The paper introduces Constraint-Referenced Strategy (CRS), a method that evaluates manipulation policies by projecting the commanded end-effector twist onto an object's joint axis, thereby distinguishing progress along the joint from off-constraint motion. CRS requires a joint model and a scripted controller but no demonstrations or learned judges. Experiments on 74 simulated articulated objects show that CRS can identify distinct signatures of policy failures and highlight significant off-axis rotation in a door-trained policy compared to a drawer-trained one.
arXiv:2609.13235v1 Announce Type: cross Abstract: Networked manipulation endpoints couple perception to actuation across compute- and bandwidth-limited links, yet commonly exchange dense geometric st...
The paper reports a 2nd place solution for the HANDS 2026 Dexterous Grasp Motion track, targeting the 12‑DoF LinkerHand O6. The method warps a single successful GraspM3 demonstration into a 12‑D trajectory, avoiding step‑by‑step policy generation, and is trained with one‑step PPO across 4,824 objects. It achieved 94.61% success on the easy track and 57.18% on the hard track of the private test set.
TacSushi is a tactile‑grounded, Cosmos3‑based world‑action policy for dexterous sushi manipulation. It encodes RGB, language, and hand state, fusing fingertip tactile data via feature‑wise gated fusion, and learns from future‑consequence predictions while excluding failed actions from imitation. Trained on 340 successful and 50 failed trials, TacSushi achieves 68.3% in‑distribution and 37.5% out‑of‑distribution success, outperforming baselines that lack future‑consequence supervision or use direct tactile concatenation.
VioLA is a generalist humanoid control policy that predicts body and hand motion latents instead of joint commands, enabling a pretrained controller to execute these latents on a robot. By aligning human and robot motion in shared latent spaces, the policy can be trained on a large dataset of 140.6 million frames, 93.2% of which are human demonstrations. As a result, VioLA can follow locomotion and manipulation instructions on a real robot zero‑shot, achieving 100% locomotion success and 88.6% manipulation success without task‑specific fine‑tuning.
arXiv:2607. 11874v1 Announce Type: cross Abstract: Recent work in humanoid whole-body control has found success with a simple recipe: retarget human motion to robot kinematic references, then train policies via reinforcement learning (RL) to track them.
arXiv:2609.01453v1 Announce Type: cross Abstract: Dexterous manipulation policies learned by imitation are typically evaluated for robustness to variation in scenes, objects, or instructions, but the...
The paper introduces DONIC, a generalist bimanual tracking controller trained on diverse operation categories and human-motion data, and BiHOI, a multi‑view dataset of 1,575 sequences with 229 objects that captures sustained bimanual interactions. DONIC employs proximal policy optimization combined with behavioral cloning, an object‑centered per‑hand reward, and Clip‑Centered Advantages to learn coordinated manipulation from a single policy. The method achieves 86.79% success on unseen trajectories of training objects and 57.32% on held‑out objects, outperforming existing bimanual tracking baselines by large margins.
BiDex-100 is a simulation benchmark for bimanual dexterous manipulation, featuring 106 tasks across five capability axes—generalization, instruction following, limited demonstrations, long horizon, and memory. Each task includes an out‑of‑domain variant to test robustness, and the benchmark supports two hand embodiments with 8,080 teleoperated demonstrations totaling 6.25 million frames. The authors evaluate seven recent Vision‑Language‑Action models under a unified protocol, providing baseline performance for capability, robustness, and data efficiency.
Screw Attention introduces a transformer layer that treats the relation between two bodies as a spatial transform rather than a graph edge, enabling each token to represent a body with a pose and its relative pose or joint screw. This design ensures equivariance to independent frame changes and allows a single layer to capture rigid‑body velocity recursion. Experiments on simulated manipulation tasks show that Screw Attention matches or outperforms other network architectures, achieving high success rates on LIBERO‑Spatial with far fewer parameters and maintaining performance under frame convention changes and pose noise.
arXiv:2606. 06218v1 Announce Type: cross Abstract: A policy tuned for one robot often behaves differently on another, whether due to the sim-to-real gap, unknown payloads, or the differing dynamics of two instances of the same robot.
The paper introduces Watch, Infer, Coordinate, a benchmark and inference method for robots to deduce a partner’s physical constraints by observing joint behavior in physically coupled manipulation tasks. It demonstrates that understanding these constraints enables zero‑shot coordination on new tasks, achieving performance close to an oracle with true constraint knowledge. The study focuses on scenarios where a robot’s limitations may be unknown to its partner, such as hardware degradation or actuator faults.
Recent work in humanoid whole-body control has found success with a simple recipe: retarget human motion to robot kinematic references, then train policies via reinforcement learning (RL) to track them. But how does this recipe transfer to dexterous manipulation?