{"work":{"id":"d9110e53-a5d4-4794-a4c5-a575e91c31ad","openalex_id":null,"doi":null,"arxiv_id":"2011.13456","raw_key":null,"title":"Score-Based Generative Modeling through Stochastic Differential Equations","authors":null,"authors_text":"Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole","year":2020,"venue":"cs.LG","abstract":"Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model.","external_url":"https://arxiv.org/abs/2011.13456","cited_by_count":null,"metadata_source":"pith","metadata_fetched_at":"2026-06-29T08:33:15.635565+00:00","pith_arxiv_id":"2011.13456","created_at":"2026-05-08T19:34:04.690178+00:00","updated_at":"2026-06-29T08:33:15.635565+00:00","title_quality_ok":true,"display_title":"Score-Based Generative Modeling through Stochastic Differential Equations","render_title":"Score-Based Generative Modeling through Stochastic Differential Equations"},"hub":{"state":{"work_id":"d9110e53-a5d4-4794-a4c5-a575e91c31ad","tier":"super_hub","tier_reason":"100+ Pith inbound or 10,000+ external citations","pith_inbound_count":387,"external_cited_by_count":null,"distinct_field_count":45,"first_pith_cited_at":"2020-10-06T06:15:51+00:00","last_pith_cited_at":"2026-06-25T14:31:17+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":"2026-06-29T08:38:30.928808+00:00","tier_text":"super_hub"},"tier":"super_hub","role_counts":[{"context_role":"background","n":71},{"context_role":"method","n":17},{"context_role":"baseline","n":3},{"context_role":"other","n":1}],"polarity_counts":[{"context_polarity":"background","n":70},{"context_polarity":"use_method","n":16},{"context_polarity":"baseline","n":3},{"context_polarity":"unclear","n":2},{"context_polarity":"support","n":1}],"runs":{"ask_index":{"job_type":"ask_index","status":"succeeded","result":{"title":"Score-Based Generative Modeling through Stochastic Differential Equations","claims":[{"claim_text":"Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Score-Based Generative Modeling through Stochastic Differential Equations because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T21:03:36.576379+00:00"},"author_expand":{"job_type":"author_expand","status":"succeeded","result":{"authors_linked":[{"id":"c36345af-089e-427a-b034-794e452cabe0","orcid":null,"display_name":"Yang Song"},{"id":"4e16e341-9f1f-4f1e-8ae6-dcc81290225a","orcid":null,"display_name":"Jascha Sohl-Dickstein"},{"id":"6b9fb90b-3116-44d5-9610-1a8263f12dbc","orcid":null,"display_name":"Diederik P Kingma"},{"id":"9c1dd117-9177-4557-8ed0-32ef0ec70e72","orcid":null,"display_name":"Ab- hishek Kumar"},{"id":"e00f90b7-aa2a-4902-8d30-19cdf0a496f9","orcid":null,"display_name":"Stefano Ermon"},{"id":"4c622740-2d53-4b78-b89b-815c032a1d6c","orcid":null,"display_name":"and Ben Poole"}]},"error":null,"updated_at":"2026-05-13T21:03:37.285073+00:00"},"context_extract":{"job_type":"context_extract","status":"succeeded","result":{"enqueued_papers":25},"error":null,"updated_at":"2026-05-13T21:03:41.078128+00:00"},"graph_features":{"job_type":"graph_features","status":"succeeded","result":{"co_cited":[{"title":"Flow Matching for Generative Modeling","work_id":"6edb71c4-5d64-40af-a394-9757ea051a36","shared_citers":67},{"title":"Denoising Diffusion Implicit Models","work_id":"8fa2128b-d18c-405c-ac92-0e669cf89ac0","shared_citers":51},{"title":"Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow","work_id":"a1989e1b-d66d-4533-be3a-fb9c5fd62290","shared_citers":45},{"title":"Classifier-Free Diffusion Guidance","work_id":"acf2c588-c088-4a6c-938e-150ad7c666d7","shared_citers":36},{"title":"Progressive Distillation for Fast Sampling of Diffusion Models","work_id":"fd04f498-ff85-4de3-bcc7-31ef072b2ceb","shared_citers":25},{"title":"SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis","work_id":"8034c587-fba6-4941-87ba-c98f2ac962cb","shared_citers":23},{"title":"Decoupled Weight Decay Regularization","work_id":"07ef7360-d385-4033-83f7-8384a6325204","shared_citers":21},{"title":"Hierarchical Text-Conditional Image Generation with CLIP Latents","work_id":"0c6a768b-70b8-4242-bb0e-459f1008c9fc","shared_citers":21},{"title":"Auto-Encoding Variational Bayes","work_id":"97d95295-30e1-42b4-bbf6-85f0fa4edb44","shared_citers":20},{"title":"Denoising Diffusion Probabilistic Models","work_id":"dc023f4e-7c79-471c-b713-deeb559ba010","shared_citers":18},{"title":"Wan: Open and Advanced Large-Scale Video Generative Models","work_id":"ad3ebc3b-4224-46c9-b61d-bcf135da0a7c","shared_citers":18},{"title":"Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets","work_id":"4f68eada-27e3-437a-a2fe-6e4ca524d0d3","shared_citers":17},{"title":"Training Diffusion Models with Reinforcement Learning","work_id":"67684dda-3930-452a-b91a-36cbb8e2e219","shared_citers":17},{"title":"Adam: A Method for Stochastic Optimization","work_id":"1910796d-9b52-4683-bf5c-de9632c1028b","shared_citers":16},{"title":"Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis","work_id":"40702548-f094-4c67-a5db-a62f426f852e","shared_citers":14},{"title":"Stochastic Interpolants: A Unifying Framework for Flows and Diffusions","work_id":"c2c7dd8f-fbfb-4591-89ec-9a3a0e6744bd","shared_citers":14},{"title":"Building Normalizing Flows with Stochastic Interpolants","work_id":"83654cbd-832b-4269-bec9-26faa87ececd","shared_citers":13},{"title":"Make-A-Video: Text-to-Video Generation without Text-Video Data","work_id":"52a801fc-a707-45a1-a8cd-0d6702f124ab","shared_citers":13},{"title":"CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer","work_id":"f38fc088-12aa-4bf4-9ecd-08d3e797ccb7","shared_citers":12},{"title":"Improving and generalizing flow-based generative models with minibatch optimal transport","work_id":"7b689852-603d-46be-aaa5-c38325ba2182","shared_citers":12},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":12},{"title":"Consistency Models","work_id":"502bf494-8fcd-434f-828f-0566ab606719","shared_citers":11},{"title":"Diffusion Posterior Sampling for General Noisy Inverse Problems","work_id":"083ab9fb-05f0-41e1-9628-982017eac344","shared_citers":11},{"title":"High-Resolution Image Synthesis with Latent Diffusion Models","work_id":"f0270d36-2952-47fb-84c1-95e3ec341126","shared_citers":11}],"time_series":[{"n":2,"year":2021},{"n":6,"year":2023},{"n":3,"year":2024},{"n":8,"year":2025},{"n":158,"year":2026}]},"error":null,"updated_at":"2026-05-13T21:03:41.164531+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":"2026-05-13T21:03:36.571536+00:00"},"role_polarity":{"job_type":"role_polarity","status":"succeeded","result":{"title":"Score-Based Generative Modeling through Stochastic Differential Equations","claims":[{"claim_text":"Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Score-Based Generative Modeling through Stochastic Differential Equations because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T21:03:42.144051+00:00"},"summary_claims":{"job_type":"summary_claims","status":"succeeded","result":{"title":"Score-Based Generative Modeling through Stochastic Differential Equations","claims":[{"claim_text":"Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Score-Based Generative Modeling through Stochastic Differential Equations because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T21:03:42.142242+00:00"}},"summary":{"title":"Score-Based Generative Modeling through Stochastic Differential Equations","claims":[{"claim_text":"Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Score-Based Generative Modeling through Stochastic Differential Equations because it crossed a citation-hub threshold.","role_counts":[]},"graph":{"co_cited":[{"title":"Flow Matching for Generative Modeling","work_id":"6edb71c4-5d64-40af-a394-9757ea051a36","shared_citers":67},{"title":"Denoising Diffusion Implicit Models","work_id":"8fa2128b-d18c-405c-ac92-0e669cf89ac0","shared_citers":51},{"title":"Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow","work_id":"a1989e1b-d66d-4533-be3a-fb9c5fd62290","shared_citers":45},{"title":"Classifier-Free Diffusion Guidance","work_id":"acf2c588-c088-4a6c-938e-150ad7c666d7","shared_citers":36},{"title":"Progressive Distillation for Fast Sampling of Diffusion Models","work_id":"fd04f498-ff85-4de3-bcc7-31ef072b2ceb","shared_citers":25},{"title":"SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis","work_id":"8034c587-fba6-4941-87ba-c98f2ac962cb","shared_citers":23},{"title":"Decoupled Weight Decay Regularization","work_id":"07ef7360-d385-4033-83f7-8384a6325204","shared_citers":21},{"title":"Hierarchical Text-Conditional Image Generation with CLIP Latents","work_id":"0c6a768b-70b8-4242-bb0e-459f1008c9fc","shared_citers":21},{"title":"Auto-Encoding Variational Bayes","work_id":"97d95295-30e1-42b4-bbf6-85f0fa4edb44","shared_citers":20},{"title":"Denoising Diffusion Probabilistic Models","work_id":"dc023f4e-7c79-471c-b713-deeb559ba010","shared_citers":18},{"title":"Wan: Open and Advanced Large-Scale Video Generative Models","work_id":"ad3ebc3b-4224-46c9-b61d-bcf135da0a7c","shared_citers":18},{"title":"Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets","work_id":"4f68eada-27e3-437a-a2fe-6e4ca524d0d3","shared_citers":17},{"title":"Training Diffusion Models with Reinforcement Learning","work_id":"67684dda-3930-452a-b91a-36cbb8e2e219","shared_citers":17},{"title":"Adam: A Method for Stochastic Optimization","work_id":"1910796d-9b52-4683-bf5c-de9632c1028b","shared_citers":16},{"title":"Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis","work_id":"40702548-f094-4c67-a5db-a62f426f852e","shared_citers":14},{"title":"Stochastic Interpolants: A Unifying Framework for Flows and Diffusions","work_id":"c2c7dd8f-fbfb-4591-89ec-9a3a0e6744bd","shared_citers":14},{"title":"Building Normalizing Flows with Stochastic Interpolants","work_id":"83654cbd-832b-4269-bec9-26faa87ececd","shared_citers":13},{"title":"Make-A-Video: Text-to-Video Generation without Text-Video Data","work_id":"52a801fc-a707-45a1-a8cd-0d6702f124ab","shared_citers":13},{"title":"CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer","work_id":"f38fc088-12aa-4bf4-9ecd-08d3e797ccb7","shared_citers":12},{"title":"Improving and generalizing flow-based generative models with minibatch optimal transport","work_id":"7b689852-603d-46be-aaa5-c38325ba2182","shared_citers":12},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":12},{"title":"Consistency Models","work_id":"502bf494-8fcd-434f-828f-0566ab606719","shared_citers":11},{"title":"Diffusion Posterior Sampling for General Noisy Inverse Problems","work_id":"083ab9fb-05f0-41e1-9628-982017eac344","shared_citers":11},{"title":"High-Resolution Image Synthesis with Latent Diffusion Models","work_id":"f0270d36-2952-47fb-84c1-95e3ec341126","shared_citers":11}],"time_series":[{"n":2,"year":2021},{"n":6,"year":2023},{"n":3,"year":2024},{"n":8,"year":2025},{"n":158,"year":2026}]},"authors":[{"id":"9c1dd117-9177-4557-8ed0-32ef0ec70e72","orcid":null,"display_name":"Ab- hishek Kumar","source":"manual","import_confidence":0.72},{"id":"4c622740-2d53-4b78-b89b-815c032a1d6c","orcid":null,"display_name":"and Ben Poole","source":"manual","import_confidence":0.72},{"id":"6b9fb90b-3116-44d5-9610-1a8263f12dbc","orcid":null,"display_name":"Diederik P Kingma","source":"manual","import_confidence":0.72},{"id":"4e16e341-9f1f-4f1e-8ae6-dcc81290225a","orcid":null,"display_name":"Jascha Sohl-Dickstein","source":"manual","import_confidence":0.72},{"id":"e00f90b7-aa2a-4902-8d30-19cdf0a496f9","orcid":null,"display_name":"Stefano Ermon","source":"manual","import_confidence":0.72},{"id":"c36345af-089e-427a-b034-794e452cabe0","orcid":null,"display_name":"Yang Song","source":"manual","import_confidence":0.72}]}}