Research map / Probabilistic and scientific modeling / Variational and generative sampling
Likelihood-Based Diverse Sampling for Trajectory Forecasting
Yecheng Jason Ma, Jeevana Priya Inala, Dinesh Jayaraman, Osbert Bastani
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 the diversity loss (ablation) | LDS-AF · nuScenes | minADE_5: 5.95 -> 2.06 | Improved |
| add the diversity loss (ablation) | LDS-AF · nuScenes | minASD_5: 0.16 -> 3.09 | Improved |
| add the likelihood loss (ablation) | LDS-AF · nuScenes | minADE_5: 8.06 -> 2.06 | Improved |
| use NLL likelihood loss instead of DLow reconstruction+KL (ablation) | LDS-AF · nuScenes | minADE_5: 2.16 -> 2.06 | Improved |
| use min instead of mean in the diversity loss (ablation) | LDS-AF · nuScenes | minADE_5: 4.85 -> 2.06 | Improved |
| use min instead of mean in the diversity loss (ablation) | LDS-AF · nuScenes | minASD_10: 0.03 -> 1.98 | Improved |
| add epsilon stochasticity to LDS training (ablation) | LDS-AF (K=10) · nuScenes | minADE_10: 1.74 -> 1.65 | Improved |
| apply LDS post-hoc to the CVAE | CVAE · nuScenes | minADE_5: 2.71 -> 2.16 | Improved |
| apply LDS post-hoc to the flow model | AF · nuScenes | minADE_5: 2.86 -> 2.06 | Improved |
| apply DSF post-hoc to CAM-NF | CAM-NF · Forking Paths (test, avg of 45-Degree and Top Down) | minADE_20: 148.0 -> 162.8 | Hurt |
| apply LDS post-hoc to CAM-NF | CAM-NF · Forking Paths (test, avg of 45-Degree and Top Down) | minADE_20: 148.0 -> 98.6 | Improved |
| use transductive neural LDS instead of particle variant | LDS-AF · nuScenes | minADE_5: 2.46 -> 2.06 | Improved |
Closest papers by meaning
- Diverse Trajectory Forecasting with Determinantal Point Processes (ICLR 2020 · similarity 0.91)
- Normalizing Trajectory Models (NeurIPS 2026 · similarity 0.87)
- MoFlow: One-Step Flow Matching for Human Trajectory Forecasting via Implicit Maximum Likelihood Estimation based Distillation (CVPR 2025 · similarity 0.86)
- Annealing Flow Generative Models Towards Sampling High-Dimensional and Multi-Modal Distributions (ICML 2025 · similarity 0.86)
- Diffusion Generative Flow Samplers: Improving learning signals through partial trajectory optimization (ICLR 2024 · similarity 0.86)
- Fast and unified path gradient estimators for normalizing flows (ICLR 2024 · similarity 0.85)
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.
