{"work":{"id":"6edb71c4-5d64-40af-a394-9757ea051a36","openalex_id":null,"doi":null,"arxiv_id":"2210.02747","raw_key":null,"title":"Flow Matching for Generative Modeling","authors":null,"authors_text":"Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le","year":2022,"venue":"cs.LG","abstract":"We introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs based on regressing vector fields of fixed conditional probability paths. Flow Matching is compatible with a general family of Gaussian probability paths for transforming between noise and data samples -- which subsumes existing diffusion paths as specific instances. Interestingly, we find that employing FM with diffusion paths results in a more robust and stable alternative for training diffusion models. Furthermore, Flow Matching opens the door to training CNFs with other, non-diffusion probability paths. An instance of particular interest is using Optimal Transport (OT) displacement interpolation to define the conditional probability paths. These paths are more efficient than diffusion paths, provide faster training and sampling, and result in better generalization. Training CNFs using Flow Matching on ImageNet leads to consistently better performance than alternative diffusion-based methods in terms of both likelihood and sample quality, and allows fast and reliable sample generation using off-the-shelf numerical ODE solvers.","external_url":"https://arxiv.org/abs/2210.02747","cited_by_count":null,"metadata_source":"pith","metadata_fetched_at":"2026-05-25T06:45:26.155146+00:00","pith_arxiv_id":"2210.02747","created_at":"2026-05-09T05:55:32.250443+00:00","updated_at":"2026-05-25T06:45:26.155146+00:00","title_quality_ok":true,"display_title":"Flow Matching for Generative Modeling","render_title":"Flow Matching for Generative Modeling"},"hub":{"state":{"work_id":"6edb71c4-5d64-40af-a394-9757ea051a36","tier":"super_hub","tier_reason":"100+ Pith inbound or 10,000+ external citations","pith_inbound_count":480,"external_cited_by_count":null,"distinct_field_count":39,"first_pith_cited_at":"2022-09-30T16:30:31+00:00","last_pith_cited_at":"2026-05-22T17:59:43+00:00","author_build_status":"needed","summary_status":"needed","contexts_status":"needed","graph_status":"needed","ask_index_status":"needed","reader_status":"not_needed","recognition_status":"not_needed","updated_at":"2026-06-04T19:47:56.220834+00:00","tier_text":"super_hub"},"tier":"super_hub","role_counts":[{"context_role":"background","n":77},{"context_role":"method","n":72},{"context_role":"baseline","n":5}],"polarity_counts":[{"context_polarity":"use_method","n":72},{"context_polarity":"background","n":69},{"context_polarity":"unclear","n":6},{"context_polarity":"baseline","n":5},{"context_polarity":"support","n":2}],"runs":{"ask_index":{"job_type":"ask_index","status":"succeeded","result":{"title":"Flow Matching for Generative Modeling","claims":[{"claim_text":"We introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs based on regressing vector fields of fixed conditional probability paths. Flow Matching is compatible with a general family of Gaussian probability paths for transforming between noise and data samples -- which subsumes existing diffusion paths as specific instances. Interestingly, we find that employing FM with diffusion paths results in a more ","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Flow Matching for Generative Modeling because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T19:03:30.625129+00:00"},"author_expand":{"job_type":"author_expand","status":"succeeded","result":{"authors_linked":[{"id":"3c280171-924e-4f5c-a774-f4e7e5c21576","orcid":null,"display_name":"Yaron Lipman"},{"id":"6b138a37-2f29-461d-af74-5f22a4748b17","orcid":null,"display_name":"Ricky TQ Chen"},{"id":"e2f5d7ac-f5d5-4b65-993f-3521139b7ed4","orcid":null,"display_name":"Heli Ben-Hamu"},{"id":"91cdb42b-abb5-478b-af3d-5806e8e824df","orcid":null,"display_name":"Maximilian Nickel"},{"id":"a12e7de5-1739-459a-9e89-126f077beeb5","orcid":null,"display_name":"and Matt Le"}]},"error":null,"updated_at":"2026-05-13T19:03:30.623115+00:00"},"context_extract":{"job_type":"context_extract","status":"succeeded","result":{"enqueued_papers":25},"error":null,"updated_at":"2026-05-13T19:03:29.961108+00:00"},"graph_features":{"job_type":"graph_features","status":"succeeded","result":{"co_cited":[{"title":"Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow","work_id":"a1989e1b-d66d-4533-be3a-fb9c5fd62290","shared_citers":93},{"title":"Score-Based Generative Modeling through Stochastic Differential Equations","work_id":"d9110e53-a5d4-4794-a4c5-a575e91c31ad","shared_citers":65},{"title":"Wan: Open and Advanced Large-Scale Video Generative Models","work_id":"ad3ebc3b-4224-46c9-b61d-bcf135da0a7c","shared_citers":57},{"title":"Denoising Diffusion Implicit Models","work_id":"8fa2128b-d18c-405c-ac92-0e669cf89ac0","shared_citers":48},{"title":"Classifier-Free Diffusion Guidance","work_id":"acf2c588-c088-4a6c-938e-150ad7c666d7","shared_citers":45},{"title":"$\\pi_0$: A Vision-Language-Action Flow Model for General Robot Control","work_id":"f790abdc-a796-482f-a40d-f8ee035ecfc2","shared_citers":34},{"title":"OpenVLA: An Open-Source Vision-Language-Action Model","work_id":"3e7e65c5-5aed-4fe9-8414-2092bcb31cc7","shared_citers":32},{"title":"CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer","work_id":"f38fc088-12aa-4bf4-9ecd-08d3e797ccb7","shared_citers":29},{"title":"Auto-Encoding Variational Bayes","work_id":"97d95295-30e1-42b4-bbf6-85f0fa4edb44","shared_citers":28},{"title":"Decoupled Weight Decay Regularization","work_id":"07ef7360-d385-4033-83f7-8384a6325204","shared_citers":28},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":28},{"title":"HunyuanVideo: A Systematic Framework For Large Video Generative Models","work_id":"881efa7e-7e73-4c66-9cc3-2803e551061c","shared_citers":27},{"title":"Building Normalizing Flows with Stochastic Interpolants","work_id":"83654cbd-832b-4269-bec9-26faa87ececd","shared_citers":26},{"title":"SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis","work_id":"8034c587-fba6-4941-87ba-c98f2ac962cb","shared_citers":26},{"title":"Stochastic Interpolants: A Unifying Framework for Flows and Diffusions","work_id":"c2c7dd8f-fbfb-4591-89ec-9a3a0e6744bd","shared_citers":26},{"title":"Improving and generalizing flow-based generative models with minibatch optimal transport","work_id":"7b689852-603d-46be-aaa5-c38325ba2182","shared_citers":24},{"title":"$\\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization","work_id":"d1ad7304-d09a-49bc-809e-846439f6aff9","shared_citers":23},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":23},{"title":"Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets","work_id":"4f68eada-27e3-437a-a2fe-6e4ca524d0d3","shared_citers":23},{"title":"Training Diffusion Models with Reinforcement Learning","work_id":"67684dda-3930-452a-b91a-36cbb8e2e219","shared_citers":23},{"title":"Flow-GRPO: Training Flow Matching Models via Online RL","work_id":"bf1e8e81-ff31-401a-a5dc-d9c49df168ab","shared_citers":22},{"title":"Mean Flows for One-step Generative Modeling","work_id":"07a52ad5-0f82-4095-9a66-559b09fea1ae","shared_citers":21},{"title":"Qwen3-VL Technical Report","work_id":"1fe243aa-e3c0-4da6-b391-4cbcfc88d5c0","shared_citers":21},{"title":"DanceGRPO: Unleashing GRPO on Visual Generation","work_id":"7404dd36-8f9c-478f-b089-ef9f8189c711","shared_citers":20}],"time_series":[{"n":1,"year":2022},{"n":6,"year":2024},{"n":12,"year":2025},{"n":237,"year":2026}]},"error":null,"updated_at":"2026-05-13T19:03:30.108453+00:00"},"identity_refresh":{"job_type":"identity_refresh","status":"succeeded","result":{"fixed":1,"items":[{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","resolver":"local_arxiv","confidence":0.98,"old_work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e"}],"errors":[],"attempted":1},"error":null,"updated_at":"2026-05-13T19:03:29.314838+00:00"},"role_polarity":{"job_type":"role_polarity","status":"succeeded","result":{"title":"Flow Matching for Generative Modeling","claims":[{"claim_text":"We introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs based on regressing vector fields of fixed conditional probability paths. Flow Matching is compatible with a general family of Gaussian probability paths for transforming between noise and data samples -- which subsumes existing diffusion paths as specific instances. Interestingly, we find that employing FM with diffusion paths results in a more ","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Flow Matching for Generative Modeling because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T19:03:29.967443+00:00"},"summary_claims":{"job_type":"summary_claims","status":"succeeded","result":{"title":"Flow Matching for Generative Modeling","claims":[{"claim_text":"We introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs based on regressing vector fields of fixed conditional probability paths. Flow Matching is compatible with a general family of Gaussian probability paths for transforming between noise and data samples -- which subsumes existing diffusion paths as specific instances. Interestingly, we find that employing FM with diffusion paths results in a more ","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Flow Matching for Generative Modeling because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T19:03:29.965073+00:00"}},"summary":{"title":"Flow Matching for Generative Modeling","claims":[{"claim_text":"We introduce a new paradigm for generative modeling built on Continuous Normalizing Flows (CNFs), allowing us to train CNFs at unprecedented scale. Specifically, we present the notion of Flow Matching (FM), a simulation-free approach for training CNFs based on regressing vector fields of fixed conditional probability paths. Flow Matching is compatible with a general family of Gaussian probability paths for transforming between noise and data samples -- which subsumes existing diffusion paths as specific instances. Interestingly, we find that employing FM with diffusion paths results in a more ","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Flow Matching for Generative Modeling because it crossed a citation-hub threshold.","role_counts":[]},"graph":{"co_cited":[{"title":"Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow","work_id":"a1989e1b-d66d-4533-be3a-fb9c5fd62290","shared_citers":93},{"title":"Score-Based Generative Modeling through Stochastic Differential Equations","work_id":"d9110e53-a5d4-4794-a4c5-a575e91c31ad","shared_citers":65},{"title":"Wan: Open and Advanced Large-Scale Video Generative Models","work_id":"ad3ebc3b-4224-46c9-b61d-bcf135da0a7c","shared_citers":57},{"title":"Denoising Diffusion Implicit Models","work_id":"8fa2128b-d18c-405c-ac92-0e669cf89ac0","shared_citers":48},{"title":"Classifier-Free Diffusion Guidance","work_id":"acf2c588-c088-4a6c-938e-150ad7c666d7","shared_citers":45},{"title":"$\\pi_0$: A Vision-Language-Action Flow Model for General Robot Control","work_id":"f790abdc-a796-482f-a40d-f8ee035ecfc2","shared_citers":34},{"title":"OpenVLA: An Open-Source Vision-Language-Action Model","work_id":"3e7e65c5-5aed-4fe9-8414-2092bcb31cc7","shared_citers":32},{"title":"CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer","work_id":"f38fc088-12aa-4bf4-9ecd-08d3e797ccb7","shared_citers":29},{"title":"Auto-Encoding Variational Bayes","work_id":"97d95295-30e1-42b4-bbf6-85f0fa4edb44","shared_citers":28},{"title":"Decoupled Weight Decay Regularization","work_id":"07ef7360-d385-4033-83f7-8384a6325204","shared_citers":28},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":28},{"title":"HunyuanVideo: A Systematic Framework For Large Video Generative Models","work_id":"881efa7e-7e73-4c66-9cc3-2803e551061c","shared_citers":27},{"title":"Building Normalizing Flows with Stochastic Interpolants","work_id":"83654cbd-832b-4269-bec9-26faa87ececd","shared_citers":26},{"title":"SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis","work_id":"8034c587-fba6-4941-87ba-c98f2ac962cb","shared_citers":26},{"title":"Stochastic Interpolants: A Unifying Framework for Flows and Diffusions","work_id":"c2c7dd8f-fbfb-4591-89ec-9a3a0e6744bd","shared_citers":26},{"title":"Improving and generalizing flow-based generative models with minibatch optimal transport","work_id":"7b689852-603d-46be-aaa5-c38325ba2182","shared_citers":24},{"title":"$\\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization","work_id":"d1ad7304-d09a-49bc-809e-846439f6aff9","shared_citers":23},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":23},{"title":"Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets","work_id":"4f68eada-27e3-437a-a2fe-6e4ca524d0d3","shared_citers":23},{"title":"Training Diffusion Models with Reinforcement Learning","work_id":"67684dda-3930-452a-b91a-36cbb8e2e219","shared_citers":23},{"title":"Flow-GRPO: Training Flow Matching Models via Online RL","work_id":"bf1e8e81-ff31-401a-a5dc-d9c49df168ab","shared_citers":22},{"title":"Mean Flows for One-step Generative Modeling","work_id":"07a52ad5-0f82-4095-9a66-559b09fea1ae","shared_citers":21},{"title":"Qwen3-VL Technical Report","work_id":"1fe243aa-e3c0-4da6-b391-4cbcfc88d5c0","shared_citers":21},{"title":"DanceGRPO: Unleashing GRPO on Visual Generation","work_id":"7404dd36-8f9c-478f-b089-ef9f8189c711","shared_citers":20}],"time_series":[{"n":1,"year":2022},{"n":6,"year":2024},{"n":12,"year":2025},{"n":237,"year":2026}]},"authors":[{"id":"a12e7de5-1739-459a-9e89-126f077beeb5","orcid":null,"display_name":"and Matt Le","source":"manual","import_confidence":0.72},{"id":"e2f5d7ac-f5d5-4b65-993f-3521139b7ed4","orcid":null,"display_name":"Heli Ben-Hamu","source":"manual","import_confidence":0.72},{"id":"91cdb42b-abb5-478b-af3d-5806e8e824df","orcid":null,"display_name":"Maximilian Nickel","source":"manual","import_confidence":0.72},{"id":"6b138a37-2f29-461d-af74-5f22a4748b17","orcid":null,"display_name":"Ricky TQ Chen","source":"manual","import_confidence":0.72},{"id":"3c280171-924e-4f5c-a774-f4e7e5c21576","orcid":null,"display_name":"Yaron Lipman","source":"manual","import_confidence":0.72}]}}