The library

Articles

Conceptual illustration of a robot policy preserving several distinct action paths instead of merging them into one averaged trajectory.

How Diffusion Policies Can Preserve Distinct Action Modes

Independent action-noise pairing can mix diffusion paths and produce averaged denoising responses that suppress distinct demonstrated behaviors. A label-free training add-on instead assigns noise to actions to preserve relatively separate routes without changing architecture or inference, although the reported results come from one preprint.

Conceptual illustration of a robot converting scene geometry, a goal, an action target and a predicted future state into one sequence of spatial tokens.

A Shared Spatial Vocabulary Lets One Robot Model Act and Forecast

A preprint proposes encoding scene contours, goals, action targets and future states as spatial sequences, then training one autoregressive Transformer for both action generation and action-conditioned forecasting. Its simulation and real-robot results are author-reported.

A generic simulation interface shows robot task configurations concentrated around the boundary between mastered and not-yet-reachable challenges.

Train Robots Where the Challenge Is Just Right

Success Guided Sampling prioritizes simulated task configurations around the boundary of a robot policy’s current abilities. The abstract reports that the method solved locomotion and assembly tasks that prior methods failed to solve and that distilled RGB-based policies transferred zero-shot to assembly tasks on real hardware.

Conceptual illustration of a robot gripper, a human hand, and moving scene points represented as trajectories in a shared three-dimensional workspace.

UNITAS Joins Robot Motion and Scene Dynamics in a Shared 3D Frame

Both occupy a shared metric frame, allowing one model interface to produce executable actions and action-conditioned forecasts of physical change.

Conceptual illustration of robot-arm frames in a strip, ending with a paler repeated frame after a divider.

What Happens to a Training Sequence at an Episode’s Edge?

A loader can complete a requested window near an episode’s end by repeating the final frame as padding. That repetition differs from both newly recorded steps and the earlier context supplied through frame stacking.

Conceptual illustration of a robot arm beside an archive of image cards and data strips.

Anatomy of a Robot Demonstration Dataset

A robot demonstration dataset can align images and other observations with recorded actions, organize them into trajectories, and supply schemas, task or environment context, statistics and subset labels. These files are inputs to learning algorithms, not learned policies themselves, and the presence of rewards or termination signals alone does not independently confirm successful task completion.