Research map / Image generation / Fast diffusion sampling
Test-time Sparsity for Extreme Fast Action Diffusion
Kangye Ji, Yuan Meng, Jianbo Zhou, Ye Li, Chen Tang, Zhi Wang
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 rollout and timestep reuse to forward-only reuse (ablation) | test-time sparsity (omnidirectional reusing) · Robomimic PH Transport (93% sparsity) | success rate (%): 4 -> 92 | Improved |
| add forward and rollout reuse to timestep-only reuse (ablation) | test-time sparsity (omnidirectional reusing) · Robomimic PH Tool (93% sparsity) | success rate (%): 0 -> 56 | Improved |
| add forward and timestep reuse to rollout-only reuse (ablation) | test-time sparsity (omnidirectional reusing) · Robomimic PH Can (93% sparsity) | success rate (%): 10 -> 94 | Improved |
| add rollout and timestep reuse to forward-only reuse (ablation) | test-time sparsity (omnidirectional reusing) · Robomimic MH Transport (93% sparsity) | success rate (%): 0 -> 62 | Improved |
| add forward and timestep reuse to rollout-only reuse (ablation) | test-time sparsity (omnidirectional reusing) · Robomimic MH Transport (93% sparsity) | success rate (%): 0 -> 62 | Improved |
| add forward and rollout reuse to timestep-only reuse (ablation) | test-time sparsity (omnidirectional reusing) · Robomimic MH Square (93% sparsity) | success rate (%): 66 -> 82 | Improved |
| increase sparsity from 93% to 95% | test-time sparsity (omnidirectional reusing) · Robomimic PH (average of 5 tasks) | speedup: 4.86x -> 5.18x | Improved |
| increase sparsity from 93% to 95% | test-time sparsity (omnidirectional reusing) · Robomimic MH (average of 4 tasks) | average success rate (%): 85 -> 80 | Hurt |
| increase sparsity from 93% to 95% | test-time sparsity (omnidirectional reusing) · Kitchen | average success rate (%): 100 -> 99 | No clear effect |
| apply to RDT-1B VLA at 90% sparsity | test-time sparsity (omnidirectional reusing) · ManiSkill Insert | success rate (%): 4 -> 16 | Improved |
| apply to DiT-XL/2 image generation at 80% sparsity vs TaylorSeer | test-time sparsity (omnidirectional reusing) · ImageNet | FID: 2.65 -> 2.23 | Improved |
Closest papers by meaning
- Faster Diffusion: Rethinking the Role of the Encoder for Diffusion Model Inference (NeurIPS 2024 · similarity 0.90)
- SparseDiT: Token Sparsification for Efficient Diffusion Transformer (NeurIPS 2025 · similarity 0.89)
- Test-Time Scaling of Diffusion Models via Noise Trajectory Search (NeurIPS 2025 · similarity 0.88)
- FASTER: Value-Guided Sampling for Fast RL (NeurIPS 2026 · similarity 0.88)
- ALTER: All-in-One Layer Pruning and Temporal Expert Routing for Efficient Diffusion Generation (NeurIPS 2025 · similarity 0.88)
- Encoder-Decoder Diffusion Language Models for Efficient Training and Inference (NeurIPS 2025 · 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.
