Research map / Image generation / Fast diffusion sampling
Leapfrog Diffusion Model for Stochastic Trajectory Prediction
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the leapfrog initializer and its remaining denoising steps instead of standard diffusion | Diffusion trajectory prediction · NBA, Total (4.0s) | minADE: 0.96 -> 0.81 | Helped |
| Use the leapfrog initializer and its remaining denoising steps instead of standard diffusion | Diffusion trajectory prediction · NFL, Total (3.2s) | minADE: 1.14 -> 0.87 | Helped |
| Use the leapfrog initializer and its remaining denoising steps instead of standard diffusion | Diffusion trajectory prediction · NBA | inference time: ∼886ms -> ∼46ms | Helped |
| Add variance estimation to the initializer (ablation) | LED · NFL, K=4 | minADE: 1.63 -> 1.47 | Helped |
| Add mean estimation to the initializer (ablation) | LED · NFL, K=4 | minADE: 1.49 -> 1.47 | Helped |
| Use correlated rather than i.i.d. sample prediction (ablation) | LED · NFL, K=4 | minADE: 1.90 -> 1.47 | Helped |
| Set leapfrog step τ to 3 rather than 5 (ablation) | LED · NBA, Total (4.0s) | minADE: 0.81 -> 0.84 | Hurt |
| Set leapfrog step τ to 10 rather than 5 (ablation) | LED · NBA, Total (4.0s) | minADE: 0.81 -> 0.82 | Hurt |
| Use standard diffusion with 100 denoising steps instead of leapfrog initialization (ablation) | Diffusion trajectory prediction · NBA, Total (4.0s) | minADE: 0.81 -> 0.94 | Hurt |
Closest papers by meaning
- SingularTrajectory: Universal Trajectory Predictor Using Diffusion Model (CVPR 2024 · similarity 0.88)
- GeoSPRINT: Geometric Redundancy-Aware Step Pruning for Inference in Diffusion Trajectories (NeurIPS 2026 · similarity 0.87)
- Neural Flow Diffusion Models: Learnable Forward Process for Improved Diffusion Modelling (NeurIPS 2024 · similarity 0.87)
- Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous and Instruction-guided Driving (CVPR 2024 · similarity 0.87)
- Dynamical Diffusion: Learning Temporal Dynamics with Diffusion Models (ICLR 2025 · similarity 0.87)
- Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion (NeurIPS 2024 · similarity 0.86)
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.