Research map / Image generation / Fast diffusion sampling
TFTF: Training-Free Targeted Flow for Conditional Sampling
Qianqian Qu, Jun S. Liu
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 intermediate SMC resampling to importance sampling (ablation) | TFTF · CIFAR-10 automobile class, K=16 | ESS: 1.37/16 -> 16.00/16 | Improved |
| increase stochasticity scale alpha(t) from 1/t to larger values (ablation) | TFTF · MNIST class-conditional | intra-batch diversity and ACC_E | Mixed |
| increase particle count K (ablation) | TFTF · MNIST class-conditional | ACC_E | Improved |
| use TFTF instead of Twisted Diffusion Sampler | MNIST class-conditional · MNIST | ACC_E and intra-batch diversity | Improved |
| use TFTF instead of FlowChef | flow matching conditional sampling · MNIST | ACC_E: 23.57% -> 97.12% | Improved |
| use TFTF instead of FlowChef | flow matching conditional sampling · CIFAR-10 | ACC_E: 50.13% -> 92.82% | Improved |
| use TFTF instead of FMPS | flow matching conditional sampling · CelebA-HQ text-to-image | ACC_E: 68.69% -> 80.99% | Improved |
| use TFTF instead of FMPS | flow matching conditional sampling · CelebA-HQ text-to-image | CLIP_L: 0.307 -> 0.295 | Hurt |
| resampling introduces inter-particle correlation | TFTF · batch samples | visual diversity within batch | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FlowChef — FlowChef: Steering of Rectified Flow Models for Controlled Generations (ICCV 2025) · better on 6 of 7 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
- Faster Inference of Flow-Based Generative Models via Improved Data-Noise Coupling (ICLR 2025 · similarity 0.90)
- CAR-Flow: Condition-Aware Reparameterization Aligns Source and Target for Better Flow Matching (NeurIPS 2025 · similarity 0.89)
- Discrete Flow Matching (NeurIPS 2024 · similarity 0.89)
- Coreset-Induced Conditional Velocity Flow Matching (NeurIPS 2026 · similarity 0.89)
- Safe Few-Step Generation via Velocity Editing (NeurIPS 2026 · similarity 0.89)
- PRECISE: SDE-Consistent Stochastic Sampling for RL Post-Training of Flow-Matching Models (NeurIPS 2026 · similarity 0.88)
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.
