Accepted Papers
29 papers were accepted to the workshop — 5 for oral presentation and 24 for poster presentation. All accepted papers are presented at the poster session, 3:30 – 4:15 PM; board assignments are listed below.
Oral Presentations (5)
Listed in presentation order. Each talk is 7 minutes: 6 minutes plus 1 minute for questions. Oral papers also present a poster.
| # | Paper | Board | Side |
|---|---|---|---|
| 1 | CrossBFM: Distilling a Shared Latent Behavior Space Across Humanoid Embodiments VinRobotics; National University of Singapore; TU Darmstadt; VinUniversity; DFKI; Hessian AI. | 2 | A |
| 2 | OmniTacTune: Policy-Agnostic Real-World RL for Tactile Residual Adaptation of Visual Policies University of Maryland, College Park; Georgia Institute of Technology. | 5 | A |
| 3 | Beyond Validation Loss: Coverage and Reliability Dynamics in Robot Policy Pretraining Stanford University; Carnegie Mellon University. | 1 | A |
| 4 | Multisensory Continual Learning: Adapting Pretrained Visuomotor Policies to Force Stanford University. | 6 | B |
| 5 | Bi-Lipschitz-Regularized Pretraining for Efficient Policy Finetuning University of Washington; Hitachi; University of California, Berkeley. | 7 | B |
Poster Presentations (24)
| # | Paper | Board | Side |
|---|---|---|---|
| 1 | MolmoAct2: Action Reasoning Models for Real-World Deployment Allen Institute for AI; University of Washington; National University of Singapore; University of Pennsylvania; Johns Hopkins University; Amazon; Cortex AI; University of Michigan; University of North Carolina at Chapel Hill. | 1 | B |
| 2 | RL²-VLA: Adaptive RL Latent Compositional Steering with Test-Time Scaling for Vision-Language-Action Models National University of Singapore; University of Toronto; Singapore Technologies Engineering. | 1 | B |
| 3 | Memory Anchors for Continual Robot Learning Stanford University; Toyota Research Institute. | 1 | A |
| 4 | Scouting the Dynamics Gap: Test-Time Policy Adaptation via Action-Outcome Feedback Carnegie Mellon University; Tsinghua University. | 2 | A |
| 5 | Support-Constrained RL Enables Real-World Policy Improvement without Additional Real-World Experience University of Washington; Amazon FAR. | 2 | B |
| 6 | Amortized Low-Rank Adaptation for Model-Based Reinforcement Learning University of Texas at Austin. | 2 | B |
| 7 | SimEX: Simulation-Integrated Robotics AutoResearch University of Texas at Austin; Amazon; Sony Europe; Carnegie Mellon University. | 3 | A |
| 8 | RARA: Robust Policy Fine-Tuning through Retrieval-Anchored Representation Alignment for Robotic Manipulation University of Virginia. | 3 | A |
| 9 | SWAP: Stepwise Action Policy Routing for Vision-Language-Action Models University of Southern California; NVIDIA; Georgia Institute of Technology. | 3 | B |
| 10 | The Gaussian Is Enough: Flow-Matching Priors Do Not Help When Fine-Tuning Large Behavior Models Toyota Research Institute; Woven by Toyota; Cornell University. | 3 | B |
| 11 | CAPABLE: Capability-Aware Policy Adaptation via Behavioral Latent Encoding University of North Texas; University of Bremen; Toyota Motor North America. | 4 | A |
| 12 | Grounded Action Model: 3D Grounding as a Foundation for Robotics Northwestern University; University of Washington; National University of Singapore. | 4 | A |
| 13 | Explicitly capturing demonstrator multi-modality for distillation into diffusion policies Massachusetts Institute of Technology. | 4 | B |
| 14 | You've Got a Golden Ticket: Improving Generative Robot Policies With A Single Noise Vector Arizona State University; The Robotics & AI Institute / The AI Institute; Northeastern University; Respite Robotics. | 4 | B |
| 15 | READ: What Makes a Pretrained VLA Teachable by Demonstration? University of Utah. | 5 | A |
| 16 | Preserving Plasticity in Continual Robot Learning from Pixels with Stiefel-Constrained Optimization University of California, Santa Barbara. | 5 | B |
| 17 | SUAVE: Unified Video-Action Models via Masked Diffusion Columbia University; Toyota Research Institute; McGill University. | 5 | B |
| 18 | QCP: Suboptimal Experience Can Be Useful for RL Post-Training University of Texas at Austin. | 6 | A |
| 19 | BOOST: Bootstrapping Offline and On-policy Successes to Train Real-World Robot Policies Toyota Research Institute; University of Washington; Stanford University. | 6 | A |
| 20 | Building Exploratory Robot Policies via Adversarial Mid Training Intel; Carnegie Mellon University; University of Washington; Toyota Research Institute; McGill University; Google. | 6 | B |
| 21 | Adapt for Planning, Not Just Prediction: Optimistic Test-Time World Model Adaptation Texas A&M University. | 7 | A |
| 22 | What Did We Learn from 20k Hours of Open-Source Robot Data? NVIDIA; Stanford University. | 7 | A |
| 23 | RoboAlign: Learning Test-Time Reasoning for Language-Action Alignment in Vision-Language-Action Models KAIST; Yonsei University; Amazon. | 7 | B |
| 24 | VLA Adaptation with Real and Simulated Data: A Case Study on Robotic Chess Charles University. | 8 | A |