arXiv subjects

Jiawei Yang

Publications and source records attributed to Jiawei Yang.

At least 19 recordsLinked to original sources

arXiv2026

ZeroGR: Filtering-First Constrained Decoding for Generative Recommendation

Generative recommendation is gaining traction in production, retrieving items by generating their identifiers one token at a time. The generated sequences must identify relevant catalog items, so constrained decoding couples relevance scoring with validity filtering. Because most possible sequences are invalid, scoring before filtering wastes computation on candidates that will be discarded. Our evaluation shows that filtering first can avoid this work with negligible impact on recommendation quality. We present ZeroGR, a filtering-first constrained decoding system for generative recommendation. ZeroGR uses a step-sliced filtering index to reduce GPU memory use by tailoring data representations to each decoding step. Its execution engine adopts shared-slot execution to accelerate data exchange across stages and hybrid projection to adapt computation to workload patterns, thereby reducing execution overhead. During index refreshes, ZeroGR applies layout-guided streaming construction to speed up updates and limit serving interference. Extensive evaluation shows that ZeroGR accelerates the filtering and scoring pipeline by up to $9.9\times$ over representative baselines while reducing GPU index memory use and accelerating index reconstruction.

cs.DB↗

arXiv2026

FedHQ: Hybrid Runtime Quantization for Federated Learning

Federated Learning (FL) must balance training efficiency and model accuracy across heterogeneous clients. Post-training quantization (PTQ) avoids the extra computation of quantization-aware training (QAT), but can incur greater accuracy loss. We propose FedHQ, a hybrid quantization framework that assigns PTQ or QAT according to clients' hardware and data characteristics. Hardware-based speed estimates and local data-distribution indicators guide coarse-grained global initialization, followed by fine-grained training-based adjustment of selected client strategies. Experiments on three image-classification datasets show that FedHQ achieves up to 2.47x training speedup over FedAQT and up to 11.15 percentage points higher accuracy than FedFQ.

cs.LG↗

arXiv2026

Projective Decomposition and Intersection Cohomology of Rigid Spaces

We prove a rigid analytic analogue of the Beilinson--Bernstein--Deligne decomposition theorem for projective morphisms over $p$-adic fields, with $\mathbf{Q}_\ell$-coefficients for every prime $\ell$, including $\ell=p$. As an application, we show that the $p$-adic intersection cohomology of every proper rigid analytic space is de Rham. We also establish an independent criterion for de Rham intersection cohomology with local coefficients in terms of the dimension of the boundary, yielding further results for curves, surfaces, and finite covers. For finite coefficients of characteristic different from $p$, we obtain isotropic analytic upper bounds for singular support and prove that singular support is analytic and Lagrangian for complexes supported in dimension at most one.

math.AG↗

arXiv2026

AdaBool: Sparse Boolean Matrix Processing with Adaptive Data Representation

Many data processing tasks revolve around determining whether logical relationships exist between two sets of entities, a pattern that can be naturally modeled as Sparse Boolean Matrix Multiplication (SpBMM). Under Boolean semantics, matrix representations no longer need to store numerical values and instead focus on encoding the positions of non-zero elements, making it highly amenable to data compression for substantially lower memory access overhead. The key challenge, however, is that the most effective compression depends heavily on sparsity distribution: different methods perform best at different sparsity levels, while real-world matrices often contain sharply varying sparsity across regions. Consequently, forcing the entire matrix into any fixed, monolithic representation leaves considerable optimization opportunities untapped. We present AdaBool, a SpBMM framework designed to reconcile this tension through an adaptive data representation. AdaBool partitions matrices into fine-grained element blocks and represents each block with the encoding best matched to its local sparsity. On top of this adaptive encoding, AdaBool introduces a heterogeneous GPU execution model that groups blocks by encoding type and maps them to matching processing paths, enabling efficient execution and scheduling across GPU resources. We implement AdaBool end to end with low-level system optimizations, evaluate it on both production and public workloads, and show that it outperforms representative baselines by over an order of magnitude.

cs.DB↗

arXiv2026

FiberTune: Preserving Action-Fiber Visual Residuals in Vision-Language-Action Fine-Tuning

Action-supervised fine-tuning of vision-language-action (VLA) policies can compress visual structure that is weakly constrained by the action objective. We formalize this as residual visual collapse along local action fibers and propose FiberTune, a training-time objective that preserves teacher-structured visual residuals without adding inference-time overhead. FiberTune uses an online action probe to filter action-predictive directions from intermediate visual-token representations. It preserves the residuals through centered directional alignment with a frozen visual teacher and effective-rank regularization. Under identical training conditions, FiberTune improves over task-loss-only fine-tuning in every one of six simulation settings spanning two benchmarks and two architectures ($π_{0.5}$ and OpenVLA-OFT), as well as on physical SO-101 pick-and-place; representative gains include +10.7 pp SR(5) on long-horizon CALVIN ABC$\rightarrow$D and physical SO-101 task success rising from 72.7% to 78.1%. Representation analyses support the preservation of visual information in the residuals.

cs.CV↗

arXiv2026

PhysWAM: Physically Consistent World Action Model for Autonomous Driving

World-action models (WAMs) jointly predict how a scene will evolve and how an agent should act, however joint generation alone does not necessarily impose a shared geometric constraint on these predictions. We present PhysWAM, a unified world-action model for autonomous driving that co-denoises multiview video, metric depth, and ego motion within a single flow-matching transformer. To ground world and action generation in measured scene geometry, we introduce Coupled Point Projection (CPP) that unprojects the generated depth into 3D points, transforms them using the generated $\mathrm{SE}(3)$ ego motion, and minimizes their distance to LiDAR points transformed using the recorded ego motion. This geometric constraint promotes physical consistency with the measured scene by jointly supervising generated depth and motion alongside their standard flow-matching objectives. At inference, trajectory selection relies only on a simple label-free consensus rule, with no learned scorer or simulator feedback. We evaluate PhysWAM across NAVSIM v1 and v2 planning, zero-shot closed-loop transfer, and future video and metric-depth prediction. Despite PhysWAM's simple selection procedure, it achieves strong planning performance and transfers zero-shot to unseen driving environments. It also generates accurate metric depth and temporally coherent video, with CPP improving both planning and depth prediction. Together, these results demonstrate that the geometric relationship between scene depth and ego motion provides a direct way to couple world and action generation within a simple unified model.

cs.RO↗

arXiv2026

FuseReg: Regularizing Layer Fusion Mitigates the Reconstruction-Generation Gap in Representation Autoencoders

Representation autoencoders (RAEs) reuse features from a pretrained visual encoder as reconstruction and diffusion latents, integrating strong visual representations into image generation. However, RAEs still need to decide which encoder layers form the shared latent space for the generator and pixel decoder. This choice involves a trade-off: shallower layers tend to preserve fine pixel details better, while deeper layers tend to yield better generation metrics. A fixed heuristic layer fusion therefore couples two stages that benefit from different information. We introduce FuseReg, which replaces heuristic feature selection with training over random subsets of encoder layers. We theoretically analyze the underlying mechanism: subset sampling preserves the full-layer latent mean in expectation while explicitly penalizing sensitivity to cross-layer disagreement. On ImageNet-256 with DINOv3-L, a single FuseReg decoder reconstructs from full, sparse, and single-layer fusions without retraining, achieving higher PSNR than decoders specialized to fixed fusions. This flexibility also benefits generation: decoder replacement alone reduces unguided gFID by 27% with an unchanged RAEv2 DiT-XL generator. Applying FuseReg to both stages also reduces unguided gFID by 29% on DiT-Base. The reconstruction and generation benefits also extend to other encoder families. FuseReg narrows the reconstruction-generation gap without additional training cost or architectural changes.

cs.CV↗

arXiv2026

Steinness of the Basic $\mathrm{GL}_n$ Local Shimura Tower in Odd Rank

We prove Steinness over $\breve{\mathbf{Q}}_p$ for the basic local Shimura tower attached to $G=\mathrm{GL}_n$, $μ=(1,1,0^{n-2})$, with basic Newton slope $2/n$, for every odd $n\ge3$, at every finite level and for every prime $p$. The proof proceeds by constructing global analytic functions from the crystalline period map, proving compactness of their sublevel sets via a perfectoid normalization of Tate lattices, and showing that the resulting function map to affine space is finite. The key geometric inputs are the Fargues--Fontaine realization of the universal cover, a normalized determinant on the rank-two locus, and an integral PEL realization over $W(\overline{\mathbf{F}}_p)$ used to produce bounded global Hodge generators. The integral constructions include the prime $p=2$.

math.NT↗

arXiv2026

The Arthur-Packet Support Equality for Real Reductive Groups

Let $ψ$ be a real Arthur parameter and let $ψ_2$ be the unipotent parameter in a fixed Jordan decomposition. Adams, Ionov, Mason-Brown, and Vogan proved that the microlocal packet of $ψ$ is contained in the support of the two-step Jordan induction of the packet of $ψ_2$, and conjectured equality. We prove the reverse inclusion. The argument first passes to a sufficiently positive translate, where the relevant connected components of the ABV spaces have a common flag-variety model. Perverse devissage reduces the problem to a singular-support implication, which follows from the incidence calculation underlying the AIMV microstalk comparison. Coherent continuation for individual packet members then returns the result to the original parameter.

math.RT↗

arXiv2026

OneLA: Scaling Linear-Attention Decoding to Large Beams in Generative Recommendation

Generative recommendation (GR) relies on large-beam decoding to generate hundreds of candidate items, creating a new scaling challenge for recurrent linear attention. Existing linear attention serving systems either materialize a full recurrent state for every beam or repeatedly replay shared history, incurring substantial memory and traffic overhead. To address this, we present OneLA, a linear-attention decoding framework that exploits the shared prompt and short divergent suffixes of GR workloads. Specifically, OneLA represents all beam states using a single shared prompt-derived state and compact, append-only records of their divergent transitions. Using this representation, OneLA computes only the state information required at each decoding step, without reconstructing a full recurrent state for every beam. Furthermore, OneLA uses a lightweight ancestry index to track the transition records that make up each beam's history, allowing beams to be updated without moving or copying existing records. A fused GPU kernel further reuses the shared state across beams. Our analysis shows that OneLA achieves 1.54-2.46x end-to-end decode speedups while substantially reducing recurrent-state memory use and data movement.

cs.AI↗

arXiv2026

JigShape: Evaluating Visual-Geometric Reasoning in VLMs through Jigsaw Puzzles

Jigsaw puzzle solving requires jointly reasoning about visual content and geometric constraints, yet existing benchmarks use rectangular cuts that create ambiguous ground truth in texture-repeated regions. We introduce \textit{\ours{}}, a benchmark with tab-and-blank interlocking pieces where geometric constraints provide strong local compatibility requirements that, combined with visual content, yield unambiguous ground truth. Across 95K instances at four grid densities (4$\times$4 to 16$\times$16), we find that \textbf{zero-shot VLMs largely lack geometric reasoning}: only one of five frontier models (GPT-5.5) exceeds random baseline on 4$\times$4 puzzles, while all others perform at chance level. While supervised fine-tuning achieves $>$97\% on 4$\times$4, \textbf{all models collapse on larger grids}: GPT-5.5 drops from 70\% to near-random on 8$\times$8, and even fine-tuned models fall below 5\% on 12$\times$12. This ``scaling cliff'' suggests current architectures cannot maintain consistent constraint satisfaction as the number of pieces increases. \ours{} establishes scalable geometric reasoning as an open challenge for vision-language models.

cs.CV↗

arXiv2026

Token Radius Attention for Efficient Video Generation

Video Diffusion Transformers (VDiTs) enable high-fidelity generation but incur quadratic cost from dense 3D self-attention. Existing head- and block-level sparse methods share computation budgets across queries, overlooking token-specific attention demand. We observe that retained density varies across queries yet correlates log-linearly with attention entropy, while dominant interactions form query-centered neighborhoods with token-dependent radii. Based on these findings, we propose Token Radius Attention (TRA), a training-free framework that maps query entropy to an analytic token budget and converts it into a temporally decayed radius without explicit key ranking. Fused entropy extraction, warm-up reuse, and block-sparse mask construction further reduce overhead. Across seven Wan2.1, Wan2.2, and HunyuanVideo T2V/I2V configurations, TRA retains only 9-19% of attention interactions and achieves 1.56x-2.05x speedup with competitive generation quality. Code is available at https://github.com/IF-LAB-PKU/Token-Radius-Attention.

cs.CV↗

arXiv2026

Degradation of Proton Exchange Membrane Water Electrolyzers in Accelerated Stress Tests of Dynamic Load Cycling

Proton exchange membrane water electrolysis (PEMWE) is a promising technology for harnessing intermittent renewable energy. This study experimentally investigates the degradation of PEMWE under fluctuating power supply, characterized by dynamic load cycling under accelerated stress test (AST). We focus on the effects of key parameters of dynamic loading, including the peak voltage and cycling frequency, in a set of simplified AST protocols designed to represent selected features of dynamic load fluctuations associated with variable renewable-energy operation. The results unveil the intricate relationship between the structural characteristics of the catalyst layer and the electrochemical performance. An elevated peak voltage accelerates the degradation in the initial phase of the AST. However, a low cycling frequency can mitigate the degradation by limiting the rise in various resistance forms, whereas a higher cycling frequency exacerbates the degradation primarily by increasing mass transport resistance, suggesting a frequency-sensitive deterioration of the system's components.

physics.chem-ph↗

arXiv2026

Conformalized Rate-Adaptive Sensing

Many high-resolution imaging systems face the same fundamental question: when have enough measurements been collected to reconstruct an image accurately? We develop Conformalized Rate-Adaptive Sensing (CoRAS), a method that adaptively chooses an acquisition or compression rate for each image while keeping the reconstruction error below a target level with high probability. As measurements are collected, an image reconstruction model gradually recovers the true image, producing a reconstruction path over acquisition rates. CoRAS uses this path up to an early decision time to estimate the target stopping time, defined as the first time at which the reconstruction error falls below the target level. It then calibrates this estimate using images with similar early reconstruction behavior, producing an upper bound on the stopping time with marginal and approximate conditional coverage guarantees. Experiments on image datasets show that CoRAS attains the target stopping-time coverage, uses fewer measurements on average than fixed-rate stopping rules, and assigns more measurements to images that are harder to reconstruct.

stat.ML↗

arXiv2026

RoleCDE:Benchmarking and Mitigating Role-Alignment Trade-offs in Role-Playing Agents

Role-playing agents(RPAs) are widely used to steer large language models(LLMs) toward role-consistent behavior, yet existing benchmarks mainly evaluate surface-level fidelity and offer limited insight into decision making under role-alignment value conflicts. To address this gap, we introduce RoleCDE, the first benchmark designed to evaluate RPAs under structured conflicts between role-specific values and alignment-oriented constraints. RoleCDE formulates role-aware decision making as cognitive dilemma scenarios, jointly evaluating role-scenario grounding, value conflict resolution, and decision tendencies. The benchmark is constructed at scale, covering approximately 8k diverse role profiles and scenarios and nearly 24k dilemma instances across three difficulty levels and eight role categories. Evaluation of several mainstream LLMs reveals a "Role Value Decoupling" phenomenon, where agents systematically default to alignment-and morality-consistent decisions rather than role-specific values when the two conflict, even under explicit role conditioning. This behavior is largely invariant to dilemma difficulty but varies substantially across role categories. We further show that RoleCDE-based fine-tuning effectively mitigates this decoupling by improving value trade-off reasoning, while preserving general role-playing fidelity and general reasoning performance. Code is available at: https://github.com/rabbitrose/RoleCDE.

cs.AI↗

arXiv2026

Pyramid Forcing: Head-Aware Pyramid KV Cache Policy for High-Quality Long Video Generation

Autoregressive video generation enables streaming and open-ended long video synthesis, but still suffers from long-term degradation caused by accumulated errors. Existing KVCache strategies usually apply unified historical-frame retention, implicitly assuming homogeneous historical dependencies across attention heads. We revisit historical-frame attention and reveal three distinct head types: Anchor Heads require broad long-range context, Wave Heads exhibit periodic temporal dependencies, and Veil Heads focus on initial and adjacent frames. Based on this finding, we propose Pyramid Forcing, a head-aware pyramidal KVCache framework that identifies head types offline, assigns behavior-specific cache policies, and supports heterogeneous cache lengths via efficient ragged-cache attention. Experiments on Self Forcing and Causal Forcing show that Pyramid Forcing consistently improves long-horizon generation quality on VBench-Long, increasing the 60-second Self Forcing score from 77.87 to 81.21 while enhancing motion dynamics, visual fidelity, and semantic consistency. Project: https://if-lab-pku.github.io/Pyramid-Forcing/.

cs.CV↗

arXiv2026

Representation Fréchet Loss for Visual Generation

We show that Fréchet Distance (FD), long considered impractical as a training objective, can in fact be effectively optimized in the representation space. Our idea is simple: decouple the population size for FD estimation (e.g., 50k) from the batch size for gradient computation (e.g., 1024). We term this approach FD-loss. Optimizing FD-loss reveals several surprising findings. First, post-training a base generator with FD-loss in different representation spaces consistently improves visual quality. Under the Inception feature space, a one-step generator achieves0.72 FID on ImageNet 256x256. Second, the same FD-loss repurposes multi-step generators into strong one-step generators without teacher distillation, adversarial training or per-sample targets. Third, FID can misrank visual quality: modern representations can yield better samples despite worse Inception FID. This motivates FDr$^k$, a multi-representation metric. We hope this work will encourage further exploration of distributional distances in diverse representation spaces as both training objectives and evaluation metrics for generative models.

cs.CV↗

arXiv2026

A micromechanical frequency reference with parts-per-trillion holdover stability

Microelectromechanical (MEMS) resonators are widely used in timekeeping applications, and recent advances in fabrication, materials, and encapsulation technology have advanced their potential as high stability frequency references. However, for holdover applications that require the highest levels of long-term frequency stability, compact vapor atomic clocks remain dominant. In this work, we demonstrate a 268 MHz MEMS clock that achieves record fractional frequency stability of ~8 parts-per-trillion at an averaging time of 8 hours, competitive with chip-scale atomic clocks. We achieved this using a single-crystal silicon electrostatic resonator that has no currently known intrinsic drift mechanism and is protected from the environment with a wafer-level encapsulation. We specifically identify gain variations in the sustaining electronics as the dominant limitation in conventional phase-locked oscillator architectures -- originating from temperature sensitivity and drifts in the electronic components -- and overcome this by implementing a frequency-locked loop architecture based on dual-frequency resonance tracking (DFRT). This novel approach removes the specific gain of the supporting electronics as a frequency determining variable in the oscillator. When combined with dual-mode tracking and ratiometric temperature stabilization of the resonator, this approach enables a dramatic enhancement to long-term frequency stability and establishes gain-insensitive DFRT locking as a general paradigm for high-stability MEMS clocks.

physics.app-ph↗