FILT3R is a training‑free latent filtering layer for streaming 3D reconstruction that treats recurrent state updates as stochastic state estimation in token space. It maintains per‑token variance and computes a Kalman‑style gain to balance memory retention with new observations, estimating process noise online from temporal drift of candidate tokens. Experiments show that FILT3R generalizes overwrite and gating policies, shrinking gains in stable regimes and increasing them during genuine scene changes, thereby improving long‑horizon stability for depth, pose, and 3D reconstruction.
Streaming 3D reconstruction relies on a compact recurrent scene state to process long image streams in linear time and bounded memory. However, repeated updates can gradually corrupt this state, causing reliable historical information to be overwritten by noisy or ambiguous observations.
arXiv:2609.21938v1 Announce Type: new Abstract: Transformer-based models have recently achieved strong performance on 3D reconstruction from images, and recent works extend them to process video stre...
arXiv:2608.27529v1 Announce Type: new Abstract: Streaming 3D reconstruction from extremely long videos requires estimating camera motion and scene geometry online under bounded memory and computation...
arXiv:2605.16981v3 Announce Type: replace Abstract: Streaming 3D reconstruction under a strict constant-memory budget hinges on how the recurrent state is updated as the stream evolves. We profile TT...
arXiv:2606. 12691v1 Announce Type: cross Abstract: Auto-regressive models have emerged as powerful tools for sequential data, from language to video.
The paper introduces AURA, a meta‑learning framework that learns a low‑dimensional latent state‑space model for the evolution of optimal model parameters under distribution shift. Online adaptation is performed via extended Kalman filtering in this latent space, followed by reconstruction of full model parameters through a learned lifting map, enabling efficient single‑step updates. Experiments on neural wireless receivers and non‑stationary image classification show that AURA improves adaptation speed, accuracy, and computational efficiency compared to existing online learning and Bayesian filtering baselines.
Feed-forward 3D Gaussian Splatting enables efficient novel-view synthesis without per-scene optimization, but most existing methods assume a fixed set of context views and process them jointly. This limits their applicability to online scenarios where calibrated views arrive sequentially and the scene must be updated causally.
arXiv:2608.29029v3 Announce Type: replace-cross Abstract: Joint-Embedding Predictive Architectures (JEPAs) provide a powerful framework for latent world modeling and planning in a reconstruction-free...
arXiv:2609.17230v1 Announce Type: new Abstract: Streaming 3D reconstruction demands both speed and temporal fidelity, goals that existing methods undermine by updating every Gaussian every frame, eve...
arXiv:2608. 19556v1 Announce Type: cross Abstract: Streaming autoregressive diffusion models enable real-time, long-horizon video generation, but their training objectives optimize local frame prediction rather than the geometry and dynamics of a coherent world: long rollouts accumulate geometric drift and degrade into static or unnatural motion.
LayerRecall is a memory router for autoregressive video diffusion that selectively retrieves and injects historical key/value states into specific layers of the model, based on the current context. It addresses the problem that existing memory mechanisms expose nonlocal history but do not guarantee effective use, by recognizing that different layers prefer current, recent, or distant context. The method, combined with Cross‑Horizon Prediction Matching, achieves state‑of‑the‑art long‑range consistency on MemoBench and MovieBench while maintaining local continuity and incurring negligible inference overhead.
arXiv:2609.23286v1 Announce Type: new Abstract: 3D reconstruction from a lengthy video stream input poses a dilemma for feed-forward reconstruction models (FFRMs), that a whole-stream inference conte...