cd /news/robotics/dock-dual-stream-inspired-self-super… · home › topics › robotics › article
[ARTICLE · art-172344] src=arxiv.org ↗ pub= topic=robotics verified=true sentiment=↑ positive

DOCK: Dual-Stream-Inspired Self-Supervised Object-Centric Representations Grounded in Robot Kinematics

A new method called DOCK, described in arXiv paper 2611.23578v1, learns object-centric robot representations from 9,951 held-object transport trajectories across 21 simulated kitchen scenes, mining 404K displacement pairs using forward kinematics on joint angles as an annotation-free proxy for object motion. The authors report that six frozen DINO and V-JEPA encoders showed rank correlation of at most 0.11 between feature distance and an object's distance to its goal, versus 0.87 for object-mask centroids, and that DOCK used as a planning cost places the object in 15 of 21 tasks, matching an oracle reading true object position, while mask geometry reached 10 and frozen feature distance 9. The representation's ranking also transferred to held-out layouts with unseen object instances and categories without adaptation.

recommendation

Before trusting latent distance as a planning cost, measure its rank correlation with true object-goal distance; frozen encoder features scored at most 0.11 versus 0.87 for object-mask centroids.

more in playbook: Evals and testing for LLM apps
by read1 min views3 publishedNov 19, 2026

arXiv:2611.23578v1 Announce Type: new Abstract: Latent-space planners select actions by comparing the embedding of a predicted future image with that of a goal image. This assumes that distance in the embedding reflects how far the scene is from the goal. We profile this assumption with simulator-executed futures, which removes prediction error. For six frozen DINO and V-JEPA encoders, feature distance barely tracks the object's distance to its goal: the rank correlation is at most 0.11, against 0.87 for object-mask centroids. It is dominated by background and robot patches, and frozen-feature readouts capture the arm far better than the object. Inspired by the ventral (what) and dorsal (where/how) visual streams, we propose DOCK, which learns an object representation whose changes agree with measured physical motion. A grasped object moves with the gripper, so forward kinematics on the joint angles gives a proxy for the object's displacement without annotation. From 9,951 held-object transport trajectories in 21 simulated kitchen scenes, we mine 404K displacement pairs and train three components: an object token pooled from patch features (what), a spatial ruler that makes differences of readout coordinates equal to measured displacements (where), and a dynamics predictor over object tokens (how). Used as a planning cost, the learned representation places the object in 15 of 21 tasks, as many as an oracle that reads the true object position, while mask geometry reaches 10 and frozen feature distance 9. It still solves the most tasks when the unseen robot arm is visible, and its ranking transfers to held-out layouts with unseen object instances and categories without any adaptation.

── more in #robotics 4 stories · sorted by recency
── more on @dock 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/dock-dual-stream-ins…] indexed:0 read:1min 2026-11-19 · —