arXiv subjects

Xingchen Liu

Publications and source records attributed to Xingchen Liu.

At least 19 recordsLinked to original sources

arXiv2026

FP4Quant: Efficient 4-Bit Communication Compression for Model-Parallel LLM Training

The rapid scaling of large language models (LLMs) has made communication overhead a critical bottleneck in distributed model-parallel training. Although existing low-precision compression methods can effectively reduce communication costs, further reducing the communication footprint to ultra-low-bit formats such as FP4 introduces new challenges, including limited numerical range, sub-byte storage overhead, and increased computational cost from fine-grained quantization. In this paper, we present \textbf{FP4Quant}, a micro-grouped FP4 compression framework for efficient model-parallel LLM training. FP4Quant integrates a micro-grouped dual-scale FP4 quantization mechanism to preserve local tensor distribution characteristics, a warp-cooperative FP4 packing scheme to enable efficient sub-byte data representation, and a hybrid two-level FWHT pipeline to mitigate quantization errors with GPU-aware execution optimization. Experiments on NVIDIA H100 GPUs with Qwen-7B demonstrate up to 1.56$\times$ end-to-end throughput improvement over BF16, revealing a trade-off between throughput and model quality.

cs.AI↗

arXiv2026

Falcons: Pushing the Speed-Ratio Limits of Lossy Compression for Parallel Scientific Applications

The extreme volume of scientific data has become a critical challenge for modern exa-scale scientific computing applications, and error-bounded lossy compression remains indispensable for scientific data management due to its high data reduction rate and controllable data fidelity. However, to ensure sufficient data reduction, existing compressors incur prohibitive computational overheads in real-world parallel scientific computing applications. To overcome the performance bottlenecks of error-bounded lossy compression, we present \textbf{Falcons}, a broadly deployable error-bounded lossy compressor that simultaneously achieves \textbf{high compression ratios with ultra-fast throughput, minimizing the overhead in parallel scientific applications}. Falcons integrates three key designs: memory-regularized interpolation with SIMD (Single Instruction Multiple Data) vectorization, distribution-adaptive lossless coding, and application-oriented optimizations for multiple parallelization schemes. Across our evaluations on cutting-edge supercomputers, Falcons clearly outperforms state-of-the-art compressors, simultaneously achieving up to 29\% compression ratio improvement and 255\% throughput improvement. In case studies of real-world applications, Falcons accelerates large-scale parallel data I/O and remote transfer by up to 44.6\% and 51.5\%, and provides 1.55$\times$ end-to-end speedup for the full WRF (Weather Research \& Forecasting) application workflow.

cs.DC↗

arXiv2026

Joint Beamforming and RIS Deployment Optimization for Outdoor-to-Indoor Coverage at Upper Mid-Band: A Dual-Band Ray-Tracing Study

Upper mid-band spectrum is promising for 5G and 6G cellular networks, yet concrete and low-emissivity glass impose over 20 dB of building penetration loss above 2 GHz, degrading outdoor-to-indoor (O2I) coverage. A reconfigurable intelligent surface (RIS) on a neighboring building bypasses this blockage by focusing energy onto indoor users. This paper introduces PRIMO (Parallel RIS Min-Rate Optimization), a training-free joint optimizer of base-station precoding, RIS phase shifts, and RIS placement, validated by calibrated ray tracing in a MetroTech Commons digital twin (Brooklyn, NY) at 6.75 and 16.95 GHz. At equal physical aperture, half-wavelength spacing packs 6.25 times as many elements per array at FR3 as at FR1(C), so FR1(C) leads at low indoor-user counts while FR3 dominates at higher counts, where the additional spatial degrees of freedom suppress inter-user interference. Above a 1 bps/Hz service floor, PRIMO supports 92% more indoor users at FR3 than at FR1(C), and raises the minimum user rate by up to 90% over a fully converged alternating-optimization baseline while running about 100 times faster.

eess.SP↗

arXiv2026

Tools-CC-Bench: a Benchmark Suite for Collective Communication with Compression in HPC and AI Workloads

Distributed HPC and LLM workloads increasingly require efficient communication for scalability, yet growing data movement has become a major performance bottleneck. Communication compression can reduce this overhead and complement execution-level optimizations, but its benefits remain difficult to assess because existing benchmarks lack support for diverse backends, realistic datasets, application-specific accuracy metrics, and overlap-induced resource contention. We present CC-Bench, a lightweight, extensible, and application-oriented benchmark suite for evaluating communication compression under realistic execution conditions. CC-Bench uses declarative application-environment modeling to decouple profiling logic from communication libraries, datasets, and fidelity metrics, enabling portable cross-library evaluation. It further combines function-level interception and hardware counter monitoring to characterize per-phase latency, hardware utilization, numerical fidelity, and computation interference. With representative datasets from HPC and LLM workloads, CC-Bench evaluates three compression-enabled communication libraries on CPU and GPU clusters, revealing accuracy-performance trade-offs and bottlenecks to guide practical deployment and optimization.

cs.DC↗

arXiv2026

UniCAR-RL: Seeing Better before Thinking Deeper in Visual Mathematics

Multimodal Large Language Models (MLLMs) often struggle with complex mathematical visual reasoning primarily due to a lack of fine-grained perception, causing initial visual hallucinations to directly trigger cascading reasoning failures. In traditional end-to-end reinforcement learning (RL), sparse rewards fail to decouple perceptual hallucinations from logical missteps, hindering targeted perception optimization. Alternatively, fine-tuning with perception-enhanced CoT data incurs high costs and hallucinations. In this paper, we address these challenges by proposing UniCAR-RL, an annotation-free RL framework. By explicitly decoupling the optimization of perception and reasoning during the training process, it achieves isolation and optimization of both capabilities. Specifically, UniCAR-RL consists of three synergistic branches: 1) a Caption-RL branch that optimizes perception capabilities through verifier-guided reasoning validation; 2) a Reasoning-RL branch that performs logical reasoning based on a gold image description to halt cascading errors; 3) a QA-RL branch that retains native end-to-end alignment to ensure robust question-answering performance. Experiments show that UniCAR-RL substantially improves MLLMs' mathematical and visual reasoning using only raw short-answer data. Furthermore, it demonstrates strong generalization across diverse architectures and scales.

cs.AI↗

arXiv2026

Map-Free Single-Anchor Position Localization Using Multipath Uncertainty at Upper Mid-Band

Future 5G-Advanced and 6G systems will exploit directional wideband channels for path-based localization, yet map-free indoor positioning remains challenging because extracted multipath components (MPCs) vary widely in localization reliability. This paper presents an uncertainty-aware map-free single-anchor localization framework using 16.95~GHz indoor directional channel measurements collected at NYU WIRELESS in Brooklyn, New York. Per-MPC angular covariance is estimated from the local consistency of neighboring pattern-de-embedded power delay profile (PDP) observations and propagated into geometric localization through weighted fusion of point-type and line-type constraints. Measurements across 20 links with TX--RX separations from 11~m to 97~m show that the proposed covariance-weighted method achieves a mean localization error of 3.86~m, a median of 2.66~m, and 70\% of links within 5~m using five retained MPCs. It outperforms the power-weighted and unweighted variants under identical point/line constraints, as well as the literature baselines. These results indicate that local directional consistency provides reliability information beyond received power alone and demonstrate a practical map-free approach for future indoor directional systems.

eess.SP↗

arXiv2026

Omnidirectional UMi Path Loss Models for 3GPP Extension Above 100 GHz Using Extensive Empirical Data Sets from 6.75 to 142 GHz

Extending standardized urban microcell (UMi) path loss (PL) models from the upper mid-band (FR3) to sub-terahertz (sub-THz) frequencies requires measurement-based comparisons spanning a wide spectral range within a unified framework. In this paper, we derive single- and multi-frequency omnidirectional PL models for UMi environments based on extensive NYU WIRELESS measurements at 6.75, 16.95, 28, 73, and 142 GHz in line-of-sight (LOS) and non-line-of-sight (NLOS) conditions. Single-frequency close-in (CI) and floating-intercept (FI) models for each band show that CI yields more stable and physically interpretable parameters, with shadow fading standard deviations within 0.98 dB of FI across all five frequencies. We then extend the multi-frequency PL analysis to the 7-24 GHz, 0.5-100 GHz, and 0.5-150 GHz frequency ranges using CI, close-in with frequency weighting (CIF), and alpha-beta-gamma (ABG) models. Across progressively wider frequency spans, CI and CIF maintain stable distance exponents that remain closely aligned with the 3rd Generation Partnership Project (3GPP) UMi interpretations, whereas ABG offers only very modest reductions in fitting error at the cost of much greater parameter sensitivity. These results support physically anchored CI/CIF formulations with a close-in free space path loss anchor point at 1 m in order to extend 3GPP-oriented UMi PL models over the entire 0.5 to 150 GHz frequency range.

eess.SP↗

arXiv2026

SPARC: Sparse Path-Aware Residual Calibrator for Wireless Ray Tracing at Upper Mid-Band

Accurate site-specific ray tracing (RT) is essential for upper mid-band network planning, yet raw RT can produce per-path multipath component (MPC) power errors on the order of 19--24~dB in cluttered indoor environments. A fixed-geometry material-sensitivity bound shows that a 30% relative-permittivity perturbation changes each surface interaction by at most 6.28~dB across the considered indoor materials. However, even MPCs with only one surface interaction exhibit a 19.2~dB mean RT--measurement bias, suggesting that missing clutter, displaced surfaces, and simplified 3D geometry dominate the per-path RT error. We propose SPARC (Sparse Path-Aware Residual Calibrator), a lightweight per-path calibration method that learns a sparse linear residual model from one completed RT simulation. SPARC uses standard RT features selected per fold by nested cross-validation, with ridge regularization and power-gated path matching; four features recur in both environments. Using measured indoor factory (InF) and indoor hotspot (InH) datasets at 6.75 and 16.95~GHz, SPARC reduces per-path power RMSE from 18.74 to 4.74~dB in InF and from 23.12 to 5.39~dB in InH. A jointly trained InF+InH model achieves 5.73~dB RMSE. When all links from one transmitter location are held out for testing, SPARC achieves 4.99~dB RMSE in InF and 5.85~dB RMSE in InH. SPARC therefore provides a practical post-processing calibration layer for site-specific per-path power prediction without ray-tracer modification or additional RT runs.

eess.SP↗

arXiv2026

The Anatomy of RF Chains: Metrics, Measures, and Operating Efficiency

In 1958, Haus and Adler [1] introduced the concept of noise measure. Noise measure is a single quantitative metric that provides a comprehensive basis for comparing devices (individual circuits or outcomes of optimization iterations) in terms of their contribution to overall system noise by incorporating both noise factor and available power gain. Unlike noise factor alone, which reflects how much a device degrades the signal-to-noise ratio, noise measure captures the trade-off between noise and gain, making it a system-aware metric. This distinction is especially important when comparing devices in multistage systems, where both parameters jointly influence the overall system noise. Building on Haus and Adler's work, this article aims to advance RF system design by extending traditional device metrics, such as noise factor and noise measure, with new system-aware metrics: linearity, dynamic range, power efficiency, and waste measures. These new measures are interpretable, computable, and cascadable, making them well-suited for comparing the impact of individual devices or tracking the convergence of circuit design iterations on overall system-level performance. Additionally, a new metric--operating efficiency--is introduced, which unifies power efficiency, dynamic range, and data rate by incorporating signal statistics and variability in communication circuits and systems. Operating efficiency enables robust evaluation of devices under realistic and transient operating conditions, including interference, modulated signals, and adaptive modulation schemes.

eess.SY↗

arXiv2026

From Understanding to Action: Feedback-Grounded Policy Discovery for Generative Recommendation

Semantic-ID-based generative recommenders enable efficient next-item generation, but their item-level supervision mainly captures behavioral co-occurrence and local transitions. Large language models (LLMs) can complement these models by reasoning over heterogeneous interaction histories to understand the user's current demand. However, LLMs are not inherently trained with recommendation-specific outcome feedback, and linguistically plausible reasoning therefore does not necessarily lead to effective recommendation decisions. We term this mismatch the Understanding-Action Gap. Accordingly, we distinguish intent knowledge, which captures the user's current demand, from policy knowledge, which specifies the recommendation direction and rejection boundary under that demand. To bridge this gap, we propose a feedback-driven agent framework that first induces task-oriented intent and then discovers recommendation policies according to their incremental utility over an intent-only baseline. Candidate policies are evaluated and refined using outcome-derived feedback rather than linguistic plausibility. We further transfer the resulting intent and policy knowledge into two latent tokens of a lightweight Semantic-ID generator through dual-space relational distillation, enabling LLM-free online inference. Experiments on public benchmarks show consistent improvements over baselines, while large-scale online A/B tests achieve gains of 4.506% in Revenue and 4.621% in ADVV.

cs.IR↗

arXiv2026

Contrastive Reinforced Policy Optimization via Privileged Self-Distillation

Recent advances in post-training Large Language Models (LLMs) increasingly rely on Reinforcement Learning with Verifiable Rewards (RLVR) or On-Policy Self-Distillation (OPSD). While OPSD provides dense, logit-level supervision, it inherently suffers from exposure bias due to the privileged information of the self-teacher. In multi-turn agentic settings, this leads to reasoning route convergence and the loss of clear optimization directions. To tackle these challenges, we introduce Contrastive Reinforced Policy Optimization (CRPO), which reformulates agentic OPSD from a contrastive learning perspective. By leveraging predictive entropy to distinguish between positive positions (reflective exploration) and negative positions (exposure bias), CRPO conducts group-wise contrast to preserve reliable, fine-grained optimization signals. Extensive evaluations across 13 challenging reasoning and deep-search benchmarks demonstrate that CRPO consistently outperforms existing reinforcement learning and self-distillation baselines, significantly enhancing training stability and generalization in long-horizon interactions.

cs.LG↗

arXiv2026

ClawTrack: Towards Trace-Level Evaluation and Improvement of Real-World Autonomous Agents

As LLM-based agents are deployed in complex, multi-step workflows, a critical evaluation gap has emerged: most existing benchmarks judge only final outcomes, unable to distinguish reliable reasoning from lucky success or attribute failures to specific process deficiencies, hindering attribution in long-horizon tasks. In this work, we present ClawTrack, a dual-assessment benchmark that simultaneously measures what an agent achieves (Task Score) and how it achieves it (Process Score). ClawTrack comprises 320 tasks across 8 domains with 25+ deterministic mock services. A Process Grader scores each reasoning turn along four dimensions (goal alignment, efficiency, information utilization, and result verification), anchored by 12,541 task-specific rubric items. Evaluating 21 models over 16,000+ trials, we find that: (1) process scores effectively attribute success and failure to specific reasoning dimensions, filtering lucky passes invisible to outcome-only evaluation; (2) the four dimensions are complementary, with result verification as the systematic bottleneck; (3) the framework is robust to evaluator choice across different judge LLMs; and (4) process-based trajectory filtering yields consistent post-training improvements across model scales.

cs.LG↗

arXiv2026

AlphaOracle: Oracle bone script decipherment via human-workflow-inspired deep learning

Approximately 3,000 of the 4,500 oracle bone script (OBS) characters remain undeciphered due to fragmentary inscriptions and sparse evidence. Current AI approaches fail to replicate expert workflows that integrate form analysis, contextual semantics, and philological reasoning. We introduce AlphaOracle, a human-workflow-inspired framework that systematizes OBS decipherment using the largest digitized corpus to date. Its multi-stage pipeline comprises: (i) rubbing parsing; (ii) radical-based morphological analysis with diachronic modeling; (iii) contextual retrieval with semantic alignment; and (iv) philological validation against classical sources. Each stage generates explicit, confidence-weighted evidence chains, culminating in interpretable reports for scholarly verification. Across multiple test characters, AlphaOracle's readings strongly agreed with expert interpretations. In a study of 86 domain specialists, it reduced analysis time by 64% and 79% of participants rated it highly useful. Notably, AlphaOracle resolves the character "Lao" as a toponymic or clan designation, offering concrete revisions to Shang administrative and social interpretations. These results suggest that computational methods aligned with philological practice can facilitate OBS research and provide a conceptual reference for studies of other undeciphered scripts.

cs.HC↗

arXiv2026

Modulation of anomalous Hall angle in a magnetic topological semimetal

The anomalous Hall angle (θA) is a measure of the efficiency of converting a longitudinal driving current to a transverse spin-polarized Hall current. For anomalous Hall sensing, a large anomalous Hall angle can improve the sensitivity of magnetic field detection. However, modulation of this angle is challenging and magnetic materials typically have low angles of 0.1 to 3°. Here, we report modulation of the anomalous Hall angle in the magnetic Weyl semimetal Co3Sn2S2. We propose that the angle parameter tanθA can be formulated as a function of the product of electrical resistivity and anomalous Hall conductivity. Our scheme was utilized to demonstrate the modulation of tanθA up to a magnitude of 0.46, corresponding to an angle of around 25°. Microfabricated anomalous Hall devices using Fe-doped Co3Sn2S2 single-crystalline nanoflakes exhibit a high Hall sensitivity of 7028 μΩucm/T and a magnetic field detectability of 23.5 nT/Hz0.5 at 1 Hz.

cond-mat.mes-hall↗

arXiv2026

EventRadar: Long-Range Visual UAV Discovery through Spatiotemporal Event Sensing

Unauthorized unmanned aerial vehicle (UAV) activity around airports, public venues, and other sensitive sites has made protected-airspace monitoring increasingly important. A practical sensing system must search a wide angular region, find small long-range targets, and return both bearing support and UAV-specific evidence before a restricted perimeter is breached. Existing UAV detection paths often rely on spatially organized evidence, such as body extent, silhouette, or track continuity. At long range, however, these cues become difficult to preserve and verify as the target footprint weakens and its image-plane support shrinks. EventRadar follows a complementary cue: propeller-induced temporal periodicity, which recent event-camera sensing studies have shown can reveal UAV-specific motion after appearance becomes weak. We extend this cue to kilometer-scale active sensing with an event-camera prototype. Scene-Anchored Geometry Evidence (SAGE) fuses scanning events with IMU pose to maintain a bearing-indexed scene memory, separating transient candidate support from persistent background clutter. Comb-guided Harmonic-Group Learned Iterative Shrinkage and Thresholding Algorithm (CHG) then treats each candidate as a weak high-rate timing signal and recovers phase-insensitive harmonic evidence with fixed compute. Compared with related event-camera baselines on 700-1500 m UAV event recordings, EventRadar achieves 0.990 mAP$_{.3}$ and 0.949 F1$_{.3}$, reduces FN$_{.3}$ to 0.009, and shows real-time feasibility in prototype profiling.

cs.CV↗

arXiv2026

OneReason Technical Report

Generative recommendation models in the OneRec family have been widely deployed in many real-world services, such as short-video, live-streaming, advertising, and e-commerce. However, these generative models can only benefit from the scaling advantage, while their reasoning ability is hard to activate, since we cannot construct meaningful Chain-of-Thought (CoT) sequences consisting of itemic tokens only. Inspired by the success of the reasoning-style ``think before answer'' paradigm in the LLM field, we conduct preliminary studies (i.e., OneRec-Think, OpenOneRec) to explore reasoning capability in generative recommendation. Nevertheless, we notice an unexpected phenomenon: the thinking mode does not show advantages over the non-thinking mode. Drawing insights from recent findings on CoT robustness in multi-modal language models, we argue that effective reasoning in recommendation rests on two factors: perception, the ability to ground itemic tokens in their underlying language semantics, and cognition, the ability to reorganize a user's behavior sequence into coherent latent interest points. We therefore propose OneReason, which includes: (1) strong itemic token perception in pre-training, (2) a three-level cognition-enhanced CoT format for recommendation tasks in SFT, and (3) a specialize-then-unify training recipe in RL to enhance the thinking ability.

cs.IR↗

arXiv2026

KVServe: Service-Aware KV Cache Compression for Communication-Efficient Disaggregated LLM Serving

LLMs are widely adopted in production, pushing inference systems to their limits. Disaggregated LLM serving (e.g., PD separation and KV state disaggregation) improves scalability and cost efficiency, but it also turns KV into an explicit payload crossing network and storage boundaries, making KV a dominant end-to-end bottleneck. Existing KV compression are typically static runtime configurations, despite production service context varies over time in workload mix, bandwidth, and SLO/quality budgets. As a result, a fixed choice can be suboptimal or even increase latency. We present \emph{KVServe}, the first service-aware and adaptive KV communication compression framework for disaggregated LLM serving: KVServe (1) unifies KV compression into a modular strategy space with new components and cross-method recomposition; (2) introduces Bayesian Profiling Engine that efficiently searches this space and distills a 3D Pareto candidate set, reducing $50\times$ offline search overhead; and (3) deploys a Service-Aware Online Controller that combines an analytical latency model with a lightweight bandit to select profiles under constraints and correct offline-to-online mismatch. Integrated into vLLM and evaluated across datasets, models, GPUs and networks, KVServe achieves up to $9.13\times$ JCT speedup in PD-separated serving and up to $32.8\times$ TTFT reduction in KV-disaggregated serving.

cs.DC↗

arXiv2026

DocSeeker: Structured Visual Reasoning with Evidence Grounding for Long Document Understanding

Existing Multimodal Large Language Models (MLLMs) suffer from significant performance degradation on the long document understanding task as document length increases. This stems from two fundamental challenges: 1) a low Signal-to-Noise Ratio (SNR), with crucial evidence buried in irrelevant pages; and 2) supervision scarcity, as datasets offering only final short answers provide a weak learning signal. In this paper, we address these challenges by proposing a paradigm that requires the model to execute a structured Analysis, Localization and Reasoning workflow. To instill this capability, we design a two-stage training framework: we first perform Supervised Fine-Tuning on high-quality data generated via an efficient knowledge distillation strategy. Subsequently, we employ an Evidence-aware Group Relative Policy Optimization which jointly optimizes for both evidence localization and answer accuracy. Additionally, we introduce a Evidence-Guided Resolution Allocation strategy to mitigate memory constraints of training on multi-pages documents. Extensive experiments demonstrate that DocSeeker achieves superior performance on both in-domain and out-of-domain tasks. We show it robustly generalizes from short-page training to ultra-long documents and is naturally synergistic with visual Retrieval-Augmented Generation systems, serving as a solid foundation for their implementation.

cs.AI↗