REVIEW 13 cited by
Multisample Flow Matching: Straightening Flows with Minibatch Couplings
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Simulation-free methods for training continuous-time generative models construct probability paths that go between noise distributions and individual data samples. Recent works, such as Flow Matching, derived paths that are optimal for each data sample. However, these algorithms rely on independent data and noise samples, and do not exploit underlying structure in the data distribution for constructing probability paths. We propose Multisample Flow Matching, a more general framework that uses non-trivial couplings between data and noise samples while satisfying the correct marginal constraints. At very small overhead costs, this generalization allows us to (i) reduce gradient variance during training, (ii) obtain straighter flows for the learned vector field, which allows us to generate high-quality samples using fewer function evaluations, and (iii) obtain transport maps with lower cost in high dimensions, which has applications beyond generative modeling. Importantly, we do so in a completely simulation-free manner with a simple minimization objective. We show that our proposed methods improve sample consistency on downsampled ImageNet data sets, and lead to better low-cost sample generation.
Forward citations
Cited by 13 Pith papers
-
Generalization and Memorization in Rectified Flow
Rectified Flow models peak in membership-inference vulnerability at the flow midpoint under uniform training; U-shaped timestep sampling suppresses memorization without harming FID.
-
From geometry to dynamics: Learning overdamped Langevin dynamics from sparse observations with geometric constraints
Geometry-guided path augmentation recovers drift functions of overdamped Langevin systems from sparse observations by steering diffusion bridges toward invariant-density geodesics.
-
GraspMeanFlow: SE(3)-Equivariant MeanFlow for Few-Step 6-DoF Grasp Generation
An SE(3)-equivariant average-velocity flow generates 6-DoF grasps in one or a few function evaluations, matching iterative flow baselines on ACRONYM.
-
Computational and Statistical Guarantees of the \textit{c}-Rectified flow
Iterative c-rectified flow converges to optimal transport under regularity assumptions, and score-based plug-in estimation yields near-optimal transport-map rates.
-
SE(3)-MeanFlow: Few-Step Protein Backbone Generation on Lie Groups
SE(3)-MeanFlow trains a protein backbone generator to predict average Lie-group velocities, reaching comparable designability to flow matching at 20–100 steps and leading at 10 steps after rectification.
-
Explorative Modeling: Unlocking a Third Pretraining Axis and End-to-End Generation
Training on the best of K generated candidates improves image, video, and language generative models, with the reported gains growing with scale and enabling single-pass end-to-end generation.
-
Heavy-Tailed Flow Matching via Random Clocks
HTFM trains flow-matching models with heavy-tailed source distributions obtained by mixing Gaussian sources over a random clock path, improving tail-statistic recovery and low-NFE sampling.
-
Transport Discrepancy as a Reliability Signal for Vision-Language-Action Models
A transport-cost-based gate that modulates observation features in flow-matching VLA policies improves reported success rates on long-horizon and distribution-shifted robot tasks.
-
Energy-Based Flow Matching for Generating 3D Molecular Structure
IDFlow trains a flow matching network to refine its own predicted 3D molecular structure, improving docking and protein backbone generation over HarmonicFlow and FrameFlow baselines.
-
NeuTSFlow: Modeling Continuous Functions Behind Time Series Forecasting
A flow-matching model with a neural-operator velocity field is proposed to forecast time series by transporting distributions over continuous functions, reporting top average rank on eight benchmarks.
-
Bridging the Last Mile of Prediction: Enhancing Time Series Forecasting with Conditional Guided Flow Matching
CGFM uses an auxiliary model's predictions as the source for flow matching to learn forecast residuals and improve time series forecasts.
-
MoDA: Multi-modal Diffusion Architecture for Talking Head Generation
MoDA uses flow matching in a compact face-motion space with a progressively fused multi-modal transformer to generate expressive, lip-synced talking-head videos from a single image and audio.
-
Optimal Self-Distillation for Rectified Flow via Linear Probing
For linear rectified flow with ridge regression on fixed interpolants, optimally mixed self-distillation strictly improves velocity risk whenever the teacher is off the ridge stationary point, with a closed-form mixin...
Discussion (0). Sign in to comment.