Research map / Image generation / Autoregressive and latent generation
Diffusion Forcing Planner: History-Annealed Planning with Time-Dependent Guidance for Autonomous Driving
Zehan Zhang, Yaoyi Li, Neng Zhang, Jia Cai
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 |
|---|---|---|---|
| Apply DFP (history-annealed diffusion forcing) to Diffusion Planner | Diffusion Planner · Val14 non-reactive | score: 87.87 -> 90.33 | Improved |
| Apply DFP | Diffusion Planner · Val14 reactive | score: 77.48 -> 79.97 | Improved |
| Apply DFP | Diffusion Planner · Test14-hard reactive | score: 61.25 -> 63.56 (reimpl.); DFP 63.56 vs original 69.22 | Mixed |
| Apply DFP | Diffusion Planner · Test14 reactive | score: 82.93 (DP orig.) vs 81.96 (DFP); reimpl. 79.61 | Mixed |
| Add per-token diffusion forcing alone (chunk length 1) (ablation) | Diffusion Planner · Val14 NR | score: 87.87 -> 85.18 | Hurt |
| Add chunking alone (ablation) | Diffusion Planner · Val14 NR | score: 87.87 -> 87.58 | No clear effect |
| Add chunk-level noise to diffusion forcing (ablation) | DFP · Val14 NR | score: 85.18 -> 88.79 | Improved |
| Add history guidance without chunking (ablation) | DFP · Val14 NR | score: 87.87 -> 86.56 | Hurt |
| Add clean history guidance with chunking (ablation) | DFP · Val14 NR | score: 88.79 -> 89.24 | Improved |
| Add annealed history (ablation) | DFP · Val14 NR | score: 89.24 -> 90.33 | Improved |
| Add annealed history (ablation) | DFP · Val14 reactive | score: 79.16 -> 79.97 | Improved |
| Use DFP in high-speed scenarios | Diffusion Planner · Val14 high-speed | Comfort: 60.61 -> 96.97 | Improved |
Closest papers by meaning
- Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous and Instruction-guided Driving (CVPR 2024 · similarity 0.90)
- Diffusion-Based Planning for Autonomous Driving with Flexible Guidance (ICLR 2025 · similarity 0.90)
- DiRecT: Safe Diffusion-Based Planning via Receding-Horizon Denoising (NeurIPS 2026 · similarity 0.89)
- DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving (CVPR 2025 · similarity 0.88)
- ReflectDrive-2: Reinforcement-Learning-Aligned Self-Editing for Discrete Diffusion Driving (NeurIPS 2026 · similarity 0.87)
- ChainFlow-VLA: Causal Flow Planning with Vision-Language Models (NeurIPS 2026 · similarity 0.87)
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.
