Meta tags:
description=DESCRIPTION META TAG;
keywords=KEYWORDS SHOULD BE PLACED HERE;

Headings (most frequently used words):

of, real, adaptation, generalization, results, experiment, setup, world, objects, overview, cross, robot, visualization, tasks, bi, adapt, few, shot, bimanual, for, novel, categories, 3d, via, semantic, correspondence, video, presentation, abstract, pipeline, more, visuals, category, affordance, experiments, demo, simulation, bibtex, manipulation, across, diverse, with, embodiment,

Text of the page (most frequently used words):
and (17), the (12), novel (11), categories (11), for (9), manipulation (8), adapt (7), #bimanual (7), objects (7), with (7), #adaptation (6), gripper (6), real (6), correspondence (5), generalization (5), tasks (5), unseen (5), world (5), affordance (5), from (4), shot (4), via (4), semantic (4), zhou (4), cross (4), across (4), demo (4), setup (4), results (4), page (3), few (3), 2026 (3), instance (3), robot (3), experiments (3), experiment (3), simulation (3), right (3), corresponding (3), first (3), two (3), contact (3), points (3), category (3), fine (3), data (3), university (3), this (2), project (2), author (2), jinxian (2), ruihai (2), liu (2), yiwei (2), hou (2), yiwen (2), xunzhe (2), checheng (2), zhong (2), licheng (2), shao (2), lin (2), icra (2), arxiv (2), embodiment (2), leaphand (2), xhand (2), overview (2), diverse (2), visualization (2), each (2), left (2), second (2), set (2), image (2), action (2), network (2), our (2), mapping (2), training (2), leveraging (2), foundation (2), tuned (2), paper (2), high (2), part, borrowed, academic, template, inproceedings, zhou2026biadapt, title, booktitle, ieee, international, conference, robotics, automation, year, eprint, 2602, 08425, archiveprefix, primaryclass, bibtex, demonstration, rollout, examples, unfolding, uncapping, opening, row, represents, task, respectively, show, progress, use, ufactory, xarm6, realsense, d435, cameras, cap, ture, rgbd, observations, test, trajectory, group, shows, source, target, inferred, highlight, intensity, indicates, more, visuals, train, learning, supporting, distribution, serve, prior, knowledge, then, make, model, after, that, pre, trained, proposes, actions, based, mapped, point, pairs, interaction, finally, networks, can, facilitate, manipulating, instances, better, performance, upper, lower, middle, pipeline, imperative, yet, challenging, robots, execute, complex, requiring, coordinated, collaboration, between, arms, however, existing, methods, often, rely, costly, collection, struggling, generalize, efficiently, present, framework, designed, efficient, achieves, strong, capability, vision, models, tuning, restricted, exhibits, notable, out, zero, manner, extensive, conducted, both, environments, validate, effectiveness, approach, demonstrate, its, efficiency, achieving, success, rate, different, benchmark, limited, abstract, video, presentation, code, equal, contribution, peking, hong, kong, national, singapore, shanghai, zhi, institute, best, award, finalist, manual, ation,

Text of the page (random words):
bi adapt few shot bimanual adaptation for novel categories of 3d objects via semantic correspondence bi adapt few shot bi manual adapt ation for novel categories of 3d objects via semantic correspondence icra 2026 best paper award finalist jinxian zhou 1 2 ruihai wu 3 yiwei liu 1 yiwen hou 1 xunzhe zhou 1 4 checheng yu 1 4 licheng zhong 2 lin shao 1 1 national university of singapore 2 shanghai qi zhi institute 3 peking university 4 the university of hong kong equal contribution corresponding author arxiv code video presentation abstract bimanual manipulation is imperative yet challenging for robots to execute complex tasks requiring coordinated collaboration between two arms however existing methods for bimanual manipulation often rely on costly data collection and training struggling to generalize to unseen objects in novel categories efficiently in this paper we present bi adapt a novel framework designed for efficient generalization for bimanual manipulation via semantic correspondence bi adapt achieves cross category affordance mapping by leveraging the strong capability of vision foundation models fine tuning with restricted data on novel categories bi adapt exhibits notable generalization to out of category objects in a zero shot manner extensive experiments conducted in both simulation and real world environments validate the effectiveness of our approach and demonstrate its high efficiency achieving a high success rate on different benchmark tasks across novel categories with limited data pipeline overview left we first train the action learning network on the supporting set the corresponding affordance and action distribution serve as our prior knowledge middle then we make contact points mapping from training categories to novel categories leveraging the foundation model lower right after that the pre trained network proposes actions based on mapped contact point pairs on novel categories and fine tuned with the interaction results upper right finally the fine tuned networks can facilitate manipulating unseen instances from novel categories with better performance more visuals cross category affordance generalization results each group shows a source image with contact points first gripper second gripper and a target image with inferred points highlight intensity indicates correspondence adaptation results affordance adaptation trajectory adaptation real experiment setup a robot setup b objects a we use two ufactory xarm6 for manipulation and two realsense d435 cameras to cap ture rgbd observations b we test on a diverse set of objects visualization of real world experiments each row represents a task from left to right we respectively show the corresponding affordance on novel categories first gripper second gripper and the manipulation progress real world demo opening uncapping unfolding simulation experiment simulation rollout examples across bimanual manipulation tasks visualization of manipulation results across tasks real world experiment robot setup real world robot setup for bimanual manipulation experiments generalization of adaptation unseen instance demo 1 unseen instance demo 2 unseen instance demo 3 overview of diverse tasks with cross embodiment gripper leaphand xhand cross embodiment demonstration across gripper leaphand and xhand generalization bibtex inproceedings zhou2026biadapt title bi adapt few shot bimanual adaptation for novel categories of 3d objects via semantic correspondence author zhou jinxian and wu ruihai and liu yiwei and hou yiwen and zhou xunzhe and yu checheng and zhong licheng and shao lin booktitle 2026 ieee international conference on robotics and automation icra year 2026 eprint 2602 08425 archiveprefix arxiv primaryclass cs ro part of this page is borrowed from academic project page template project page