AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Carré du champ flow matching: better quality-generalisation tradeoff in generative models

Jacob Bamberger, Iolo Jones, Dennis Duncan, Michael M. Bronstein, Pierre Vandergheynst, Adam Gosztolai

ICLR 2026 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Flow matching and flow sampling · 8 citations (OpenAlex)

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 changedWhereMeasuredEffect
Replace isotropic FM noise with CDC geometric noiseFlow matching (MLP) · CITE-seq single-cell interpolationEarth mover distance: I-FM 48.276 -> 46.657; OT-FM 45.393 -> 44.410Improved
Replace isotropic FM noise with CDC geometric noiseFlow matching (MLP) · Multiomics single-cell interpolationEarth mover distance: I-FM 57.262 -> 54.419; OT-FM 54.814 -> 52.043Improved
Use CDC-FM in latent space on 1000 CelebA-HQ imagesLatent flow matching · CelebA-HQ subsetFID: 13.10 -> 10.85 at epoch 5000Improved
Use CDC-FM in latent space on 1000 CelebA-HQ imagesLatent flow matching · CelebA-HQ subsetNLL: 6.80 -> 7.18 at epoch 1000; 6.68 -> 6.48 at epoch 5000Mixed
Increase sigma_min of FM as naive regularisation (ablation)Flow matching · Fruit fly motion capturequality-generalisation frontierNo clear effect
Add CDC regularisation (gamma>0, strongest near 0.3) (ablation)Flow matching (transformer) · Fruit fly motion capturequality, generalisation, memorisationImproved
Increase rank of G (d_cdc) (ablation)CDC-FM · Fruit fly motion capturegeneralisation / qualityMixed
Add CDC-FM on CIFAR-10 small subsets (<10k) late in trainingFlow matching (UNet) · CIFAR-10 subsetsmemorisation, FIDImproved
Add CDC-FM with larger CIFAR-10 training setsFlow matching (UNet) · CIFAR-10 large subsetsFID, generalisationNo clear effect
Add CDC-FM on high-dimensional tori at fixed data sizeFlow matching · Torus T^d syntheticsample qualityHurt
Add CDC-FM on high-dimensional toriFlow matching · Torus T^d syntheticmemorisationImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy