Modality-Autoregressive World-Action Models
Overview
ModAR, a new world-action model, outperforms existing WAMs by denoising multiple future modalities before predicting actions, achieving 75% success with fewer FLOPs and no pretraining. Evaluations show benefits from predicting point tracks, DINO features, and depth maps. arXiv:2609.17524v2 Latest: ModAR achieves 75% average success rate with $20 imes$ fewer training FLOPs and no pretraining, outperforming existing WAM formulations in real-world bimanual tasks.
Generated from attributed reports · Updated 3 hours ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
No current evidence-backed claims. Missing information remains not reported.
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsModality-Autoregressive World-Action Models
The paper presents ModAR, the first world-action model (WAM) to autoregressively denoise multiple future modalities before predicting actions, enabling each prediction to condition on previously generated modalities. Evaluations show that ModAR outperforms existing WAM formulations, achieving a higher average success rate with fewer training FLOPs and no pretraining, particularly in real-world bimanual tasks.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).