Research map / Image generation / Autoregressive and latent generation
DiffuseBot: Breeding Soft Robots With Physics-Augmented Generative Diffusion Models
Tsun-Hsuan Wang, Juntian Zheng, Pingchuan Ma, Yilun Du, Byungchul Kim, Andrew Everett Spielberg, Joshua B. Tenenbaum, Chuang Gan, Daniela Rus
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add embedding optimization (ablation) | diffusion with random embedding · Balancing | task performance: 0.081 -> 0.556 | Improved |
| add embedding optimization (ablation) | diffusion with random embedding · Landing | task performance: 0.832 -> 0.955 | Improved |
| add embedding optimization (ablation) | diffusion with random embedding · Crawling | task performance: 0.011 -> 0.048 | Improved |
| add embedding optimization (ablation) | diffusion with random embedding · Gripping | task performance: 0.014 -> 0.025 | Improved |
| add diffusion as co-design on top of embedding optimization (ablation) | DiffuseBot · Balancing | task performance: 0.556 -> 0.653 | Improved |
| add diffusion as co-design on top of embedding optimization (ablation) | DiffuseBot · Crawling | task performance: 0.048 -> 0.081 | Improved |
| add diffusion as co-design on top of embedding optimization (ablation) | DiffuseBot · Hurdling | task performance: 0.019 -> 0.035 | Improved |
| add diffusion as co-design on top of embedding optimization (ablation) | DiffuseBot · Gripping | task performance: 0.025 -> 0.027 | Improved |
| replace embedding optimization with manually designed text prompt (ablation) | DiffuseBot · Crawling | task performance: 0.048 -> 0.016 | Hurt |
| replace embedding optimization with finetuning the diffusion model (ablation) | DiffuseBot · Crawling | task performance: 0.048 -> 0.031 | Hurt |
| run co-design after diffusion instead of interleaved during sampling (ablation) | DiffuseBot · Crawling | task performance: 0.081 -> 0.064 | Hurt |
| DiffuseBot vs implicit function baseline | soft robot design representation · Hurdling | task performance: 0.044 -> 0.031 | Hurt |
Closest papers by meaning
- MorphVAE: Advancing Morphological Design of Voxel-Based Soft Robots with Variational Autoencoders (AAAI 2024 · similarity 0.87)
- DittoGym: Learning to Control Soft Shape-Shifting Robots (ICLR 2024 · similarity 0.85)
- One-Step Diffusion Policy: Fast Visuomotor Policies via Diffusion Distillation (ICML 2025 · similarity 0.85)
- Two-Steps Diffusion Policy for Robotic Manipulation via Genetic Denoising (NeurIPS 2025 · similarity 0.84)
- Constrained Synthesis with Projected Diffusion Models (NeurIPS 2024 · similarity 0.84)
- Embracing Evolution: A Call for Body-Control Co-Design in Embodied Humanoid Robot (NeurIPS 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
