ResearchResearch paperRobotics & Embodied AI · Image, Video & 3D Generation · Computer Vision1 source · Oct 13, 2026

YUBI-Real2Sim: 3DGS Scene Verification for Reliable Policy Evaluation

To address this, we present YUBI-Real2Sim, a unified framework for 3DGS-based scene reconstruction and real-to-sim policy evaluation under dual fisheye wrist views.

Key points

  • Evaluating VLA policies on real robots is costly, time-consuming, and difficult to reproduce, making real-to-sim simulation essential for scalable evaluation.
  • While 3D Gaussian Splatting (3DGS) enables rapid real-to-sim scene generation, the reconstruction quality required for reliable policy evaluation remains unquantified.
  • We develop an integrated pipeline combining 200-degree fisheye camera capture, 3DGS training, 3D foundation models for object modeling, and MuJoCo-based physical simulation, reproducing the observed real-world policy ordering in eight of nine scene-task comparisons.
  • This systematically measures real-to-sim rank agreement and evaluates policy sensitivity against counterfactual 3DGS artifacts (ceiling holes, floaters, warping) and capture disturbances.

Sources (1)

  • [1]YUBI-Real2Sim: 3DGS Scene Verification for Reliable Policy Evaluation
    arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 12:33 AM
    To address this, we present YUBI-Real2Sim, a unified framework for 3DGS-based scene reconstruction and real-to-sim policy evaluation under dual fisheye wrist views.
    Evaluating VLA policies on real robots is costly, time-consuming, and difficult to reproduce, making real-to-sim simulation essential for scalable evaluation.

Extractive summary: sentences quoted from the sources.

Before this

  1. Oct 13, 2026WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction
  2. Oct 12, 2026Listen Before Act: Audio-Conditioned Robot Manipulation with ListenVLA
  3. Oct 12, 2026AUGUstus: Autonomous Driving VLA Model with Unified Generation and Understanding
  4. Oct 11, 2026$Δ$ Flow Matching for Robot Foundation Models
  5. Oct 11, 2026A Persian Vision-Language-Action Dataset and Benchmark for Ambiguity- and Infeasibility-Aware Instruction-to-Plan Grounding
  6. Oct 10, 2026Barrier-Constrained Guidance For Real-time Asynchronous VLA Execution

Related