. To enable disaggregation, we develop XCCL, a memory-semantic communication layer providing microsecond-level point-to-point and scalable all-to-all primitives, and we extend FlowServe with decentralized DP groups and techniques to mitigate stragglers and synchronization variance. In a peak decoding configuration, xDeepServe reaches 2400 tokens/s per Ascend 910C chip at ~50ms time-per-output-token (TPOT).

cs.DC↗

arXiv2026

ReviveMoE: Fast Recovery for Hardware Failures in Large-Scale MoE LLM Inference Deployments

As LLM deployments scale over more hardware, the probability of a single failure in a system increases significantly, and cloud operators must consider robust countermeasures to handle these inevitable failures. A common recovery approach is to simply restart the LLM serving instance; however, this is costly in model-as-a-service (MaaS) inference settings, where reloading model weights and recompiling computation graphs can introduce significant delays to incoming requests. We propose ReviveMoE, a method for rapid failure recovery in large-scale LLM deployments without restarting the serving instance. ReviveMoE is designed to support both the traditional LLM architecture, which collocates MoE and attention on the same hardware, and the disaggregated architectures, which separate MoE from attention. Integrated into Huawei Cloud's MaaS, ReviveMoE is built on top of Huawei's xDeepServe serving platform and the XCCL communications library.

cs.DC↗

arXiv2026

ReFilter: Improving Robustness of Retrieval-Augmented Generation via Gated Filter

Retrieval-augmented generation (RAG) has become a dominant paradigm for grounding large language models (LLMs) with external evidence in knowledge-intensive question answering. A core design choice is how to fuse retrieved samples into the LLMs, where existing internal fusion approaches broadly fall into query-based fusion, parametric fusion, and latent-based fusion. Despite their effectiveness at modest retrieval scales, these methods often fail to scale gracefully as the number of retrieved candidates k increases: Larger k improves evidence coverage, yet realistic top-k retrieval inevitably contains irrelevant or redundant content and increases the inference cost. To address these limitations, we propose ReFilter, a novel latent-based fusion framework that performs token-level filtering and fusion. ReFilter consists of three key components: a context encoder for encoding context features, a gated filter for weighting each token, and a token fusion module for integrating the weighted token feature into the LLM's hidden states. Our experiments across four general-domain QA benchmarks show that ReFilter consistently achieves the best average performance under both in-domain adaptation and out-of-domain transfer. ReFilter further generalizes to five biomedical QA benchmarks in zero-shot transfer without domain fine-tuning, reaching 70.01% average accuracy with Qwen2.5-14B-Instruct.

cs.CL↗

arXiv2026

DECKBench: Benchmarking Multi-Agent Frameworks for Academic Slide Generation and Editing

Automatically generating and iteratively editing academic slide decks requires more than document summarization. It demands faithful content selection, coherent slide organization, layout-aware rendering, and robust multi-turn instruction following. However, existing benchmarks and evaluation protocols do not adequately measure these challenges. To address this gap, we introduce the Deck Edits and Compliance Kit Benchmark (DECKBench), an evaluation framework for multi-agent slide generation and editing. DECKBench is built on a curated dataset of paper to slide pairs augmented with realistic, simulated editing instructions. Our evaluation protocol systematically assesses slide-level and deck-level fidelity, coherence, layout quality, and multi-turn instruction following. We further implement a modular multi-agent baseline system that decomposes the slide generation and editing task into paper parsing and summarization, slide planning, HTML creation, and iterative editing. Experimental results demonstrate that the proposed benchmark highlights strengths, exposes failure modes, and provides actionable insights for improving multi-agent slide generation and editing systems. Overall, this work establishes a standardized foundation for reproducible and comparable evaluation of academic presentation generation and editing. Code and data are publicly available at https://github.com/morgan-heisler/DeckBench .

cs.AI↗

arXiv2026

Kramers Dichroism in PT Symmetric Magnets

Superpositions between states in doubly degenerate Kramers pairs can act as an internal degree of freedom. Here we uncover a Kramers dichroism in PT symmetric magnets: interband transitions induced by circularly polarized light irradiation produce a coherent superposition between Kramers partnered states. This allows to optically control the Kramers degree of freedom. In contrast, Kramers pairs optically excited by linearly polarized light remain in a completely mixed state. Strikingly, we find a class of second-order nonlinear responses that directly track the coherence between Kramers partnered states. Such Kramers nonlinearities can be pronounced producing large second-order nonlinear layer polarization responses activated by Kramers degeneracy in layered antiferromagnets. Together with Kramers dichroism, these render optical responses a novel means for accessing the Kramers degree of freedom and diagnosing their quantum coherent state.

cond-mat.mes-hall↗

arXiv2025

MEPIC: Memory Efficient Position Independent Caching for LLM Serving

Modern LLM applications such as deep-research assistants, coding agents, and Retrieval-Augmented Generation (RAG) systems, repeatedly process long prompt histories containing shared document or code chunks, creating significant pressure on the Key Value (KV) cache, which must operate within limited memory while sustaining high throughput and low latency. Prefix caching partially alleviates some of these costs by reusing KV cache for previously processed tokens, but limited by strict prefix matching. Position-independent caching (PIC) enables chunk-level reuse at arbitrary positions, but requires selective recomputation and positional-encoding (PE) adjustments. However, because these operations vary across queries, KV for the same chunk diverges across requests. Moreover, without page alignment, chunk KV layouts diverge in memory, preventing page sharing. These issues result in only modest HBM savings even when many requests reuse the same content. We present MEPIC, a memory-efficient PIC system that enables chunk KV reuse across positions, requests, and batches. MEPIC aligns chunk KV to paged storage, shifts recomputation from token- to block-level so only the first block is request-specific, removes positional encodings via Rotary Position Embedding (RoPE) fusion in the attention kernel, and makes remaining blocks fully shareable. These techniques eliminate most duplicate chunk KV in HBM, reducing usage by up to 2x over state-of-the-art PIC at comparable latency and accuracy, and up to 5x for long prompts, without any model changes.

cs.LG↗

arXiv2025

Airport Passenger Flow Forecasting via Deformable Temporal-Spectral Transformer Approach

Accurate forecasting of passenger flows is critical for maintaining the efficiency and resilience of airport operations. Recent advances in patch-based Transformer models have shown strong potential in various time series forecasting tasks. However, most existing methods rely on fixed-size patch embedding, making it difficult to model the complex and heterogeneous patterns of airport passenger flows. To address this issue, this paper proposes a deformable temporal-spectral transformer named DTSFormer that integrates a multiscale deformable partitioning module and a joint temporal-spectral filtering module. Specifically, the input sequence is dynamically partitioned into multiscale temporal patches via a novel window function-based masking, enabling the extraction of heterogeneous trends across different temporal stages. Then, within each scale, a frequency-domain attention mechanism is designed to capture both high- and low-frequency components, thereby emphasizing the volatility and periodicity inherent in airport passenger flows. Finally, the resulting multi-frequency features are subsequently fused in the time domain to jointly model short-term fluctuations and long-term trends. Comprehensive experiments are conducted on real-world passenger flow data collected at Beijing Capital International Airport from January 2023 to March 2024. The results indicate that the proposed method consistently outperforms state-of-the-art forecasting models across different prediction horizons. Further analysis shows that the deformable partitioning module aligns patch lengths with dominant periods and heterogeneous trends, enabling superior capture of sudden high-frequency fluctuations.

cs.LG↗

arXiv2025

ElasticMoE: An Efficient Auto Scaling Method for Mixture-of-Experts Models

Mixture-of-Experts (MoE) models promise efficient scaling of large language models (LLMs) by activating only a small subset of experts per token, but their parallelized inference pipelines make elastic serving challenging. Existing strategies fall short: horizontal scaling provisions entire replicas of the current configuration, often tens to hundreds of accelerators, leading to coarse granularity, long provisioning delays, and costly overprovisioning. Vertical scaling offers finer adjustments but typically requires instance restarts, incurring downtime. These limitations make current approaches ill-suited for the bursty, short-lived traffic patterns common in cloud deployments. We present ElasticMoE, an elastic scaling framework for MoE LLMs that achieves fine-grained, low-latency, and zero-downtime scaling. ElasticMoE decouples inference execution from memory operations, enabling scaling steps to proceed concurrently with serving. An HBM Management Module (HMM) reuses weights and KV caches via zero-copy remapping, while high-bandwidth peer-to-peer transfers bring newly added accelerators online without interrupting service. A virtual memory based expert redistribution mechanism migrates MoE experts without costly buffer reallocations, reducing peak memory usage during expert parallelism reconfiguration. Our evaluation on Ascend NPUs with three popular MoE LLMs shows that ElasticMoE achieves up to 9x lower scale-up latency, up to 2x better throughput during scaling, and significantly improves SLO attainment compared to baselines. By enabling fine-grained, concurrent scaling with minimal disruption, ElasticMoE advances the practicality of deploying massive MoE LLMs in dynamic cloud environments.

cs.DC↗

arXiv2025

Distinct dimensions for attractors of bi-Lipschitz iterated function systems

In this paper, we construct an iterated function system on the line consisting of two bi-Lipschitz contractions whose attractor has distinct lower, Hausdorff, lower box, upper box, and Assouad dimensions, thereby providing negative answers to certain folklore questions. Furthermore, as a by-product of our study of bi-Lipschitz IFSs, we construct IFSs within this family that exhibit interesting fractal behaviour. In particular, we prove the following two statements: (i) There exists a bi-Lipschitz IFS for which the pushforward of any ergodic measure with positive entropy is not exact dimensional; (ii) There exists a bi-Lipschitz IFS whose attractor has empty interior yet positive Lebesgue measure.

math.DS↗