arXiv subjects

Pengcheng Wang

Publications and source records attributed to Pengcheng Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

DiscreteRTC: Discrete Diffusion Policies are Natural Asynchronous Executors

Unlike chatbots, physical AI must act while the world keeps evolving. Therefore, the inter-chunk pause of synchronous executors are fatal for dynamic tasks regardless of how fast the inference is. Asynchronous execution -- thinking while acting -- is therefore a structural requirement, and real-time chunking (RTC) makes it viable by recasting chunk transitions as inpainting: freezing committed actions and consistently generating the remainder. However, RTC with flow-matching policy is structurally suboptimal: its inpainting comes from inference-time corrections rather than the base policy, yielding little pre-training benefit, specific fine-tuning, heuristic guidance, and extra computation that inflates the latency. In this work, we observe that discrete diffusion policies, which generate actions by iteratively unmasking, are natural asynchronous executors that resolve all limitations at once: they are fine-tuning free since inpainting is their native operation, while early stopping further provides adaptive guidance and reduces inference cost. We propose DiscreteRTC, which replaces external corrections with native unmasking, and show on dynamic simulated benchmarks and real-world dynamic manipulation tasks that it achieves higher success rates than continuous RTC and other baselines. In summary, DiscreteRTC is simpler to implement with 0 lines of additional code to enable async inpainting, faster at inference with only ~0.7 computation compared with generating actions from scratch, and better at execution with 65% higher success rate in real-world hockey defend task compared with flow-matching RTC, and 30% higher compared with training-time flow-matching RTC. More visualizations are on https://outsider86.github.io/DiscreteRTCSite/.

cs.RO↗

arXiv2026

WarmServe: Enabling One-for-Many GPU Prewarming for Multi-LLM Serving

Deploying multiple models within shared GPU clusters is a key strategy to improve resource efficiency in large language model (LLM) serving. Existing multi-LLM serving systems improve GPU utilization at the cost of degraded inference performance, particularly time-to-first-token (TTFT). We attribute this degradation to the lack of awareness regarding future workload characteristics. In contrast, recent analyses have shown the strong periodicity and long-term predictability of real-world LLM serving workloads. In this paper, we propose one-for-many GPU prewarming, which proactively loads parameters from multiple models onto GPUs based on workload forecasts. These prewarmed weights enable the system to promptly instantiate serving instances upon encountering request bursts. We design and implement WarmServe, a multi-LLM serving system incorporating three key techniques: (1) a model placement algorithm that optimizes prewarming decisions to minimize cross-model prewarming interference, (2) a KV cache reservation strategy that repurposes idle KV cache space on running GPUs for prewarming new models, and (3) an efficient GPU memory switching mechanism for tensor management. Evaluation on real-world datasets shows that WarmServe reduces tail TTFT by up to 50.8$\times$ compared to the state-of-the-art autoscaling-based system, while supporting up to 2.5$\times$ higher request throughput than the GPU-sharing system.

cs.DC↗

arXiv2026

CaptchaMind: Training CAPTCHA Solvers via Reinforcement Learning with Explicit Reasoning Supervision

CAPTCHAs are widely deployed as human verification mechanisms and frequently block intelligent agents from completing end-to-end automation in real-world web environments. Solving modern CAPTCHAs requires robust multi-step visual reasoning and interaction capabilities, yet training-based approaches have remained absent due to the lack of large-scale training data and process-level annotations. We introduce CaptchaBench, the first CAPTCHA benchmark designed to support large-scale training, comprising 16,000 programmatically generated samples across eight task categories with detailed region and process-level annotations. Systematic evaluation on CaptchaBench reveals that existing methods fail consistently on tasks requiring fine-grained visual detail capture and region-level comparison. We therefore present CaptchaMind, an RL-based solver trained with explicit reasoning process supervision, achieving 82.9% average success rate across eight tasks and 71.0% on real-world instances, substantially outperforming all existing methods without closed-source APIs.

cs.CV↗

arXiv2026

Geometric and Spectral Alignment for Deep Neural Network I

Deep residual architectures are modeled as products of near-identity Jacobians. This paper proves deterministic quotient-geometric estimates for singular spectra of Frobenius-normalized layer factors, emphasizing a normalized top-radial Cartan coordinate and fitted power-law chart. Full-rank factors are mapped from $\mathrm{GL}(d)$ to the positive cone by $A\mapsto A^\top A$, then to ordered eigenvalue data. Under Frobenius normalization, exact power-law spectra form a trace-normalized Cartan orbit. This orbit is a Gibbs family on ranks, a Fisher information line, and a Bures--Wasserstein curve with line element $d/4$ times Fisher information. The main rigidity theorem is a slack-aware margin inequality: interface radial amplitude, non-backtracking slack, and signed residual variation control displacement of the fitted Cartan coordinate. In the exact-chart zero-slack case, a depth-$L$ budget gives exponent drift of order $(\log M)/L$; generally, slack and residual increments augment the bound. We separate scalar top-radial from full-Cartan spectral control, which also needs Bures/Hellinger residual variation. We prove approximate-power-law and metric-chart versions, converse lower bounds, Fisher--KL/Bures action estimates, and near-identity expansions for normalized residual chains. Near-identity results verify transport budgets; chart quality remains measurable. Effective rank is a spectral-energy quantile, giving finite-width power-law tail bounds and robust rank-window transition estimates. Empirical static-weight exponent profiles serve as diagnostics; full verification also requires interface budgets, slacks, and residuals for the same operator chain.

cs.LG↗

arXiv2026

Geometric and Spectral Alignment for Deep Neural Network II

This paper develops the angular and static-channel component of Geometric and Spectral Alignment for residual Jacobian chains. Starting from Cartan-coordinate rigidity and fitted effective-rank windows, we study how dominant singular subspaces are transported across adjacent layers and how the resulting finite matrices can be displayed in physical channel coordinates. The main results are deterministic, margin-verified results. We bound the error between full interface transport and its dominant-window truncation, add fitted-tail errors so that empirical spectra can be certified against the Gibbs--Cartan tail model, and distinguish source-mode incidence from fully physical input-output channel incidence. Given row groups and active supports, the Physical Alignment Matrix decomposes orthogonally as core plus overlap plus noise. Active-column gaps, pairwise overlap margins, and noise bounds combine into a static certificate radius under which the full transport and the truncated transport induce the same active supports, pairwise incidence graph, SRS sets, hub columns, and core/overlap/noise masks. The finer SC/SA/ST labels of the Invariant Channel Mapping require additional row-energy and profile-correlation margins, stated as explicit perturbation tests. The empirical section reports the matrices and block-energy heatmaps that measure these certificate quantities across CNNs, language models, and vision/diffusion backbones. The figures are interpreted as finite-dimensional measurements; complete membership in the Physical GSA certificate domain requires checking the numerical margin protocol stated in Section 10.

cs.LG↗

arXiv2026

AgentSPEX: An Agent SPecification and EXecution Language

Language-model agent systems commonly rely on reactive prompting, in which a single instruction guides the model through an open-ended sequence of reasoning and tool-use steps, leaving control flow and intermediate state implicit and making agent behavior potentially difficult to control. Orchestration frameworks such as LangGraph, DSPy, and CrewAI impose greater structure through explicit workflow definitions, but tightly couple workflow logic with Python, making agents difficult to maintain and modify. In this paper, we introduce AgentSPEX, an Agent SPecification and EXecution Language for specifying LLM-agent workflows with explicit control flow and modular structure, along with a customizable agent harness. AgentSPEX supports typed steps, branching and loops, parallel execution, reusable submodules, and explicit state management, and these workflows execute within an agent harness that provides tool access, a sandboxed virtual environment, and support for checkpointing, verification, and logging. Furthermore, we provide a visual editor with synchronized graph and workflow views for authoring and inspection. We include ready-to-use agents for deep research and scientific research, and we evaluate AgentSPEX on 7 benchmarks. Finally, we show through a user study that AgentSPEX provides a more interpretable and accessible workflow-authoring paradigm than a popular existing agent framework.

cs.CL↗

arXiv2026

LiteCache: A Query Similarity-Driven, GPU-Centric KVCache Subsystem for Efficient LLM Inference

During LLM inference, KVCache memory usage grows linearly with sequence length and batch size and often exceeds GPU capacity. Recent proposals offload KV states to host memory and reduce transfers using top-k attention. But their CPU-centric management of the on-GPU cache and CPU-GPU data movement incurs high overhead and fragments the bulk GPU execution that CUDA Graph relies on. To close this gap, we observe that adjacent queries within the same attention head exhibit strong directional similarity and retrieve highly overlapping top-k KV states. This insight enables a simple head granularity cache algorithm, QSAC, in which each head reuses its previously cached KV states whenever the current query is sufficiently similar to the prior one. QSAC further simplifies cache management primitives and cuts CPU involvement almost entirely. We develop LiteCache, a KVCache subsystem that incorporates QSAC. LiteCache introduces a GPU-centric synchronization controller and speculative sparse prefetching, enabling fully overlapped data movement and computation. These mechanisms produce a stable and predictable execution pattern that remains compatible with the bulk execution mode required by CUDA Graphs. Evaluation on two widely-used LLMs indicates that LiteCache achieves comparable accuracy to baselines, while sharply minimizing CPU overhead, fully utilizing PCIe bandwidth, thus improving decoding throughput by 10.7-224.2% on both H100 and A40 GPUs and easily supporting sequence lengths beyond 1M. We opensource LiteCache at https://anonymous.4open.science/r/LiteCache-888D.

cs.LG↗

arXiv2026

Mean Flow Policy with Instantaneous Velocity Constraint for One-step Action Generation

Learning expressive and efficient policy functions is a promising direction in reinforcement learning (RL). While flow-based policies have recently proven effective in modeling complex action distributions with a fast deterministic sampling process, they still face a trade-off between expressiveness and computational burden, which is typically controlled by the number of flow steps. In this work, we propose mean velocity policy (MVP), a new generative policy function that models the mean velocity field to achieve the fastest one-step action generation. To ensure its high expressiveness, an instantaneous velocity constraint (IVC) is introduced on the mean velocity field during training. We theoretically prove that this design explicitly serves as a crucial boundary condition, thereby improving learning accuracy and enhancing policy expressiveness. Empirically, our MVP achieves state-of-the-art success rates across several challenging robotic manipulation tasks from Robomimic and OGBench. It also delivers substantial improvements in training and inference speed over existing flow-based policy baselines.

cs.LG↗

arXiv2026

ReCast: Reliability-aware Codebook Assisted Lightweight Time Series Forecasting

Time series forecasting is crucial for applications in various domains. Conventional methods often rely on global decomposition into trend, seasonal, and residual components, which become ineffective for real-world series dominated by local, complex, and highly dynamic patterns. Moreover, the high model complexity of such approaches limits their applicability in real-time or resource-constrained environments. In this work, we propose a novel \textbf{RE}liability-aware \textbf{C}odebook-\textbf{AS}sisted \textbf{T}ime series forecasting framework (\textbf{ReCast}) that enables lightweight and robust prediction by exploiting recurring local shapes. ReCast encodes local patterns into discrete embeddings through patch-wise quantization using a learnable codebook, thereby compactly capturing stable regular structures. To compensate for residual variations not preserved by quantization, ReCast employs a dual-path architecture comprising a quantization path for efficient modeling of regular structures and a residual path for reconstructing irregular fluctuations. A central contribution of ReCast is a reliability-aware codebook update strategy, which incrementally refines the codebook via weighted corrections. These correction weights are derived by fusing multiple reliability factors from complementary perspectives by a distributionally robust optimization (DRO) scheme, ensuring adaptability to non-stationarity and robustness to distribution shifts. Extensive experiments demonstrate that ReCast outperforms state-of-the-art (SOTA) models in accuracy, efficiency, and adaptability to distribution shifts.

cs.LG↗

arXiv2026

Towards Realistic Detection Pipelines of Taiji: New Challenges in Data Analysis and High-Fidelity Simulations of Space-Based Gravitational Wave Antenna

Taiji, a Chinese space-based gravitational wave (GW) detection project, aims to explore the millihertz GW universe with unprecedented sensitivity. By observing astrophysical and cosmological sources, including Galactic binaries, massive black hole binaries, extreme mass-ratio inspirals, and stochastic gravitational wave backgrounds, etc., Taiji is expected to deliver transformative insights into astrophysics, cosmology, and fundamental physics. However, Taiji's data analysis faces unique challenges compared to ground-based detectors like LIGO-Virgo-KAGRA, such as the overlap of numerous signals, extended data durations, more rigorous accuracy requirements for the waveform templates, incompletely characterized noise spectra, non-stationary noises, and various data anomalies. Taking Taiji as a representative example, this paper reviews the data characteristics and data analysis challenges of space-based GW detection, and introduces the second round of Taiji Data Challenge, a collection of simulation datasets designed as a shared platform for resolving these critical issues. This platform distinguishes itself from previous works by the systematic integration of orbital dynamics based on a full drag-free and attitude control simulation, extended noise sources, more complicated and overlapping GW signals, second-generation time-delay interferometry, and the coupling effect of time-varying arm-lengths, etc. Concurrently released is the open-source toolkit Triangle, which offers the capabilities for customized simulation of signals, noises, and other instrumental effects. By taking a step further towards realistic detection, Taiji Data Challenge II and Triangle altogether serve as a new testbed, supporting the development of Taiji's global analysis and end-to-end pipelines, and ultimately bridging the gaps between observation and scientific objectives.

gr-qc↗

arXiv2025

What's on Your Plate? Inferring Chinese Cuisine Intake from Wearable IMUs

Accurate food intake detection is vital for dietary monitoring and chronic disease prevention. Traditional self-report methods are prone to recall bias, while camera-based approaches raise concerns about privacy. Furthermore, existing wearable-based methods primarily focus on a limited number of food types, such as hamburgers and pizza, failing to address the vast diversity of Chinese cuisine. To bridge this gap, we propose CuisineSense, a system that classifies Chinese food types by integrating hand motion cues from a smartwatch with head dynamics from smart glasses. To filter out irrelevant daily activities, we design a two-stage detection pipeline. The first stage identifies eating states by distinguishing characteristic temporal patterns from non-eating behaviors. The second stage then conducts fine-grained food type recognition based on the motions captured during food intake. To evaluate CuisineSense, we construct a dataset comprising 27.5 hours of IMU recordings across 11 food categories and 10 participants. Experiments demonstrate that CuisineSense achieves high accuracy in both eating state detection and food classification, offering a practical solution for unobtrusive, wearable-based dietary monitoring.The system code is publicly available at https://github.com/joeeeeyin/CuisineSense.git.

cs.CV↗

arXiv2025

Mind Your Entropy: From Maximum Entropy to Trajectory Entropy-Constrained RL

Maximum entropy has become a mainstream off-policy reinforcement learning (RL) framework for balancing exploitation and exploration. However, two bottlenecks still limit further performance improvement: (1) non-stationary Q-value estimation caused by jointly injecting entropy and updating its weighting parameter, i.e., temperature; and (2) short-sighted local entropy tuning that adjusts temperature only according to the current single-step entropy, without considering the effect of cumulative entropy over time. In this paper, we extends maximum entropy framework by proposing a trajectory entropy-constrained reinforcement learning (TECRL) framework to address these two challenges. Within this framework, we first separately learn two Q-functions, one associated with reward and the other with entropy, ensuring clean and stable value targets unaffected by temperature updates. Then, the dedicated entropy Q-function, explicitly quantifying the expected cumulative entropy, enables us to enforce a trajectory entropy constraint and consequently control the policy long-term stochasticity. Building on this TECRL framework, we develop a practical off-policy algorithm, DSAC-E, by extending the state-of-the-art distributional soft actor-critic with three refinements (DSAC-T). Empirical results on the OpenAI Gym benchmark demonstrate that our DSAC-E can achieve higher returns and better stability.

cs.LG↗

arXiv2025

Inter-Spacecraft Tilt-to-Length Noise Reduction Algorithm for Taiji Mission

The Taiji mission for space-based gravitational wave (GW) detection employs laser interferometry to measure picometer-scale distance variations induced by GWs. The tilt-to-length (TTL) coupling noise in the inter-spacecraft interferometers, which originates from the angular jitters of the spacecrafts and the movable optical subassemblies, is predicted to be one of the main noise sources that might reduce Taiji's sensitivity to GWs. Since these angular jitters can be read out through the differential wavefront sensors, it is possible to suppress TTL noise during the data processing stage by fitting and subtracting it. This paper proposes an improved algorithm for TTL noise suppression, which addresses the issue of unknown noise floor required for optimal estimation in the practical detection scenario, and the design of this algorithm takes into account the presence of GW signals. The algorithm is validated via numerical simulation, which is built on a spacecraft dynamics simulation incorporating Taiji's drag-free and attitude control system. We also demonstrate the robustness of this algorithm by varying TTL coefficients at different levels, indicating that our algorithm is applicable to a range of payload statuses, and ultimately providing a critical advancement toward realizing Taiji's full sensitivity.

gr-qc↗

arXiv2025

Entropy-Regularized Process Reward Model

Large language models (LLMs) have shown promise in performing complex multi-step reasoning, yet they continue to struggle with mathematical reasoning, often making systematic errors. A promising solution is reinforcement learning (RL) guided by reward models, particularly those focusing on process rewards, which score each intermediate step rather than solely evaluating the final outcome. This approach is more effective at guiding policy models towards correct reasoning trajectories. In this work, we propose an entropy-regularized process reward model (ER-PRM) that integrates KL-regularized Markov Decision Processes (MDP) to balance policy optimization with the need to prevent the policy from shifting too far from its initial distribution. We derive a novel reward construction method based on the theoretical results. Our theoretical analysis shows that we could derive the optimal reward model from the initial policy sampling. Our empirical experiments on the MATH and GSM8K benchmarks demonstrate that ER-PRM consistently outperforms existing process reward models, achieving 1% improvement on GSM8K and 2-3% improvement on MATH under best-of-N evaluation, and more than 1% improvement under RLHF. These results highlight the efficacy of entropy-regularization in enhancing LLMs' reasoning capabilities.

cs.LG↗

arXiv2025

Adaptive extended Kalman filter and laser link acquisition in the detection of gravitational waves in space

An alternative, new laser link acquisition scheme for the triangular constellation of spacecraft (SCs) in deep space in the detection of gravitational waves is considered. In place of a wide field CCD camera in the initial stage of laser link acquisition adopted in the conventional scheme, an extended Kalman filter based on precision orbit determination is incorporated in the point ahead angle mechanism (PAAM) to steer the laser beam in such a way to narrow the uncertainty cone and at the same time avoids the heating problem generated by the CCD camera.A quadrant photodetector (QPD) based on the Differential Power Sensing (DPS) technique, which offers a higher dynamic range than differential wavefront sensing (DWS), is employed as the readout of the laser beam spot. The conventional two stages (coarse acquisition and fine acquisition) are integrated into a single control loop. The payload structure of the ATP control loop is simplified and numerical simulations, based on a colored measurement noise model that closely mimics the prospective on-orbit conditions, demonstrate that the AEKF significantly reduces the initial uncertainty region by predicting the point ahead angle (PAA) even when the worst case scenario in SC position (navigation) error is considered.

astro-ph.IM↗

arXiv2025

RAG-Boost: Retrieval-Augmented Generation Enhanced LLM-based Speech Recognition

In this paper, we propose RAG-Boost (ST-ShinozakiLab Task I system), which enhances the baseline LLM-based ASR system of the MLC-SLM Challenge (task I) with a retrieval-augmented generation (RAG) module on the fly. Each partial ASR hypothesis queries a vector store of audio-text pairs and domain terms, and the retrieved results are fused with the live ASR hypotheses to fix recognition errors. The fused hypotheses are passed to the LLM, yielding improved responses.

eess.AS↗

arXiv2025

Interactive Hybrid Rice Breeding with Parametric Dual Projection

Hybrid rice breeding crossbreeds different rice lines and cultivates the resulting hybrids in fields to select those with desirable agronomic traits, such as higher yields. Recently, genomic selection has emerged as an efficient way for hybrid rice breeding. It predicts the traits of hybrids based on their genes, which helps exclude many undesired hybrids, largely reducing the workload of field cultivation. However, due to the limited accuracy of genomic prediction models, breeders still need to combine their experience with the models to identify regulatory genes that control traits and select hybrids, which remains a time-consuming process. To ease this process, in this paper, we proposed a visual analysis method to facilitate interactive hybrid rice breeding. Regulatory gene identification and hybrid selection naturally ensemble a dual-analysis task. Therefore, we developed a parametric dual projection method with theoretical guarantees to facilitate interactive dual analysis. Based on this dual projection method, we further developed a gene visualization and a hybrid visualization to verify the identified regulatory genes and hybrids. The effectiveness of our method is demonstrated through the quantitative evaluation of the parametric dual projection method, identified regulatory genes and desired hybrids in the case study, and positive feedback from breeders.

cs.HC↗

arXiv2025

Red Teaming for Generative AI, Report on a Copyright-Focused Exercise Completed in an Academic Medical Center

Background: Generative artificial intelligence (AI) deployment in academic medical settings raises copyright compliance concerns. Dana-Farber Cancer Institute implemented GPT4DFCI, an internal generative AI tool utilizing OpenAI models, that is approved for enterprise use in research and operations. Given (1) the exceptionally broad adoption of the tool in our organization, (2) our research mission, and (3) the shared responsibility model required to benefit from Customer Copyright Commitment in Azure OpenAI Service products, we deemed rigorous copyright compliance testing necessary. Case Description: We conducted a structured red teaming exercise in Nov. 2024, with 42 participants from academic, industry, and government institutions. Four teams attempted to extract copyrighted content from GPT4DFCI across four domains: literary works, news articles, scientific publications, and access-restricted clinical notes. Teams successfully extracted verbatim book dedications and near-exact passages through various strategies. News article extraction failed despite jailbreak attempts. Scientific article reproduction yielded only high-level summaries. Clinical note testing revealed appropriate privacy safeguards. Discussion: The successful extraction of literary content indicates potential copyrighted material presence in training data, necessitating inference-time filtering. Differential success rates across content types suggest varying protective mechanisms. The event led to implementation of a copyright-specific meta-prompt in GPT4DFCI; this mitigation has been in production since Jan. 2025. Conclusion: Systematic red teaming revealed specific vulnerabilities in generative AI copyright compliance, leading to concrete mitigation strategies. Academic medical institutions deploying generative AI should implement continuous testing protocols to ensure legal and ethical compliance.

cs.CY↗