vecId
stringlengths
12
23
id
stringlengths
2
13
conference
stringclasses
11 values
year
float64
2.02k
2.03k
title
stringlengths
6
189
abstract
stringlengths
10
4.74k
author
stringlengths
0
7.45k
aff
stringlengths
0
7.16k
status
stringclasses
11 values
track
stringclasses
4 values
keywords
stringlengths
0
804
github
stringlengths
0
141
site
stringlengths
0
193
gsCitation
float64
-1
11.1k
arxiv
stringlengths
0
12
text
stringlengths
58
4.82k
vector
list
corl_2025_TFbT7kHD89
TFbT7kHD89
corl
2,025
Constrained Style Learning from Imperfect Demonstrations under Task Optimality
Learning from demonstration has proven effective in robotics for acquiring natural behaviors, such as stylistic motions and lifelike agility, particularly when explicitly defining style-oriented reward functions is challenging. Synthesizing stylistic motions for real-world tasks usually requires balancing task performa...
Kehan Wen;Chenhao Li;Junzhe He;Marco Hutter
ETHZ - ETH Zurich;ETHZ - ETH Zurich;ETHZ - ETH Zurich;ETHZ - ETH Zurich
Poster
main
Constrained Markov Decision Process;Imitation Learning;Legged Robots
https://openreview.net/forum?id=TFbT7kHD89
-1
Constrained Style Learning from Imperfect Demonstrations under Task Optimality Learning from demonstration has proven effective in robotics for acquiring natural behaviors, such as stylistic motions and lifelike agility, particularly when explicitly defining style-oriented reward functions is challenging. Synthesizing ...
[ -0.04552294313907623, -0.04650763422250748, -0.019135164096951485, 0.009046833030879498, -0.03503221645951271, 0.016379928216338158, -0.01697642169892788, -0.013340646401047707, 0.004201498813927174, 0.014647253789007664, -0.03480498120188713, -0.018443988636136055, -0.0047766901552677155, ...
corl_2025_Th1kFSnjUW
Th1kFSnjUW
corl
2,025
MoTo: A Zero-shot Plug-in Interaction-aware Navigation for General Mobile Manipulation
Mobile manipulation is the fundamental challenge for robotics in assisting humans with diverse tasks and environments in everyday life. Conventional mobile manipulation approaches often struggle to generalize across different tasks and environments due to the lack of large-scale training. However, recent advances in ma...
Zhenyu Wu;Angyuan Ma;Xiuwei Xu;Hang Yin;Yinan Liang;Ziwei Wang;Jiwen Lu;Haibin Yan
Beijing University of Posts and Telecommunications;Tsinghua University+Tsinghua University;Tsinghua University;Tsinghua University;Tsinghua University;Nanyang Technological University;Tsinghua University;Beijing University of Posts and Telecommunications
Poster
main
Mobile Manipulation;VLA;VLM
https://openreview.net/forum?id=Th1kFSnjUW
-1
MoTo: A Zero-shot Plug-in Interaction-aware Navigation for General Mobile Manipulation Mobile manipulation is the fundamental challenge for robotics in assisting humans with diverse tasks and environments in everyday life. Conventional mobile manipulation approaches often struggle to generalize across different tasks a...
[ -0.016970863565802574, -0.05014374107122421, -0.01787097565829754, -0.0025807900819927454, 0.0029581808485090733, -0.03354792669415474, 0.003141016000881791, -0.010173141956329346, 0.008138513192534447, 0.010876353830099106, -0.016286402940750122, -0.01368920411914587, 0.016877101734280586, ...
corl_2025_Tl7girqoLi
Tl7girqoLi
corl
2,025
DemoSpeedup: Accelerating Visuomotor Policies via Entropy-Guided Demonstration Acceleration
Imitation learning has shown great promise in robotic manipulation, but the policy’s execution is often unsatisfactorily slow due to commonly tardy demonstrations collected by human operators. In this work, we present DemoSpeedup, a self-supervised method to accelerate visuomotor policy execution via entropy-guided dem...
Lingxiao Guo;Zhengrong Xue;Zijing Xu;Huazhe Xu
Shanghai Jiaotong University;Tsinghua University;University of Electronic Science and Technology of China;Tsinghua University
Oral
main
Imitation learning;Manipulation;Demonstration Acceleration
https://openreview.net/forum?id=Tl7girqoLi
-1
DemoSpeedup: Accelerating Visuomotor Policies via Entropy-Guided Demonstration Acceleration Imitation learning has shown great promise in robotic manipulation, but the policy’s execution is often unsatisfactorily slow due to commonly tardy demonstrations collected by human operators. In this work, we present DemoSpeedu...
[ -0.0139797767624259, -0.05356381833553314, 0.00024469950585626066, 0.009022274054586887, -0.058274395763874054, -0.0274657029658556, 0.001127784256823361, 0.020589780062437057, 0.03426564857363701, 0.03572820872068405, -0.017227794975042343, -0.03838740289211273, -0.001641816576011479, 0.0...
corl_2025_TqevdDMqrK
TqevdDMqrK
corl
2,025
CUPID: Curating Data your Robot Loves with Influence Functions
In robot imitation learning, policy performance is tightly coupled with the quality and composition of the demonstration data. Yet, developing a precise understanding of how individual demonstrations contribute to downstream outcomes—such as closed-loop task success or failure—remains a persistent challenge. Inspired b...
Christopher Agia;Rohan Sinha;Jingyun Yang;Rika Antonova;Marco Pavone;Haruki Nishimura;Masha Itkina;Jeannette Bohg
Stanford University;Stanford University;Stanford University;;NVIDIA+Stanford University;;Toyota Research Institute;Stanford University
Poster
main
Imitation Learning;Data Curation;Influence Functions
https://openreview.net/forum?id=TqevdDMqrK
-1
CUPID: Curating Data your Robot Loves with Influence Functions In robot imitation learning, policy performance is tightly coupled with the quality and composition of the demonstration data. Yet, developing a precise understanding of how individual demonstrations contribute to downstream outcomes—such as closed-loop tas...
[ -0.08784329146146774, -0.04170868173241615, -0.00653104716911912, 0.02499895542860031, -0.034807246178388596, -0.04223379120230675, -0.013108978979289532, 0.003124869428575039, -0.002731037326157093, 0.04632214456796646, -0.03349447250366211, 0.007576577365398407, -0.018603872507810593, 0....
corl_2025_Tx54fkQ3Cq
Tx54fkQ3Cq
corl
2,025
Humanoid Policy ~ Human Policy
Training manipulation policies for humanoid robots with diverse data enhances their robustness and generalization across tasks and platforms. However, learning solely from robot demonstrations is labor-intensive, requiring expensive tele-operated data collection,n which is difficult to scale. This paper investigates a ...
Ri-Zhao Qiu;Shiqi Yang;Xuxin Cheng;Chaitanya Chawla;Jialong Li;Tairan He;Ge Yan;David J. Yoon;Ryan Hoque;Lars Paulsen;Ge Yang;Jian Zhang;Sha Yi;Guanya Shi;Xiaolong Wang
University of California, San Diego;University of California, San Diego;University of California, San Diego;School of Computer Science, Carnegie Mellon University;University of California, San Diego;NVIDIA+Carnegie Mellon University;Department of Computer Science, University of Washington;Apple;Apple;University of Cali...
Poster
main
Robot Manipulation;Humanoid;Learning from Human
https://openreview.net/forum?id=Tx54fkQ3Cq
-1
Humanoid Policy ~ Human Policy Training manipulation policies for humanoid robots with diverse data enhances their robustness and generalization across tasks and platforms. However, learning solely from robot demonstrations is labor-intensive, requiring expensive tele-operated data collection,n which is difficult to sc...
[ -0.04231160879135132, -0.06871937960386276, -0.04164586588740349, 0.0020319002214819193, -0.029422102496027946, -0.051816925406455994, -0.0133333345875144, -0.002135922433808446, 0.00855293683707714, 0.03373093158006668, -0.01573740318417549, -0.014858993701636791, 0.02576052024960518, 0.0...
corl_2025_U9zcbQVDGa
U9zcbQVDGa
corl
2,025
Text2Touch: Tactile In-Hand Manipulation with LLM-Designed Reward Functions
Large language models (LLMs) are beginning to automate reward design for dexterous manipulation. However, no prior work has considered tactile sensing, which is known to be critical for human-like dexterity. We present Text2Touch, bringing LLM-crafted rewards to the challenging task of multi-axis in-hand object rotatio...
Harrison Field;Max Yang;Yijiong Lin;Efi Psomopoulou;David A.W. Barton;Nathan F. Lepora
University of Bristol;University of Bristol;;University of Bristol;University of Bristol;University of Bristol
Poster
main
Tactile Sensing;Reinforcement Learning;Large Language Models
https://openreview.net/forum?id=U9zcbQVDGa
-1
Text2Touch: Tactile In-Hand Manipulation with LLM-Designed Reward Functions Large language models (LLMs) are beginning to automate reward design for dexterous manipulation. However, no prior work has considered tactile sensing, which is known to be critical for human-like dexterity. We present Text2Touch, bringing LLM-...
[ -0.07405933737754822, 0.01581922359764576, 0.013095630332827568, 0.022814752534031868, -0.027907496318221092, -0.015660658478736877, -0.032645802944898605, -0.0012358768144622445, 0.008506563492119312, 0.008772393688559532, 0.001126280170865357, 0.010819751769304276, 0.0058715809136629105, ...
corl_2025_UTPBM4dEUS
UTPBM4dEUS
corl
2,025
Sampling-based System Identification with Active Exploration for Legged Sim2Real Learning
Sim-to-real discrepancies hinder learning-based policies from achieving high-precision tasks in the real world. While Domain Randomization (DR) is commonly used to bridge this gap, it often relies on heuristics and can lead to overly conservative policies with degrading performance when not properly tuned. System Ident...
Nikhil Sobanbabu;Guanqi He;Tairan He;Yuxiang Yang;Guanya Shi
Carnegie Mellon University;Carnegie Mellon University;NVIDIA+Carnegie Mellon University;Department of Computer Science, u+Google;Carnegie Mellon University
Oral
main
System Identification;Sim2Real;Legged Robots
https://openreview.net/forum?id=UTPBM4dEUS
-1
Sampling-based System Identification with Active Exploration for Legged Sim2Real Learning Sim-to-real discrepancies hinder learning-based policies from achieving high-precision tasks in the real world. While Domain Randomization (DR) is commonly used to bridge this gap, it often relies on heuristics and can lead to ove...
[ -0.07592116296291351, -0.025912223383784294, -0.021089212968945503, 0.03348600119352341, -0.03322926163673401, -0.05626235157251358, 0.0029387359973043203, -0.011314820498228073, 0.005703256465494633, 0.04247184097766876, -0.028424590826034546, 0.018494322896003723, -0.028296221047639847, ...
corl_2025_VVhAhzr2WV
VVhAhzr2WV
corl
2,025
Real2Render2Real: Scaling Robot Data Without Dynamics Simulation or Robot Hardware
Scaling robot learning requires vast and diverse datasets. Yet the prevailing data collection paradigm—human teleoperation—remains costly and constrained by manual effort and physical robot access. We introduce Real2Render2Real (R2R2R), a novel approach for generating robot training data without relying on object dynam...
Justin Yu;Letian Fu;Huang Huang;Karim El-Refai;Rares Andrei Ambrus;Richard Cheng;Muhammad Zubair Irshad;Ken Goldberg
University of California, Berkeley;University of California, Berkeley;Stanford University+Meta Facebook+University of California, Berkeley;University of California, Berkeley;Toyota Research Institute;Toyota Research Institute;Toyota Research Institute;University of California, Berkeley
Oral
main
Robot Datasets;Imitation Learning;Data Augmentation
https://openreview.net/forum?id=VVhAhzr2WV
-1
Real2Render2Real: Scaling Robot Data Without Dynamics Simulation or Robot Hardware Scaling robot learning requires vast and diverse datasets. Yet the prevailing data collection paradigm—human teleoperation—remains costly and constrained by manual effort and physical robot access. We introduce Real2Render2Real (R2R2R), ...
[ -0.07259619981050491, -0.02181190811097622, -0.022638116031885147, 0.027448486536741257, -0.028293054550886154, 0.03881344571709633, -0.02331744320690632, 0.005388716235756874, -0.006247055251151323, -0.00396809633821249, -0.04865450784564018, -0.006412297021597624, 0.009046983905136585, 0...
corl_2025_VmCkEvRULX
VmCkEvRULX
corl
2,025
Generative Visual Foresight Meets Task-Agnostic Pose Estimation in Robotic Table-top Manipulation
Robotic manipulation in unstructured environments requires systems that can generalize across diverse tasks while maintaining robust and reliable performance. We introduce GVF-TAPE, a closed-loop framework that combines generative visual foresight with task-agnostic pose estimation to enable scalable robotic manipulati...
Chuye Zhang;Xiaoxiong Zhang;Linfang Zheng;Wei Pan;Wei Zhang
Southern University of Science and Technology;Southern University of Science and Technology;University of Hong Kong;Southern University of Science and Technology;Southern University of Science and Technology of China
Poster
main
robotic manipulation;action-label-free learning;generative visual foresight
https://openreview.net/forum?id=VmCkEvRULX
-1
Generative Visual Foresight Meets Task-Agnostic Pose Estimation in Robotic Table-top Manipulation Robotic manipulation in unstructured environments requires systems that can generalize across diverse tasks while maintaining robust and reliable performance. We introduce GVF-TAPE, a closed-loop framework that combines ge...
[ -0.06343333423137665, -0.03890770673751831, -0.004625724628567696, 0.04050169140100479, 0.023583723232150078, -0.007924637757241726, 0.0022743712179362774, 0.023529382422566414, 0.009210693649947643, 0.04604440927505493, -0.021826718002557755, -0.028329448774456978, 0.0004828368837479502, ...
corl_2025_Vo1tL9dhpk
Vo1tL9dhpk
corl
2,025
Uncertainty-aware Accurate Elevation Modeling for Off-road Navigation via Neural Processes
Terrain elevation modeling for off-road navigation aims to accurately estimate changes in terrain geometry in real-time and quantify the corresponding uncertainties. Having precise estimations and uncertainties plays a crucial role in planning and control algorithms to explore safe and reliable maneuver strategies. How...
Sanghun Jung;Daehoon Gwak;Byron Boots;James Hays
University of Washington;Korea Advanced Institute of Science & Technology;;Overland AI+Georgia Institute of Technology
Poster
main
Robot perception;Terrain modeling;Neural processes
https://openreview.net/forum?id=Vo1tL9dhpk
-1
Uncertainty-aware Accurate Elevation Modeling for Off-road Navigation via Neural Processes Terrain elevation modeling for off-road navigation aims to accurately estimate changes in terrain geometry in real-time and quantify the corresponding uncertainties. Having precise estimations and uncertainties plays a crucial ro...
[ -0.04834410175681114, -0.036015234887599945, -0.028580553829669952, 0.015429763123393059, -0.0006012657540850341, -0.04359935596585274, 0.038256846368312836, 0.006664126180112362, -0.00648199487477541, 0.024993974715471268, -0.02766522951424122, 0.014626518823206425, 0.0050389571115374565, ...
corl_2025_VqmAvBkFhw
VqmAvBkFhw
corl
2,025
LocoFormer: Generalist Locomotion via Long-context Adaptation
Humans and animals exhibit flexible locomotion strategies, such as learning to walk within minutes, and efficient adaptation to changes in morphology. In contrast, modern locomotion controllers are manually tuned for specific embodiments. In this paper, we present LocoFormer, a generalist policy that can control previo...
Min Liu;Deepak Pathak;Ananye Agarwal
Skild AI;Skild AI+Carnegie Mellon University;Carnegie Mellon University
Oral
main
Cross-Embodied Learning;Legged Locomotion;Online Adaptation
https://openreview.net/forum?id=VqmAvBkFhw
-1
LocoFormer: Generalist Locomotion via Long-context Adaptation Humans and animals exhibit flexible locomotion strategies, such as learning to walk within minutes, and efficient adaptation to changes in morphology. In contrast, modern locomotion controllers are manually tuned for specific embodiments. In this paper, we p...
[ -0.03895764797925949, -0.027470137923955917, -0.029208987951278687, 0.018341172486543655, -0.028820522129535675, 0.00008078601968009025, -0.050611648708581924, -0.00002348430825804826, 0.019330836832523346, -0.010146377608180046, -0.03851368650794029, -0.005669762380421162, 0.014391762204468...
corl_2025_VrNSv02Xfu
VrNSv02Xfu
corl
2,025
SAVOR: Skill Affordance Learning from Visuo-Haptic Perception for Robot-Assisted Bite Acquisition
Robot-assisted feeding requires reliable bite acquisition, a challenging task due to the complex interactions between utensils and food with diverse physical properties. These interactions are further complicated by the temporal variability of food properties—for example, steak becomes firm as it cools even during a me...
Zhanxin Wu;Bo Ai;Tom Silver;Tapomayukh Bhattacharjee
Cornell University;University of California, San Diego;Princeton University+Cornell University;Cornell University
Oral
main
Assistive Robotics;Visuo-Haptic Perception;Affordance Learning
https://openreview.net/forum?id=VrNSv02Xfu
-1
SAVOR: Skill Affordance Learning from Visuo-Haptic Perception for Robot-Assisted Bite Acquisition Robot-assisted feeding requires reliable bite acquisition, a challenging task due to the complex interactions between utensils and food with diverse physical properties. These interactions are further complicated by the te...
[ -0.023190902546048164, -0.028904331848025322, -0.0016847591614350677, 0.004254631698131561, -0.008471798151731491, 0.010181143879890442, 0.020324822515249252, 0.014133712276816368, 0.011764044873416424, 0.036003969609737396, 0.019294532015919685, -0.011286363936960697, -0.02272258885204792, ...
corl_2025_XF69ltYlMU
XF69ltYlMU
corl
2,025
Planning from Point Clouds over Continuous Actions for Multi-object Rearrangement
Multi-object rearrangement is a challenging task that requires robots to reason about a physical 3D scene and the effects of a sequence of actions. While traditional task planning methods are shown to be effective for long-horizon manipulation, they require discretizing the continuous state and action space into symbol...
Kallol Saha;Amber Li;Angela Rodriguez-Izquierdo;Lifan Yu;Ben Eisner;Maxim Likhachev;David Held
Carnegie Mellon University;Carnegie Mellon University;;;Carnegie Mellon University;;Carnegie Mellon University
Oral
main
Robot Learning;Robot Planning
https://openreview.net/forum?id=XF69ltYlMU
-1
Planning from Point Clouds over Continuous Actions for Multi-object Rearrangement Multi-object rearrangement is a challenging task that requires robots to reason about a physical 3D scene and the effects of a sequence of actions. While traditional task planning methods are shown to be effective for long-horizon manipul...
[ 0.0018708534771576524, -0.012349928729236126, -0.032596465200185776, -0.005412849597632885, -0.01832747831940651, 0.040952179580926895, 0.008892866782844067, 0.004483161959797144, 0.0012062988243997097, 0.01583912782371044, -0.017666365951299667, -0.05097903683781624, -0.0051603419706225395,...
corl_2025_XaJkbK02Vm
XaJkbK02Vm
corl
2,025
LocoTouch: Learning Dynamic Quadrupedal Transport with Tactile Sensing
Quadrupedal robots have demonstrated remarkable agility and robustness in traversing complex terrains. However, they struggle with dynamic object interactions, where contact must be precisely sensed and controlled. To bridge this gap, we present LocoTouch, a system that equips quadrupedal robots with tactile sensing to...
Changyi Lin;Yuxin Ray Song;Boda Huo;Mingyang Yu;Yikai Wang;Shiqi Liu;Yuxiang Yang;Wenhao Yu;Tingnan Zhang;Jie Tan;Yiyue Luo;Ding Zhao
Carnegie Mellon University;Massachusetts Institute of Technology+Department of Computer Science, University of Washington;Carnegie Mellon University;Carnegie Mellon University;Carnegie Mellon University;Carnegie Mellon University;Department of Computer Science, u+Google;Google;Google;Google;University of Washington;Car...
Poster
main
Tactile Quadrupedal Policy;Tactile Sim-to-Real;Legged Robots
https://openreview.net/forum?id=XaJkbK02Vm
-1
LocoTouch: Learning Dynamic Quadrupedal Transport with Tactile Sensing Quadrupedal robots have demonstrated remarkable agility and robustness in traversing complex terrains. However, they struggle with dynamic object interactions, where contact must be precisely sensed and controlled. To bridge this gap, we present Loc...
[ -0.06259877979755402, -0.01898457109928131, -0.0034513168502599, 0.016767531633377075, -0.005016286391764879, -0.01464364305138588, 0.013758690096437931, 0.011998100206255913, -0.027554642409086227, 0.012277558445930481, -0.010526282712817192, 0.008174171671271324, 0.004042837768793106, -0...
corl_2025_XjjXLxfPou
XjjXLxfPou
corl
2,025
ReWiND: Language-Guided Rewards Teach Robot Policies without New Demonstrations
We introduce ReWiND, a framework for learning robot manipulation tasks solely from language instructions without per-task demonstrations. Standard reinforcement learning (RL) and imitation learning methods require expert supervision through human-designed reward functions or demonstrations for every new task. In contra...
Jiahui Zhang;Yusen Luo;Abrar Anwar;Sumedh Anand Sontakke;Joseph J Lim;Jesse Thomason;Erdem Biyik;Jesse Zhang
University of Southern California;University of Southern California;NVIDIA+University of Southern California;Amazon;Korea Advanced Institute of Science & Technology;University of Southern California+Amazon;University of Southern California;University of Washington+NVIDIA+University of Southern California
Oral
main
Reinforcement Learning;Offline Reinforcement Learning;Reward Learning;Reward Modeling;Language
https://openreview.net/forum?id=XjjXLxfPou
-1
ReWiND: Language-Guided Rewards Teach Robot Policies without New Demonstrations We introduce ReWiND, a framework for learning robot manipulation tasks solely from language instructions without per-task demonstrations. Standard reinforcement learning (RL) and imitation learning methods require expert supervision through...
[ -0.09636193513870239, -0.022857433184981346, 0.021053867414593697, -0.01470458135008812, -0.02762399986386299, -0.021679595112800598, 0.011355102062225342, -0.0029791039414703846, -0.008875199593603611, 0.00016031374980229884, -0.05410696938633919, -0.061910148710012436, -0.00582017982378602...
corl_2025_XoRtWWjXuC
XoRtWWjXuC
corl
2,025
AgentWorld: An Interactive Simulation Platform for Scene Construction and Mobile Robotic Manipulation
We introduce AgentWorld, an interactive simulation platform for developing household mobile manipulation capabilities. Our platform combines automated scene construction that encompasses layout generation, semantic asset placement, visual material configuration, and physics simulation, with a dual-mode teleoperation sy...
Yizheng Zhang;Zhenjun Yu;JiaXin Lai;Cewu Lu;Lei Han
Tencent Robotics X;Shanghai Jiaotong University;IEG;Shanghai Jiaotong University;Tencent Robotics X
Poster
main
Simulation Platform;scene construction;teleoperation;simulation dataset
https://openreview.net/forum?id=XoRtWWjXuC
-1
AgentWorld: An Interactive Simulation Platform for Scene Construction and Mobile Robotic Manipulation We introduce AgentWorld, an interactive simulation platform for developing household mobile manipulation capabilities. Our platform combines automated scene construction that encompasses layout generation, semantic ass...
[ -0.028117593377828598, -0.02319563739001751, -0.011294787749648094, -0.01227734237909317, -0.022350823506712914, -0.04808090254664421, -0.003519290592521429, 0.026758546009659767, 0.017098288983106613, 0.0367126502096653, 0.03123972937464714, -0.02284669317305088, -0.022461017593741417, 0....
corl_2025_XrgRvBklWu
XrgRvBklWu
corl
2,025
DexUMI: Using Human Hand as the Universal Manipulation Interface for Dexterous Manipulation
We present DexUMI - a data collection and policy learning framework that uses the human hand as the natural interface to transfer dexterous manipulation skills to various robot hands. DexUMI incorporates hardware and software adaptations to minimize the embodiment gap between the human hand and various robot hands. The...
Mengda Xu;Han Zhang;Yifan Hou;Zhenjia Xu;Linxi Fan;Manuela Veloso;Shuran Song
Stanford University+Columbia University;Stanford University+Tsinghua University;Stanford University;NVIDIA;NVIDIA;School of Computer Science, Carnegie Mellon University;Stanford University
Oral
main
Dexterous Manipulation;Learning from Human;Imitation Learning
https://openreview.net/forum?id=XrgRvBklWu
-1
DexUMI: Using Human Hand as the Universal Manipulation Interface for Dexterous Manipulation We present DexUMI - a data collection and policy learning framework that uses the human hand as the natural interface to transfer dexterous manipulation skills to various robot hands. DexUMI incorporates hardware and software ad...
[ -0.005011967848986387, -0.02634650655090809, -0.04578123614192009, -0.00651230663061142, -0.05778394639492035, -0.017075063660740852, -0.007320538628846407, -0.019639110192656517, 0.05588878318667412, 0.009350408799946308, -0.02119983360171318, 0.004830812104046345, 0.03740163519978523, 0....
corl_2025_YvsUD8C9QS
YvsUD8C9QS
corl
2,025
Mechanistic Interpretability for Steering Vision-Language-Action Models
Vision-Language-Action (VLA) models are a promising path to realizing generalist embodied agents that can quickly adapt to new tasks, modalities, and environments. However, methods for interpreting and steering VLAs fall far short of classical robotics pipelines, which are grounded in explicit models of kinematics, dyn...
Bear Häon;Kaylene Caswell Stocking;Ian Chuang;Claire Tomlin
University of California, Berkeley;University of California, Berkeley;;
Poster
main
Mechanistic Interpretability;Vision-Language-Action Models;Foundation Models for Robotics
https://openreview.net/forum?id=YvsUD8C9QS
-1
Mechanistic Interpretability for Steering Vision-Language-Action Models Vision-Language-Action (VLA) models are a promising path to realizing generalist embodied agents that can quickly adapt to new tasks, modalities, and environments. However, methods for interpreting and steering VLAs fall far short of classical robo...
[ -0.014621482230722904, -0.020073087885975838, 0.016802124679088593, 0.026242252439260483, -0.02704368531703949, 0.005968807730823755, 0.03861786052584648, 0.016466641798615456, 0.047303151339292526, 0.01601932942867279, -0.013288868591189384, -0.03438704460859299, 0.04286731779575348, 0.00...
corl_2025_Yy9EVIajH5
Yy9EVIajH5
corl
2,025
GraphEQA: Using 3D Semantic Scene Graphs for Real-time Embodied Question Answering
In Embodied Question Answering (EQA), agents must explore and develop a semantic understanding of an unseen environment in order to answer a situated question with confidence. This remains a challenging problem in robotics, due to the difficulties in obtaining useful semantic representations, updating these representat...
Saumya Saxena;Blake Buchanan;Chris Paxton;Peiqi Liu;Bingqing Chen;Narunas Vaskevicius;Luigi Palmieri;Jonathan Francis;Oliver Kroemer
Carnegie Mellon University;Neya Systems;Hello Robot;Hello Robot Inc;Bosch;Robert Bosch GmbH, Bosch;Bosch;;Carnegie Mellon University
Poster
main
Embodied Question Answering;Vision Language Models;Robot Planning;Real-time 3D Scene Graphs;Guided Exploration
https://openreview.net/forum?id=Yy9EVIajH5
-1
GraphEQA: Using 3D Semantic Scene Graphs for Real-time Embodied Question Answering In Embodied Question Answering (EQA), agents must explore and develop a semantic understanding of an unseen environment in order to answer a situated question with confidence. This remains a challenging problem in robotics, due to the di...
[ -0.08302539587020874, -0.0033516623079776764, 0.025131816044449806, -0.024553243070840836, 0.00124642055016011, -0.06682532280683517, 0.03796892985701561, 0.0017119919648393989, 0.00044268835335969925, 0.01596503145992756, -0.008588210679590702, 0.020430902019143105, 0.02113603800535202, -...
corl_2025_ZA8iXa45P2
ZA8iXa45P2
corl
2,025
Tool-as-Interface: Learning Robot Policies from Observing Human Tool Use
Tool use is essential for enabling robots to perform complex real-world tasks, but learning such skills requires extensive datasets. While teleoperation is widely used, it is slow, delay-sensitive, and poorly suited for dynamic tasks. In contrast, human videos provide a natural way for data collection without specializ...
Haonan Chen;Cheng Zhu;Shuijing Liu;Yunzhu Li;Katherine Rose Driggs-Campbell
University of Illinois Urbana-Champaign;University of Illinois, Urbana Champaign;, University of Texas at Austin;Columbia University;
Poster
main
Tool Use;Data Collection;Learning from Video
https://openreview.net/forum?id=ZA8iXa45P2
-1
Tool-as-Interface: Learning Robot Policies from Observing Human Tool Use Tool use is essential for enabling robots to perform complex real-world tasks, but learning such skills requires extensive datasets. While teleoperation is widely used, it is slow, delay-sensitive, and poorly suited for dynamic tasks. In contrast,...
[ -0.06454218924045563, -0.03563803434371948, -0.014084107242524624, 0.030560027807950974, -0.031111985445022583, -0.030320847406983376, 0.005027409642934799, 0.0077090016566216946, 0.021526329219341278, 0.06078888103365898, -0.01998084969818592, -0.03598760813474655, 0.058801837265491486, 0...
corl_2025_ZPJo9RJL15
ZPJo9RJL15
corl
2,025
Unsupervised Skill Discovery as Exploration for Learning Agile Locomotion
Exploration is crucial for legged robots to learn agile locomotion behaviors capable of overcoming diverse obstacles. For example, a robot may need to try different contact patterns and momentum profiles to successfully jump over an obstacle—but encouraging such diverse exploration is inherently challenging. As a resu...
Seungeun Rho;Kartik Garg;Morgan Byrd;Sehoon Ha
Georgia Institute of Technology;Georgia Institute of Technology;Georgia Institute of Technology;Georgia Institute of Technology
Poster
main
Unsupervised Reinforcement Learning;Locomotion;Quadruped;Skill Discovery
https://openreview.net/forum?id=ZPJo9RJL15
-1
Unsupervised Skill Discovery as Exploration for Learning Agile Locomotion Exploration is crucial for legged robots to learn agile locomotion behaviors capable of overcoming diverse obstacles. For example, a robot may need to try different contact patterns and momentum profiles to successfully jump over an obstacle—but...
[ -0.08307940512895584, -0.024006471037864685, -0.004297768697142601, 0.009820829145610332, -0.004642237443476915, 0.009085653349757195, 0.010801061987876892, -0.012974221259355545, -0.009598889388144016, -0.013982197269797325, -0.021657608449459076, -0.02594844251871109, -0.001673794700764119...
corl_2025_ZmASpafbOc
ZmASpafbOc
corl
2,025
Subteaming and Adaptive Formation Control for Coordinated Multi-Robot Navigation
Coordinated multi-robot navigation is essential for robots to operate as a team in diverse environments. During navigation, robot teams usually need to maintain specific formations, such as circular formations to protect human teammates at the center. However, in complex scenarios such as narrow corridors, rigidly pr...
Zihao Deng;Peng Gao;Williard Joshua Jose;Maggie Wigness;John G. Rogers III;Brian Reily;Christopher M. Reardon;Hao Zhang
University of Massachusetts at Amherst+University of Massachusetts at Amherst;North Carolina State University;University of Massachusetts at Amherst+Samsung Research;CCDC Army Research Laboratory+DEVCOM Army Research Laboratory;DEVCOM Army Research Laboratory;DEVCOM U.S. Army Research Laboratory;The MITRE Corporation;
Poster
main
Coordinated multi-robot navigation;Hierarchical learning
https://openreview.net/forum?id=ZmASpafbOc
-1
Subteaming and Adaptive Formation Control for Coordinated Multi-Robot Navigation Coordinated multi-robot navigation is essential for robots to operate as a team in diverse environments. During navigation, robot teams usually need to maintain specific formations, such as circular formations to protect human teammates a...
[ -0.03614165633916855, -0.05145972594618797, -0.027377773076295853, 0.015589600428938866, 0.005060768220573664, -0.02421303652226925, -0.010149626061320305, -0.005294846370816231, 0.016057755798101425, 0.019662559032440186, -0.02593585103750229, -0.028183002024888992, -0.00411275215446949, ...
corl_2025_ZqBXnR6ppz
ZqBXnR6ppz
corl
2,025
Generalist Robot Manipulation beyond Action Labeled Data
Recent advances in generalist robot manipulation leverage pre-trained Vision–Language Models (VLMs) and large-scale robot demonstrations to tackle diverse tasks in a zero-shot manner. A key challenge remains: scaling high-quality, action-labeled robot demonstration data, which existing methods rely on for robustness an...
Alexander Spiridonov;Jan-Nico Zaech;Nikolay Nikolov;Luc Van Gool;Danda Pani Paudel
ETHZ - ETH Zurich;;Institute for Computer Science, Artificial Intelligence and Technology;Sofia Un. St. Kliment Ohridski;
Poster
main
Vision-Language-Action Models;Learning from Videos
https://openreview.net/forum?id=ZqBXnR6ppz
-1
Generalist Robot Manipulation beyond Action Labeled Data Recent advances in generalist robot manipulation leverage pre-trained Vision–Language Models (VLMs) and large-scale robot demonstrations to tackle diverse tasks in a zero-shot manner. A key challenge remains: scaling high-quality, action-labeled robot demonstrati...
[ -0.047454845160245895, -0.023176901042461395, -0.024516500532627106, 0.006583304610103369, 0.0014680540189146996, 0.0019669628236442804, -0.01256103627383709, -0.0012891348451375961, -0.022057510912418365, 0.006482375785708427, -0.050978172570466995, -0.03303121402859688, -0.0070237205363810...
corl_2025_a2RMXJbkJ8
a2RMXJbkJ8
corl
2,025
ARCH: Hierarchical Hybrid Learning for Long-Horizon Contact-Rich Robotic Assembly
Generalizable long-horizon robotic assembly requires reasoning at multiple levels of abstraction. While end-to-end imitation learning (IL) is a promising approach, it typically requires large amounts of expert demonstration data and often struggles to achieve the high precision demanded by assembly tasks. Reinforcement...
Jiankai Sun;Aidan Curtis;Yang You;Yan Xu;Michael Koehle;Qianzhong Chen;Suning Huang;Leonidas Guibas;Sachin Chitta;Mac Schwager;Hui Li
Stanford University;Massachusetts Institute of Technology;Stanford University;University of Michigan - Ann Arbor;;Stanford University;;Stanford University;Autodesk;Stanford University;
Poster
main
Long-horizon learning;hybrid learning;robotic assembly
https://openreview.net/forum?id=a2RMXJbkJ8
-1
ARCH: Hierarchical Hybrid Learning for Long-Horizon Contact-Rich Robotic Assembly Generalizable long-horizon robotic assembly requires reasoning at multiple levels of abstraction. While end-to-end imitation learning (IL) is a promising approach, it typically requires large amounts of expert demonstration data and often...
[ -0.01692875102162361, -0.024186551570892334, 0.0007931051077321172, 0.01980067417025566, -0.013974247500300407, -0.03605961427092552, -0.006523762363940477, -0.016121309250593185, -0.013065875507891178, 0.014433021657168865, -0.02809530310332775, -0.022534966468811035, 0.00815699715167284, ...
corl_2025_a5LFUOlkIj
a5LFUOlkIj
corl
2,025
ATK: Automatic Task-driven Keypoint Selection for Robust Policy Learning
Learning visuamotor policy through imitation learning often suffers from perceptual challenges, where visual differences between training and evaluation environments degrade policy performance. Policies relying on state estimations like 6D pose, require task-specific tracking and are difficult to scale, while raw senso...
Yunchu Zhang;Shubham Mittal;Zhengyu Zhang;Liyiming Ke;Siddhartha Srinivasa;Abhishek Gupta
;;University of Washington;Physical Intelligence;Cruise+University of Washington;University of Washington
Poster
main
Imitation learning;Sim2real;Representation Learning
https://openreview.net/forum?id=a5LFUOlkIj
-1
ATK: Automatic Task-driven Keypoint Selection for Robust Policy Learning Learning visuamotor policy through imitation learning often suffers from perceptual challenges, where visual differences between training and evaluation environments degrade policy performance. Policies relying on state estimations like 6D pose, r...
[ -0.02664804272353649, -0.019533274695277214, -0.010866192169487476, 0.004647699650377035, -0.03492704778909683, 0.0016008230159059167, 0.02594580687582493, -0.0033194555435329676, 0.028034036979079247, 0.04220813512802124, 0.004315061029046774, -0.006029073614627123, -0.0034465049393475056, ...
corl_2025_a9RXjOt5bU
a9RXjOt5bU
corl
2,025
The Sound of Simulation: Learning Multimodal Sim-to-Real Robot Policies with Generative Audio
Robots must integrate multiple sensory modalities to act effectively in the real world. Yet, learning such multimodal policies at scale remains challenging. Simulation offers a viable solution, but while vision has benefited from high-fidelity simulators, other modalities (e.g. sound) can be notoriously difficult to si...
Renhao Wang;Haoran Geng;Tingle Li;Philipp Wu;Feishi Wang;Gopala Anumanchipalli;Trevor Darrell;Boyi Li;Pieter Abbeel;Jitendra Malik;Alexei A Efros
University of California, Berkeley;University of California, Berkeley;University of California, Berkeley;University of California, Berkeley;University of California, Berkeley+Peking University;University of California, Berkeley;;NVIDIA Research+University of California, Berkeley;Amazon+University of California, Berkele...
Oral
main
generative modeling;real2sim;sim2real;multimodal learning
https://openreview.net/forum?id=a9RXjOt5bU
-1
The Sound of Simulation: Learning Multimodal Sim-to-Real Robot Policies with Generative Audio Robots must integrate multiple sensory modalities to act effectively in the real world. Yet, learning such multimodal policies at scale remains challenging. Simulation offers a viable solution, but while vision has benefited f...
[ -0.05023432523012161, -0.014339413493871689, 0.007072317413985729, 0.007345935329794884, -0.03418827801942825, -0.024096893146634102, 0.009803855791687965, 0.027565807104110718, 0.01960771158337593, -0.0010179500095546246, -0.02836347185075283, -0.010137761943042278, 0.0015014183009043336, ...
corl_2025_aSUNzvEJIf
aSUNzvEJIf
corl
2,025
Fabrica: Dual-Arm Assembly of General Multi-Part Objects via Integrated Planning and Learning
Multi-part assembly poses significant challenges for robotic systems to execute long-horizon, contact-rich manipulation with generalization across complex geometries. We present a dual-arm robotic system capable of end-to-end planning and control for autonomous assembly of general multi-part objects. For planning over ...
Yunsheng Tian;Joshua Jacob;Yijiang Huang;Jialiang Zhao;Edward Li Gu;Pingchuan Ma;Annan Zhang;Farhad Javid;Branden Romero;Sachin Chitta;Shinjiro Sueda;Hui Li;Wojciech Matusik
Massachusetts Institute of Technology;Carnegie Mellon University+Massachusetts Institute of Technology;ETHZ - ETH Zurich;Massachusetts Institute of Technology;;OpenAI+Massachusetts Institute of Technology;Massachusetts Institute of Technology;;Massachusetts Institute of Technology;Autodesk;Texas A&M University - Colleg...
Oral
main
Assembly;Planning;Reinforcement Learning;Benchmark
https://openreview.net/forum?id=aSUNzvEJIf
-1
Fabrica: Dual-Arm Assembly of General Multi-Part Objects via Integrated Planning and Learning Multi-part assembly poses significant challenges for robotic systems to execute long-horizon, contact-rich manipulation with generalization across complex geometries. We present a dual-arm robotic system capable of end-to-end ...
[ -0.0050520095974206924, -0.00020881678210571408, -0.025417057797312737, 0.025583302602171898, -0.04026830568909645, -0.03823641687631607, 0.00530599569901824, 0.020836075767874718, -0.012034311890602112, 0.03670326620340347, -0.012117434293031693, 0.02090996317565441, 0.017880605533719063, ...
corl_2025_aZwWRycAXi
aZwWRycAXi
corl
2,025
GraspQP: Differentiable Optimization of Force Closure for Diverse and Robust Dexterous Grasping
Dexterous robotic hands enable versatile interactions through the flexibility and adaptability of a multi-finger setup, allowing for a wise range of task-specific grasp configurations in diverse environments. However, access to diverse and high-quality grasp data is essential to fully exploit the capabilities of dexter...
René Zurbrügg;Andrei Cramariuc;Marco Hutter
ETHZ - ETH Zurich;ETHZ - ETH Zurich;ETHZ - ETH Zurich
Poster
main
Grasping;Manipulation;Optimization
https://openreview.net/forum?id=aZwWRycAXi
-1
GraspQP: Differentiable Optimization of Force Closure for Diverse and Robust Dexterous Grasping Dexterous robotic hands enable versatile interactions through the flexibility and adaptability of a multi-finger setup, allowing for a wise range of task-specific grasp configurations in diverse environments. However, access...
[ -0.025079522281885147, -0.0333220511674881, -0.047454752027988434, -0.001058679074048996, -0.03437783569097519, -0.0471954345703125, 0.05275220051407814, 0.0014077131636440754, 0.008423126302659512, 0.011002389714121819, -0.017902037128806114, 0.023801466450095177, -0.018541064113378525, -...
corl_2025_b24y5SENo5
b24y5SENo5
corl
2,025
Latent Theory of Mind: A Decentralized Diffusion Architecture for Cooperative Manipulation
We present Latent Theory of Mind (LatentToM), a decentralized diffusion policy architecture for collaborative robot manipulation. Our policy allows multiple manipulators with their own perception and computation to collaborate with each other towards a common task goal with or without explicit communication. Our key in...
Chengyang He;Gadiel Mark Sznaier Camps;Xu Liu;Mac Schwager;Guillaume Adrien Sartoretti
Stanford University+National University of Singapore;Stanford University;Stanford University;Stanford University;National University of Singapore
Oral
main
Cooperative Manipulation;Diffusion Policy;Consensus Learning
https://openreview.net/forum?id=b24y5SENo5
-1
Latent Theory of Mind: A Decentralized Diffusion Architecture for Cooperative Manipulation We present Latent Theory of Mind (LatentToM), a decentralized diffusion policy architecture for collaborative robot manipulation. Our policy allows multiple manipulators with their own perception and computation to collaborate wi...
[ -0.03369418904185295, -0.03627753257751465, -0.010379507206380367, 0.011634274385869503, -0.0196518674492836, -0.020482229068875313, -0.022345926612615585, 0.00851119589060545, 0.013783985748887062, 0.008266701363027096, -0.021718543022871017, -0.027937021106481552, -0.018277158960700035, ...
corl_2025_b2mXmmGX8E
b2mXmmGX8E
corl
2,025
IRIS: An Immersive Robot Interaction System
This paper introduces IRIS, an Immersive Robot Interaction System leveraging Extended Reality (XR). Existing XR-based systems enable efficient data collection but are often challenging to reproduce and reuse due to their specificity to particular robots, objects, simulators, and environments. IRIS addresses these issue...
Xinkai Jiang;Qihao Yuan;Enes Ulas Dincer;Hongyi Zhou;Ge Li;Xueyin Li;Xiaogang Jia;Timo Schnizer;Nicolas Schreiber;Weiran Liao;Julius Haag;Kailai Li;Gerhard Neumann;Rudolf Lioutikov
Karlsruher Institut für Technologie;University of Groningen;Karlsruher Institut für Technologie;Karlsruher Institut für Technologie;Karlsruhe Institute of Technology;Karlsruher Institut für Technologie;Karlsruher Institut für Technologie;Karlsruher Institut für Technologie;Karlsruher Institut für Technologie;;Karlsruhe...
Poster
main
Human-Robot Interaction;Extended Reality;Robot Learning: Imitation Learning
https://openreview.net/forum?id=b2mXmmGX8E
-1
IRIS: An Immersive Robot Interaction System This paper introduces IRIS, an Immersive Robot Interaction System leveraging Extended Reality (XR). Existing XR-based systems enable efficient data collection but are often challenging to reproduce and reuse due to their specificity to particular robots, objects, simulators, ...
[ -0.03153252974152565, -0.003741698805242777, -0.0035278876312077045, 0.04302256554365158, -0.06544487178325653, -0.025843286886811256, 0.03123505227267742, -0.010634791105985641, 0.020879143849015236, 0.03629215434193611, -0.03173704445362091, -0.014353250153362751, 0.018081003800034523, 0...
corl_2025_b86nyIOJWq
b86nyIOJWq
corl
2,025
exUMI: Extensible Robot Teaching System with Action-aware Task-agnostic Tactile Representation
Tactile-aware robot learning faces critical challenges in data collection and representation due to data scarcity and sparsity, and the absence of force feedback in existing systems. To address these limitations, we introduce a tactile robot learning system with both hardware and algorithm innovations. We present exUMI...
Yue Xu;Litao Wei;Pengyu An;Qingyu Zhang;Yong-Lu Li
Shanghai Jiaotong University;Shanghai Jiaotong University;Shanghai Jiaotong University;;Shanghai Jiaotong University
Poster
main
Tactile Sensing;Robot Data Collection System;Imitation Learning
https://openreview.net/forum?id=b86nyIOJWq
-1
exUMI: Extensible Robot Teaching System with Action-aware Task-agnostic Tactile Representation Tactile-aware robot learning faces critical challenges in data collection and representation due to data scarcity and sparsity, and the absence of force feedback in existing systems. To address these limitations, we introduce...
[ -0.04739982634782791, -0.05924978107213974, -0.007834035903215408, 0.04011320695281029, -0.03039771318435669, -0.03148334473371506, -0.01530005969107151, 0.02918327786028385, 0.054355233907699585, 0.04574377462267876, -0.00021706902771256864, -0.0012500875163823366, 0.009347480721771717, 0...
corl_2025_b8RqTaDyb3
b8RqTaDyb3
corl
2,025
From Tabula Rasa to Emergent Abilities: Discovering Robot Skills via Real-World Unsupervised Quality-Diversity
Autonomous skill discovery aims to enable robots to acquire diverse be-haviors without explicit supervision. Learning such behaviors directly on physical hardware remains challenging due to safety and data efficiency constraints. Existing methods, including Quality-Diversity Actor-Critic (QDAC), require manually define...
Luca Grillotti;Lisa Coiffard;Oscar Pang;Maxence Faldor;Antoine Cully
Sakana AI+Imperial College London;Merantix Momentum;;Imperial College London;Imperial College London
Poster
main
Quality-Diversity;Reinforcement Learning;Real-World Learning;Robotics
https://openreview.net/forum?id=b8RqTaDyb3
-1
From Tabula Rasa to Emergent Abilities: Discovering Robot Skills via Real-World Unsupervised Quality-Diversity Autonomous skill discovery aims to enable robots to acquire diverse be-haviors without explicit supervision. Learning such behaviors directly on physical hardware remains challenging due to safety and data eff...
[ -0.0646454393863678, 0.02505897730588913, 0.00639078626409173, 0.023863913491368294, -0.034059297293424606, 0.02235141210258007, 0.0027122327592223883, 0.012669535353779793, 0.006120029836893082, 0.01571321114897728, -0.026758207008242607, 0.01614268682897091, -0.027953270822763443, 0.0149...
corl_2025_bILubVwPoD
bILubVwPoD
corl
2,025
Few-Shot Neuro-Symbolic Imitation Learning for Long-Horizon Planning and Acting
Imitation learning enables intelligent systems to acquire complex behaviors with minimal supervision. However, existing methods often focus on short-horizon skills, require large datasets, and struggle to solve long-horizon tasks or generalize across task variations and distribution shifts. We propose a novel neuro-sym...
Pierrick Lorang;Hong Lu;Johannes Huemer;Patrik Zips;Matthias Scheutz
Tufts University;;AIT Austrian Institute Of Technology;AIT Austrian Institute Of Technology;Tufts University
Poster
main
Neuro-symbolic;Imitation Learning;Task and Motion Planning;Symbolic Planning;Skill Learning;Human-Robot Interaction
https://openreview.net/forum?id=bILubVwPoD
-1
Few-Shot Neuro-Symbolic Imitation Learning for Long-Horizon Planning and Acting Imitation learning enables intelligent systems to acquire complex behaviors with minimal supervision. However, existing methods often focus on short-horizon skills, require large datasets, and struggle to solve long-horizon tasks or general...
[ -0.07210133969783783, -0.017035236582159996, 0.006407876964658499, 0.01734984666109085, 0.0028731347993016243, -0.018885886296629906, -0.0022635767236351967, 0.02102338895201683, 0.00080965954111889, 0.019302282482385635, -0.02981397695839405, -0.004520213697105646, -0.008258527144789696, ...
corl_2025_bOVF8Rj33i
bOVF8Rj33i
corl
2,025
VT-Refine: Learning Bimanual Assembly with Visuo-Tactile Feedback via Simulation Fine-Tuning
Humans excel at bimanual assembly tasks by adapting to rich tactile feedback—a capability that remains difficult to replicate in robots through behavioral cloning alone, due to the suboptimality and limited diversity of human demonstrations. In this work, we present VT-Refine, a visuo-tactile policy learning framework ...
Binghao Huang;Jie Xu;Iretiayo Akinola;Wei Yang;Balakumar Sundaralingam;Rowland O'Flaherty;Dieter Fox;Xiaolong Wang;Arsalan Mousavian;Yu-Wei Chao;Yunzhu Li
Columbia University;NVIDIA;NVIDIA;NVIDIA;NVIDIA;NVIDIA;NVIDIA Research+Department of Computer Science;University of California, San Diego;NVIDIA;NVIDIA;Columbia University
Poster
main
Tactile Simulation;Bimanual Manipulation;RL Fine-Tuning
https://openreview.net/forum?id=bOVF8Rj33i
-1
VT-Refine: Learning Bimanual Assembly with Visuo-Tactile Feedback via Simulation Fine-Tuning Humans excel at bimanual assembly tasks by adapting to rich tactile feedback—a capability that remains difficult to replicate in robots through behavioral cloning alone, due to the suboptimality and limited diversity of human d...
[ -0.06300436705350876, -0.02601131983101368, -0.019939513877034187, 0.03240170702338219, -0.021851006895303726, -0.0006031983648426831, 0.041153356432914734, 0.02209462970495224, 0.011215976439416409, 0.056707676500082016, 0.000758390175178647, 0.008006726391613483, 0.010916133411228657, 0....
corl_2025_bU15EK0oqk
bU15EK0oqk
corl
2,025
CASPER: Inferring Diverse Intents for Assistive Teleoperation with Vision Language Models
Assistive teleoperation, where control is shared between a human and a robot, enables efficient and intuitive human-robot collaboration in diverse and unstructured environments. A central challenge in real-world assistive teleoperation is for the robot to infer a wide range of human intentions from user control inputs ...
Huihan Liu;Rutav Shah;Shuijing Liu;Jack Pittenger;Mingyo Seo;Yuchen Cui;Yonatan Bisk;Roberto Martín-Martín;Yuke Zhu
;University of Texas at Austin;, University of Texas at Austin;;University of Texas at Austin;University of California, Los Angeles;Carnegie Mellon University;Amazon+University of Texas at Austin;Computer Science Department, University of Texas, Austin
Poster
main
Assistive Teleoperation;Mobile Manipulation;Vision Language Models
https://openreview.net/forum?id=bU15EK0oqk
-1
CASPER: Inferring Diverse Intents for Assistive Teleoperation with Vision Language Models Assistive teleoperation, where control is shared between a human and a robot, enables efficient and intuitive human-robot collaboration in diverse and unstructured environments. A central challenge in real-world assistive teleoper...
[ -0.02526962198317051, -0.031069813296198845, -0.007508846465498209, 0.022517303004860878, -0.021612176671624184, 0.0301646888256073, 0.007601206190884113, -0.0037105523515492678, 0.021778425201773643, 0.026082389056682587, -0.013244385831058025, -0.024032002314925194, 0.013004250824451447, ...
corl_2025_bi8o9p6h2R
bi8o9p6h2R
corl
2,025
Improving Efficiency of Sampling-based Motion Planning via Message-Passing Monte Carlo
Sampling-based motion planning methods, while effective in high-dimensional spaces, often suffer from inefficiencies due to irregular sampling distributions, leading to suboptimal exploration of the configuration space. In this paper, we propose an approach that enhances the efficiency of these methods by utilizing low...
Makram Chahine;T. Konstantin Rusch;Zach J Patterson;Daniela Rus
Massachusetts Institute of Technology;Max Planck Institute for Intelligent Systems, Max-Planck Institute+ELLIS Institute Tübingen+Massachusetts Institute of Technology;Case Western Reserve University;Massachusetts Institute of Technology
Poster
main
Graph Neural Networks;Discrepancy Theory;Motion Planning
https://openreview.net/forum?id=bi8o9p6h2R
-1
Improving Efficiency of Sampling-based Motion Planning via Message-Passing Monte Carlo Sampling-based motion planning methods, while effective in high-dimensional spaces, often suffer from inefficiencies due to irregular sampling distributions, leading to suboptimal exploration of the configuration space. In this paper...
[ -0.08490514755249023, -0.021153781563043594, -0.05042829364538193, -0.003004489466547966, 0.007685692515224218, -0.013232442550361156, -0.003971998579800129, -0.01324150525033474, -0.028495067730545998, 0.015425764955580235, -0.04350392147898674, -0.003004489466547966, -0.018833572044968605,...
corl_2025_bmpDAqsJov
bmpDAqsJov
corl
2,025
Latent Adaptive Planner for Dynamic Manipulation
This paper presents Latent Adaptive Planner (LAP), a novel approach for dynamic nonprehensile manipulation tasks that formulates planning as latent space inference, effectively learned from human demonstration videos. Our method addresses key challenges in visuomotor policy learning through a principled variationa...
Donghun Noh;Deqian Kong;Minglu Zhao;Andrew Lizarraga;Jianwen Xie;Ying Nian Wu;Dennis Hong
Dexterity, Inc;University of California, Los Angeles;University of California, Los Angeles;University of California, Los Angeles;;;
Poster
main
Imitation Learning;Robotics;Dynamic Nonprehensile Manipulation;Latent Space Planning;Classical Variational Bayes;Video-based Skill Acquisition
https://openreview.net/forum?id=bmpDAqsJov
-1
Latent Adaptive Planner for Dynamic Manipulation This paper presents Latent Adaptive Planner (LAP), a novel approach for dynamic nonprehensile manipulation tasks that formulates planning as latent space inference, effectively learned from human demonstration videos. Our method addresses key challenges in visuomoto...
[ -0.028449416160583496, -0.024229826405644417, -0.010108263231813908, 0.028449416160583496, -0.039270322769880295, 0.015162395313382149, 0.022410713136196136, 0.012208681553602219, 0.0039617265574634075, 0.057573966681957245, -0.011524170637130737, 0.004554813262075186, -0.03038105182349682, ...
corl_2025_brTSiML1nh
brTSiML1nh
corl
2,025
AimBot: A Simple Auxiliary Visual Cue to Enhance Spatial Awareness of Visuomotor Policies
In this paper, we propose AimBot, a lightweight visual augmentation technique that provides explicit spatial cues to improve visuomotor policy learning in robotic manipulation. AimBot overlays shooting lines and scope reticles onto multi-view RGB images, offering auxiliary visual guidance that encodes the end-effector'...
Yinpei Dai;Jayjun Lee;Yichi Zhang;Ziqiao Ma;Jianing Yang;Amir Zadeh;Chuan Li;Nima Fazeli;Joyce Chai
University of Michigan - Ann Arbor;University of Michigan - Ann Arbor+University of Michigan - Ann Arbor;University of Michigan;University of Michigan+International Business Machines+Adobe Research;Meta+University of Michigan - Ann Arbor;;Lambda;University of Michigan;
Poster
main
Robotic Manipulation;Visuomotor Policy;Imitation Learning
https://openreview.net/forum?id=brTSiML1nh
-1
AimBot: A Simple Auxiliary Visual Cue to Enhance Spatial Awareness of Visuomotor Policies In this paper, we propose AimBot, a lightweight visual augmentation technique that provides explicit spatial cues to improve visuomotor policy learning in robotic manipulation. AimBot overlays shooting lines and scope reticles ont...
[ -0.04372942075133324, -0.03573370724916458, 0.009614983573555946, -0.006252941209822893, -0.04043599218130112, -0.030610594898462296, 0.019431231543421745, 0.046473946422338486, 0.0278477743268013, 0.024279892444610596, -0.04687647521495819, -0.012890116311609745, 0.01056641899049282, 0.03...
corl_2025_bt1Tovn0SW
bt1Tovn0SW
corl
2,025
Training Strategies for Efficient Embodied Reasoning
Robot chain-of-thought reasoning (CoT) -- wherein a model predicts helpful intermediate representations before choosing actions -- provides an effective method for improving the generalization and performance of robot policies, especially vision-language-action models (VLAs). While such approaches have been shown to im...
William Chen;Suneel Belkhale;Suvir Mirchandani;Karl Pertsch;Danny Driess;Oier Mees;Sergey Levine
University of California, Berkeley;Stanford University;Stanford University;University of California, Berkeley+Stanford University;Physical Intelliigence;Electrical Engineering & Computer Science Department, University of California, Berkeley;University of California, Berkeley
Oral
main
robot reasoning;vision-language-action models
https://openreview.net/forum?id=bt1Tovn0SW
-1
Training Strategies for Efficient Embodied Reasoning Robot chain-of-thought reasoning (CoT) -- wherein a model predicts helpful intermediate representations before choosing actions -- provides an effective method for improving the generalization and performance of robot policies, especially vision-language-action model...
[ -0.03361395373940468, -0.02899635210633278, 0.015770742669701576, 0.03690444678068161, -0.023378876969218254, -0.023106183856725693, 0.0066309841349720955, 0.024978676810860634, 0.009289740584790707, 0.014861767180263996, 0.005417500156909227, -0.015434421598911285, 0.00042352627497166395, ...
corl_2025_cUeY476ohd
cUeY476ohd
corl
2,025
MirrorDuo: Reflection-Consistent Visuomotor Learning from Mirrored Demonstration Pairs
Image-based behaviour cloning leverages demonstrations captured from ubiquitous RGB cameras, enabling impressive visuomotor performance. However, it remains constrained by the cost of collecting sufficiently diverse demonstrations, especially for generalizing across workspace variations. We propose MirrorDuo, a mirrori...
Zheyu Zhuang;Ruiyu Wang;Giovanni Luca Marchetti;Florian T. Pokorny;Danica Kragic
KTH Royal Institute of Technology;;KTH Royal Institute of Technology;;KTH
Poster
main
Behavior Cloning;Data Efficiency;Robotic Manipulation
https://openreview.net/forum?id=cUeY476ohd
-1
MirrorDuo: Reflection-Consistent Visuomotor Learning from Mirrored Demonstration Pairs Image-based behaviour cloning leverages demonstrations captured from ubiquitous RGB cameras, enabling impressive visuomotor performance. However, it remains constrained by the cost of collecting sufficiently diverse demonstrations, e...
[ -0.07484128326177597, -0.029865091666579247, -0.029056910425424576, -0.005577385891228914, -0.010055270045995712, 0.029282450675964355, 0.0016234092181548476, 0.018945256248116493, 0.022422311827540398, -0.0030165810603648424, 0.0032186261378228664, -0.006456047296524048, -0.0039210855029523...
corl_2025_cpmwi3Xwcr
cpmwi3Xwcr
corl
2,025
RoboArena: Distributed Real-World Evaluation of Generalist Robot Policies
Comprehensive, unbiased, and comparable evaluation of modern generalist policies is uniquely challenging: existing approaches for robot benchmarking typically rely on heavy standardization, either by specifying fixed evaluation tasks and environments, or by hosting centralized "robot challenges", and do not readily sca...
Pranav Atreya;Karl Pertsch;Tony Lee;Moo Jin Kim;Arhan Jain;Artur Kuramshin;Cyrus Neary;Edward S. Hu;Kanav Arora;Kirsty Ellis;Luca Macesanu;Matthew Leonard;Meedeum Cho;Ozgur Aslan;Shivin Dass;Tony Wang;Xingfang Yuan;Abhishek Gupta;Dinesh Jayaraman;Glen Berseth;Kostas Daniilidis;Roberto Martín-Martín;Youngwoon Lee;Percy ...
University of California, Berkeley;University of California, Berkeley+Stanford University;Stanford University;Stanford University;;Université de Montréal;University of British Columbia+Mila - Quebec Artificial Intelligence Institute;University of Pennsylvania;University of Washington;;University of Texas at Austin;Univ...
Oral
main
Generalist Robot Policy Evaluation;Crowd-Sourced Evaluation;Robot Foundation Models
https://openreview.net/forum?id=cpmwi3Xwcr
-1
RoboArena: Distributed Real-World Evaluation of Generalist Robot Policies Comprehensive, unbiased, and comparable evaluation of modern generalist policies is uniquely challenging: existing approaches for robot benchmarking typically rely on heavy standardization, either by specifying fixed evaluation tasks and environm...
[ -0.058996714651584625, -0.02207738719880581, -0.017439279705286026, -0.02202172949910164, -0.02888612635433674, 0.016205543652176857, -0.02217014878988266, 0.031335048377513885, 0.026585625484585762, 0.005287441425025463, -0.024693278595805168, 0.025880634784698486, 0.007119493559002876, 0...
corl_2025_dBaSaa7qi4
dBaSaa7qi4
corl
2,025
CoRI: Communication of Robot Intent for Physical Human-Robot Interaction
Clear communication of robot intent fosters transparency and interpretability in physical human-robot interaction (pHRI), particularly during assistive tasks involving direct human-robot contact. We introduce CoRI, a pipeline that automatically generates natural language communication of a robot's upcoming actions dire...
Junxiang Wang;Emek Barış Küçüktabak;Rana Soltani Zarrin;Zackory Erickson
Carnegie Mellon University;;Honda Research Institution US;Carnegie Mellon University
Poster
main
Robot intent generation;Human-robot communication;Assistive robotics
https://openreview.net/forum?id=dBaSaa7qi4
-1
CoRI: Communication of Robot Intent for Physical Human-Robot Interaction Clear communication of robot intent fosters transparency and interpretability in physical human-robot interaction (pHRI), particularly during assistive tasks involving direct human-robot contact. We introduce CoRI, a pipeline that automatically ge...
[ -0.04284623637795448, -0.03183969482779503, 0.01231552753597498, 0.023469559848308563, -0.025331638753414154, 0.011080287396907806, -0.03772091120481491, -0.02339581400156021, 0.010490321554243565, 0.025331638753414154, -0.010942013934254646, -0.02929546870291233, 0.035158250480890274, 0.0...
corl_2025_dT45OMevL5
dT45OMevL5
corl
2,025
3DS-VLA: A 3D Spatial-Aware Vision Language Action Model for Robust Multi-Task Manipulation
Recently, 2D vision-language-action (VLA) models have made significant strides in multi-task manipulation. However, these models struggle to reason about 3D spatial relationships from 2D image inputs. Although an increasing number of 3D approaches explicitly integrate 3D information, they encounter challenges such as l...
Xiaoqi Li;Liang Heng;Jiaming Liu;Yan Shen;Chenyang Gu;Zhuoyang Liu;Hao Chen;Nuowei Han;Renrui Zhang;Hao Tang;Shanghang Zhang;Hao Dong
;;Peking University;;Peking University;Peking University;Department of Computer Science and Engineering, The Chinese University of Hong Kong;Beijing University of Posts and Telecommunications;MMLab of CUHK & Shanghai AI Laboratory;Peking University;;Peking University+Peking University
Poster
main
Vision-Language-Action; Robotic Manipulation; Imitation Learning
https://openreview.net/forum?id=dT45OMevL5
-1
3DS-VLA: A 3D Spatial-Aware Vision Language Action Model for Robust Multi-Task Manipulation Recently, 2D vision-language-action (VLA) models have made significant strides in multi-task manipulation. However, these models struggle to reason about 3D spatial relationships from 2D image inputs. Although an increasing numb...
[ -0.032806530594825745, -0.001169905997812748, -0.006847357843071222, 0.0014928312739357352, -0.03797333687543869, 0.04895056411623955, -0.004337702877819538, 0.01593843288719654, 0.03312833979725838, -0.00173306989017874, 0.011737051419913769, 0.0043332334607839584, 0.014776348136365414, 0...
corl_2025_dmXFboqSnX
dmXFboqSnX
corl
2,025
Learning Impact-Rich Rotational Maneuvers via Centroidal Velocity Rewards and Sim-to-Real Techniques: A One-Leg Hopper Flip Case Study
Dynamic rotational maneuvers, such as front flips, inherently involve large angular momentum generation and intense impact forces, presenting major challenges for reinforcement learning and sim-to-real transfer. In this work, we propose a general framework for learning and deploying impact-rich, rotation-intensive beha...
Dongyun Kang;Gijeong Kim;JongHun Choe;Hajun Kim;Hae-Won Park
Korea Advanced Institute of Science & Technology;Korea Advanced Institute of Science & Technology;;Korea Advanced Institute of Science & Technology;Korea Advanced Institute of Science & Technology
Poster
main
Reinforcement Learning;Sim-to-Real Transfer;One-Leg Hopper
https://openreview.net/forum?id=dmXFboqSnX
-1
Learning Impact-Rich Rotational Maneuvers via Centroidal Velocity Rewards and Sim-to-Real Techniques: A One-Leg Hopper Flip Case Study Dynamic rotational maneuvers, such as front flips, inherently involve large angular momentum generation and intense impact forces, presenting major challenges for reinforcement learning...
[ -0.09924199432134628, -0.04567345604300499, -0.000956910545937717, 0.005958209279924631, -0.029957063496112823, -0.0071480064652860165, -0.012340687215328217, 0.048403531312942505, -0.0015794788487255573, -0.026452235877513885, -0.04058222845196724, -0.011206229217350483, -0.0210843123495578...
corl_2025_eLeCrM5PEO
eLeCrM5PEO
corl
2,025
Self-supervised perception for tactile skin covered dexterous hands
We present PercepSkin, a pre-trained encoder for magnetic skin sensors distributed across the fingertips, phalanges, and palm of a dexterous robot hand. Magnetic tactile skins offer a flexible form factor for hand-wide coverage with fast response times, in contrast to vision-based tactile sensors that are restricted t...
Akash Sharma;Carolina Higuera;Chaithanya Krishna Bodduluri;Zixi Liu;Taosha Fan;Tess Hellebrekers;Mike Lambeta;Byron Boots;Michael Kaess;Tingfan Wu;Francois Robert Hogan;Mustafa Mukadam
Meta Facebook+Carnegie Mellon University;University of Washington;Meta Facebook;;;;Meta;;Carnegie Mellon University;;Meta Facebook;Amazon Robotics
Poster
main
Robot Perception;Sensing & Vision;Representation learning;Foundation models;Tactile sensing
https://openreview.net/forum?id=eLeCrM5PEO
-1
Self-supervised perception for tactile skin covered dexterous hands We present PercepSkin, a pre-trained encoder for magnetic skin sensors distributed across the fingertips, phalanges, and palm of a dexterous robot hand. Magnetic tactile skins offer a flexible form factor for hand-wide coverage with fast response time...
[ -0.043144747614860535, -0.01266566850244999, -0.032552991062402725, 0.0017753230640664697, 0.011563903652131557, -0.034145455807447433, 0.01590615324676037, -0.025442440062761307, 0.02040579915046692, 0.0641060620546341, -0.012702702544629574, 0.010258451104164124, -0.0019755386747419834, ...
corl_2025_ejV8YHTpHR
ejV8YHTpHR
corl
2,025
Leveraging Correlation Across Test Platforms for Variance-Reduced Metric Estimation
Learning-based robotic systems demand rigorous validation to assure reliable performance, but extensive real‐world testing is often prohibitively expensive and if conducted may still yield insufficient data for high-confidence guarantees. In this work, we introduce a general estimation framework that leverages *paired*...
Rachel Luo;Heng Yang;Michael Watson;Apoorva Sharma;Sushant Veer;Edward Schmerling;Marco Pavone
NVIDIA;Harvard University+NVIDIA;;NVIDIA;NVIDIA;NVIDIA;NVIDIA+Stanford University
Poster
main
metric estimation;sample efficiency;control variates
https://openreview.net/forum?id=ejV8YHTpHR
-1
Leveraging Correlation Across Test Platforms for Variance-Reduced Metric Estimation Learning-based robotic systems demand rigorous validation to assure reliable performance, but extensive real‐world testing is often prohibitively expensive and if conducted may still yield insufficient data for high-confidence guarantee...
[ -0.05415214225649834, -0.0005724139628000557, -0.027296651154756546, 0.0003136357699986547, -0.010459120385348797, -0.019263898953795433, -0.0056201689876616, 0.03198395296931267, 0.00032656031544320285, 0.009898481890559196, -0.04165266826748848, 0.01025692280381918, 0.028932610526680946, ...
corl_2025_etSYDtRO0Z
etSYDtRO0Z
corl
2,025
ManiFlow: A General Robot Manipulation Policy via Consistency Flow Training
Generative models based on flow matching offer significant potential for learning robot policies, particularly in generating high-dimensional, dexterous behaviors that are conditioned on diverse observations. In this work, we introduce ManiFlow, an advanced flow matching model specifically designed to support dexterous...
Ge Yan;Jiyue Zhu;Yuquan Deng;Shiqi Yang;Ri-Zhao Qiu;Xuxin Cheng;Marius Memmel;Ranjay Krishna;Ankit Goyal;Xiaolong Wang;Dieter Fox
Department of Computer Science, University of Washington;University of California, San Diego;;University of California, San Diego;University of California, San Diego;University of California, San Diego;University of Washington;University of Washington;NVIDIA;University of California, San Diego;NVIDIA Research+Departmen...
Poster
main
Imitation Learning;Dexterous Manipulation
https://openreview.net/forum?id=etSYDtRO0Z
-1
ManiFlow: A General Robot Manipulation Policy via Consistency Flow Training Generative models based on flow matching offer significant potential for learning robot policies, particularly in generating high-dimensional, dexterous behaviors that are conditioned on diverse observations. In this work, we introduce ManiFlow...
[ -0.05768699571490288, -0.033595066517591476, -0.035061370581388474, -0.006807176861912012, -0.029474565759301186, -0.04640202224254608, -0.004876852501183748, 0.015962297096848488, 0.019099073484539986, 0.018031826242804527, -0.004519557114690542, -0.05271269753575325, 0.010737428441643715, ...
corl_2025_f2Y549UzM5
f2Y549UzM5
corl
2,025
Cost-aware Discovery of Contextual Failures using Bayesian Active Learning
Ensuring the robustness of robotic systems is crucial for their deployment in safety-critical domains. Failure discovery, or falsification, is a widely used approach for evaluating robustness, with recent advancements focusing on improving sample efficiency and generalization through probabilistic sampling techniques a...
Anjali Parashar;Joseph Zhang;Yingke Li;Chuchu Fan
Massachusetts Institute of Technology;Massachusetts Institute of Technology+Massachusetts Institute of Technology;Massachusetts Institute of Technology;Massachusetts Institute of Technology
Poster
main
Failure discovery;Testing;Contextual failures
https://openreview.net/forum?id=f2Y549UzM5
-1
Cost-aware Discovery of Contextual Failures using Bayesian Active Learning Ensuring the robustness of robotic systems is crucial for their deployment in safety-critical domains. Failure discovery, or falsification, is a widely used approach for evaluating robustness, with recent advancements focusing on improving sampl...
[ -0.03312820568680763, -0.04655246064066887, -0.0010146922431886196, 0.028916673734784126, 0.010068191215395927, 0.01490017119795084, 0.06159364432096481, 0.030383188277482986, 0.019168106839060783, 0.03777217119932175, -0.04046078026294708, -0.035008352249860764, 0.01249358244240284, 0.023...
corl_2025_fBRqCMqVyS
fBRqCMqVyS
corl
2,025
Enabling Long(er) Horizon Imitation for Manipulation Tasks by Modeling Subgoal Transitions
Imitation-based policy training for long-horizon manipulation tasks involving multi-step object interactions is often susceptible to compounding action errors. Contemporary methods discover semantic subgoals embedded within the overall task, decomposing the overall task into tractable shorter-horizon goal-conditioned p...
Shivam Jain;Sachit Sachdeva;Rohan Paul
Indian Institute of Technology, Delhi;;Indian Institute of Technology, Delhi
Poster
main
Long Horizon Task Execution;Goal Conditioned Policy Learning;Imitation Learning;Learning from Demonstration;Robot Manipulation
https://openreview.net/forum?id=fBRqCMqVyS
-1
Enabling Long(er) Horizon Imitation for Manipulation Tasks by Modeling Subgoal Transitions Imitation-based policy training for long-horizon manipulation tasks involving multi-step object interactions is often susceptible to compounding action errors. Contemporary methods discover semantic subgoals embedded within the o...
[ -0.038385871797800064, -0.04270428419113159, -0.015151347033679485, 0.030727151781320572, -0.040489714592695236, -0.01636935956776142, 0.010814481414854527, -0.027386844158172607, -0.004184613469988108, 0.0006943598273210227, -0.01784573867917061, -0.0019850386306643486, 0.014007152058184147...
corl_2025_gD6YV5OuW3
gD6YV5OuW3
corl
2,025
ScrewSplat: An End-to-End Method for Articulated Object Recognition
Articulated object recognition -- the task of identifying both the geometry and kinematic joints of objects with movable parts -- is essential for enabling robots to interact with everyday objects such as doors and laptops. However, existing approaches often rely on strong assumptions, such as a known number of articul...
Seungyeon Kim;Junsu HA;Young Hun Kim;Yonghyeon Lee;Frank C. Park
Seoul National University;Seoul National University;Seoul National University;Korea Institute for Advanced Study+Massachusetts Institute of Technology;Seoul National University
Oral
main
Articulated objects;Gaussian splatting;Screw theory
https://openreview.net/forum?id=gD6YV5OuW3
-1
ScrewSplat: An End-to-End Method for Articulated Object Recognition Articulated object recognition -- the task of identifying both the geometry and kinematic joints of objects with movable parts -- is essential for enabling robots to interact with everyday objects such as doors and laptops. However, existing approaches...
[ -0.032831206917762756, -0.0007914329180493951, 0.019536549225449562, 0.020969731733202934, -0.00914597138762474, -0.02954997308552265, 0.033792950212955475, -0.0431840680539608, 0.005067999474704266, 0.0461258664727211, 0.00349809811450541, 0.015604724176228046, 0.02517499215900898, 0.0293...
corl_2025_gyihSZwQbR
gyihSZwQbR
corl
2,025
Dexplore: Scalable Neural Control for Dexterous Manipulation from Reference Scoped Exploration
Hand–object motion-capture (MoCap) repositories provide abundant, contact-rich human demonstrations for scaling dexterous manipulation on robots. Yet demonstration inaccuracy and embodiment gaps between human and robot hands challenge direct policy learning. Existing pipelines adapt a three-stage workflow: retargeting,...
Sirui Xu;Yu-Wei Chao;Liuyu Bian;Arsalan Mousavian;Yu-Xiong Wang;Liangyan Gui;Wei Yang
University of Illinois, Urbana Champaign+NVIDIA;NVIDIA;University of Illinois, Urbana Champaign;NVIDIA;School of Computer Science, Carnegie Mellon University+Department of Computer Science, University of Illinois Urbana-Champaign;UIUC;NVIDIA
Poster
main
Dexterous Manipulation;Reinforcement Learning;Learning from Demonstrations
https://openreview.net/forum?id=gyihSZwQbR
-1
Dexplore: Scalable Neural Control for Dexterous Manipulation from Reference Scoped Exploration Hand–object motion-capture (MoCap) repositories provide abundant, contact-rich human demonstrations for scaling dexterous manipulation on robots. Yet demonstration inaccuracy and embodiment gaps between human and robot hands ...
[ -0.04409986734390259, -0.03442394733428955, -0.029772063717246056, 0.00505659868940711, -0.02469220571219921, 0.01966816931962967, 0.01228097639977932, 0.0029841840732842684, 0.019761206582188606, 0.03237712010741234, -0.03129788115620613, -0.001043766620568931, 0.02186385914683342, -0.010...
corl_2025_h2K52fhsDU
h2K52fhsDU
corl
2,025
Pointing3D: A Benchmark for 3D Object Referral via Pointing Gestures
Pointing gestures provide a natural and efficient way to communicate spatial information in human-machine interaction, yet their potential for 3D object referral remains largely under-explored. To fill this gap, we introduce the task of pointing-based 3D segmentation. In this task, given an image of a person pointing a...
Mert Arslanoglu;Kadir Yilmaz;Cemhan Kaan Özaltan;Timm Linder;Bastian Leibe
Rheinisch Westfälische Technische Hochschule Aachen;Rheinisch Westfälische Technische Hochschule Aachen;Rheinisch Westfälische Technische Hochschule Aachen;;RWTH Aachen University
Poster
main
Object Referral;Pointing Gesture;3D Segmentation
https://openreview.net/forum?id=h2K52fhsDU
-1
Pointing3D: A Benchmark for 3D Object Referral via Pointing Gestures Pointing gestures provide a natural and efficient way to communicate spatial information in human-machine interaction, yet their potential for 3D object referral remains largely under-explored. To fill this gap, we introduce the task of pointing-based...
[ -0.02059824764728546, -0.01179953571408987, -0.047790899872779846, -0.0006199618801474571, -0.013827870599925518, 0.02432149276137352, 0.014337269589304924, -0.02515505440533161, -0.024895723909139633, 0.008256896398961544, -0.03725096583366394, 0.011456849053502083, 0.02934138849377632, 0...
corl_2025_hg9YtHV8MJ
hg9YtHV8MJ
corl
2,025
KoopMotion: Learning Almost Divergence Free Koopman Flow Fields for Motion Planning
In this work, we propose a novel flow field-based motion planning method that drives a robot from any initial state to a desired reference trajectory such that it converges to the trajectory's end point. Despite demonstrated efficacy in using Koopman operator theory for modeling dynamical systems, Koopman does not inh...
Alice Kate Li;Thales C. Silva;Victoria Edwards;Vijay Kumar;M. Ani Hsieh
University of Pennsylvania;;University of Pennsylvania;;
Poster
main
motion planning;koopman operator theory;dynamical systems;learning from demonstrations
https://openreview.net/forum?id=hg9YtHV8MJ
-1
KoopMotion: Learning Almost Divergence Free Koopman Flow Fields for Motion Planning In this work, we propose a novel flow field-based motion planning method that drives a robot from any initial state to a desired reference trajectory such that it converges to the trajectory's end point. Despite demonstrated efficacy i...
[ -0.06728191673755646, -0.032420337200164795, -0.05093303695321083, -0.00510902376845479, -0.026169298216700554, 0.013491521589457989, -0.0045195212587714195, 0.002690906636416912, 0.028222154825925827, 0.03726582229137421, -0.02110188640654087, 0.005122894421219826, 0.05740600824356079, 0....
corl_2025_hh9afiQMb2
hh9afiQMb2
corl
2,025
Real-Time Out-of-Distribution Failure Prevention via Multi-Modal Reasoning
Foundation models can provide robust high-level reasoning on appropriate safety interventions in hazardous scenarios beyond a robot's training data, i.e. out-of-distribution (OOD) failures. However, due to the high inference latency of Large Vision and Language Models, current methods rely on manually defined intervent...
Milan Ganai;Rohan Sinha;Christopher Agia;Daniel Morton;Luigi Di Lillo;Marco Pavone
Stanford University;Stanford University;Stanford University;Stanford University;Swiss Re;NVIDIA+Stanford University
Oral
main
Multi-modal Reasoning in Robotics;OOD Safety;Fallback Synthesis
https://openreview.net/forum?id=hh9afiQMb2
-1
Real-Time Out-of-Distribution Failure Prevention via Multi-Modal Reasoning Foundation models can provide robust high-level reasoning on appropriate safety interventions in hazardous scenarios beyond a robot's training data, i.e. out-of-distribution (OOD) failures. However, due to the high inference latency of Large Vis...
[ -0.016652259975671768, 0.00010897773609030992, -0.02219686284661293, 0.0040362123399972916, -0.03830480948090553, -0.024152696132659912, 0.018866410478949547, 0.03511274233460426, 0.03520499914884567, 0.02470623329281807, -0.04465204104781151, -0.017796237021684647, 0.017759334295988083, 0...
corl_2025_htgNQHa6Ta
htgNQHa6Ta
corl
2,025
TWIST: Teleoperated Whole-Body Imitation System
Teleoperating humanoid robots in a whole-body manner marks a fundamental step toward developing general-purpose robotic intelligence, with human motion providing an ideal interface for controlling all degrees of freedom. Yet, most current humanoid teleoperation systems fall short of enabling coordinated whole-body beha...
Yanjie Ze;Zixuan Chen;Joao Pedro Araujo;Zi-ang Cao;Xue Bin Peng;Jiajun Wu;Karen Liu
Stanford University;University of California, San Diego;Computer Science Department, Stanford University;Stanford University;Simon Fraser University;Stanford University;Computer Science Department, Stanford University
Poster
main
humanoid robots;whole-body teleoperation;learning-based control
https://openreview.net/forum?id=htgNQHa6Ta
-1
TWIST: Teleoperated Whole-Body Imitation System Teleoperating humanoid robots in a whole-body manner marks a fundamental step toward developing general-purpose robotic intelligence, with human motion providing an ideal interface for controlling all degrees of freedom. Yet, most current humanoid teleoperation systems fa...
[ -0.031889285892248154, -0.015401607379317284, -0.017096253111958504, 0.02454894594848156, -0.012798783369362354, 0.02069152146577835, -0.018519379198551178, -0.0005310394917614758, 0.004386415239423513, 0.029380088672041893, -0.04040932655334473, -0.02142181061208248, 0.011871878057718277, ...
corl_2025_iQQy1BKlGv
iQQy1BKlGv
corl
2,025
MEReQ: Max-Ent Residual-Q Inverse RL for Sample-Efficient Alignment from Intervention
Aligning robot behavior with human preferences is crucial for deploying embodied AI agents in human-centered environments. A promising solution is interactive imitation learning from human intervention, where a human expert observes the policy's execution and provides interventions as feedback. However, existing method...
Yuxin Chen;Chen Tang;Jianglan Wei;Chenran Li;Thomas Tian;Xiang Zhang;Wei Zhan;Peter Stone;Masayoshi Tomizuka
University of California, Berkeley;University of Texas at Austin;Huazhong University of Science and Technology+University of California, Berkeley;;University of California, Berkeley;;;Sony AI+University of Texas, Austin;
Poster
main
Interactive imitation learning;Learning from human feedback;Inverse reinforcement learning
https://openreview.net/forum?id=iQQy1BKlGv
-1
MEReQ: Max-Ent Residual-Q Inverse RL for Sample-Efficient Alignment from Intervention Aligning robot behavior with human preferences is crucial for deploying embodied AI agents in human-centered environments. A promising solution is interactive imitation learning from human intervention, where a human expert observes t...
[ -0.07344350218772888, -0.027808884158730507, 0.009687899611890316, -0.007884104736149311, -0.019707955420017242, 0.0012628869153559208, 0.05159498378634453, 0.0008355813333764672, -0.001672315993346274, 0.030761387199163437, -0.014725606888532639, -0.0005368710844777524, -0.03236681222915649...
corl_2025_iVbCWUDyBF
iVbCWUDyBF
corl
2,025
Search-TTA: A Multi-Modal Test-Time Adaptation Framework for Visual Search in the Wild
To perform autonomous visual search for environmental monitoring, a robot may leverage satellite imagery as a prior map. This can help inform coarse, high level search and exploration strategies, even when such images lack sufficient resolution to allow fine-grained, explicit visual recognition of targets. However, the...
Derek Ming Siang Tan;Shailesh;Boyang Liu;Alok Raj;Qi Xuan Ang;Weiheng Dai;Tanishq Duhan;Jimmy Chiun;Yuhong Cao;Florian Shkurti;Guillaume Adrien Sartoretti
national university of singaore, National University of Singapore;;;;;national university of singaore, National University of Singapore;national university of singaore, National University of Singapore;;National University of Singapore;University of Toronto;National University of Singapore
Poster
main
Test-Time Adaptation;Visual Search;VLM;Ecological Monitoring
https://openreview.net/forum?id=iVbCWUDyBF
-1
Search-TTA: A Multi-Modal Test-Time Adaptation Framework for Visual Search in the Wild To perform autonomous visual search for environmental monitoring, a robot may leverage satellite imagery as a prior map. This can help inform coarse, high level search and exploration strategies, even when such images lack sufficient...
[ -0.035350218415260315, 0.0007535106269642711, -0.041132815182209015, 0.019438982009887695, -0.029567619785666466, -0.008437500335276127, 0.0015786217991262674, 0.00852387584745884, 0.0202390905469656, 0.017456896603107452, -0.014783809892833233, -0.019802667200565338, 0.002361681777983904, ...
corl_2025_iWMM4oxMBu
iWMM4oxMBu
corl
2,025
Learn from What We HAVE: History-Aware VErifier that Reasons about Past Interactions Online
We introduce a novel History-Aware VErifier (HAVE) to disambiguate uncertain scenarios online by leveraging past interactions. Robots frequently encounter visually ambiguous objects whose manipulation outcomes remain uncertain until physically interacted with. While generative models alone could theoretically adapt to ...
Yishu Li;Xinyi Mao;Ying Yuan;Kyutae Sim;Ben Eisner;David Held
Computer Science Department, School of Computer Science;Tsinghua University;Carnegie Mellon University;Carnegie Mellon University;Carnegie Mellon University;Carnegie Mellon University
Poster
main
Ambiguities;Multi-modality;Generation-Verification
https://openreview.net/forum?id=iWMM4oxMBu
-1
Learn from What We HAVE: History-Aware VErifier that Reasons about Past Interactions Online We introduce a novel History-Aware VErifier (HAVE) to disambiguate uncertain scenarios online by leveraging past interactions. Robots frequently encounter visually ambiguous objects whose manipulation outcomes remain uncertain u...
[ -0.02112503908574581, -0.04931073635816574, -0.02209303341805935, 0.03858688473701477, -0.028375500813126564, -0.0475645512342453, 0.008284887298941612, 0.029039811342954636, -0.0031791948713362217, 0.043730538338422775, -0.02748342975974083, -0.027293626219034195, -0.034316327422857285, -...
corl_2025_iq0wUf6dUy
iq0wUf6dUy
corl
2,025
Reflective Planning: Vision-Language Models for Multi-Stage Long-Horizon Robotic Manipulation
Solving complex long-horizon robotic manipulation problems requires sophisticated high-level planning capabilities, the ability to reason about the physical world, and reactively choose appropriate motor skills. Vision-language models (VLMs) pretrained on Internet data could in principle offer a framework for tackling ...
Yunhai Feng;Jiaming Han;Zhuoran Yang;Xiangyu Yue;Sergey Levine;Jianlan Luo
;The Chinese University of Hong Kong+ByteDance Inc.;Yale University;The Chinese University of Hong Kong;University of California, Berkeley;University of California, Berkeley
Poster
main
Vision-Language Models;Long-Horizon Manipulation;Manipulation Planning
https://openreview.net/forum?id=iq0wUf6dUy
-1
Reflective Planning: Vision-Language Models for Multi-Stage Long-Horizon Robotic Manipulation Solving complex long-horizon robotic manipulation problems requires sophisticated high-level planning capabilities, the ability to reason about the physical world, and reactively choose appropriate motor skills. Vision-languag...
[ -0.04772680625319481, -0.02060847170650959, 0.03009897470474243, 0.02483256906270981, -0.02029760740697384, 0.012123705819249153, -0.005435552913695574, -0.003026355057954788, -0.010770532302558422, 0.022400513291358948, -0.014327185228466988, -0.017097534611821175, -0.011081396602094173, ...
corl_2025_irh5o90Mj1
irh5o90Mj1
corl
2,025
Long-VLA: Unleashing Long-Horizon Capability of Vision Language Action Model for Robot Manipulation
Vision-Language-Action (VLA) models have become a cornerstone in robotic policy learning, leveraging large-scale multimodal data for robust and scalable control. However, existing VLA frameworks primarily address short-horizon tasks, and their effectiveness on long-horizon, multi-step robotic manipulation remains limit...
Yiguo Fan;Shuanghao Bai;Xinyang Tong;Pengxiang Ding;Yuyang Zhu;Hongchao Lu;Fengqi Dai;Wei Zhao;Yang Liu;Siteng Huang;Zhaoxin Fan;Badong Chen;Donglin Wang
Westlake University;Xi'an Jiaotong University;Westlake University;Zhejiang University+Westlake University;Westlake University+University of Electronic Science and Technology of China;Westlake University;;Westlake University;Westlake University;Alibaba Group;Beihang University;Xi'an Jiaotong University;Westlake Universi...
Poster
main
VLA; Manipulation
https://openreview.net/forum?id=irh5o90Mj1
-1
Long-VLA: Unleashing Long-Horizon Capability of Vision Language Action Model for Robot Manipulation Vision-Language-Action (VLA) models have become a cornerstone in robotic policy learning, leveraging large-scale multimodal data for robust and scalable control. However, existing VLA frameworks primarily address short-h...
[ -0.018952153623104095, -0.002124980092048645, 0.0066350954584777355, 0.031365904957056046, -0.025361627340316772, 0.02705608494579792, 0.012247989885509014, 0.025951003655791283, -0.014025329612195492, 0.00791054405272007, 0.0032093417830765247, 0.005170861724764109, 0.03740701824426651, 0...
corl_2025_isrcFrgwZp
isrcFrgwZp
corl
2,025
AutoEval: Autonomous Evaluation of Generalist Robot Manipulation Policies in the Real World
Scalable and reproducible policy evaluation has been a long-standing challenge in robot learning: evaluations are critical to assess progress and build better policies, but evaluation in the real world, especially at a scale that would provide statistically reliable results, is costly in terms of human time and hard to...
Zhiyuan Zhou;Pranav Atreya;You Liang Tan;Karl Pertsch;Sergey Levine
University of California, Berkeley;University of California, Berkeley;NVIDIA;University of California, Berkeley+Stanford University;University of California, Berkeley
Poster
main
Robot evaluation;benchmark;generalist robot policy;manipulation
https://openreview.net/forum?id=isrcFrgwZp
-1
AutoEval: Autonomous Evaluation of Generalist Robot Manipulation Policies in the Real World Scalable and reproducible policy evaluation has been a long-standing challenge in robot learning: evaluations are critical to assess progress and build better policies, but evaluation in the real world, especially at a scale tha...
[ -0.0647677555680275, -0.014790802262723446, 0.00769233051687479, 0.0020193527452647686, -0.023605898022651672, -0.000756242370698601, 0.02445957064628601, 0.0364852175116539, 0.019801488146185875, 0.0276701208204031, -0.011441065929830074, -0.01876223459839821, 0.027131935581564903, 0.0203...
corl_2025_jPHhft5tNo
jPHhft5tNo
corl
2,025
JaxRobotarium: Training and Deploying Multi-Robot Policies in 10 Minutes
Multi-agent reinforcement learning (MARL) has emerged as a promising solution for learning complex and scalable coordination behaviors in multi-robot systems. However, established MARL platforms (e.g., SMAC and MPE) lack robotics relevance and hardware deployment, leaving multi-robot learning researchers to develop bes...
Shalin Jain;Jiazhen Liu;Siva Kailas;Harish Ravichandar
Georgia Institute of Technology;Georgia Institute of Technology;College of Computing, Georgia Institute of Technology;Georgia Institute of Technology
Poster
main
Multi-Robot Learning;Sim2Real Deployment;Benchmarking;Jax
https://openreview.net/forum?id=jPHhft5tNo
-1
JaxRobotarium: Training and Deploying Multi-Robot Policies in 10 Minutes Multi-agent reinforcement learning (MARL) has emerged as a promising solution for learning complex and scalable coordination behaviors in multi-robot systems. However, established MARL platforms (e.g., SMAC and MPE) lack robotics relevance and har...
[ -0.015379420481622219, -0.029916634783148766, -0.046211495995521545, -0.018729938194155693, -0.045442525297403336, 0.05822209268808365, 0.009319196455180645, -0.034585386514663696, -0.008504453115165234, 0.01163526438176632, -0.0037601767107844353, -0.041817378252744675, 0.02121993899345398,...
corl_2025_jU7AbGq3se
jU7AbGq3se
corl
2,025
Steering Your Diffusion Policy with Latent Space Reinforcement Learning
Robotic control policies learned from human demonstrations have achieved impressive results in many real-world applications. However, in scenarios where initial performance is not satisfactory, as is often the case in novel open-world settings, such behavioral cloning (BC)-learned policies typically require collecting ...
Andrew Wagenmaker;Yunchu Zhang;Mitsuhiko Nakamoto;Seohong Park;Waleed Yagoub;Anusha Nagabandi;Abhishek Gupta;Sergey Levine
University of California, Berkeley;;University of California, Berkeley;University of California, Berkeley;Department of Computer Science, University of Washington;University of California, Berkeley;University of Washington;University of California, Berkeley
Oral
main
diffusion policies;reinforcement learning;finetuning
https://openreview.net/forum?id=jU7AbGq3se
-1
Steering Your Diffusion Policy with Latent Space Reinforcement Learning Robotic control policies learned from human demonstrations have achieved impressive results in many real-world applications. However, in scenarios where initial performance is not satisfactory, as is often the case in novel open-world settings, suc...
[ -0.06362798064947128, -0.005652861203998327, -0.02375028282403946, 0.0194704569876194, -0.03978585824370384, -0.021913448348641396, 0.006006452254951, 0.004353300202637911, 0.03385287895798683, 0.045994359999895096, -0.029573051258921623, -0.011406748555600643, -0.004830877296626568, 0.029...
corl_2025_jZRz7Hvsyd
jZRz7Hvsyd
corl
2,025
Morphologically Symmetric Reinforcement Learning for Ambidextrous Bimanual Manipulation
Humans naturally exhibit bilateral symmetry in their gross manipulation skills, effortlessly mirroring simple actions between left and right hands. Bimanual robots—which also feature bilateral symmetry—should similarly exploit this property to perform tasks with either hand. Unlike humans, who often favor a dominant ha...
Zechu Li;Yufeng Jin;Daniel Ordonez-Apraez;Claudio Semini;Puze Liu;Georgia Chalvatzaki
Columbia University+Massachusetts Institute of Technology+Technische Universität Darmstadt;Technische Universität Darmstadt;Università degli Studi di Genova, Istituto Italiano di Tecnologia;Istituto Italiano di Tecnologia;German Research Center for AI;Technische Universität Darmstadt+Technische Universität Darmstadt
Poster
main
Bimanual Dexterous Manipulation;Sim-to-Real Transfer;Reinforcement Learning
https://openreview.net/forum?id=jZRz7Hvsyd
-1
Morphologically Symmetric Reinforcement Learning for Ambidextrous Bimanual Manipulation Humans naturally exhibit bilateral symmetry in their gross manipulation skills, effortlessly mirroring simple actions between left and right hands. Bimanual robots—which also feature bilateral symmetry—should similarly exploit this ...
[ -0.06660884618759155, -0.013728968799114227, -0.007863803766667843, -0.022134460508823395, -0.013990473002195358, -0.02631852775812149, 0.04561379924416542, 0.023460660129785538, 0.05644754320383072, 0.026187775656580925, -0.053421564400196075, 0.012206641025841236, 0.007392162457108498, -...
corl_2025_jedBaI1fgU
jedBaI1fgU
corl
2,025
BranchOut: Capturing Realistic Multimodality in Autonomous Driving Decisions
Modeling the nuanced, multimodal nature of human driving remains a core challenge for autonomous systems, as existing methods often fail to capture the diversity of plausible behaviors in complex real-world scenarios. In this work, we introduce a novel benchmark and end-to-end planner for modeling realistic multimodali...
Hee Jae Kim;Zekai Yin;Lei Lai;Jason Lee;Eshed Ohn-Bar
Boston University;Boston University;;Allen Institute for Artificial Intelligence+University of Washington;Boston University
Poster
main
Autonomous Driving;Human-in-the-Loop Simulation;Multi-modal Planning and Evaluation
https://openreview.net/forum?id=jedBaI1fgU
-1
BranchOut: Capturing Realistic Multimodality in Autonomous Driving Decisions Modeling the nuanced, multimodal nature of human driving remains a core challenge for autonomous systems, as existing methods often fail to capture the diversity of plausible behaviors in complex real-world scenarios. In this work, we introduc...
[ -0.09193573892116547, -0.07566162943840027, 0.00786829087883234, -0.016913581639528275, -0.051676783710718155, -0.021093320101499557, 0.002336622681468725, -0.0053011332638561726, -0.00013213735655881464, 0.04018481448292732, -0.06224197894334793, -0.0017817183397710323, 0.02581985481083393,...
corl_2025_jnpILGz9gQ
jnpILGz9gQ
corl
2,025
Streaming Flow Policy: Simplifying diffusion/flow-matching policies by treating action trajectories as flow trajectories
Recent advances in diffusion$/$flow-matching policies have enabled imitation learning of complex, multi-modal action trajectories. However, they are computationally expensive because they sample a *trajectory of trajectories*—a diffusion$/$flow trajectory of action trajectories. They discard intermediate action traject...
Sunshine Jiang;Xiaolin Fang;Nicholas Roy;Tomás Lozano-Pérez;Leslie Pack Kaelbling;Siddharth Ancha
Massachusetts Institute of Technology;Massachusetts Institute of Technology;Massachusetts Institute of Technology;Massachusetts Institute of Technology;Massachusetts Institute of Technology;Massachusetts Institute of Technology
Oral
main
imitation learning;diffusion policy;flow matching
https://openreview.net/forum?id=jnpILGz9gQ
-1
Streaming Flow Policy: Simplifying diffusion/flow-matching policies by treating action trajectories as flow trajectories Recent advances in diffusion$/$flow-matching policies have enabled imitation learning of complex, multi-modal action trajectories. However, they are computationally expensive because they sample a *t...
[ -0.05673641338944435, -0.036378487944602966, -0.019398169592022896, 0.0075304023921489716, -0.02284960448741913, -0.014128733426332474, -0.01653735525906086, 0.029586361721158028, 0.020782435312867165, 0.05400479957461357, 0.0064322189427912235, -0.04595760256052017, 0.015919050201773643, ...
corl_2025_jr1Gjpjmqc
jr1Gjpjmqc
corl
2,025
Sample-Efficient Online Control Policy Learning with Real-Time Recursive Model Updates
Data-driven control methods need to be sample-efficient and lightweight, especially when data acquisition and computational resources are limited---such as during learning on hardware. Most modern data-driven methods require large datasets and struggle with real-time updates of models, limiting their performance in dyn...
Zixin Zhang;James Avtges;Todd Murphey
Northwestern University;Northwestern University;
Poster
main
Koopman Operator; Control
https://openreview.net/forum?id=jr1Gjpjmqc
-1
Sample-Efficient Online Control Policy Learning with Real-Time Recursive Model Updates Data-driven control methods need to be sample-efficient and lightweight, especially when data acquisition and computational resources are limited---such as during learning on hardware. Most modern data-driven methods require large da...
[ -0.10702663660049438, 0.016129186376929283, -0.029067378491163254, -0.010728338733315468, -0.013470024801790714, -0.02017294429242611, 0.02827879972755909, 0.05813475698232651, 0.007528176065534353, 0.03190993145108223, -0.04107945039868355, 0.0004467275575734675, 0.024097498506307602, 0.0...
corl_2025_kUA2ec94LI
kUA2ec94LI
corl
2,025
Non-conflicting Energy Minimization in Reinforcement Learning based Robot Control
Efficient robot locomotion often requires balancing task performance with energy expenditure. A common approach in reinforcement learning (RL) is to penalize energy use directly in the reward function. This requires carefully weighting the reward terms to avoid undesirable trade-offs where energy minimization harms tas...
Skand Peri;Akhil Perincherry;Bikram Pandit;Stefan Lee
;;;
Oral
main
Energy efficient locomotion;Reinforcement Learning
https://openreview.net/forum?id=kUA2ec94LI
-1
Non-conflicting Energy Minimization in Reinforcement Learning based Robot Control Efficient robot locomotion often requires balancing task performance with energy expenditure. A common approach in reinforcement learning (RL) is to penalize energy use directly in the reward function. This requires carefully weighting th...
[ -0.07633665204048157, -0.031624142080545425, 0.005451965145766735, -0.010172736831009388, -0.03959415480494499, -0.02904668264091015, 0.014843238517642021, 0.014541621319949627, 0.030801547691226006, -0.011059309355914593, -0.031551022082567215, -0.028900442644953728, 0.019157283008098602, ...
corl_2025_kXhOmN3x18
kXhOmN3x18
corl
2,025
ControlVLA: Few-shot Object-centric Adaptation for Pre-trained Vision-Language-Action Models
Learning real-world robotic manipulation is challenging, particularly when limited demonstrations are available. Existing methods for few-shot manipulation often rely on simulation-augmented data or pre-built modules like grasping and pose estimation, which struggle with sim-to-real gaps and lack extensibility. While l...
Puhao Li;Yingying Wu;Ziheng Xi;Wanlin Li;Yuzhe Huang;Zhiyuan Zhang;Yinghan Chen;Jianan Wang;Song-Chun Zhu;Tengyu Liu;Siyuan Huang
Tsinghua University;Tsinghua University ;Tsinghua University;BIGAI:Beijing Institute for General Artificial Intelligence;Beihang University;University of California, Los Angeles;University of Cambridge+Peking University;Astribot;Beijing Institute for General Artificial Intelligence+Peking University;Beijing Institute o...
Poster
main
Robotic manipulation;Imitation learning;Few-shot learning
https://openreview.net/forum?id=kXhOmN3x18
-1
ControlVLA: Few-shot Object-centric Adaptation for Pre-trained Vision-Language-Action Models Learning real-world robotic manipulation is challenging, particularly when limited demonstrations are available. Existing methods for few-shot manipulation often rely on simulation-augmented data or pre-built modules like grasp...
[ -0.05286891758441925, 0.009211869910359383, -0.010824060998857021, 0.021081799641251564, -0.02305479347705841, 0.003950553946197033, 0.013673940673470497, 0.04223667457699776, 0.01055003423243761, 0.014313336461782455, -0.036043666303157806, -0.026927707716822624, 0.0026238071732223034, 0....
corl_2025_kbrb6W7bUL
kbrb6W7bUL
corl
2,025
SDS – See it, Do it, Sorted: Quadruped Skill Synthesis from Single Video Demonstration
Imagine a robot learning locomotion skills from any single video, without labels or reward engineering. We introduce SDS ("See it. Do it. Sorted."), an automated pipeline for skill acquisition from unstructured video demonstrations. Using GPT-4o, SDS applies novel prompting techniques, in the form of spatio-temporal gr...
Maria Stamatopoulou;Jeffrey Li;Dimitrios Kanoulas
University College London, University of London;;University College London, University of London+Athena Research and Innovation Centre
Poster
main
Skill Imitation;Robot Skill Learning;Quadrupedal Robot
https://openreview.net/forum?id=kbrb6W7bUL
-1
SDS – See it, Do it, Sorted: Quadruped Skill Synthesis from Single Video Demonstration Imagine a robot learning locomotion skills from any single video, without labels or reward engineering. We introduce SDS ("See it. Do it. Sorted."), an automated pipeline for skill acquisition from unstructured video demonstrations. ...
[ -0.055495455861091614, 0.0026937713846564293, -0.031967341899871826, -0.022247159853577614, -0.014410734176635742, 0.021644357591867447, -0.021776221692562103, 0.01807464100420475, 0.010238214395940304, 0.021926920861005783, -0.009607156738638878, -0.03449157625436783, 0.012997917830944061, ...
corl_2025_ksOrtEgIC0
ksOrtEgIC0
corl
2,025
AirExo-2: Scaling up Generalizable Robotic Imitation Learning with Low-Cost Exoskeletons
Scaling up robotic imitation learning for real-world applications requires efficient and scalable demonstration collection methods. While teleoperation is effective, it depends on costly and inflexible robot platforms. In-the-wild demonstrations offer a promising alternative, but existing collection devices have key li...
Hongjie Fang;Chenxi Wang;Yiming Wang;Jingjing Chen;Shangning Xia;Jun Lv;Zihao He;Xiyan Yi;Yunhan Guo;Xinyu Zhan;Lixin Yang;Weiming Wang;Cewu Lu;Hao-Shu Fang
Shanghai Jiaotong University;;;;Shanghai Jiaotong University;Shanghai Jiaotong University;University of California, San Diego+Shanghai Jiaotong University;;Shanghai Jiaotong University;Shanghai Jiaotong University;Shanghai Jiaotong University;Shanghai Jiaotong University;Shanghai Jiaotong University;Massachusetts Insti...
Oral
main
Robotic Manipulation;Scalable Data Collection;Generalizable Imitation Policy
https://openreview.net/forum?id=ksOrtEgIC0
-1
AirExo-2: Scaling up Generalizable Robotic Imitation Learning with Low-Cost Exoskeletons Scaling up robotic imitation learning for real-world applications requires efficient and scalable demonstration collection methods. While teleoperation is effective, it depends on costly and inflexible robot platforms. In-the-wild ...
[ -0.05190890654921532, -0.016405869275331497, -0.02427699975669384, 0.021677866578102112, 0.00001648939542064909, 0.017567183822393417, -0.02027691714465618, 0.04759545251727104, -0.0007379184826277196, 0.04648943990468979, -0.036332547664642334, 0.00259913201443851, -0.008147633634507656, ...
corl_2025_kto4zVmo4w
kto4zVmo4w
corl
2,025
One View, Many Worlds: Single-Image to 3D object Meets Generative Domain Randomization for One-Shot 6D Pose Estimation
Estimating the 6D pose of arbitrary objects from a single reference image is a critical yet challenging task in robotics, especially considering the long-tail distribution of real-world instances. While category-level and model-based approaches have achieved notable progress, they remain limited in generalizing to unse...
Zheng Geng;Nan Wang;Shaocong Xu;Chongjie Ye;Bohan Li;Zhaoxi Chen;Sida Peng;Hao Zhao
Beijing Institute of Technology;Beijing Academy of Artificial Intelligence+Tongji University;Beijing Academy of Artificial Intelligence;The Chinese University of Hong Kong, Shenzhen;Shanghai Jiaotong University;Nanyang Technological University;Zhejiang University;Tsinghua University+Peking University
Oral
main
Unseen Object Pose Estimation;Generative Model;Robot Manipulation
https://openreview.net/forum?id=kto4zVmo4w
-1
One View, Many Worlds: Single-Image to 3D object Meets Generative Domain Randomization for One-Shot 6D Pose Estimation Estimating the 6D pose of arbitrary objects from a single reference image is a critical yet challenging task in robotics, especially considering the long-tail distribution of real-world instances. Whil...
[ -0.038370661437511444, 0.003856359049677849, -0.02406691014766693, -0.0118809062987566, -0.0120603758841753, 0.00388327962718904, -0.009709320962429047, 0.0006713295006193221, -0.000011199341315659694, -0.013074380345642567, -0.02417459338903427, -0.003396467538550496, 0.023420820012688637, ...
corl_2025_lJWUourMTT
lJWUourMTT
corl
2,025
Train-Once Plan-Anywhere Kinodynamic Motion Planning via Diffusion Trees
Kinodynamic motion planning is concerned with computing collision-free trajectories while abiding by the robot's dynamic constraints. This critical problem is often tackled using sampling-based planners (SBPs) that explore the robot's high-dimensional state space by constructing a search tree via action propagations. A...
Yaniv Hassidof;Tom Jurgenson;Kiril Solovey
Technion - Israel Institute of Technology;Technion;Technion - Israel Institute of Technology, Technion
Poster
main
Kinodynamic motion planning;diffusion models;nonlinear systems
https://openreview.net/forum?id=lJWUourMTT
-1
Train-Once Plan-Anywhere Kinodynamic Motion Planning via Diffusion Trees Kinodynamic motion planning is concerned with computing collision-free trajectories while abiding by the robot's dynamic constraints. This critical problem is often tackled using sampling-based planners (SBPs) that explore the robot's high-dimensi...
[ -0.05788205936551094, -0.014388400129973888, -0.040765613317489624, 0.025546938180923462, -0.016222305595874786, 0.0015168493846431375, 0.008197830058634281, -0.039378780871629715, -0.0034533985890448093, 0.04083860293030739, -0.018776999786496162, -0.02213459648191929, -0.002303786342963576...
corl_2025_lRKGIidrYZ
lRKGIidrYZ
corl
2,025
Point Policy: Unifying Observations and Actions with Key Points for Robot Manipulation
Building robotic agents capable of operating across diverse environments and object types remains a significant challenge, often requiring extensive data collection. This is particularly restrictive in robotics, where each data point must be physically executed in the real world. Consequently, there is a critical need ...
Siddhant Haldar;Lerrel Pinto
New York University;New York University
Poster
main
Robot Learning;Imitation Learning;Robot Perception;Sensing & Vision
https://openreview.net/forum?id=lRKGIidrYZ
-1
Point Policy: Unifying Observations and Actions with Key Points for Robot Manipulation Building robotic agents capable of operating across diverse environments and object types remains a significant challenge, often requiring extensive data collection. This is particularly restrictive in robotics, where each data point...
[ -0.003422053065150976, -0.03385157510638237, -0.031730085611343384, 0.014638269320130348, -0.03012513555586338, -0.0017686756327748299, -0.0004571576719172299, 0.02073524333536625, 0.017313189804553986, 0.02776382677257061, -0.018844351172447205, -0.04198702052235603, 0.04405316710472107, ...
corl_2025_mjYKNIRqpy
mjYKNIRqpy
corl
2,025
GC-VLN: Instruction as Graph Constraints for Training-free Vision-and-Language Navigation
In this paper, we propose a training-free framework for vision-and-language navigation (VLN). Existing zero-shot VLN methods are mainly designed for discrete environments or involve unsupervised training in continuous simulator environments, which makes it challenging to generalize and deploy them in real-world scenari...
Hang Yin;Haoyu Wei;Xiuwei Xu;Wenxuan Guo;Jie Zhou;Jiwen Lu
Tsinghua University;Beihang University;Tsinghua University;Tsinghua University;Tsinghua University;Tsinghua University
Poster
main
Vision-Language Navigation;Graph Constraint
https://openreview.net/forum?id=mjYKNIRqpy
-1
GC-VLN: Instruction as Graph Constraints for Training-free Vision-and-Language Navigation In this paper, we propose a training-free framework for vision-and-language navigation (VLN). Existing zero-shot VLN methods are mainly designed for discrete environments or involve unsupervised training in continuous simulator en...
[ -0.06527978926897049, -0.02287334017455578, -0.005949791520833969, -0.013832924887537956, 0.013842001557350159, -0.011554667726159096, 0.006993614602833986, -0.008849804289638996, 0.04120831936597824, 0.007501911371946335, -0.008990493603050709, -0.029354121536016464, 0.024398230016231537, ...
corl_2025_nMiyWyNhQx
nMiyWyNhQx
corl
2,025
Human-like Navigation in a World Built for Humans
When navigating in a man-made environment they haven’t visited before—like an office building—humans employ behaviors such as reading signs and asking others for directions. These behaviors help humans reach their destinations efficiently by reducing the need to search through large areas. Existing robot navigation sys...
Bhargav Chandaka;Gloria Xinyue Wang;Haozhe Chen;Henry Che;Albert J. Zhai;Shenlong Wang
University of Illinois, Urbana Champaign;;University of Illinois, Urbana Champaign;UIUC+Waabi.ai;;University of Illinois, Urbana Champaign
Poster
main
navigation;reasoning;vison-language model
https://openreview.net/forum?id=nMiyWyNhQx
-1
Human-like Navigation in a World Built for Humans When navigating in a man-made environment they haven’t visited before—like an office building—humans employ behaviors such as reading signs and asking others for directions. These behaviors help humans reach their destinations efficiently by reducing the need to search ...
[ -0.06081627309322357, -0.019832225516438484, 0.006497587542980909, 0.002370256930589676, -0.028438301756978035, 0.019392360001802444, -0.01906724087893963, -0.006473681889474392, -0.013157734647393227, 0.011857260949909687, 0.007434694096446037, -0.016150737181305885, 0.0021132698748260736, ...
corl_2025_najxw1MlZH
najxw1MlZH
corl
2,025
QuaDreamer: Controllable Panoramic Video Generation for Quadruped Robots
Panoramic cameras, capturing comprehensive 360-degree environmental data, are suitable for quadruped robots in surrounding perception and interaction with complex environments. However, the scarcity of high-quality panoramic training data — caused by inherent kinematic constraints and complex sensor calibration challen...
Sheng Wu;Fei Teng;Hao Shi;Qi Jiang;Kai Luo;Kaiwei Wang;Kailun Yang
Hunan University;;Zhejiang University;Zhejiang University;;Zhejiang University;Hunan University
Poster
main
Panoramic Video Generation;World Model;Quadruped Robots
https://openreview.net/forum?id=najxw1MlZH
-1
QuaDreamer: Controllable Panoramic Video Generation for Quadruped Robots Panoramic cameras, capturing comprehensive 360-degree environmental data, are suitable for quadruped robots in surrounding perception and interaction with complex environments. However, the scarcity of high-quality panoramic training data — caused...
[ -0.0705864205956459, -0.008367639034986496, -0.02923610620200634, 0.01855793222784996, -0.01772945187985897, 0.0057211085222661495, 0.006637038197368383, 0.04558476060628891, -0.02395225130021572, 0.03903057053685188, -0.003019345924258232, -0.04138713330030441, -0.014940239489078522, -0.0...
corl_2025_nryBWao01j
nryBWao01j
corl
2,025
Constraint-Aware Diffusion Guidance for Robotics: Real-Time Obstacle Avoidance for Autonomous Racing
Diffusion models hold great potential in robotics due to their ability to capture complex, high-dimensional data distributions. However, their lack of constraint-awareness limits their deployment in safety-critical applications. We propose Constraint-Aware Diffusion Guidance (CoDiG), a data-efficient and general-purpos...
Hao Ma;Sabrina Bodmer;Andrea Carron;Melanie Zeilinger;Michael Muehlebach
ETHZ - ETH Zurich;ETHZ - ETH Zurich;ETHZ - ETH Zurich;ETHZ - ETH Zurich;Max-Planck Institute
Poster
main
Diffusion Guidance;Constraint-Aware Sampling;Real-Time Obstacle Avoidance;Autonomous Racing;Safe Control
https://openreview.net/forum?id=nryBWao01j
-1
Constraint-Aware Diffusion Guidance for Robotics: Real-Time Obstacle Avoidance for Autonomous Racing Diffusion models hold great potential in robotics due to their ability to capture complex, high-dimensional data distributions. However, their lack of constraint-awareness limits their deployment in safety-critical appl...
[ -0.06315553188323975, -0.035793155431747437, -0.023259904235601425, 0.02647790126502514, -0.017040327191352844, -0.022563613951206207, -0.025217048823833466, -0.004822290036827326, 0.0227141622453928, 0.02743765525519848, -0.060332730412483215, -0.019778447225689888, 0.010914841666817665, ...
corl_2025_o0LBjJxUeS
o0LBjJxUeS
corl
2,025
Learning Long-Context Diffusion Policies via Past-Token Prediction
Reasoning over long sequences of observations and actions is essential for many robotic tasks. Yet, learning effective long-context policies from demonstrations remains challenging. As context length increases, training becomes increasingly expensive due to rising memory demands, and policy performance often degrades...
Marcel Torne Villasevil;Andy Tang;Yuejiang Liu;Chelsea Finn
Stanford University;Computer Science Department, Stanford University;Stanford University;Physical Intelligence+Stanford University
Poster
main
history-conditioned policy learning
https://openreview.net/forum?id=o0LBjJxUeS
-1
Learning Long-Context Diffusion Policies via Past-Token Prediction Reasoning over long sequences of observations and actions is essential for many robotic tasks. Yet, learning effective long-context policies from demonstrations remains challenging. As context length increases, training becomes increasingly expensive ...
[ -0.0564233660697937, -0.027808398008346558, -0.0067733703181147575, 0.04450810328125954, -0.019596029072999954, -0.01249269861727953, -0.00513273011893034, 0.025645319372415543, 0.011292005889117718, 0.020860880613327026, -0.0514739491045475, -0.028358332812786102, -0.02509538270533085, 0....
corl_2025_o2w2iiMyEU
o2w2iiMyEU
corl
2,025
LaDi-WM: A Latent Diffusion-Based World Model for Predictive Manipulation
Predictive manipulation has recently gained considerable attention in the Embodied AI community due to its potential to improve robot policy performance by leveraging predicted states. However, generating accurate future visual states of robot-object interactions from world models remains a well-known challenge, partic...
Yuhang Huang;Jiazhao Zhang;Shilong Zou;Xinwang Liu;Ruizhen Hu;Kai Xu
National University of Defense Technology;Peking University;National University of Defense Technology;National University of Defense Technology;Shenzhen University;
Poster
main
Robot Learning: Model Learning;Control & Dynamics;Grasping & Manipulation
https://openreview.net/forum?id=o2w2iiMyEU
-1
LaDi-WM: A Latent Diffusion-Based World Model for Predictive Manipulation Predictive manipulation has recently gained considerable attention in the Embodied AI community due to its potential to improve robot policy performance by leveraging predicted states. However, generating accurate future visual states of robot-ob...
[ -0.08974771201610565, -0.020279886201024055, -0.012269331142306328, 0.023967137560248375, -0.01666637882590294, 0.006738453172147274, -0.025718582794070244, 0.024833641946315765, 0.022141948342323303, 0.038384292274713516, -0.04133409261703491, 0.012868509627878666, 0.015053206123411655, 0...
corl_2025_oDUbsdc0Ru
oDUbsdc0Ru
corl
2,025
Diffusion Dynamics Models with Generative State Estimation for Cloth Manipulation
Cloth manipulation is challenging due to its highly complex dynamics, near-infinite degrees of freedom, and frequent self-occlusions, which complicate both state estimation and dynamics modeling. Inspired by recent advances in generative models, we hypothesize that these expressive models can effectively capture intric...
Tongxuan Tian;Haoyang Li;Bo Ai;Xiaodi Yuan;Zhiao Huang;Hao Su
University of Virginia, Charlottesville;University of California, San Diego;University of California, San Diego;University of California, San Diego;;University of California, San Diego
Poster
main
Model Learning;Deformable Object Manipulation;Perception
https://openreview.net/forum?id=oDUbsdc0Ru
-1
Diffusion Dynamics Models with Generative State Estimation for Cloth Manipulation Cloth manipulation is challenging due to its highly complex dynamics, near-infinite degrees of freedom, and frequent self-occlusions, which complicate both state estimation and dynamics modeling. Inspired by recent advances in generative ...
[ -0.03693026304244995, -0.046988654881715775, -0.03152990713715553, 0.023512886837124825, -0.017973346635699272, -0.04387092590332031, 0.016572223976254463, 0.021601419895887375, 0.013092612847685814, 0.05533972010016441, -0.042794566601514816, 0.009107300080358982, -0.013714303262531757, 0...
corl_2025_oGcC8nMOit
oGcC8nMOit
corl
2,025
Cross-Sensor Touch Generation
Today's visuo-tactile sensors come in many shapes and sizes, making it challenging to develop general-purpose tactile representations. This is because most models are tied to a specific sensor design. To address this challenge, we propose two approaches to cross-sensor image generation. The first is an end-to-end metho...
Samanta Rodriguez;Yiming Dou;Miquel Oller;Andrew Owens;Nima Fazeli
University of Michigan - Ann Arbor;Cornell University+University of Michigan - Ann Arbor;University of Michigan - Ann Arbor;Cornell University+University of Michigan;University of Michigan
Oral
main
Tactile Sensing;Cross-Modal Generation;Manipulation;Representation Learning;Soft Bubbles;GelSlims;Digits
https://openreview.net/forum?id=oGcC8nMOit
-1
Cross-Sensor Touch Generation Today's visuo-tactile sensors come in many shapes and sizes, making it challenging to develop general-purpose tactile representations. This is because most models are tied to a specific sensor design. To address this challenge, we propose two approaches to cross-sensor image generation. Th...
[ -0.054838672280311584, -0.0033398298546671867, -0.05380606651306152, 0.03713688254356384, -0.009920378215610981, -0.023989614099264145, -0.019877634942531586, 0.02470874972641468, -0.015249354764819145, 0.005398123990744352, -0.005900596734136343, 0.02476406656205654, 0.0430743545293808, 0...
corl_2025_oOCa85Z1Ho
oOCa85Z1Ho
corl
2,025
Steerable Scene Generation with Post Training and Inference-Time Search
Training robots in simulation requires diverse 3D scenes that reflect the specific challenges of downstream tasks. However, scenes that satisfy strict task requirements, such as high-clutter environments with plausible spatial arrangement, are rare and costly to curate manually. Instead, we generate large-scale scene d...
Nicholas Ezra Pfaff;Hongkai Dai;Sergey Zakharov;Shun Iwase;Russ Tedrake
Massachusetts Institute of Technology;Toyota Research Institute;Toyota Research Institute;Carnegie Mellon University;Toyota Research Institute+Massachusetts Institute of Technology
Poster
main
Scene Generation;Simulation;Diffusion;MCTS
https://openreview.net/forum?id=oOCa85Z1Ho
-1
Steerable Scene Generation with Post Training and Inference-Time Search Training robots in simulation requires diverse 3D scenes that reflect the specific challenges of downstream tasks. However, scenes that satisfy strict task requirements, such as high-clutter environments with plausible spatial arrangement, are rare...
[ -0.03783432021737099, -0.022027578204870224, -0.035087693482637405, 0.02955806441605091, 0.006102603394538164, -0.027302555739879608, -0.0533682219684124, 0.024173948913812637, 0.020626980811357498, 0.03645191341638565, -0.02026318944990635, -0.0021611491683870554, 0.012223396450281143, 0....
corl_2025_oRwcxFuN25
oRwcxFuN25
corl
2,025
Robot Trains Robot: Automatic Real-World Policy Adaptation and Learning for Humanoids
Simulation-based reinforcement learning (RL) has significantly advanced humanoid locomotion tasks, yet direct real-world RL from scratch or starting from pretrained policies remains rare, limiting the full potential of humanoid robots. Real-world training, despite being crucial for overcoming the sim-to-real gap, faces...
Kaizhe Hu;Haochen Shi;Yao He;Weizhuo Wang;Karen Liu;Shuran Song
Tsinghua University+Stanford University;Stanford University;Stanford University;Stanford University;Computer Science Department, Stanford University;Stanford University
Poster
main
Humanoid Robots;Sim-to-Real Adaptation;Real-World RL
https://openreview.net/forum?id=oRwcxFuN25
-1
Robot Trains Robot: Automatic Real-World Policy Adaptation and Learning for Humanoids Simulation-based reinforcement learning (RL) has significantly advanced humanoid locomotion tasks, yet direct real-world RL from scratch or starting from pretrained policies remains rare, limiting the full potential of humanoid robots...
[ -0.06440922617912292, -0.016451755538582802, 0.021114159375429153, 0.020911846309900284, -0.03237014263868332, -0.05028407275676727, -0.004319850355386734, 0.023339606821537018, 0.018612831830978394, 0.02611681818962097, -0.039837345480918884, -0.028599755838513374, 0.0029864213429391384, ...
corl_2025_oYC10hiFua
oYC10hiFua
corl
2,025
Learning Smooth State-Dependent Traversability from Dense Point Clouds
A key open challenge in off-road autonomy is that the traversability of terrain often depends on the vehicle's state. In particular, some obstacles are only traversable from some orientations. However, learning this interaction by encoding the angle of approach as a model input demands a large and diverse training data...
Zihao Dong;Alan Papalia;Leonard Jung;Alenna Spiro;Philip R Osteen;Christa S. Robison;Michael Everett
Northeastern University;Northeastern University+Massachusetts Institute of Technology+Woods Hole Oceanographic Institution;Northeastern University;Northeastern University;U.S. Army Research Laboratory;DEVCOM Army Research Laboratory;Northeastern University
Poster
main
Off-road Autonomy;Traversability;Geometric Deep Learning
https://openreview.net/forum?id=oYC10hiFua
-1
Learning Smooth State-Dependent Traversability from Dense Point Clouds A key open challenge in off-road autonomy is that the traversability of terrain often depends on the vehicle's state. In particular, some obstacles are only traversable from some orientations. However, learning this interaction by encoding the angle...
[ -0.0011646051425486803, -0.0007071846630424261, -0.03479394689202309, -0.007387945894151926, -0.0049652899615466595, -0.00808936171233654, 0.0003547459782566875, 0.009782914072275162, 0.038097988814115524, 0.022814493626356125, -0.04562898725271225, 0.020636409521102905, 0.029367199167609215...
corl_2025_pJ5FONkM9N
pJ5FONkM9N
corl
2,025
Efficient Evaluation of Multi-Task Robot Policies With Active Experiment Selection
Evaluating learned robot control policies to determine their performance costs the experimenter time and effort. As robots become more capable in accomplishing diverse tasks, evaluating across all these tasks becomes more difficult as it is impractical to test every policy on every task multiple times. Rather than cons...
Abrar Anwar;Rohan Gupta;Zain Merchant;Sayan Ghosh;Willie Neiswanger;Jesse Thomason
NVIDIA+University of Southern California;;University of Southern California;University of Southern California;University of Southern California;University of Southern California+Amazon
Poster
main
robot evaluation;active testing;manipulation
https://openreview.net/forum?id=pJ5FONkM9N
-1
Efficient Evaluation of Multi-Task Robot Policies With Active Experiment Selection Evaluating learned robot control policies to determine their performance costs the experimenter time and effort. As robots become more capable in accomplishing diverse tasks, evaluating across all these tasks becomes more difficult as it...
[ -0.02846967615187168, -0.029645029455423355, -0.005942065734416246, 0.03305915370583534, -0.01786351017653942, -0.002999017247930169, -0.013880366459488869, 0.004766711965203285, 0.03020472265779972, 0.02274215966463089, -0.01934669353067875, -0.018115371465682983, 0.00382689549587667, 0.0...
corl_2025_pukgxvcOwL
pukgxvcOwL
corl
2,025
“Stack It Up!”: 3D Stable Structure Generation from 2D Hand-drawn Sketch
Imagine a child sketching the Eiffel Tower and asking a robot to bring it to life. Today’s robot manipulation systems can’t act on such sketches directly—they require precise 3D block poses as goals, which in turn demand structural analysis and expert tools like CAD. We present *StackItUp*, a system that enables non-ex...
Yiqing Xu;Linfeng Li;Cunjun Yu;David Hsu
National University of Singapore;national university of singaore, National University of Singapore;National University of Singapore;National University of Singapore
Oral
main
Goal Specification;Diffusion Models
https://openreview.net/forum?id=pukgxvcOwL
-1
“Stack It Up!”: 3D Stable Structure Generation from 2D Hand-drawn Sketch Imagine a child sketching the Eiffel Tower and asking a robot to bring it to life. Today’s robot manipulation systems can’t act on such sketches directly—they require precise 3D block poses as goals, which in turn demand structural analysis and ex...
[ -0.028151826933026314, -0.011922791600227356, -0.02634965442121029, 0.0025159271899610758, 0.001585200079716742, -0.0011352498549968004, 0.018486492335796356, 0.030636927112936974, 0.011552872136235237, 0.008199883624911308, -0.012510868720710278, 0.03401362895965576, 0.0008779661147855222, ...
corl_2025_qAGHsCsA1O
qAGHsCsA1O
corl
2,025
Predictive Red Teaming: Breaking Policies Without Breaking Robots
Visuomotor policies trained via imitation learning are capable of performing challenging manipulation tasks, but are often extremely brittle to lighting, visual distractors, and object locations. These vulnerabilities can depend unpredictably on the specifics of training, and are challenging to expose without time-cons...
Anirudha Majumdar;Mohit Sharma;Dmitry Kalashnikov;Sumeet Singh;Pierre Sermanet;Vikas Sindhwani
Google+Princeton University;Google DeepMind+Carnegie Mellon University;Google;Google Brain Robotics;Google;Google
Poster
main
red teaming;policy evaluation;targeted data collection
https://openreview.net/forum?id=qAGHsCsA1O
-1
Predictive Red Teaming: Breaking Policies Without Breaking Robots Visuomotor policies trained via imitation learning are capable of performing challenging manipulation tasks, but are often extremely brittle to lighting, visual distractors, and object locations. These vulnerabilities can depend unpredictably on the spec...
[ -0.07946843653917313, -0.037248510867357254, -0.021258363500237465, 0.0632556900382042, -0.022946419194340706, -0.0004907069378532469, -0.0002581066219136119, 0.031238293275237083, 0.02383682131767273, 0.012706788256764412, -0.05038195103406906, -0.033798202872276306, 0.01770602911710739, ...
corl_2025_qTYD21Wpxb
qTYD21Wpxb
corl
2,025
ReCoDe: Reinforcement Learning-based Dynamic Constraint Design for Multi-Agent Coordination
Constraint-based optimization is a cornerstone of robotics, enabling the design of controllers that reliably encode task and safety requirements such as collision avoidance or formation adherence. However, handcrafted constraints can fail in multi-agent settings that demand complex coordination. We introduce ReCoDe—Rei...
Michael Amir;Guang Yang;Zhan Gao;Keisuke Okumura;Heedo Woo;Amanda Prorok
University of Cambridge;;;University of Cambridge+AIST, National Institute of Advanced Industrial Science and Technology;;
Poster
main
multi-agent reinforcement learning;multi-robot systems
https://openreview.net/forum?id=qTYD21Wpxb
-1
ReCoDe: Reinforcement Learning-based Dynamic Constraint Design for Multi-Agent Coordination Constraint-based optimization is a cornerstone of robotics, enabling the design of controllers that reliably encode task and safety requirements such as collision avoidance or formation adherence. However, handcrafted constraint...
[ -0.008713291957974434, -0.018240565434098244, -0.01810181885957718, -0.009120282717049122, -0.003480229526758194, 0.021570485085248947, 0.028322825208306313, 0.012958941049873829, 0.0039380937814712524, 0.030690768733620644, -0.06918834894895554, -0.043251968920230865, 0.01429090928286314, ...
corl_2025_qoKo2caB9B
qoKo2caB9B
corl
2,025
Capability-Aware Shared Hypernetworks for Flexible Heterogeneous Multi-Robot Coordination
Recent advances have enabled heterogeneous multi-robot teams to learn complex and effective coordination skills. However, existing neural architectures that support heterogeneous teaming tend to force a trade-off between expressivity and efficiency. Shared-parameter designs prioritize sample efficiency by enabling a s...
Kevin Fu;Shalin Jain;Pierce Howell;Harish Ravichandar
Georgia Institute of Technology;Georgia Institute of Technology;Naval Information Warfare Center Pacific;Georgia Institute of Technology
Poster
main
Multi-Robot Learning;Heterogeneous Teams;Parameter Sharing
https://openreview.net/forum?id=qoKo2caB9B
-1
Capability-Aware Shared Hypernetworks for Flexible Heterogeneous Multi-Robot Coordination Recent advances have enabled heterogeneous multi-robot teams to learn complex and effective coordination skills. However, existing neural architectures that support heterogeneous teaming tend to force a trade-off between expressi...
[ -0.025250181555747986, -0.05028063803911209, -0.02872917801141739, 0.013439908623695374, 0.005914292298257351, -0.007347088772803545, 0.025744566693902016, 0.04826648533344269, -0.009521461091935635, 0.022247260436415672, -0.024499451741576195, -0.008660867810249329, 0.007777385879307985, ...
corl_2025_r29CIl3ePP
r29CIl3ePP
corl
2,025
Multimodal Fused Learning for Solving the Generalized Traveling Salesman Problem in Robotic Task Planning
Effective and efficient task planning is essential for mobile robots, especially in applications like warehouse retrieval and environmental monitoring. These tasks often involve selecting one location from each of several target clusters, forming a Generalized Traveling Salesman Problem (GTSP) that remains challenging ...
Jiaqi Cheng;Mingfeng Fan;Xuefeng Zhang;Jingsong Liang;Yuhong Cao;Guohua Wu;Guillaume Adrien Sartoretti
Central South University;National University of Singapore;;National University of Singapore;National University of Singapore;;National University of Singapore
Poster
main
Generalized Traveling Salesman Problem;Robotic Task Planning;Multimodal Learning
https://openreview.net/forum?id=r29CIl3ePP
-1
Multimodal Fused Learning for Solving the Generalized Traveling Salesman Problem in Robotic Task Planning Effective and efficient task planning is essential for mobile robots, especially in applications like warehouse retrieval and environmental monitoring. These tasks often involve selecting one location from each of ...
[ -0.06662452965974808, -0.053934141993522644, -0.022318847477436066, -0.011924905702471733, -0.016259558498859406, 0.007806985639035702, -0.036263518035411835, -0.024366280063986778, 0.028645595535635948, 0.041723333299160004, -0.01369565725326538, -0.014000005088746548, -0.00794532522559166,...
corl_2025_rJRFFDVTnf
rJRFFDVTnf
corl
2,025
O$^3$Afford: One-Shot 3D Object-to-Object Affordance Grounding for Generalizable Robotic Manipulation
Grounding object affordance is fundamental to robotic manipulation as it establishes the critical link between perception and action among interacting objects. However, prior works predominantly focus on predicting single-object affordance, overlooking the fact that most real-world interactions involve relationships be...
Tongxuan Tian;Xuhui Kang;Yen-Ling Kuo
University of Virginia, Charlottesville;University of Virginia, Charlottesville;University of Virginia, Charlottesville
Poster
main
Affordance Grounding;Few-shot Learning;Vision Foundation Models
https://openreview.net/forum?id=rJRFFDVTnf
-1
O$^3$Afford: One-Shot 3D Object-to-Object Affordance Grounding for Generalizable Robotic Manipulation Grounding object affordance is fundamental to robotic manipulation as it establishes the critical link between perception and action among interacting objects. However, prior works predominantly focus on predicting sin...
[ -0.035806119441986084, 0.008914425037801266, -0.03504547104239464, -0.012866084463894367, -0.010732559487223625, -0.04422890394926071, 0.018608050420880318, 0.0007919552735984325, -0.0003182895015925169, -0.01133551262319088, -0.047234393656253815, -0.017865953966975212, 0.028088323771953583...
corl_2025_rbMoMEK4m2
rbMoMEK4m2
corl
2,025
Sequence Modeling for Time-Optimal Quadrotor Trajectory Optimization with Sampling-based Robustness Analysis
Time-optimal trajectories drive quadrotors to their dynamic limits, but computing such trajectories involves solving non-convex problems via iterative nonlinear optimization, making them prohibitively costly for real-time applications. In this work, we investigate learning-based models that imitate a model-based time-o...
Katherine Mao;Hongzhan Yu;Ruipeng Zhang;Igor Spasojevic;Sicun Gao;Vijay Kumar
University of Pennsylvania;University of California, San Diego;University of California, San Diego;;University of California, San Diego;
Poster
main
Trajectory Planning;Imitation Learning;Robustness Analysis;Aerial Robotics
https://openreview.net/forum?id=rbMoMEK4m2
-1
Sequence Modeling for Time-Optimal Quadrotor Trajectory Optimization with Sampling-based Robustness Analysis Time-optimal trajectories drive quadrotors to their dynamic limits, but computing such trajectories involves solving non-convex problems via iterative nonlinear optimization, making them prohibitively costly for...
[ -0.027499809861183167, -0.019438572227954865, -0.01340180542320013, 0.02222336269915104, -0.025557784363627434, -0.012815534137189388, -0.0011364740785211325, -0.002958840224891901, -0.0009114693384617567, 0.04712159186601639, -0.04217492416501045, -0.002800821792334318, 0.005771112628281116...
corl_2025_sA2Yv4QKMr
sA2Yv4QKMr
corl
2,025
PrioriTouch: Adapting to User Contact Preferences for Whole-Arm Physical Human-Robot Interaction
Many robot caregiving tasks, such as bathing, dressing, and transferring, require a robot arm to make contact with a human body at multiple points rather than solely at the end effector. However, varied human touch preferences can lead to unsafe or uncomfortable multi-contact interactions. To address this, we introduce...
Rishabh Madan;Jiawei Lin;Mahika Goel;Amber Li;Angchen Xie;Xiaoyu Liang;Marcus Lee;Justin Guo;Pranav N. Thakkar;Rohan Banerjee;Jose Barreiros;Kate Tsui;Tom Silver;Tapomayukh Bhattacharjee
Cornell University;Cornell University;;;Shanghai Jiaotong University;Cornell University;;;Cornell University;Cornell University;Toyota Research Institute;;Princeton University+Cornell University;Cornell University
Poster
main
Physical Human-Robot Interaction;Online Preference Learning;Assistive Robotics
https://openreview.net/forum?id=sA2Yv4QKMr
-1
PrioriTouch: Adapting to User Contact Preferences for Whole-Arm Physical Human-Robot Interaction Many robot caregiving tasks, such as bathing, dressing, and transferring, require a robot arm to make contact with a human body at multiple points rather than solely at the end effector. However, varied human touch preferen...
[ -0.020921053364872932, -0.03493594750761986, -0.05617007985711098, 0.0205895584076643, -0.02272586151957512, -0.02887694723904133, -0.0008805351098999381, -0.008931963704526424, 0.02158404514193535, 0.010635482147336006, -0.017246976494789124, -0.017329851165413857, -0.018084924668073654, ...
corl_2025_sMs4pJYhWi
sMs4pJYhWi
corl
2,025
Tactile Beyond Pixels: Multisensory Touch Representations for Robot Manipulation
We present TacX, the first multisensory touch representations across four tactile modalities: image, audio, motion, and pressure. Trained on ~1M contact-rich interactions collected with the Digit 360 sensor, TacX captures complementary touch signals at diverse temporal and spatial scales. By leveraging self-supervised ...
Carolina Higuera;Akash Sharma;Taosha Fan;Chaithanya Krishna Bodduluri;Byron Boots;Michael Kaess;Mike Lambeta;Tingfan Wu;Zixi Liu;Francois Robert Hogan;Mustafa Mukadam
University of Washington;Meta Facebook+Carnegie Mellon University;;Meta Facebook;;Carnegie Mellon University;Meta;;;Meta Facebook;Amazon Robotics
Oral
main
Multi-sensory Touch;Self-Supervised Learning;Tactile Adaptation
https://openreview.net/forum?id=sMs4pJYhWi
-1
Tactile Beyond Pixels: Multisensory Touch Representations for Robot Manipulation We present TacX, the first multisensory touch representations across four tactile modalities: image, audio, motion, and pressure. Trained on ~1M contact-rich interactions collected with the Digit 360 sensor, TacX captures complementary tou...
[ -0.01301268208771944, -0.0084771029651165, -0.018491210415959358, 0.023026790469884872, -0.023290816694498062, 0.009585066698491573, 0.034964513033628464, -0.0007113363826647401, 0.0004352294490672648, 0.04375278577208519, -0.015464347787201405, -0.00045939249685034156, 0.011334234848618507,...
corl_2025_sUWOSP6SUJ
sUWOSP6SUJ
corl
2,025
Adapt3R: Adaptive 3D Scene Representation for Domain Transfer in Imitation Learning
Imitation Learning can train robots to perform complex and diverse manipulation tasks, but learned policies are brittle with observations outside of the training distribution. 3D scene representations that incorporate observations from calibrated RGBD cameras have been proposed as a way to mitigate this, but in our eva...
Albert Wilcox;Mohamed Ghanem;Masoud Moghani;Pierre Barroso;Benjamin Joffe;Animesh Garg
Georgia Institute of Technology;Georgia Institute of Technology;NVIDIA+University of Toronto;Georgia Institute of Technology;Georgia Institute of Technology;Georgia Institute of Technology
Poster
main
Imitation Learning;3D Perception;Cross-Embodiment Learning
https://openreview.net/forum?id=sUWOSP6SUJ
-1
Adapt3R: Adaptive 3D Scene Representation for Domain Transfer in Imitation Learning Imitation Learning can train robots to perform complex and diverse manipulation tasks, but learned policies are brittle with observations outside of the training distribution. 3D scene representations that incorporate observations from ...
[ -0.04102754965424538, -0.039859507232904434, -0.0021307680290192366, 0.011297177523374557, -0.014746557921171188, -0.0339827835559845, 0.0017281122272834182, -0.007368715945631266, 0.007806732784956694, 0.010631026700139046, -0.04778030514717102, -0.0252407044172287, 0.006748192477971315, ...
corl_2025_sVWKm4UiTL
sVWKm4UiTL
corl
2,025
Multi-critic Learning for Whole-body End-effector Twist Tracking
Learning whole-body control for locomotion and arm motions in a single policy has challenges, as the two tasks have conflicting goals. For instance, efficient locomotion typically favors a horizontal base orientation, while end-effector tracking may benefit from base tilting to extend reachability. Additionally, curren...
Aravind Elanjimattathil Vijayan;Andrei Cramariuc;Mattia Risiglione;Christian Gehring;Marco Hutter
ETHZ - ETH Zurich;ETHZ - ETH Zurich;ETHZ - ETH Zurich;;ETHZ - ETH Zurich
Poster
main
Loco-Manipulation;Multi-critic Reinforcement Learning;Whole-Body Control
https://openreview.net/forum?id=sVWKm4UiTL
-1
Multi-critic Learning for Whole-body End-effector Twist Tracking Learning whole-body control for locomotion and arm motions in a single policy has challenges, as the two tasks have conflicting goals. For instance, efficient locomotion typically favors a horizontal base orientation, while end-effector tracking may benef...
[ -0.03502250835299492, -0.018820419907569885, -0.011076842434704304, 0.027167512103915215, -0.011736067943274975, -0.008936681784689426, -0.02950729802250862, -0.00524594821035862, 0.023806391283869743, -0.010324768722057343, -0.022079406306147575, -0.02508770301938057, -0.004317461978644133,...
corl_2025_sXoaNAECCK
sXoaNAECCK
corl
2,025
Reactive In-Air Clothing Manipulation with Confidence-Aware Dense Correspondence and Visuotactile Affordance
Manipulating clothing is challenging due to their complex, variable configurations and frequent self-occlusion. While prior systems often rely on flattening garments, humans routinely identify keypoints in highly crumpled and suspended states. We present a novel, task-agnostic, visuotactile framework that operates dire...
Neha Sunil;Megha Tippur;Arnau Saumell Portillo;Edward H Adelson;Alberto Rodriguez Garcia
Massachusetts Institute of Technology;Massachusetts Institute of Technology;Universidad Politécnica de Cataluna;;Massachusetts Institute of Technology
Oral
main
Deformable Object Manipulation;Dense Correspondence Learning;Confidence-Aware Planning;Visuotactile Perception
https://openreview.net/forum?id=sXoaNAECCK
-1
Reactive In-Air Clothing Manipulation with Confidence-Aware Dense Correspondence and Visuotactile Affordance Manipulating clothing is challenging due to their complex, variable configurations and frequent self-occlusion. While prior systems often rely on flattening garments, humans routinely identify keypoints in highl...
[ -0.06004716828465462, 0.003065147204324603, -0.04041488841176033, 0.03259265050292015, -0.007707203738391399, 0.008857532404363155, 0.0367913544178009, 0.013324643485248089, 0.03172990679740906, 0.03834429755806923, -0.04401925206184387, -0.007855787873268127, 0.009648383595049381, 0.03192...
corl_2025_seSw6ssEid
seSw6ssEid
corl
2,025
Residual Neural Terminal Constraint for MPC-based Collision Avoidance in Dynamic Environments
In this paper, we propose a hybrid MPC local planner that uses a learning-based approximation of a time-varying safe set, derived from local observations and applied as the MPC terminal constraint. This set can be represented as a zero-superlevel set of the value function computed via Hamilton-Jacobi (HJ) reachability ...
Bojan Derajic;Mohamed-Khalil Bouzidi;Sebastian Bernhard;Wolfgang Hönig
Technische Universität Berlin+Continental AG;Freie Universität Berlin+Continental AG;Continental AG;Technische Universität Berlin
Poster
main
MPC;Obstacle Avoidance;Learning for Control
https://openreview.net/forum?id=seSw6ssEid
-1
Residual Neural Terminal Constraint for MPC-based Collision Avoidance in Dynamic Environments In this paper, we propose a hybrid MPC local planner that uses a learning-based approximation of a time-varying safe set, derived from local observations and applied as the MPC terminal constraint. This set can be represented ...
[ -0.06307803094387054, -0.00861073937267065, -0.030491391196846962, 0.020088661462068558, -0.01377534493803978, 0.008086927235126495, 0.028929144144058228, 0.015254884958267212, 0.03583979234099388, 0.02789989858865738, -0.024260781705379486, 0.03481054678559303, 0.010568143799901009, 0.022...
corl_2025_t2asRJv2SD
t2asRJv2SD
corl
2,025
Belief-Conditioned One-Step Diffusion: Real-Time Trajectory Planning with Just-Enough Sensing
Robots equipped with rich sensor suites can localize reliably in partially-observable environments---but powering every sensor continuously is wasteful and often infeasible. Belief-space planners address this by propagating pose-belief covariance through analytic models and switching sensors heuristically--a brittle, r...
Gokul Puthumanaillam;Aditya Penumarti;Manav Vora;Paulo Padrao;Jose Fuentes;Leonardo Bobadilla;Jane Shin;Melkior Ornik
University of Illinois, Urbana Champaign;University of Florida;University of Illinois, Urbana Champaign;;Florida International University;Florida International University;University of Florida;
Oral
main
Navigation;Planning Under Uncertainty;Multi-Modal Sensing
https://openreview.net/forum?id=t2asRJv2SD
-1
Belief-Conditioned One-Step Diffusion: Real-Time Trajectory Planning with Just-Enough Sensing Robots equipped with rich sensor suites can localize reliably in partially-observable environments---but powering every sensor continuously is wasteful and often infeasible. Belief-space planners address this by propagating po...
[ -0.07103730738162994, -0.014093240723013878, -0.032828960567712784, 0.012601015157997608, -0.02365453727543354, 0.014940678142011166, 0.010814029723405838, 0.025699440389871597, 0.0015037396224215627, 0.053204286843538284, -0.021130651235580444, -0.01768563501536846, 0.04893025755882263, -...