# DSO: Aligning 3D Generators with Simulation Feedback for Physical Soundness

arXiv: https://arxiv.org/abs/2503.22677 · HTML: https://mlsanity.com/paper/2503.22677/

## Key facts

- Cited by 25 papers in the map (data through 14 Oct 2026).
- Builds on 98 of its 119 references that are in the map.
- First posted on arXiv 28 Mar 2025.

## Builds on (most cited first, 30 of 98)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Score-Based Generative Modeling through Stochastic Differential Equations](https://mlsanity.com/paper/2011.13456/) (2020)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Deep Unsupervised Learning using Nonequilibrium Thermodynamics](https://mlsanity.com/paper/1503.03585/) (2015)
- [3D Gaussian Splatting for Real-Time Radiance Field Rendering](https://mlsanity.com/paper/2308.04079/) (2023)
- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow](https://mlsanity.com/paper/2209.03003/) (2022)
- [Elucidating the Design Space of Diffusion-Based Generative Models](https://mlsanity.com/paper/2206.00364/) (2022)
- [DreamFusion: Text-to-3D using 2D Diffusion](https://mlsanity.com/paper/2209.14988/) (2022)
- [Trust Region Policy Optimization](https://mlsanity.com/paper/1502.05477/) (2015)
- [pixelNeRF: Neural Radiance Fields from One or Few Images](https://mlsanity.com/paper/2012.02190/) (2020)
- [Objaverse: A Universe of Annotated 3D Objects](https://mlsanity.com/paper/2212.08051/) (2022)
- [Zero-1-to-3: Zero-shot One Image to 3D Object](https://mlsanity.com/paper/2303.11328/) (2023)
- [Magic3D: High-Resolution Text-to-3D Content Creation](https://mlsanity.com/paper/2211.10440/) (2022)
- [ProlificDreamer: High-Fidelity and Diverse Text-to-3D Generation with Variational Score Distillation](https://mlsanity.com/paper/2305.16213/) (2023)
- [Efficient Geometry-aware 3D Generative Adversarial Networks](https://mlsanity.com/paper/2112.07945/) (2021)
- [Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning](https://mlsanity.com/paper/2108.10470/) (2021)
- [Building Normalizing Flows with Stochastic Interpolants](https://mlsanity.com/paper/2209.15571/) (2022)
- [Training Diffusion Models with Reinforcement Learning](https://mlsanity.com/paper/2305.13301/) (2023)
- [MVDream: Multi-view Diffusion for 3D Generation](https://mlsanity.com/paper/2308.16512/) (2023)
- [LRM: Large Reconstruction Model for Single Image to 3D](https://mlsanity.com/paper/2311.04400/) (2023)
- [Diffusion Model Alignment Using Direct Preference Optimization](https://mlsanity.com/paper/2311.12908/) (2023)
- [GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields](https://mlsanity.com/paper/2011.12100/) (2020)
- [Objaverse-XL: A Universe of 10M+ 3D Objects](https://mlsanity.com/paper/2307.05663/) (2023)
- [Structured 3D Latents for Scalable and Versatile 3D Generation](https://mlsanity.com/paper/2412.01506/) (2024)

## Built on by (most cited first, 25 of 25)

- [Epipolar Geometry Improves Video Generation Models](https://mlsanity.com/paper/2510.21615/) (2025)
- [Particulate: Feed-Forward 3D Object Articulation](https://mlsanity.com/paper/2512.11798/) (2025)
- [Reconstructing 4D Spatial Intelligence: A Survey](https://mlsanity.com/paper/2507.21045/) (2025)
- [Is Single-View Mesh Reconstruction Ready for Robotics?](https://mlsanity.com/paper/2505.17966/) (2025)
- [SimuScene: Simulation-Ready Compositional 3D Scene Reconstruction from a Single Image](https://mlsanity.com/paper/2606.03994/) (2026)
- [Instruct-Particulate: Scaling Feed-Forward 3D Object Articulation with Kinematic Control](https://mlsanity.com/paper/2606.14699/) (2026)
- [Fast3Dcache: Training-free 3D Geometry Synthesis Acceleration](https://mlsanity.com/paper/2511.22533/) (2025)
- [MotionCrafter: Dense Geometry and Motion Reconstruction with a 4D VAE](https://mlsanity.com/paper/2602.08961/) (2026)
- [CADKnitter: Compositional CAD Generation from Text and Geometry Guidance](https://mlsanity.com/paper/2512.11199/) (2025)
- [Stroke3D: Lifting 2D strokes into rigged 3D model via latent diffusion models](https://mlsanity.com/paper/2602.09713/) (2026)
- [RnG: A Unified Transformer for Complete 3D Modeling from Partial Observations](https://mlsanity.com/paper/2603.01194/) (2026)
- [HSImul3R: Physics-in-the-Loop Reconstruction of Simulation-Ready Human-Scene Interactions](https://mlsanity.com/paper/2603.15612/) (2026)
- [AnyHand: A Large-Scale Synthetic Dataset for RGB(-D) Hand Pose Estimation](https://mlsanity.com/paper/2603.25726/) (2026)
- [A Cross-Model VLM-Judge Protocol for Single-Image 3D Mesh Quality (and Why Cheap Proxies Fall Short)](https://mlsanity.com/paper/2606.18451/) (2026)
- [DensiCrafter: Physically-Constrained Generation and Fabrication of Self-Supporting Hollow Structures](https://mlsanity.com/paper/2511.09298/) (2025)
- [DeDPO: Debiased Direct Preference Optimization for Diffusion Models](https://mlsanity.com/paper/2602.06195/) (2026)
- [3D Generation for Embodied AI and Robotic Simulation: A Survey](https://mlsanity.com/paper/2604.26509/) (2026)
- [DiagGen: Agentic Generation of Deformable Assets with Sim-based Diagnostics for Robotic Simulation](https://mlsanity.com/paper/2609.23103/) (2026)
- [Prompt-to-Parts: Generative AI for Physical Assembly and Scalable Instructions](https://mlsanity.com/paper/2512.15743/) (2025)
- [Generating Physically Sound Designs from Text and a Set of Physical Constraints](https://mlsanity.com/paper/2602.02213/) (2026)
- [Physically Grounded 3D Generative Reconstruction under Hand Occlusion using Proprioception and Multi-Contact Touch](https://mlsanity.com/paper/2604.09100/) (2026)
- [Judging to Improve: A De-biased VLM-as-3D-Judge Protocol for Single-Image 3D Generation](https://mlsanity.com/paper/2606.20364/) (2026)
- [SNAP3D: Physically Grounded 3D Parts for Assembly from a Single Image](https://mlsanity.com/paper/2609.13146/) (2026)
- [ESG: Generating Physically Consistent Dynamic 3D Scenes from Text Descriptions](https://mlsanity.com/paper/2609.15392/) (2026)
- [Towards Physically Grounded Articulated 3D Assets: A Multi-Agent Pipeline and Benchmark](https://mlsanity.com/paper/2610.13947/) (2026)

## Abstract (arXiv)

Most 3D object generators prioritize aesthetic quality, often neglecting the physical constraints necessary for practical applications. One such constraint is that a 3D object should be self-supporting, i.e., remain balanced under gravity. Previous approaches to generating stable 3D objects relied on differentiable physics simulators to optimize geometry at test time, which is slow, unstable, and prone to local optima. Inspired by the literature on aligning generative models with external feedback, we propose Direct Simulation Optimization (DSO). This framework leverages feedback from a (non-differentiable) simulator to increase the likelihood that the 3D generator directly outputs stable 3D objects. We construct a dataset of 3D objects labeled with stability scores obtained from the physics simulator. This dataset enables fine-tuning of the 3D generator using the stability score as an alignment metric, via direct preference optimization (DPO) or direct reward optimization (DRO) - a novel objective we introduce to align diffusion models without requiring pairwise preferences. Our experiments demonstrate that the fine-tuned feed-forward generator, using either the DPO or DRO objective, is significantly faster and more likely to produce stable objects than test-time optimization. Notably, the DSO framework functions even without any ground-truth 3D objects for training, allowing the 3D generator to self-improve by automatically collecting simulation feedback on its own outputs.