AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

Trajectory-Level Continuous Action Representation for Robotic Manipulation

Tong Yang, Jingkai Jia, Yuecheng Xu, Xueyao Chen, Chi Zhang, Wenqiang Zhang

ECCV 2026 · Vision-language and multimodal › Robot action and navigation · approach: Memory and subgoal-based manipulation · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace timestep-level action prediction with CAT latentsSmolVLA · LIBEROsuccess rate (%): 87.3 -> 90.8Improved
replace timestep-level action prediction with CAT latentspi0 · LIBEROsuccess rate (%): 94.2 -> 96.5Improved
use CAT instead of FAST or VQ-VLA discretizationSmolVLA · LIBEROsuccess rate (%): 84.5 -> 90.8Improved
replace timestep-level prediction with CAT latentsDiffusion Policy (flow matching) · MimicGen 8 taskssuccess rate (%): 85.8 -> 86.9Improved
replace timestep-level prediction with CAT latentsDiffusion Policy (flow matching) · MimicGen Stack 3success rate (%): 76 -> 85Improved
add CAT across control frequenciesSmolVLA · RoboTwin 2.0 at 10Hzsuccess rate (%): 35.3 -> 45.6Improved
add CAT across control frequenciesSmolVLA · RoboTwin 2.0 at 16.7Hzsuccess rate (%): 42.1 -> 46.1Improved
add CATSmolVLA · RoboTwin 2.0 Adjust Bottle (smooth task)success rate (%): 75.0 -> 76.0No clear effect
add CATSmolVLA · Real-world long-horizon tasksnormalized task score: 45.0 -> 60.5Improved
use 4 register tokens instead of 2 (ablation)CAT (SmolVLA) · LIBEROsuccess rate (%): 88.0 -> 90.8Improved
use 4 register tokens instead of 8 (ablation)CAT (SmolVLA) · LIBEROsuccess rate (%): 87.3 -> 90.8Improved
add contrastive regularization (lambda 0.1) instead of none (ablation)CAT (SmolVLA) · LIBEROsuccess rate (%): 89.3 -> 90.8Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy