Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Beijing University of Technology
3
0
4
12
Decomposing robotic manipulation into coarse trajectory anchors and fine-grained residual refinement boosts real-world success rates by 18%, suggesting a more human-like approach to VLA control.
Forget modality-specific architectures: this work achieves state-of-the-art camouflaged object detection by learning modality-agnostic prompts for SAM, unlocking efficient adaptation to new modalities.
Forget brute-force scaling: injecting topological awareness into smaller VLMs drastically improves performance on Vision-Language Navigation, surpassing larger models.