Research map / Probabilistic and scientific modeling / Molecules and proteins
Flow matching generative modeling
Develops flow matching theory and variants, including equivariant and spectral versions.
97 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Molecules and proteins, keyword cluster “flow · trajectory · matching”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Few recent reports None of the tracked benchmarks has a report from 2025–26.
16 papers in 2023–24 and 64 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 3.3% to 6.9%.
No longer reported by papers here: Sintel Clean test (EPE, last in 2024); Sintel Final test (EPE, last in 2024).
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Efficient Coarse-To-Fine PatchMatch for Large Displacement Optical Flow (CVPR 2016 · 193 citations)
- 2017 · FlowNet 2.0: Evolution of Optical Flow Estimation With Deep Networks (CVPR 2017 · 3,371 citations)
- 2018 · Twin Networks: Matching the Future for Sequence Generation (ICLR 2018 · 57 citations)
- 2019 · Iterative Residual Refinement for Joint Optical Flow and Occlusion Estimation (CVPR 2019 · 304 citations)
- 2020 · WaveFlow: A Compact Flow-based Model for Raw Audio (ICML 2020 · 53 citations)
- 2021 · Combining Physics and Machine Learning for Network Flow Estimation (ICLR 2021 · 4 citations)
- 2022 · Global Matching With Overlapping Attention for Optical Flow Estimation (CVPR 2022 · 95 citations)
- 2023 · Tangentially Elongated Gaussian Belief Propagation for Event-Based Incremental Optical Flow Estimation (CVPR 2023 · 7 citations)
- 2024 · Lorentz-Equivariant Geometric Algebra Transformers for High-Energy Physics (NeurIPS 2024 · 14 citations)
- 2025 · CellFlux: Simulating Cellular Morphology Changes via Flow Matching (ICML 2025 · 2 citations)
- 2026 · scDFM: Distributional Flow Matching Model for Robust Single-Cell Perturbation Prediction (ICLR 2026 · 1 citation)
Earliest papers here
- Efficient Coarse-To-Fine PatchMatch for Large Displacement Optical Flow (CVPR 2016 · 193 citations)
- Full Flow: Optical Flow Estimation By Global Optimization Over Regular Grids (CVPR 2016 · 165 citations)
- FlowNet 2.0: Evolution of Optical Flow Estimation With Deep Networks (CVPR 2017 · 3,371 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
Sintel Clean test · EPE (lower is better)
4 papers, 2017–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2017 | 6.64 | Optical Flow Estimation Using a Spatial Pyramid Network (CVPR 2017 · 1,433 citations) |
| 2019 | 3.84 | Iterative Residual Refinement for Joint Optical Flow and Occlusion Estimation (CVPR 2019 · 304 citations) |
| 2024 | 1.17 | Context-Aware Iteration Policy Network for Efficient Optical Flow Estimation (AAAI 2024 · 2 citations) |
Sintel Final test · EPE (lower is better)
4 papers, 2017–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2017 | 8.36 | Optical Flow Estimation Using a Spatial Pyramid Network (CVPR 2017 · 1,433 citations) |
| 2019 | 4.58 | Iterative Residual Refinement for Joint Optical Flow and Occlusion Estimation (CVPR 2019 · 304 citations) |
| 2024 | 2.21 | Context-Aware Iteration Policy Network for Efficient Optical Flow Estimation (AAAI 2024 · 2 citations) |
What has been tried
45 papers here report 466 measured changes (a component added, removed or swapped, compared with a baseline): helped 69%, no clear effect 5%, hurt 24%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use Riemannian flow matching probability paths instead of Euclidean paths | ttbar + 0j test set · NLL | -30.11 ± 0.98 -> -32.65 ± 0.01 | Helped | Lorentz-Equivariant Geometric Algebra Transformers for High-Energy Physics (NeurIPS 2024 · 14 citations) |
| Add the proposed lattice base distribution and unconstrained angle transformation | MP-20 · Match Rate (%) | 43.21 ± 0.47 -> 56.82 ± 0.42 | Helped | FlowMM: Generating Materials with Riemannian Flow Matching (ICML 2024 · 9 citations) |
| Set Gaussian radius r to 3 rather than 1, 2, or 4 | FCDN and VBOF (Fuji2) · EPE | FCDN: 1.27, 1.17, 1.08, 1.21; VBOF(Fuji2): 14.22, 13.99, 13.94, 14.15 | Helped | CEDFlow: Latent Contour Enhancement for Dark Optical Flow Estimation (AAAI 2024 · 6 citations) |
| replace the deep decoder with an inner-product decoder | Beijing · RMSE | 39.437 -> 45.425 | Hurt | Explainable Origin-Destination Crowd Flow Interpolation via Variational Multi-Modal Recurrent Graph Auto-Encoder (AAAI 2024 · 5 citations) |
| Add perturbation condition | BBBC021 · FID_o | 45.0 -> 18.7 | Helped | CellFlux: Simulating Cellular Morphology Changes via Flow Matching (ICML 2025 · 2 citations) |
| Replace the positional interpolant with linear SDE with γ | MP-20 · Novelty rate (%) | 73.31; 66.45; 55.21; 76.25 | Mixed | Open Materials Generation with Stochastic Interpolants (ICML 2025 · 2 citations) |
| replace max(0,x) resource preference loss with absolute-distance L1 loss | Sintel-Clean, Sintel-Final, KITTI · EPE | 1.56 vs 1.73; 2.74 vs 2.97; 5.52 vs 5.84 | Hurt | Context-Aware Iteration Policy Network for Efficient Optical Flow Estimation (AAAI 2024 · 2 citations) |
| add Gaussian mixture paths | Alanine dipeptide (exact) · MinMax Energy | -14.67 -> -15.54; -15.95 | Helped | Doob's Lagrangian: A Sample-Efficient Variational Approach to Transition Path Sampling (NeurIPS 2024 · 2 citations) |
| apply DPO to the VP path | CDR-H3 · Cα-w RMSD | 5.01 -> 4.42 | Helped | 3D Structure Prediction of Atomic Systems with Flow-based Direct Preference Optimization (NeurIPS 2024 · 1 citation) |
| Add the improved architecture to retrained GMSF | F3D_o-all · EPE_3D | 0.039 -> 0.037 | Helped | DiffSF: Diffusion Models for Scene Flow Estimation (NeurIPS 2024 · 1 citation) |
| Add trajectory guidance | OC22 IS2RS validation · ADwT | 58.54 -> 59.39 | Helped | Bridging Geometric States via Geometric Diffusion Bridge (NeurIPS 2024 · 1 citation) |
| Remove equivariance enforcement | Molecular generation tasks · Atom and molecule stability | Hurt | Controlled Generation with Equivariant Variational Flow Matching (ICML 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 30 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- perov-5 · 2 of 30 papers
- MP-20 · 2 of 30 papers
- MPTS-52 · 2 of 30 papers
- Alex-MP-20 · 2 of 30 papers
Baselines
- FlowMM · 3 of 30 papers
- MIOFlow · 2 of 30 papers
- DiffCSP · 2 of 30 papers
- CDVAE · 2 of 30 papers
- OMatG · 2 of 30 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 80 papers; from OpenAlex and the papers' affiliations).
- Pohang University of Science and Technology · 3 papers
- Technical University Munich · 3 papers
- Meta · 3 papers
- Massachusetts Institute of Technology · 3 papers
- Tsinghua University · 3 papers
- Korea Advanced Institute of Science and Technology · 2 papers
- Technische Universität München · 2 papers
- University of Amsterdam · 2 papers
- New York University · 2 papers
- University of British Columbia · 2 papers
Use it
Download BibTeX of 37 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Multimodal trajectory prediction (95 papers)
- Spatiotemporal forecasting and imputation (30 papers)
Elsewhere in the topic
- Generative protein structure design (142 papers)
- Prediction-powered and conformal inference (129 papers)
- Statistical analysis of ML practice (122 papers)
- Neuro-symbolic probabilistic layers (120 papers)
- Structure-based 3D molecule generation (107 papers)
- Sparse graphical and tensor models (97 papers)
- Machine learning for physical simulation (96 papers)
- Probabilistic embeddings and relational inference (93 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).