| ZimaBlue: Evolving Generalizable World Action Models through Scalable Video Pre-training | ✓ Link | 94.7 | ZimaBlue (Slow System) | 2026-08-31 |
| Long-WAM: Scaling the Context of World-Action Models | ✓ Link | 94.7 | Long-WAM (IDM) | 2026-10-07 |
| World Action Learning via Interaction-Centric Spectral Latent Guidance | | 94.56 | WING | 2026-10-02 |
| ABot-M0.5: Unified Mobility-and-Manipulation World Action Model | | 94.0 | ABot-M0.5 (RoboTwin 2.0 Clean) | 2026-07-01 |
| OpenWAM: An Open, Modular Exploration Towards Systematic World-Action Model Pretraining | ✓ Link | 93.74 | OpenWAM-alpha (extra data) | 2026-09-07 |
| Qwen-RobotManip Technical Report: Alignment Unlocks Scale for Robotic Manipulation Foundation Models | ✓ Link | 93.7 | Qwen-RobotManip-Context (RoboTwin Easy) | 2026-06-17 |
| From Foundation to Application: Improving VLA Models in Practice | ✓ Link | 93.52 | LingBot-VLA 2.0 | 2026-07-07 |
| Rolling-WAM: World Action Models with Rolling Imagination | ✓ Link | 93.5 | Rolling-WAM | 2026-09-24 |
| InternVLA-A1.5: Unifying Understanding, Latent Foresight, and Action for Compositional Generalization | ✓ Link | 93.3 | InternVLA-A1.5-RoboTwin (Clean) | 2026-07-06 |
| ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing? | ✓ Link | 93.20 | ImageWAM (FLUX.2 4B) | 2026-06-17 |
| InternW0: A Foundational Physical World Model for Efficient Real-World Interactions | | 93.2 | InternW0 (extra data) | 2026-09-23 |
| Beyond Data Scaling: Representation-Centric Continued Pre-training for Vision-Language-Action Models | ✓ Link | 93 | VLAct (PI, data scaling) | 2026-08-27 |
| RealtimeWAM: One-Step Asynchronous World Action Models | ✓ Link | 92.98 | RealtimeWAM† | 2026-10-05 |
| World-Language-Action Model for Unified World Modeling, Language Reasoning, and Action Synthesis | ✓ Link | 92.94 | WLA-0 (RoboTwin 2.0 Clean) | 2026-06-04 |
| UNITAS: A 3D-Native World Action Model for Embodied Manipulation | ✓ Link | 92.94 | UNITAS | 2026-10-08 |
| Causal World Modeling for Robot Control | ✓ Link | 92.93 | LingBot-VA | 2026-01-29 |
| Acting Through Registers | ✓ Link | 92.9 | RegisterWAM | 2026-10-13 |
| AnyStep-WAM: Budget-Aligned Distillation and Adaptive Inference for World Action Models | ✓ Link | 92.74 | AnyStep-LingBotVA | 2026-09-27 |
| Enfold: Folding World Model Imagination into Predictive Representations for Ultra-Efficient Embodied Control | ✓ Link | 91.96 | Enfold-Flash | 2026-08-06 |
| Fast-WAM: Do World Action Models Need Test-time Future Imagination? | ✓ Link | 91.88 | Fast-WAM | 2026-03-17 |
| WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction | ✓ Link | 91.46 | WAVE-0 | 2026-10-13 |
| ACE-Ego-0: Unifying Egocentric Human and Robotic Data for VLA Pretraining | ✓ Link | 91.12 | ACE-Ego-0 | 2026-06-15 |
| Dream4ACT: A Shared Visual Action Interface for Multi-Embodiment Video-Action Modeling | | 90.5 | Dream4ACT | 2026-09-30 |
| VLA-Δ: Learning Physical State Transition for Vision-Language-Action Models | ✓ Link | 90.06 | VLA-Δ | 2026-10-15 |
| BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries | ✓ Link | 89.84 | BridgeWAM | 2026-10-10 |
| InternVLA-A1: Unifying Understanding, Generation and Action for Robotic Manipulation | ✓ Link | 89.4 | InternVLA-A1 (3B, RoboTwin Easy) | 2026-01-05 |
| StarVLA: A Lego-like Codebase for Vision-Language-Action Model Developing | ✓ Link | 88.7 | StarVLA (generalist) | 2026-04-06 |
| Less is More: Reducing Complexity in Vision-Language-Action Systems | ✓ Link | 88.7 | StarVLA-α (Generalist) | 2026-09-08 |
| Motus: A Unified Latent Action World Model | | 88.66 | Motus (RoboTwin 2.0 Clean) | 2025-12-15 |
| A Pragmatic VLA Foundation Model | ✓ Link | 88.56 | LingBot-VLA (RoboTwin 2.0 Clean) (extra data) | 2026-01-26 |
| Native Action-Prior Learning from Videos for World Action Models | | 88.5 | NAVA-WAM | 2026-10-02 |
| Flash-WAM: Modality-Aware Distillation for World Action Models | ✓ Link | 88.42 | Flash-WAM (1v/2a) | 2026-06-03 |
| TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAM | ✓ Link | 88.02 | TurboVLA (DINOv3 ViT-L) | 2026-07-29 |
| GreenWAM: A World Foundation Model for Robotics | ✓ Link | 86.98 | GreenWAM | 2026-10-13 |
| ABot-M0: VLA Foundation Model for Robotic Manipulation with Action Manifold Learning | ✓ Link | 86.06 | ABot-M0 (RoboTwin 2.0 clean) | 2026-02-11 |
| REKI: Registering Knowledge in VLAs | ✓ Link | 85.9 | REKI | 2026-10-15 |
| PerturBot: Breaking Shortcut Priors in Vision-Language-Action Models with Perturbative Training | ✓ Link | 85.6 | Perturbot (extra data) | 2026-10-03 |
| EventVLA: Event-Driven Visual Evidence Memory for Long-Horizon Vision-Language-Action Policies | ✓ Link | 83.8 | EventVLA (RoboTwin 2.0 Easy) | 2026-06-18 |
| Rethinking World-Action Model for Compositional and In-Context Robotic Manipulation | ✓ Link | 82.0 | ViGAR | 2026-10-01 |
| PINT: A Unified Model with Shared Predictive Intent for Robot Manipulation | ✓ Link | 71.9 | PINT-VA | 2026-10-12 |
| X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Model | ✓ Link | 70.0 | X-VLA-0.9B (RoboTwin-2.0 Easy) | 2025-10-11 |
| VLA-Rolling: Fewer Denoising Steps, Better Control via Adaptive Action Reuse | ✓ Link | 67.26 | VLA-Rolling (pi0.5) | 2026-10-18 |
| GigaBrain-0.7: Scaling Embodied Foundation Models to Emergent Capabilities with a Three-System Architecture | ✓ Link | 66.8 | GigaBrain-0.7 3.5B (Co-Train) | 2026-08-16 |