#01Oct 21, 2026
cs.CV
Looking Ahead: Modeling Future Evolution in Autoregressive Video Generation
Ruopeng Gao, Yukun Huang, Jiarui Hu and 4 more
Autoregressive video diffusion models generate videos chunk by chunk, yet producing plausible continuations does not guarantee that they capture how scenes evolve. Under teacher forcing, strong temporal redundancy may allow local visual continuity to serve as a shortcut for current-chunk denoising, reducing the pressure to model visual evolution. We introduce Future Chunk Prediction (FCP), an auxiliary objective jointly optimized with teacher-forced denoising that predicts temporally separated chunks without access to the intervening visual context. Drawing supervision directly from training videos, FCP requires no additional pretrained visual encoder and leaves autoregressive inference unchanged. Our method achieves superior physical plausibility over standard teacher-forced autoregressive training with a lower overall training compute budget, highlighting the effectiveness and efficiency of future-chunk supervision for modeling visual evolution. Intriguingly, our visualizations suggest that FCP fosters a more structured representation space, which may be valuable for building autoregressive world models.
#02Oct 21, 2026
cs.CV
Attn-Forcing: Learning Full-Context Attention Proxies for Streaming Video Generation
Junyao Gao, Sibo Liu, Jiaxing Li and 5 more
Increasing the attention context length significantly improves streaming video generation quality, while incurring prohibitive memory and computational costs as the context grows. Existing methods attempt to capture this contextual information through attention selection or compression, but remain limited by bounded representations and lack explicit attention-level supervision to retain historical contributions to current queries. To address this, we introduce Attn-Forcing, a novel framework that preserves full-context attention contributions for streaming video generation under bounded memory and computation. We achieve this by introducing attention proxies that continuously approximate the attention responses induced by the historical context. Specifically, we implement the proxies as learnable visual latents and employ a Proxy Update Module to incorporate visual context from historical latents. Moreover, we propose an Attention Matching Loss to explicitly constrain the attention behavior of the learned proxies using full-context attention supervision. Extensive qualitative and quantitative experiments demonstrate that Attn-Forcing consistently improves streaming video generation quality and achieves state-of-the-art performance over existing baseline methods.
#03Oct 21, 2026
cs.CV
ClunoGS: Compact Language 3D Gaussian Splatting via Unified Optimization
Jian Xu, Yi Li, Ying Huang and 4 more
Language 3D Gaussian Splatting enables open-vocabulary scene understanding by associating Gaussian primitives with language features. These semantic attributes increase model storage, while most existing 3DGS compression methods are designed mainly to preserve RGB rendering quality. One straightforward approach is to apply sparsification and quantization sequentially. However, pruning based only on rendering quality can remove Gaussians that have small RGB contributions but are important for semantic prediction. Once removed, their semantic information cannot be recovered by later quantization. We propose ClunoGS, a compact framework for Language 3DGS that combines semantics-aware Gaussian sparsification and vector quantization in a unified constrained objective. Semantic supervision is used during sparsification, and the objective jointly considers rendering quality, semantic accuracy, and compactness. We optimize the unified objective with alternating primal, sparsification, quantization, and dual updates. We evaluate ClunoGS on open-vocabulary localization and segmentation using the LERF and ScanNet-10 benchmarks. Under aggressive compression, our unified, semantics-aware formulation substantially alleviates the degradation of semantic representations caused by sequential compression and semantics-agnostic Gaussian sparsification. Ablations further show that jointly coordinating sparsification and quantization better preserves semantic accuracy than applying them independently.
#04Oct 21, 2026
cs.CV
GuidedSAM-Plume: An Expert-Guided Segmentation Framework for Volcanic Plumes
Taddeo D'Adamo, Simon Thivet, Catherine Pothier and 6 more
Quantitative analysis of environmental phenomena increasingly relies on image-based measurements. In volcanology, the geometric characterization of volcanic plumes and clouds from RGB images remains hindered by the scarcity of dedicated segmentation solutions and partially handcrafted annotations remain widely prevalent. We present GuidedSAM-Plume, an open-source expert-guided framework tailored for the interactive segmentation of volcanic plumes and clouds. It leverages the zero-shot, spatial prompt-based capabilities of SAM2. The framework was thoroughly validated on a challenging dataset through dedicated segmentation metrics and field-relevant volcanological descriptors. From minimal user inputs, it is able to significantly increase performance compared to fully automatic dedicated solutions while preserving the reliability of derived volcanological parameters. The absence of retraining requirements lowers adoption barriers within the volcanological community, and the framework's demonstrated suitability for complex, diffuse plume-shaped phenomena extends its applicability to broader environmental science domains. The re-annotated dataset, produced under the guidance of expert volcanologists within the proposed framework, ultimately yields a more standardized release. The framework and accompanying dataset are made available at https://github.com/tad-damo/GuidedSAM-Plume.
#05Oct 21, 2026
cs.CV
GeoMem-VLA: Internalizing 3D Scene Understanding into Temporal Memory for Vision-Language-Action Models
Qinfan Zhang, Yifan Luo, Haoyuan Shi and 6 more
Geometric structure is essential for robotic manipulation and understanding the 3D world. Most VLA models rely on 2D representations of current observations, while geometry-aware approaches often require additional 3D information. However, geometry inferred independently from individual observations is often discarded rather than accumulated into a persistent task representation. Our key insight is that action generation needs a compact, task-relevant geometric abstraction that can be continuously updated, rather than a complete, renderable 3D reconstruction at every step. We introduce GeoMem-VLA, a VLA framework that internalizes explicit 3D supervision into dedicated geometric tokens and maintains them as temporal memory for action generation, without directly aligning the VLM's native visual embeddings with the teacher feature space. During training, a frozen geometry foundation model supervises learnable GeoPlan tokens through a geometric flow-matching objective over the current scene point map and future end-effector tracks. These tokens are integrated with the memory of observation representations, where historical information is retrieved, adaptively fused with current observations, and consolidated to retain event frames as the task progresses. At inference, the policy retrieves historical representations and fuses them with the current working representation to predict actions, without invoking the 3D teacher or explicitly reconstructing the scene. After benchmark-specific downstream adaptation, GeoMem-VLA achieves 98.9% average success on LIBERO and 91.10% / 91.24% on RoboTwin's clean/randomized settings. Across five real-world tasks, GeoMem-VLA achieves 78.0% average success. These results suggest that internalizing and remembering 3D scene understanding is a promising approach to robot control.