#01Aug 25, 2026
cs.CV
On-Policy Self-Distillation in Diffusion Models
Wei Zhou, Xiongwei Zhu, Lingdong Kong and 14 more
Reinforcement learning can align diffusion models with human preferences and task-specific objectives, but endpoint rewards do not specify how an intermediate denoising prediction should change. We introduce DiffusionOPSD as an on-policy self-distillation framework that converts image-level reward guidance into explicit targets for clean-output predictions at sampled queries. At each outer iteration, a frozen behavior policy generates trajectories and supplies query states and anchors. Reward gradients construct bounded positive and negative targets around each anchor. The trainable policy fits these targets as detached supervision through finite fitting before an exponential moving average update refreshes the behavior policy. This setup lets us measure target construction and finite realization separately. Controlled same-query experiments show that larger target-construction gains do not necessarily translate into larger realized gains after a single fitting update. Across SD 3.5-M and the step-distilled Z-Image-Turbo, our approach achieves the best final held-out scores in 19 of 20 reward-matched settings across two backbones and ten evaluators. It outperforms the strongest competing method by up to 44.0% and reduces training GPU-hours relative to DiffusionNFT by 40% on SD 3.5-M and 63% on Z-Image-Turbo. These results support on-policy self-distillation as an efficient and analyzable approach to diffusion post-training by converting image-level reward guidance into explicit and continually refreshed intermediate supervision, thereby opening a path toward more efficient and diagnosable alignment.
#02Aug 25, 2026
cs.CV
ReGround-Surg: Reliability-Guided Anchor Grounding for Referring Surgical Video Segmentation
Jiaxin Wen, Ming Yin, Lu Liu and 1 more
Referring surgical video segmentation requires segmenting a target instrument or tissue region across video frames according to a natural language expression. Recent Segment Anything Model 2 (SAM2) based two-stage methods (e.g., ReSurgSAM2) first ground the referred target in an initial or selected frame, then propagate the selected mask via tracking. Although effective, their performance is highly sensitive to the quality of the initial grounded mask: once an incorrect anchor is selected, subsequent tracking tends to propagate the error. This issue is especially challenging in surgical videos due to visually similar instruments, occlusion, and complex tissue-tool interactions. To address this issue, we propose ReGround-Surg, a lightweight reliability-guided anchor grounding framework to improve SAM2-based referring surgical video segmentation. It first predicts a text-conditioned spatial reliability map from the referring expression and current-frame visual features. The map is then reused in two complementary branches: a Gated Side Adapter enhances expression-relevant visual regions before text-to-vision fusion, while a Reliability-Weighted Vision-to-Text Attention module suppresses off-target visual evidence during prompt-token aggregation. Experiments on Ref-EndoVis17 and Ref-EndoVis18 show consistent improvements over state-of-the-art methods across three evaluation splits with negligible speed reduction. Code is publicly available at https://github.com/JiaxinWen1/ReGround-Surg.
#03Aug 25, 2026
cs.CV
KLTNet: Learning Sparse Feature Tracking for Robust and Accurate Monocular Visual-Inertial Odometry
Renbiao Jin, Danping Zou, Wenxian Yu
Many feature-based visual-inertial odometry (VIO) systems rely on sparse feature tracking, whose accuracy and robustness directly affect state estimation. Classical KLT trackers rely primarily on local image patches and can become unreliable under rapid motion or in low-texture environments. We propose KLTNet, a lightweight learning-based, plug-and-play sparse feature tracker designed to replace classical KLT trackers in KLT-based VIO front ends. KLTNet follows a coarse-to-fine, dense-to-sparse architecture that combines low-resolution dense optical flow for robust global motion initialization with triplet-patch refinement for accurate and temporally consistent tracking. A fixed reference patch provides a stable anchor throughout each feature track and helps reduce accumulated tracking drift. In addition, KLTNet predicts anisotropic confidence weights supervised through differentiable multi-view triangulation, which can be used as observation weights in compatible VIO estimators. Experiments with VINS-Mono and OpenVINS on public benchmarks and a self-collected low-texture dataset demonstrate improved tracking and odometry accuracy over classical KLT, while maintaining real-time performance on an embedded platform.
#04Aug 25, 2026
cs.CV
Human-Inspired Social Engagement Analysis via Interpretable Mutual Visual Attention
Urwa Fatima, Mohammad Zohaib, Francesca Odone and 1 more
Understanding social interactions from non-verbal visual data is important for behavior analysis and activity monitoring. We propose an interpretable computational model of social engagement inspired by psychological theories of mutual visual attention. Rather than learning interaction patterns end-to-end, our framework explicitly models dyadic visual attention and aggregates these cues into interpretable measures of individual and group engagement. The resulting modular framework combines state-of-the-art head orientation estimation with lightweight geometric reasoning, producing explanations that remain accessible to non-technical users. We evaluate the proposed approach on a variety of data through quantitative experiments and demonstrate its practical usefulness with qualitative visualizations designed to support teachers, caregivers, and social workers in understanding group interaction dynamics.
#05Aug 25, 2026
cs.CV
TurboT2VA: Fast Large-Scale Text-to-Video-Audio Generation via Score-Regularized Consistency Distillation
Xiaoda Yang, Yuxiang Liu, Kaiwen Zheng and 10 more
Joint text-to-video-audio generation produces synchronized visual and acoustic content, but the long sampling trajectories and heterogeneous multimodal computation of large models make inference prohibitively expensive. We present TurboT2VA, a distillation and inference framework for accelerating a 19B-parameter joint video-audio model. Large-scale T2VA distillation is challenged by modality-imbalanced optimization, the difficulty of continuous-time consistency training at scale, and the quality--diversity trade-off. TurboT2VA addresses these issues with per-modality normalization and a progressive curriculum comprising discrete consistency warm-up, continuous consistency refinement, and joint consistency--distribution matching. The curriculum first establishes a stable, diverse generation trajectory and only then introduces distribution-level refinement. On LTX-2, four-step distillation reduces generator latency from 50.52s to 2.51s at the standard evaluation resolution of 512$\times$768, achieving a 20.1$\times$ speedup while maintaining strong visual quality, audio fidelity, diversity, and video-audio synchronization. We further develop an architecture-aware inference stack that combines guarded W8A8 and fused operators, padded-text compaction, and modality-aware sparse attention while preserving dense cross-modal and text-conditioning paths. Under the high-resolution deployment setting at 1024$\times$1792, the complete stack reduces generator latency from 318.74s to 5.83s on one NVIDIA H20, achieving a 54.67$\times$ generator-only speedup. Inference code and generation demos are available at https://github.com/thu-ml/TurboDiffusion/tree/main/turbot2va.