DexForesight lifts dexterous policies with implicit spatial foresight
On 11 DexJoCo tasks success rises from 49.4% to 60.5%; three real-robot tasks reach 58.3% versus 50.0%.
Stories per day · 30 days
2 RSS
Relatedrobotics6reinforcement-learning3behavior-cloning2manipulation2vision-language-action2autodagger1data-collection1demonstration-curation1dexterous-manipulation1diffusion-policy1distillation1embodied-ai1fleet-learning1hindsight-relabeling1holoboxing1humanoid-robots1mobile-manipulation1neural-ode1
35 stories, newest first · one card per story
Stories tagged #imitation-learning
MOMASPLAT augments mobile manipulation demonstrations
MOMASPLAT more than doubles success rate on tasks with unseen base poses.
GUILD guided data collection improves imitation learning sample efficiency
GUILD alternates vision-language model task proposals with predictive data attribution evaluation to close the collection loop.
ODE Policy learns robot action dynamics rather than discrete actions
After one epoch it reports 5.9 times flow matching's success and the highest rate on 14 real-world robot tasks.
CuraEval finds demo curation rarely beats full robot training data
On robomimic, no curated subset significantly beat full-data training after Holm correction, though intervals still allow gains.
Underwater VLA policies beat baseline; ACT beats SmolVLA at sea
A 14-model DynamicVLA study and an ROV case study at a Malta wreck also open-source the BLUEMBODIED KIT.
Quality-controllable robot policies use iterative hindsight relabeling
Speed control raises conveyor throughput, and stability control improves success while carrying wine glasses.
HoloBoxing trains humanoid boxing in about seven minutes on one GPU
A frozen whole-body controller executes joints while a compact planner outputs future body-point targets via a 32-dimensional decoder.
DiscoDemo generates diverse robot demos to improve imitation learning
Authors report downstream success rising from 40% to 62% in simulation and from 39% to 78% on a real robot.
Collective experience learning improves robot policies from fleet failures
Held-out robots reach 93.3–99.2% success on five tasks; a 96-robot phone-packaging run rises from 79.2% to 98.3%.
TILDE distills embodied foundation models without action pretraining
AutoDAgger collected recovery trajectories more than twice as fast as a human-in-the-loop method.
VioLA: Learning Generalist Humanoid Control Policies from Human Data
VioLA trains a generalist humanoid policy mostly on human motion and follows real-robot instructions without task fine-tuning.
Dex-One2Many: Learning Dexterous Manipulation from a Single Human Demonstration
Dex-One2Many learns a generalizable dexterous manipulation policy from one human video via scene-graph-guided reinforcement learning.
EgoLAP: Learning from Egocentric Human Data through Language-Action Reasoning
EgoLAP pretrains robot policies on human and robot data via language action reasoning, reaching 80.1% task progress.
Conditional Trajectory Peaks: Single-Pass Multimodal Policies over Action Chunks
Conditional Trajectory Peaks models multimodal action chunks in one pass, matching pi0.5 at lower latency.
BRACE tracks humanoid motion under commanded force and slope
A flow-matching student distilled by DAgger runs on proprioception alone and was tested on the Unitree-G1.
InterMimicGen: Scaling Humanoid Loco-Manipulation through Self-Evolving Motion Imitation
InterMimicGen grows humanoid loco-manipulation data by retargeting human demos and self-evolving a physics tracker.
CoDance: Learning Reactive and Compliant Human-Humanoid Interaction from Video
CoDance learns compliant two-hand humanoid dancing from one partnered-dance video, including on a physical robot.
EyeRobot 2.0: Active Gaze for Precise Manipulation without Wrist Cameras
EyeRobot 2.0 achieves wrist-camera-level bimanual manipulation using only active gaze from a single stereo camera, outperforming passive stereo by 40% in real trials.
World Action Learning via Interaction-Centric Spectral Latent Guidance
WING transfers interaction knowledge from egocentric human video to robot policies via spectral latent guidance.
Understanding Multimodality in Generative Behavioral Cloning
Analysis shows generative behavioral cloning faces distinct bottlenecks when one observation has several valid actions.
In-context Robot Learning Made Simple: A Democratized Recipe for Manipulation Tasks
Researchers define robot in-context learning and introduce SimpleICL, a minimal reproducible manipulation framework.
Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation
PRISM turns a few real videos into counterfactual data so a humanoid carries novel objects without real-world fine-tuning.
Provable Benefits of Regularization: Fast Rates for Adversarial Imitation Learning
Theory paper proves fast finite-sample rates for dually regularized adversarial imitation learning.