Eccv 2026 Longvqubench

Summary

Title: HitMem: Hierarchical Temporal 3D Memory with Multi-Modal Context-Aware Retrieval for Dynamic Environments Abstract: ... Video for paper: VLA Knows Its Limits. Project link: hatchetproject.github.io/autohorizon/ Diffusion to Obfuscation: Time-Adaptive Synthesized Generation Against Gradient Leakage Attacks in Federated Learning. Continuous 6-DoF pose estimation is essential for autonomous UAV operations. Yet, existing visual odometry and SLAM methods ... Explainability-Aware Frustum Attack: Exposing Structural Vulnerabilities in LiDAR-Based 3D Object Detectors. Paper: RankT2I: A Submodular Framework for Discovering Interpretable and Diverse Semantics in Text-to-Image Models. OmniSch: A Multimodal PCB Schematic Benchmark For Structured Diagram Visual Reasoning. TabletopGen: Tabletop Scene Generation and Interactive Simulation for Robotic Manipulation Paper: RCEdit-500K: Reference Completion for Image-Conditioned Image Editing We introduce RCEdit-500K, a large-scale ...