AI Paper Atlas

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

When would Vision-Proprioception Policies Fail in Robotic Manipulation?

Jingxian Lu, Wenke Xia, Yuxuan Wu, Zhiwu Lu, Di Hu

ICLR 2026 · Vision-language and multimodal › Robot action and navigation · approach: Memory and subgoal-based manipulation · 4 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
add GAP gradient adjustment to vision-proprioception policyConcatenation policy · RoboSuite stack (MLP head)success rate (%): 54.6 -> 77.6Improved
naively concatenate proprioception with visionVision-only policy · Meta-World assemblysuccess rate (%): 82.6 -> 74.6Hurt
add GAP to concatenation policy on out-of-distribution placementConcatenation policy · RoboSuite threading OODsuccess rate (%): 28 -> 49Improved
apply GAP during Octo fine-tuning with proprioceptionOcto-VP · RoboSuite threadingsuccess rate (%): 57 -> 78Improved
apply GAP to a UNet diffusion policy headConcatenation policy · RoboSuite threadingsuccess rate (%): 40 -> 63Improved
add GAP on top of Summation and FiLM fusionSummation / FiLM fusion · Meta-World and RoboSuite (8 tasks)success rate (%): e.g. FiLM disassemble 47 -> 85Improved
replace CPD motion-distance phases with HDBSCAN clustering (ablation)GAP · RoboSuite threadingsuccess rate (%): 37 (HDBSCAN) -> 52 (CPD)Improved
use motion distance instead of cosine distance for CPD (ablation)GAP · RoboSuite threadingsuccess rate (%): 50 (CoTPC) -> 52Improved
predict transition indicator with LSTM instead of fixed rho (ablation)GAP · put hammer into drawer / threading / cubesuccess rate (%): fixed rho=0.5: 89/48/75 -> LSTM: 91/52/85Improved
increase adjustment strength lambda too high (0.8) or too low (0.1) (ablation)GAP · Meta-World assemblysuccess rate (%): 94 (0.3) -> 70 (0.8), 77 (0.1)Hurt
apply GAP for too many epochs (90 vs 50) (ablation)GAP · Meta-World assemblysuccess rate (%): 94 -> 72Hurt
add GAP to the vision branch before linear probing (ablation)vision-proprioception policy · assembly OOD (linear probe)success rate (%): 45 -> 69Improved

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