Research map / Video / Video generation
ActionPlan: Future-Aware Streaming Motion Synthesis via Frame-Level Action Planning
Eric Nazarenus, Chuqiao Li, Yannan He, Xianghui Xie, Jan Eric Lenssen, Gerard Pons-Moll
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 frame-level text co-generation (D vs C) (ablation) | ActionPlan offline, full data · HumanML3D-272 | FID: 6.952 -> 6.093 | Improved |
| generate action plan first, then motion (E vs D) (ablation) | ActionPlan offline · HumanML3D-272 | FID: 6.093 -> 5.522 | Improved |
| add action plan generation (B vs A, partial data) (ablation) | ActionPlan offline · HumanML3D-272 | FID: 9.674 -> 7.863 | Improved |
| add action plan generation (B vs A, partial data) (ablation) | ActionPlan streaming · HumanML3D-272 | FID: 10.693 -> 8.018 | Improved |
| add action plan generation (E vs C, full data) (ablation) | ActionPlan streaming · HumanML3D-272 | FID: 8.648 -> 5.878 | Improved |
| add action plan generation (E vs C, full data) (ablation) | ActionPlan offline · HumanML3D-272 | FID: 6.952 -> 5.522 | Improved |
| train on full HumanML3D data with masked text loss instead of BABEL-overlap subset (C vs A) (ablation) | ActionPlan offline · HumanML3D-272 | FID: 9.674 -> 6.952 | Improved |
| train on full data instead of partial (E vs B) (ablation) | ActionPlan streaming · HumanML3D-272 | FID: 8.018 -> 5.878 | Improved |
| activate latents with overlapping window K=2 versus fully overlapping parallel (ablation) | ActionPlan progressive sampling · HumanML3D-272 | FID: 5.420 -> 5.522 | No clear effect |
| use 2-step non-overlap instead of 5-steps non-overlap (ablation) | ActionPlan progressive sampling · HumanML3D-272 | FID: 5.721 -> 5.522 | Improved |
| switch from offline to streaming sampling | ActionPlan · HumanML3D-272 | FID: 5.522 -> 5.735 | Hurt |
| replace MotionStreamer with ActionPlan streaming | streaming T2M · HumanML3D-272 | FID: 11.790 -> 5.735 | Improved |
Closest papers by meaning
- MotionStream: Real-Time Video Generation with Interactive Motion Controls (ICLR 2026 · similarity 0.88)
- MotionStreamer: Streaming Motion Generation via Diffusion-based Autoregressive Model in Causal Latent Space (ICCV 2025 · similarity 0.88)
- AdaState: Self-Evolving Anchors for Streaming Video Generation (NeurIPS 2026 · similarity 0.87)
- Enhanced Fine-Grained Motion Diffusion for Text-Driven Human Motion Synthesis (AAAI 2024 · similarity 0.86)
- FASTER: Rethinking Real-Time Flow VLAs (NeurIPS 2026 · similarity 0.86)
- AR-Diffusion: Asynchronous Video Generation with Auto-Regressive Diffusion (CVPR 2025 · 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.
