# Latent Flow Matching

3 related articles

Subscribe to updates

bioRxiv Nov 13, 2026
Single-cell perturbation prediction drops the pretraining crutch: GFM wires knowledge graphs into flow matching and trains up to 30x faster
Across several large Perturb-seq datasets, a generative framework called Graph-guided Flow Matching (GFM) combines three biological knowledge graphs, covering Gene Ontology, protein interactions and perturbation similarity, with latent flow matching to predict single-cell transcriptomes after unseen perturbations. It outperforms a linear baseline and the deep-learning models CellFlow, GEARS and scGPT on unseen single-gene perturbations, trains up to 30 times faster, and keeps its advantage as training data shrink; on two-gene perturbations it does better when neither gene was seen in training.
Across several large Perturb-seq datasets, a generative framework called Graph-guided Flow Matching (GFM) combines three biological knowledge graphs, covering Gene Ontology, protein interactions and perturbation similarity, with latent flow matching to predict single-cell transcriptomes after unseen perturbations. It outperforms a linear baseline and the deep-learning models CellFlow, GEARS and scGPT on unseen single-gene perturbations, trains up to 30 times faster, and keeps its advantage as training data shrink; on two-gene perturbations it does better when neither gene was seen in training.
University Medical Center of the Johannes Gutenberg University Mainz German Research Centre for Artificial Intelligence Johannes Gutenberg University Mainz Translationale Onkologie an der Universitätsmedizin der Johannes Gutenberg-Universität Mainz
Huai-Yu Liou（First author）Johnny Kim（Corresponding author）
GFM Latent Flow Matching Graph Attention Network In Silico Perturbation
arXiv Oct 6, 2026
FLaRe reasons in a latent space with flow matching, beating prior latent methods on five probes and reaching 97% of explicit CoT accuracy at a quarter of the latency
The authors define five requirements for latent reasoning (useful, diverse, explainable, refinable, efficient) and present Flow-based Latent Reasoning (FLaRe), a recipe built on flow matching in a learned latent space: train a VAE on symbolic CoTs, corrupt inputs and latent codes, train the flow mostly near pure noise, let the answer reader see both noised codes and the model's own predicted endpoints, and self-train on verified rollouts; probes show FLaRe improves on Coconut, CODI and PCCoT on all five requirements, compares favorably on arithmetic benchmarks, and reaches 97% of explicit CoT accuracy at about a quarter of its latency.
The authors define five requirements for latent reasoning (useful, diverse, explainable, refinable, efficient) and present Flow-based Latent Reasoning (FLaRe), a recipe built on flow matching in a learned latent space: train a VAE on symbolic CoTs, corrupt inputs and latent codes, train the flow mostly near pure noise, let the answer reader see both noised codes and the model's own predicted endpoints, and self-train on verified rollouts; probes show FLaRe improves on Coconut, CODI and PCCoT on all five requirements, compares favorably on arithmetic benchmarks, and reaches 97% of explicit CoT accuracy at about a quarter of its latency.
Gheorghe Asachi Technical University of Iași
Yassine Ouali（First author）
FLaRe (Flow-based Latent Reasoning)Latent Flow Matching Latent Reasoning Flow Matching Variational Autoencoder
arXiv Sep 27, 2026
Unite-Audio jointly trains audio representation learning with latent flow matching, reaching competitive text-to-audio generation with a compact model
The work introduces Unite-Audio, which jointly learns continuous audio representations and latent flow matching for text-to-audio generation; by coupling reconstruction with self-supervised generative prediction it lets the generative objective directly shape the latent space, and it applies Flow-GRPO post-training to improve text-conditioned generation; experiments show competitive text-to-audio performance with a compact latent flow model, and ablation studies confirm the benefit of jointly learning the audio representation and the generative model.
The work introduces Unite-Audio, which jointly learns continuous audio representations and latent flow matching for text-to-audio generation; by coupling reconstruction with self-supervised generative prediction it lets the generative objective directly shape the latent space, and it applies Flow-GRPO post-training to improve text-conditioned generation; experiments show competitive text-to-audio performance with a compact latent flow model, and ablation studies confirm the benefit of jointly learning the audio representation and the generative model.
Ant Group (China)Wuhan University Tsinghua University The University of Tokyo Institute of Science Tokyo
Runwu Shi（First author）
Unite-Audio Latent Flow Matching Flow-GRPO Text-to-Audio Generation

You’re all caught up

Single-cell perturbation prediction drops the pretraining crutch: GFM wires knowledge graphs into flow matching and trains up to 30x faster

FLaRe reasons in a latent space with flow matching, beating prior latent methods on five probes and reaching 97% of explicit CoT accuracy at a quarter of the latency

Unite-Audio jointly trains audio representation learning with latent flow matching, reaching competitive text-to-audio generation with a compact model