ReVeal: A Reconstruction-Aware Real-to-Sim Framework for VLA Policy Evaluation
On the map: Vision-language-action models → Failure recovery and robustness also fits: 3D and spatial-aware VLA
Abstract (arXiv)
Simulation-based evaluation provides a scalable and repeatable alternative to real-world evaluation of vision-language-action (VLA) policies. However, reconstruction errors can cause simulated policy performance to diverge from real-world performance, motivating the need to assess reconstructed environments for downstream VLA policy evaluation. We present ReVeal, a real-to-sim assessment framework combining workspace reconstruction, reconstruction-level assessment, and matched closed-loop policy evaluation. Novel-View Mesh Fidelity (NVMF) and Annotated Planar Geometry Fidelity (APGF) assess observation and planar geometric fidelity, respectively. We also develop PGSR-D, a reconstruction pipeline incorporating monocular depth supervision to improve geometry where multi-view visual cues are limited. Across 8 assessment scenes, NVMF and APGF consistently distinguish the fidelity of 2DGS, PGSR, and PGSR-D. Matched evaluations of GR00T, SmolVLA, and pi0.5 across 8 humanoid manipulation tasks show consistent ordering between reconstruction fidelity and real-sim performance agreement across pipelines. Further analysis of the evaluation workspaces shows that higher fidelity is associated with stronger real-sim agreement.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Failure recovery and robustness, one of 13 directions in Vision-language-action models. Builds on 37 of its 45 references that are in the map. First posted on arXiv 20 Sep 2026.
Builds on
37 of 45 references.
Closest ancestors · 1
Same direction, newest first
Other references · 36
- 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
- 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
- 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,141
- 2023Sigmoid Loss for Language Image Pre-Trainingcited by 3,048
- 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
- 2021Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fieldscited by 2,013
- 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
- 2024Octo: An Open-Source Generalist Robot Policycited by 1,427
- 2024Depth Anything V2cited by 1,385
- 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,167
- 20242D Gaussian Splatting for Geometrically Accurate Radiance Fieldscited by 1,061
- 2023ScanNet++: A High-Fidelity Dataset of 3D Indoor Scenescited by 747
24 more
- 2024Structured 3D Latents for Scalable and Versatile 3D Generationcited by 742
- 2025SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Roboticscited by 507
- 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 484
- 2024PGSR: Planar-based Gaussian Splatting for Efficient and High-Fidelity Surface Reconstructioncited by 280
- 2024DepthSplat: Connecting Gaussian Splatting and Depthcited by 275
- 2025SAM 3D: 3Dfy Anything in Imagescited by 230
- 2024DN-Splatter: Depth and Normal Priors for Gaussian Splatting and Meshingcited by 136
- 2024SplatSim: Zero-Shot Sim2Real Transfer of RGB Manipulation Policies Using Gaussian Splattingcited by 63
- 2025RoboArena: Distributed Real-World Evaluation of Generalist Robot Policiescited by 63
- 2025WorldGym: World Model as An Environment for Policy Evaluationcited by 61
- 2025AutoEval: Autonomous Evaluation of Generalist Robot Manipulation Policies in the Real Worldcited by 46
- 2025Re3Sim: Generating High-Fidelity Simulation Data via 3D-Photorealistic Real-to-Sim for Robotic Manipulationcited by 38
- 2025PolaRiS: Scalable Real-to-Sim Evaluations for Generalist Robot Policiescited by 29
- 2025Real-to-Sim Robot Policy Evaluation with Gaussian Splatting Simulation of Soft-Body Interactionscited by 27
- 2024RL-GSBridge: 3D Gaussian Splatting Based Real2Sim2Real Method for Robotic Manipulation Learningcited by 21
- 2025MeshSplatting: Differentiable Rendering with Opaque Meshescited by 19
- 2025MonoInstance: Enhancing Monocular Priors via Multi-view Instance Alignment for Neural Rendering and Reconstructioncited by 15
- 2023ContraNeRF: Generalizable Neural Radiance Fields for Synthetic-to-real Novel View Synthesis via Contrastive Learningcited by 14
- 2023MuSHRoom: Multi-Sensor Hybrid Room Dataset for Joint 3D Reconstruction and Novel View Synthesiscited by 12
- 2025High-Fidelity Simulated Data Generation for Real-World Zero-Shot Robotic Manipulation Learning with Gaussian Splattingcited by 12
- 2025GS-2DGS: Geometrically Supervised 2DGS for Reflective Object Reconstructioncited by 11
- 2026SimFoundry: Modular and Automated Scene Generation for Policy Learning and Evaluationcited by 11
- 2025FIORD: A Fisheye Indoor-Outdoor Dataset with LIDAR Ground Truth for 3D Scene Reconstruction and Benchmarkingcited by 7
- 2025A Data-Centric Revisit of Pre-Trained Vision Models for Robot Learningcited by 6
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026A Practical Recipe Towards Improving Sim-and-Real Correlation for VLA Evaluation
- 2026VLA-REPLICA: A Low-Cost, Reproducible Benchmark for Real-World Evaluation of Vision-Language-Action Models
- 2026VLAW: Iterative Co-Improvement of Vision-Language-Action Policy and World Model
- 2026A Sim-to-Real Integration Pipeline for Training and Deployment of Chunk-Based VLA Manipulation Policies
- 2026Reflective VLA: In-Context Action Consequences Make VLAs Generalize
- 2026R2S-Eval: Robot Evaluation with Real-to-Sim Calibration via Vision-Language Models
- 2026VeriSpace: Spatially Grounded Action Verification for Vision-Language-Action Models
- 2025Eva-VLA: Evaluating Vision-Language-Action Models' Robustness Under Real-World Physical Variations