AI Paper Atlas

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

Dual-Stream Diffusion for World-Model Augmented Vision-Language-Action Model

John Won, Kyungmin Lee, Huiwon Jang, Dongyoung Kim, Jinwoo Shin

ICML 2026 · Vision-language and multimodal › Robot action and navigation · approach: Vision-language-action foundation models · 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
Use dual-stream MMDiT with decoupled noise (DUST) instead of GR00T-N1.5GR00T-N1.5 · RoboCasa (100 demos/task)average success rate (%): 41.7 -> 50.1Improved
Use DUST instead of FLARE world modelingGR00T-N1.5 · GR-1 (1,000 demos/task)average success rate (%): 36.3 -> 42.0Improved
Use DUST on real Franka robot tasksGR00T-N1.5 · Real-world Franka Research 3average success rate (%): 46.5 -> 59.9Improved
Add decoupled (independent per-modality) noise to MMDiT (ablation)DUST · RoboCasa (100 demos/task)average success rate (%): 38.2 -> 50.1Improved
Use dual-stream MMDiT instead of single-stream DiT (ablation)DUST · RoboCasa (100 demos/task)average success rate (%): 42.5 -> 50.1Improved
Pretrain on action-free BridgeV2 video before finetuningDUST · RoboCasa (100 demos/task)average success rate (%): 50.1 -> 58.5Improved
Joint-train with GR-1 and EgoDex human data mixtureDUST · RoboCasa (300 demos/task)average success rate (%): 58.5 -> 64.4Improved
Joint-train with GR-1 and EgoDex data mixtureGR00T-N1.5 · RoboCasa (300 demos/task)average success rate (%): 45.0 -> 46.9Improved
Increase vision denoising steps from 4 to 64 (asynchronous sampling)DUST · RoboCasa (1,000 demos/task)average success rate (%): 66.3 -> 69.7Improved
Increase vision denoising steps from 4 to 32DUST · GR-1 (1,000 demos/task)average success rate (%): 42.0 -> 47.1Improved
Increase vision denoising steps from 32 to 64DUST · GR-1 (1,000 demos/task)average success rate (%): 47.1 -> 45.0Hurt
Set world-modeling loss weight to 0.2 instead of 1.0 (ablation)DUST · RoboCasa (100 demos/task)average success rate (%): 50.1 -> 34.3Hurt

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