AI Paper Atlas

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

The Lie We Tell: Correcting the Euclidean Fallacy in Vision Language Action Policies via Score Matching on Tangent Space

Bing-Cheng Chuang, I-Hsuan Chu, Bor-Jiun Lin, YuanFu Yang, Min Sun, Chun-Yi Lee

ICML 2026 · Vision-language and multimodal › Robot action and navigation · approach: Memory and subgoal-based manipulation · 0 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 Euclidean diffusion with Lie group diffusion on SE(3) plus GAT encoder3D Diffuser Actor · CALVIN ABC->Daverage task length: 3.27 -> 3.512Improved
add GAT encoder (keeping Euclidean diffusion) (ablation)3D Diffuser Actor · CALVIN ABC->Daverage task length: 3.27 -> 3.474Improved
add Lie space diffusion (without GAT encoder) (ablation)3D Diffuser Actor · CALVIN ABC->Daverage task length: 3.27 -> 3.368Improved
add Lie space diffusion to the GAT-encoder model (ablation)Lie Diffuser Actor · CALVIN ABC->Daverage task length: 3.474 -> 3.512Improved
add GAT encoder and Lie diffusion3D Diffuser Actor · CALVIN ABCD->Daverage task length: 3.288 -> 3.584Improved
add GAT encoder only (Euclidean diffusion, 400K) (ablation)3D Diffuser Actor · CALVIN ABCD->Daverage task length: 3.288 -> 3.239Hurt
add Lie diffusion without GAT encoder (ablation)3D Diffuser Actor · CALVIN ABCD->Daverage task length: 3.288 -> 3.404Improved
add SE(3) score matchingOpenVLA-OFT · LIBERO Longsuccess rate: 92.20 -> 94.13Improved
use Lie diffusion on real robot3D Diffuser Actor · Real robot: Sort Blockssuccess rate (%): 55 -> 75Improved
use Lie diffusion on real robot3D Diffuser Actor · Real robot: Put Block in Boxsuccess rate (%): 80 -> 75No clear effect
predict in so(3) and map via exponential map instead of SVD orthogonalization3D Diffuser Actor · Orthogonality violation, medianorthogonality violation: 5.7% reduction at medianImproved
diffuse on Lie group instead of Euclidean space3D Diffuser Actor · Geodesic jitter during denoisinggeodesic jitter: nearly an order of magnitude lowerImproved

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