{"paper":{"title":"Real2Sim: A Physics-driven and Editable Gaussian Splatting Framework for Autonomous Driving Scenes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A framework fuses 4D Gaussian Splatting with a physics solver to reconstruct and edit dynamic driving scenes while preserving realistic interactions.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Kaicong Huang, Ruimin Ke, Talha Azfar, Weisong Shi","submitted_at":"2026-05-13T14:26:25Z","abstract_excerpt":"Reliable autonomous driving relies on large-scale, well-labeled data and robust models. However, manual data collection is resource-intensive, and traditional simulation suffers from a persistent reality gap. While recent generative frameworks and radiance-field methods improve visual fidelity, they still struggle with temporal and spatial consistency and cannot ensure physics-aware behavior, limiting their applicability to driving scenario generation. To address these challenges, we propose Real2Sim, an unified framework that combines 4D Gaussian Splatting (4DGS) with a differentiable Materia"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Real2Sim explicitly reconstructs dynamic driving scenes as temporally continuous Gaussian primitives, supports instance-level editing, and simulates realistic object-object and object-environment interactions. This framework enables physics-aware, high-fidelity synthesis of diverse, editable scenarios, including challenging corner cases such as collisions and post-impact trajectories.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the differentiable MPM solver can be tightly coupled with 4D Gaussian Splatting without introducing visual artifacts, temporal inconsistencies, or loss of physical accuracy when applied to complex real-world driving data.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Real2Sim reconstructs editable dynamic driving scenes as temporally continuous Gaussians integrated with a differentiable MPM physics solver for high-fidelity simulation of interactions and collisions.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A framework fuses 4D Gaussian Splatting with a physics solver to reconstruct and edit dynamic driving scenes while preserving realistic interactions.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"eec23c91314e99fc4215488e65569f4e695d113bd0f1123a269709f93f2b6112"},"source":{"id":"2605.13591","kind":"arxiv","version":1},"verdict":{"id":"a2759f60-d943-40e2-b3d8-fdd7e39df208","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:27:01.406120Z","strongest_claim":"Real2Sim explicitly reconstructs dynamic driving scenes as temporally continuous Gaussian primitives, supports instance-level editing, and simulates realistic object-object and object-environment interactions. This framework enables physics-aware, high-fidelity synthesis of diverse, editable scenarios, including challenging corner cases such as collisions and post-impact trajectories.","one_line_summary":"Real2Sim reconstructs editable dynamic driving scenes as temporally continuous Gaussians integrated with a differentiable MPM physics solver for high-fidelity simulation of interactions and collisions.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the differentiable MPM solver can be tightly coupled with 4D Gaussian Splatting without introducing visual artifacts, temporal inconsistencies, or loss of physical accuracy when applied to complex real-world driving data.","pith_extraction_headline":"A framework fuses 4D Gaussian Splatting with a physics solver to reconstruct and edit dynamic driving scenes while preserving realistic interactions."},"references":{"count":27,"sample":[{"doi":"","year":2020,"title":"Scalability in perception for autonomous driving: Waymo open dataset","work_id":"cf5ae3fc-bc37-46a5-87ce-fb9a52cded92","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Surfelgan: Synthesizing realistic sensor data for autonomous driving,","work_id":"35ce85bd-54df-4f8e-8e16-783f5310491f","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"GAIA-2: A Controllable Multi-View Generative World Model for Autonomous Driving","work_id":"1339e674-d09b-48b4-8e6f-efe55dcab22e","ref_index":3,"cited_arxiv_id":"2503.20523","is_internal_anchor":true},{"doi":"","year":2021,"title":"Nerf: Representing scenes as neural radiance fields for view synthesis","work_id":"b42b10ce-3678-4f2e-bf7e-da503030acf8","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"3d gaussian splatting for real-time radiance field rendering","work_id":"698160da-5666-42e0-b122-4a7c98e53e9b","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":27,"snapshot_sha256":"e77d70e3be50e577b4bb7c39383f43f4f5e944f3bd11083244b48b3ddc2864ab","internal_anchors":1},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}