arXiv subjects

Kewei Zhang

Publications and source records attributed to Kewei Zhang.

At least 19 recordsLinked to original sources

arXiv2026

PyRUA-DataFarm: LLM Agent as a Cost-Efficient Teleoperator for Real-Ready Robot Data

Teleoperation remains the default way to collect robot demonstrations, but every episode costs robot and operator time, and every new task starts from scratch. We ask whether a coding agent can take over the teleoperator's job. In our DataFarm, the agent receives a data request and one calibration episode of the robot in the target scene and rebuilds the task in simulation: it writes the scene, an expert controller, and a success judge. The simulator supplies exact actions, and a video model conditioned on real images renders each rollout as photographic video, generating all camera views together consistently. The output is real-ready data: synchronized camera views labelled with the robot's actions, on which a policy can be trained directly and deployed on the real robot. Independent judges certify the expert and every rendered episode before admission. On five real-robot tasks, policies trained only on farm data match policies trained on as many teleoperated episodes (93 vs. 84 successes in 100 trials), and under an unseen tablecloth and coloured light they keep 87 successes where the teleoperation-trained policies fall to 46. Adapting a task is fully automatic and costs \$21 of API usage on average, paid once per task, plus \$5 for a render contract; every further episode costs \$0.80 on average for simulation, rendering, and review, with no robot or operator. One hour of three-view data for a task costs \$107 on average, these costs included, slightly below a quote of \$117 for teleoperation in China and below \$290 in labour alone in North America.

cs.RO↗

arXiv2026

Understanding entropy using Aubin--Yau deformations

In this paper we investigate three types of entropy that frequently appear in the research of Kähler geometry: the classical Kähler entropy, the radial entropy and the non-Archimedean entropy. We give a unifying interpretation for these entropies by showing that they can all be expressed as the derivative of the Monge--Ampère energy along certain Aubin--Yau type deformations. Such considerations also lead to a novel algebraic criterion for the existence of constant scalar curvature Kähler metrics.

math.DG↗

arXiv2026

AV-FABLE: Audio-Visual Manipulation Forensics in Long-Form Videos

Localized AI manipulations in long-form videos often affect brief events within otherwise authentic content, making video-level classification insufficient for forensic analysis. Existing audio-visual deepfake benchmarks largely focus on short talking-head clips, while long-form temporal forensics remains predominantly visual. We introduce AV-FABLE, a dual-timeline benchmark for sparse manipulations in open-scene long-form audio-visual videos. AV-FABLE contains 16,650 pristine--manipulated twin pairs (998 hours) created with 27 editing configurations, with separate audio and visual timelines that enable modality-specific temporal localization and expose errors hidden by joint-timeline evaluation. On AV-FABLE, zero-shot prompting of audio-visual large language models (AV-LLMs) yields near-chance detection and poor temporal localization. In contrast, supervised readouts of their frozen perception features recover strong forensic signals, suggesting manipulation evidence is encoded in these representations but not effectively exposed through language outputs. Motivated by this observation, we propose NFSR(Native Forensics via Separated Readouts), which reads frozen AV-LLM features with separate audio and visual temporal decoders. NFSR achieves the best audio-visual localization performance in-domain and across all three out-of-distribution settings, including transfer to held-out videos from an external long-video forensic benchmark. AV-FABLE, code, and model weights will be publicly released.

cs.CV↗

arXiv2026

Fewer Tokens, Better Action: GPT-6 Astra Robot Agents with 14% Higher Success Rate but 65% Fewer Tokens

Vision-language model (VLM) agents can control robots through visual feedback and action primitives, but repeated LLM calls and redundant observations incur substantial token overhead. We introduce PyRUA-Lean, an interactive code-execution framework in which the agent composes robot primitives, including vision-language-action (VLA) policies, into Python cells that check and retry locally and return only the images and state they request. Across 700 simulated task instances from LIBERO-PRO, RoboTwin 2.0, and RoboCasa365, it raises success from 63.1% to 71.7% over tool calling with the same GPT-6 Astra planner and primitives, using 49% fewer LLM calls and 65% fewer input tokens on jointly solved instances. On RoboDojo, without any VLA policy or privileged information, it reaches 51.6% average success rate (Score 58.2), 19.4 points above the top entry of the public leaderboard. Based on the PyRUA-Lean framework, we train PyRUA-Lean Coder v0.1-4B, an open 4B model, with GRPO on 45 LIBERO-PRO instances; it improves from 18.8% to 44.1% on unseen task instances, surpassing the Qwen3.5-27B base model.

cs.CV↗

arXiv2026

SAGE: Pushing 2-Bit Vector-Quantized LLMs toward Real-World Usability

Extreme low-bit weight-only quantization is an effective approach to reducing the deployment costs of large language models. Below 3 bits, scalar quantization often suffers severe accuracy loss, while vector quantization (VQ) better preserves accuracy through more expressive joint encoding. However, existing VQ methods still exhibit substantial degradation on real-world tasks, and two design challenges remain underexplored: (1) protecting sensitive weight channels within jointly encoded vectors; and (2) allocating a fixed bit budget among representation components, such as direction and radius, to suit different weight matrices. To address these challenges, we propose SAGE: Split Adaptation and Grouping by Energy for weight-only vector quantization. First, Energy-Stratified Grouping (ESG) couples channel grouping with activation-weighted fitting, grouping channels from different energy tiers within each subvector to prioritize sensitive channels. Second, Adaptive Bit Split (ABS) balances direction-codebook capacity and radius precision per matrix, selecting the split with the lowest activation-weighted reconstruction error under a fixed budget. Across benchmarks covering knowledge, mathematics, code generation, instruction following, and tool use, SAGE achieves state-of-the-art performance in the 2-bit quantization setting. On Qwen3-30B-A3B, it retains 95.3% of the BF16 model's average benchmark score. With our custom-designed CUDA kernel, SAGE delivers up to a 2.55-fold non-streaming end-to-end speedup over FP16. The code is available at https://github.com/Zincwolf/sage.

cs.AI↗

arXiv2026

ContourVLA: A Closed-Loop Perception-Action Contour Policy for Generalized Referring Expression Segmentation

Generalized referring expression segmentation (GRES) requires dynamically balancing high-level semantics for identifying a variable number of language-specified referents with fine-grained visual evidence for precise boundary delineation. This requirement challenges existing cascaded vision-language architectures, which typically rely on static feature interfaces and single-pass mask prediction, limiting adaptive perception and geometric correction. We introduce ContourVLA, a vision-language-action policy that recasts GRES as a closed-loop visuomotor process, in which editable contours serve as explicit policy states that condition multimodal perception and are updated by geometric action chunks. Evolution-Aware Semantic Scheduling (EASS) couples contour-guided bidirectional boundary sampling with state-conditioned routing of multilevel multimodal features, adapting perception to each contour state. Following supervised initialization, Dustbin-Augmented Entropic Credit Transport GRPO (DECT-GRPO) jointly optimizes discrete grounding and continuous contour actions with instance-level credits. Its rollout rewards and credits are derived from soft prediction-target correspondences that account for false positives and missed targets. ContourVLA improves gIoU over the strongest evaluated baselines by 8.7, 2.8, and 2.7 points on gRefCOCO val, testA, and testB, respectively, and achieves the highest mIoU across all eight RefCOCO, RefCOCO+, and RefCOCOg splits.

cs.CV↗

arXiv2026

SPHQuant: Efficient extreme low bit weight quantization for Vision-Language Models

Recent foundation models are moving toward native multimodal Vision-Language Models (VLMs), making VLMs a central form of next-generation foundation models. However, their large language backbones make edge deployment difficult due to high memory footprint and memory-bound autoregressive decoding. Weight-only post-training quantization is a practical solution, but pushing VLMs to extreme low bit-widths remains challenging: existing rotation-free methods suffer from outliers at 2-3 bits, while rotation-based methods improve accuracy at the cost of additional runtime overhead. We propose SPHQuant, a rotation-free spherical weight-only quantization framework for VLMs. Instead of quantizing weights directly in Cartesian coordinates, SPHQuant decomposes each 8D weight vector into coordinate signs, radius, and a positive unit direction. This representation isolates outlier magnitude into the radius while keeping directions bounded and statistically regular. Based on this insight, SPHQuant allocates extra precision to the radius to mitigate accuracy degradation induced by outliers. It further uses a compact positive-direction codebook and fine-tunes codebook entries through angular parameterization to preserve the unit-sphere constraint. We also design a hardware-friendly GEMV kernel that keeps the direction codebook small enough for shared-memory lookup and packs radial bits efficiently. Experiments show that SPHQuant matches the performance of state-of-the-art extreme low-bit quantization methods while improving decode throughput over QTIP by 30.3% on RTX A6000. Code will be released in https://github.com/Pushazf/SPHQuant.

cs.CV↗

arXiv2026

Demailly-Kollár continuity on klt pairs, and applications to alpha and delta invariants

We establish a Demailly-Kollár type continuity theorem for plurisubharmonic functions with respect to adapted measures on normal complex analytic klt pairs. As applications, we prove the equality of the analytic and divisorial versions of the alpha and delta invariants on compact normal Kähler klt pairs, thereby completing a program initiated by the second author. We further show that the two local alpha invariants introduced by Guedj and Trusiani for an isolated log terminal singularity coincide and that their common value is the $n$th root of Li's normalized volume. These identities have geometric consequences. The equality for the delta invariant yields a Yau-Tian-Donaldson type divisorial criterion for the solvability of twisted Kähler-Einstein equations in big cohomology classes, without a semipositivity assumption on the twist. The local alpha identity determines algebraically the critical exponent governing the existence of positively curved KE metrics near an isolated log terminal singularity, thus confirming a prediction of Guedj and Trusiani.

math.DG↗

arXiv2026

Quantized Volume Comparison for Fano Manifolds, II

In this note, the second author's quantized volume comparison conjecture is solved: If $X$ is a $K$-semistable Fano manifold of dimension $n$, then for every integer $m\geq1$, \[ h^0(X,-mK_X)\leq h^0(\mathbb P^n,-mK_{\mathbb P^n}) =\binom{n+m(n+1)}{n}, \] and equality for one $m$ characterizes projective space. Somewhat surprisingly, the same statement actually holds whenever $T_X$ is slope semistable with respect to $-K_X$. The idea is to apply a jet-dimension counting trick to a filtration of subsheaves induced by $H^0(X,-mK_X)$. Then the slope semistability condition yields the desired dimension bound.

math.AG↗

arXiv2026

A counterexample to the bounded mass property

A compact complex manifold has the bounded mass property if, for one (equivalently, every) Hermitian form $ω$, the masses $\int_X(ω+\mathrm{dd}^{\mathrm{c}}φ)^n$ are uniformly bounded over all smooth $φ$ with $ω+\mathrm{dd}^{\mathrm{c}}φ>0$. We prove that this property fails on the Hopf threefold $(\mathbb C^3\setminus\{0\})/\langle z\mapsto\mathrm{e}^{-1}z\rangle$, answering a question of Boucksom--Guedj--Lu.

math.CV↗

arXiv2026

The sharp volume gap for Kähler manifolds with positive Ricci curvature

We prove a sharp volume gap estimate: if an $n$-dimensional compact Kähler manifold $(X, ω)$ satisfies $\mathrm{Ric}(ω)\ge (n+1)ω$ and $X\not\cong \mathbb{P}^n$, then $\mathrm{vol}(X, ω)\le \frac{2n^n}{(n+1)^n}\mathrm{vol}(\mathbb{P}^n,ω_{\mathrm{FS}})=\frac{2^{n+1} \, π^n \, n^n}{(n+1)^n}$. Moreover $\mathrm{vol}(X, ω)= \frac{2^{n+1} \, π^n \, n^n}{(n+1)^n}$ occurs if and only if $(X, ω)$ is biholomorphically isometric to the Kähler-Einstein metric on the quadric hypersurface $Q^n$ or on the product $\mathbb{P}^1\times \mathbb{P}^{n-1}$. We also obtain sharp volume gap estimates for K-semistable toric log Fano pairs.

math.DG↗

arXiv2026

HumanScale: Egocentric Human Video Can Outperform Real-Robot Data for Embodied Pretraining

Embodied foundation models are expected to benefit from data scaling like large language models, but face a much tighter data bottleneck. Teleoperated real-robot trajectories remain the dominant pretraining source due to their precise action supervision and embodiment alignment, yet their scalability is limited by high collection cost, acquisition difficulty, and low behavioral and environmental diversity. These limitations have sparked interest in egocentric human video as a scalable, substantially lower-cost, and more diverse alternative for embodied model pretraining. However, its effectiveness compared to teleoperated real-robot data remains underexplored. To address this question, we conduct a systematic study comparing egocentric human video and teleoperated real-robot trajectories as pretraining data sources for embodied foundation models, under fixed post-training and validation protocols. Surprisingly, we find that egocentric data, when processed through a carefully designed filtering and labeling pipeline, is not merely a viable substitute for model pretraining but can lead to superior performance. With the same amount of pretraining data, models pretrained on egocentric data achieve a 24% lower validation loss on real-robot action prediction, as well as 52.5% and 90% higher success rates on in-distribution and out-of-distribution real-robot task execution, respectively. This finding verifies a scalable paradigm for embodied foundation models: pretrain on egocentric human video to learn diverse world representations, then adapt with a small amount of labeled real-robot data for action-space alignment. We hope this study encourages broader exploration of egocentric data and offers guidance for data quality assessment before costly robot data collection.

cs.CV↗

arXiv2026

Fast-dDrive: Efficient Block-Diffusion VLM for Autonomous Driving

End-to-end autonomous driving via Vision-Language-Action (VLA) models demands a precarious balance between high-fidelity trajectory planning and efficient inference. Existing paradigms typically fall short: autoregressive (AR) VLAs are memory-bandwidth-bound on edge hardware and prone to exposure-bias drift, while full-sequence diffusion models preclude KV-cache reuse and suffer from "logical leakage" that violates the fundamental perceive-then-plan causality. We present Fast-dDrive, a block-diffusion VLA that performs bidirectional refinement within semantic units while enforcing strict causal ordering across them. Leveraging the observation that driving VLAs often emit structured JSON-like outputs, Fast-dDrive freezes structural tokens into a section scaffold and employs a section-aware training recipe that prioritizes safety-critical planning. We further introduce Scaffold Speculative Decoding to achieve AR-equivalent quality at significantly higher throughput. Finally, we propose a low-overhead test-time scaling scheme: by forking $N$ stochastic trajectory rollouts from a single shared-prefix KV cache and averaging them, we effectively suppress prediction variance at a fractional computational cost. Empirical results demonstrate that Fast-dDrive redefines the speed-accuracy frontier for driving agents. On the WOD-E2E test set, Fast-dDrive achieves SOTA ADE@3s and ADE@5s, alongside the highest RFS among diffusion-based VLAs; on nuScenes, it reduces average L2 error to $0.32$m (a $22\%$ improvement). When integrated with SGLang, our framework delivers $12\times$ throughput speedup over the AR baseline, narrowing the gap between high-capacity VLAs and the efficiency demands of real-time on-vehicle deployment.

cs.CL↗

arXiv2026

MHLA: Restoring Expressivity of Linear Attention via Token-Level Multi-Head

While the Transformer architecture dominates many fields, its quadratic self-attention complexity hinders its use in large-scale applications. Linear attention offers an efficient alternative, but its direct application often degrades performance, with existing fixes typically re-introducing computational overhead through extra modules (e.g., depthwise separable convolution) that defeat the original purpose. In this work, we identify a key failure mode in these methods: global context collapse, where the model loses representational diversity. To address this, we propose Multi-Head Linear Attention (MHLA), which preserves this diversity by computing attention within divided heads along the token dimension. We prove that MHLA maintains linear complexity while recovering much of the expressive power of softmax attention, and verify its effectiveness across multiple domains, achieving a 3.6\% improvement on ImageNet classification, a 6.3\% gain on NLP, a 12.6\% improvement on image generation, and a 41\% enhancement on video generation under the same time complexity.

cs.CV↗

arXiv2026

Transcendental Okounkov bodies

We show that the volume of transcendental big $(1,1)$-classes on compact Kähler manifolds can be realized by convex bodies, thus answering questions of Lazarsfeld-Mustaţă and Deng. In our approach we use an approximation process by partial Okounkov bodies together with properties of the restricted volume, and we study the extension of Kähler currents, as well as the bimeromorphic behavior of currents with analytic singularities. We also establish a connection between transcendental Okounkov bodies and toric degenerations.

math.DG↗

arXiv2025

DOVE: Efficient One-Step Diffusion Model for Real-World Video Super-Resolution

Diffusion models have demonstrated promising performance in real-world video super-resolution (VSR). However, the dozens of sampling steps they require, make inference extremely slow. Sampling acceleration techniques, particularly single-step, provide a potential solution. Nonetheless, achieving one step in VSR remains challenging, due to the high training overhead on video data and stringent fidelity demands. To tackle the above issues, we propose DOVE, an efficient one-step diffusion model for real-world VSR. DOVE is obtained by fine-tuning a pretrained video diffusion model (i.e., CogVideoX). To effectively train DOVE, we introduce the latent-pixel training strategy. The strategy employs a two-stage scheme to gradually adapt the model to the video super-resolution task. Meanwhile, we design a video processing pipeline to construct a high-quality dataset tailored for VSR, termed HQ-VSR. Fine-tuning on this dataset further enhances the restoration capability of DOVE. Extensive experiments show that DOVE exhibits comparable or superior performance to multi-step diffusion-based VSR methods. It also offers outstanding inference efficiency, achieving up to a 28$\times$ speed-up over existing methods such as MGLD-VSR. Code is available at: https://github.com/zhengchen1999/DOVE.

cs.CV↗

arXiv2025

A YTD correspondence for constant scalar curvature metrics

Given a compact Kähler manifold, to better understand Mabuchi's $K$ energy we introduce a family of $K^β$ energies, whose favorable properties are similar to those of the Ding energy from the Fano case. The construction uses Berman's transcendental quantization, and we show that the slope of the $K^β$ energies along test configurations can be computed using intersection theory. With these ingredients in place we provide a uniform Yau-Tian-Donaldson correspondence that characterizes the existence of a unique constant scalar curvature Kähler metric using test configurations. Combining our techniques with the non-Archimedean approach to $K$-stability pioneered by Boucksom--Jonsson, we show that the properness of the classical $K$ energy can be tested by checking its slope along a distinguished subclass of Chi Li-type models, called log discrepancy models, thus yielding another $G$-uniform Yau--Tian--Donaldson correspondence.

math.DG↗

arXiv2025

QuantDemoire: Quantization with Outlier Aware for Image Demoiréing

Demoiréing aims to remove moiré artifacts that often occur in images. While recent deep learning-based methods have achieved promising results, they typically require substantial computational resources, limiting their deployment on edge devices. Model quantization offers a compelling solution. However, directly applying existing quantization methods to demoiréing models introduces severe performance degradation. The main reasons are distribution outliers and weakened representations in smooth regions. To address these issues, we propose QuantDemoire, a post-training quantization framework tailored to demoiréing. It contains two key components. **First}, we introduce an outlier-aware quantizer to reduce errors from outliers. It uses sampling-based range estimation to reduce activation outliers, and keeps a few extreme weights in FP16 with negligible cost. **Second**, we design a frequency-aware calibration strategy. It emphasizes low- and mid-frequency components during fine-tuning, which mitigates banding artifacts caused by low-bit quantization. Extensive experiments validate that our QuantDemoire achieves large reductions in parameters and computation while maintaining quality. Meanwhile, it outperforms existing quantization methods by over **4 dB** on W4A4. Code is released at: https://github.com/zhengchen1999/QuantDemoire.

cs.CV↗