Research map / Image generation / Fast diffusion sampling
FlowCast: Trajectory Forecasting for Scalable Zero-Cost Speculative Flow Matching
Divya Jyoti Bajpai, Shubham Agarwal, Apoorv Saxena, Kuldeep Kulkarni, Subrata Mitra, Manjesh Kumar Hanawal
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 |
|---|---|---|---|
| use FlowCast speculative generation at 50 steps instead of full sampling | BAGEL · GenEval | Overall / speedup: 0.78 -> 0.78 at 2.5x speedup | Improved |
| use FlowCast speculative generation at 50 steps instead of full sampling | FLUX · GenEval | Overall / speedup: 0.65 -> 0.65 at 2.4x speedup | Improved |
| FlowCast versus TeaCache | BAGEL · GenEval | Overall / speedup: 0.75 -> 0.78 overall; 1.8x -> 2.5x | Improved |
| apply FlowCast with very few steps (5) | BAGEL/FLUX · GenEval | Overall: 0.57 -> 0.57 (BAGEL); 0.44 -> 0.43 (FLUX) | No clear effect |
| add FlowCast on top of PerFlow | PerFlow · GenEval | Overall / speedup: 0.57 -> 0.57; 1.0x -> 1.9x | Improved |
| add FlowCast to TeaCache (ablation) | Step-1X-Edit · GEdit | G_O: 6.91 -> 6.90; speedup 1.8x -> 2.4x | No clear effect |
| FlowCast versus reducing steps statically | Step-1X-Edit · GEdit | G_O: 6.54 -> 6.50 at 10 steps (3.0x -> 3.6x) | Improved |
| increase MSE threshold epsilon (ablation) | FlowCast · Multi-turn editing (BAGEL) | speedup vs PSNR | Mixed |
| speculative drafting with parallel verification | FlowCast · FLUX latency | Memory (GB): 48 -> 79 GB at 50 steps | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- PerFlow — PeRFlow: Piecewise Rectified Flow as Universal Plug-and-Play Accelerator (NeurIPS 2024) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- The Velocity Deficit: Initial Energy Injection for Flow Matching (ICML 2026 · similarity 0.90)
- Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation (NeurIPS 2026 · similarity 0.88)
- FlowTurbo: Towards Real-time Flow-Based Image Generation with Velocity Refiner (NeurIPS 2024 · similarity 0.88)
- Accelerating Rectified Flow Models via Trajectory-Aware Caching (NeurIPS 2026 · similarity 0.87)
- MESSENGER: Memory-Enhanced Sequential Scene Flow Estimation via Autoregressive Next-Frame Forecasting (NeurIPS 2026 · similarity 0.87)
- Faster Inference of Flow-Based Generative Models via Improved Data-Noise Coupling (ICLR 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.
