arXiv subjects

Zetian Song

Publications and source records attributed to Zetian Song.

10 recordsLinked to original sources

arXiv2026

RIFT: Residual-Informed Flow Transfer for One-Step Generative Video Compression

Recent advances in video generation have enabled generative video compression to reconstruct details lost during coding using pretrained spatiotemporal priors. However, a substantial gap remains between codec-induced degradation and the perturbation trajectories learned by flow-matching models. In this paper, we identify a compression-dependent directional mismatch that degradation-strength-based timestep calibration alone cannot resolve. Compressed latents remain strongly aligned across codec operating points, while their reconstruction residuals vary substantially in magnitude and direction. To bridge this gap, we propose RIFT, a Residual-Informed Flow Transfer framework for one-step generative video compression. At its core, RIFT introduces a Residual Geometry Inducer (RGI) that converts the pretrained flow prediction into a content-adaptive restoration update using decoder-visible codec cues. RGI integrates flow-timestep conditioning with a learned full-band residual-strength estimate and content-conditioned direction and gain modulation, aligning the orientation and magnitude of the update with the compression-dependent residual geometry. Extensive experiments on three public video compression benchmarks show that RIFT outperforms competing generative video compression methods. The code is publicly available on RIFT.

cs.CV↗

arXiv2026

Rethinking Video Storage for AI Training: Breaking Cross-Frame Dependencies for Efficient Random Access

Large-scale AI training increasingly relies on massive video datasets with random temporal sampling patterns. However, existing video storage paradigms are optimized for sequential playback and employ sophisticated inter-frame reference structures to maximize compression efficiency. When applied to the random-sampling workloads of AI training, these paradigms face substantial decoding-efficiency bottlenecks. Workarounds such as independent JPEG frames and short groups of pictures (GOPs) reduce decoding work at the cost of larger storage and transfer volumes. In this paper, we identify two sources of decode amplification: sequential reconstruction of unrelated frames and reconstruction of the reference chains required by sampled targets. We address both sources by co-designing the decoding schedule and video reference structure. Dependency-aware sparse decoding reconstructs only requested frames and their required references, skipping unrelated reconstruction without transcoding the source videos. Optional offline Dependency-Decoupled Random Access (DDRA) encoding further reduces reference reconstruction by restricting each interior frame to two bounding intra-coded anchors. Together, sparse decoding and DDRA retain inter-frame prediction while bounding reconstruction to at most three frames per requested target, independent of the intra period. Experiments on a large-scale robotics corpus show that, at comparable reconstruction quality, our method achieves $1.45\times$ the loading throughput of LeRobot v3 with 58.7% less media payload.

cs.CV↗

arXiv2026

GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed

Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens. We view grounding as visual evidence extraction: objects, locations, and spatial relations are jointly constrained by the image and query, yet their dependencies do not imply an intrinsic left-to-right generation order. This distinction makes bidirectional diffusion a natural fit, allowing spatial hypotheses to emerge in parallel and be jointly refined through iterative denoising. We introduce GroundAnything, a 4B-parameter grounding foundation model that reconciles fast parallel decoding with precise localization through blockwise denoising. Training combines grounding pretraining from public datasets and dedicated data engines, direct AR-to-diffusion conversion with joint AR and diffusion objectives, supervised fine-tuning, and GRPO-based reinforcement post-training. Across 30 grounding benchmarks, our autoregressive variant, GroundAnything-VLM, establishes a new overall state of the art among similarly sized models at 72.42%, remaining competitive with GPT-6 Astra (71.35%). With entropy-guided decoding, GroundAnything also surpasses the prior state of the art at this scale, averaging 61.75% versus 53.32% for the fast MTP-based LocateAnything model. We further explore decoding strategies, showing that an optional self-speculative mode achieves a $4.51\times$ speedup over the AR counterpart with a 0.74 percentage-point drop in COCO F1mIoU. Infrastructure experiments show that progressive inference optimizations translate parallel decoding into practical speedups. These support efficient visual grounding in latency-sensitive real-world systems.

cs.CV↗

arXiv2026

GroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual Primitives

Precise grounding matters. It specifies which object is the target and where that object is, even in clutter and for tiny objects, and it has to be fast enough for closed-loop control. Yet vision-language-action (VLA) and world-action models (WAMs) take perception from general-purpose vision-language and video-generation backbones, which still fail in these settings. We introduce GroundingPI, a 4B grounding foundation model that generates points and boxes as quantized coordinates in a shared vocabulary. Training combines multimodal and spatial pretraining, supervised fine-tuning, and reinforcement learning with GRPO, using supervision from public datasets and dedicated data engines. Against 44 baselines across 34 grounding benchmarks spanning 11 perceptual capabilities, GroundingPI establishes a new state of the art, averaging 73.68%, above the larger GPT-6 Astra (71.54%). As a downstream visual backbone, GroundingPI improves performance on robotic manipulation and autonomous driving. On RoboTwin 2.0, it outperforms every mainstream backbone we evaluate in all four out-of-distribution settings, by up to 24.8% relative to the strongest backbone. On RoboCasa-GR1, GroundingPI trained with 50% of the demonstrations outperforms those baselines trained with 75%. On nuScenes, used as the visual backbone, GroundingPI attains an average open-loop L2 error of 0.296 m. We systematically analyze GroundingPI's pretraining in scale and data composition. Downstream autonomous driving and robotic manipulation improve as the pretraining is scaled. Analyzing the data recipe across these 11 perceptual capabilities shows dense grounding's substantial benefits for both, and OCR's potential as a catalyst for perceptual learning. These results support grounding as a perceptual foundation, and dedicated perceptual pretraining as a promising direction for foundation models of physical intelligence.

cs.CV↗

arXiv2026

AUGUstus: Autonomous Driving VLA Model with Unified Generation and Understanding

Vision-language-action (VLA) models show strong potential for end-to-end autonomous driving by leveraging world knowledge and understanding to excel in long-tail scenarios. However, current VLAs have two limitations: 1) Degeneration into vision-action (VA) models. For low latency, regression/flow-based methods directly generate actions but underuse VLM understanding, causing unsafe, unexplainable actions. 2) Autoregressive high latency. Autoregressive methods sequentially generate chain-of-thought (CoT) and action tokens, fully leveraging VLM understanding, but high latency hinders deployment. To enable low-latency deployment while using CoT to assist action generation, we propose AUGUstus, an Autonomous-driving VLA model with Unified Generation and Understanding. We unify diffusion-based action generation and autoregressive CoT understanding in one model, where understanding only serves generation. Joint CoT supervision preserves VLM spatial understanding in shared representation for safer, more accurate trajectories. AUGUstus-mode orders action generation before CoT, so CoT supervises reliable actions during training while inference needs only action output, enabling low latency. AUGUstus-GRPO, a unified GRPO algorithm, jointly optimizes action and CoT, improving action performance. Experiments show AUGUstus achieves 94.6 PDMS on NAVSIMv1 and 90.7 EPDMS on NAVSIMv2, SOTA planning without scorer module. On nuScenes, it achieves average L2 error 0.27 m and average collision rate 0.05%, best among compared methods. To validate AUGUstus practicality, we train on in-house data with 512 NVIDIA H800 GPUs for ~2 days, evaluate on an internal 1,000 km dataset, and preliminarily deploy on a real vehicle, achieving 12 Hz operation. Project page: https://augustus-ad.github.io

cs.CV↗

arXiv2026

RoboAware: Learning to Coordinate Embodied Skills from Counterfactual Outcomes

Embodied coding agents can combine modular robot skills with frozen end-to-end policies, yet effective composition requires anticipating which policy family will succeed in the current physical state. We present RoboAware, which builds on coding agents' skill orchestration by learning only a state-conditioned responsibility coordinator from counterfactual outcomes. Inspired by the success of REPL, we propose the $P^5$ schema and formulate a hierarchical MDP based on it. $P^5$ organizes skills uniformly into five semantic stages, defining where responsibility can be compared. To address the lack of counterfactual branch outcomes in existing work, we introduce State-Locked Counterfactual Branching (SCB), which restores the same training state to generate and execute a code block from each admissible family, exposing outcomes that selected-branch experience leaves unobserved. Building on this, we propose Execution-Aware Learning (EAL), which combines Monte Carlo tree search with Q-learning to distill these outcomes into family-conditioned values. At deployment, the coordinator selects the policy family according to observable context, and the frozen coding agent generates the next local code block. Comprehensive single-episode evaluations on 100 tasks show that RoboAware reaches a 77.0% overall success rate, with SOTA averages of 90.0% on RoboSuite, 73.8% on diverse LIBERO-Pro task clusters, and 90.0% on challenging RoboTwin bimanual tasks, outperforming existing code-as-policy and VLA-harness baselines.

cs.RO↗

arXiv2026

L2D2-GS: Learning to Densify for Feedforward Dynamic Gaussian Scene Reconstruction

High-fidelity reconstruction of dynamic urban environments is a cornerstone of autonomous driving simulation and large-scale world modeling. While 3D Gaussian Splatting (3DGS) has established a new standard for real-time rendering, its reliance on expensive per-scene optimization limits scalability. Conversely, recent feedforward methods that infer Gaussian parameters offer faster speed but face fundamental bottlenecks: they are memory-prohibitive at high resolutions and struggle to fuse dense multi-view observations consistently. This paper presents L2D2-GS, a unified framework that reformulates generalizable reconstruction not as a one-shot regression, but as a robust iterative process of optimization and densification. To resolve the ambiguity of supervision in primitive generation, we propose a self-supervised densification policy that derives explicit reward signals from global reconstruction gains to guide local densification. Furthermore, we mitigate irreversible early-stage artifacts through a geometric regularization mechanism, utilizing reparameterization to constrain the optimization manifold and prevent convergence to poor local optima. Extensive experiments on the PandaSet and Waymo datasets demonstrate that our method achieves state-of-the-art reconstruction fidelity and strong zero-shot generalization, while using fewer primitives than competing baselines.

cs.CV↗

arXiv2025

TinySplat: Feedforward Approach for Generating Compact 3D Scene Representation

The recent development of feedforward 3D Gaussian Splatting (3DGS) presents a new paradigm to reconstruct 3D scenes. Using neural networks trained on large-scale multi-view datasets, it can directly infer 3DGS representations from sparse input views. Although the feedforward approach achieves high reconstruction speed, it still suffers from the substantial storage cost of 3D Gaussians. Existing 3DGS compression methods relying on scene-wise optimization are not applicable due to architectural incompatibilities. To overcome this limitation, we propose TinySplat, a complete feedforward approach for generating compact 3D scene representations. Built upon standard feedforward 3DGS methods, TinySplat integrates a training-free compression framework that systematically eliminates key sources of redundancy. Specifically, we introduce View-Projection Transformation (VPT) to reduce geometric redundancy by projecting geometric parameters into a more compact space. We further present Visibility-Aware Basis Reduction (VABR), which mitigates perceptual redundancy by aligning feature energy along dominant viewing directions via basis transformation. Lastly, spatial redundancy is addressed through an off-the-shelf video codec. Comprehensive experimental results on multiple benchmark datasets demonstrate that TinySplat achieves over 100x compression for 3D Gaussian data generated by feedforward methods. Compared to the state-of-the-art compression approach, we achieve comparable quality with only 6% of the storage size. Meanwhile, our compression framework requires only 25% of the encoding time and 1% of the decoding time.

cs.CV↗

arXiv2025

ProSplat: Improved Feed-Forward 3D Gaussian Splatting for Wide-Baseline Sparse Views

Feed-forward 3D Gaussian Splatting (3DGS) has recently demonstrated promising results for novel view synthesis (NVS) from sparse input views, particularly under narrow-baseline conditions. However, its performance significantly degrades in wide-baseline scenarios due to limited texture details and geometric inconsistencies across views. To address these challenges, in this paper, we propose ProSplat, a two-stage feed-forward framework designed for high-fidelity rendering under wide-baseline conditions. The first stage involves generating 3D Gaussian primitives via a 3DGS generator. In the second stage, rendered views from these primitives are enhanced through an improvement model. Specifically, this improvement model is based on a one-step diffusion model, further optimized by our proposed Maximum Overlap Reference view Injection (MORI) and Distance-Weighted Epipolar Attention (DWEA). MORI supplements missing texture and color by strategically selecting a reference view with maximum viewpoint overlap, while DWEA enforces geometric consistency using epipolar constraints. Additionally, we introduce a divide-and-conquer training strategy that aligns data distributions between the two stages through joint optimization. We evaluate ProSplat on the RealEstate10K and DL3DV-10K datasets under wide-baseline settings. Experimental results demonstrate that ProSplat achieves an average improvement of 1 dB in PSNR compared to recent SOTA methods.

cs.CV↗

arXiv2024

SPC-NeRF: Spatial Predictive Compression for Voxel Based Radiance Field

Representing the Neural Radiance Field (NeRF) with the explicit voxel grid (EVG) is a promising direction for improving NeRFs. However, the EVG representation is not efficient for storage and transmission because of the terrific memory cost. Current methods for compressing EVG mainly inherit the methods designed for neural network compression, such as pruning and quantization, which do not take full advantage of the spatial correlation of voxels. Inspired by prosperous digital image compression techniques, this paper proposes SPC-NeRF, a novel framework applying spatial predictive coding in EVG compression. The proposed framework can remove spatial redundancy efficiently for better compression performance.Moreover, we model the bitrate and design a novel form of the loss function, where we can jointly optimize compression ratio and distortion to achieve higher coding efficiency. Extensive experiments demonstrate that our method can achieve 32% bit saving compared to the state-of-the-art method VQRF on multiple representative test datasets, with comparable training time.

cs.CV↗