{
 "id": "epc43pn0g3",
 "slug": "score-distillation",
 "title": "Score distillation",
 "updated": "2026-10-15",
 "topic_path": [
  {
   "id": "technology",
   "label": "Technology and the built world",
   "api_url": "https://www.edgepedia.org/api/v1/topics/technology"
  },
  {
   "id": "technology.computing",
   "label": "Computing and digital systems",
   "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing"
  },
  {
   "id": "technology.computing.modern_ai",
   "label": "Modern AI: foundation models, generative AI, and the AI industry",
   "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai"
  },
  {
   "id": "technology.computing.modern_ai.methods",
   "label": "Foundation-model methods and training",
   "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai.methods"
  },
  {
   "id": "technology.computing.modern_ai.methods.generative_media_methods",
   "label": "Generative media methods: diffusion, flow, and autoregressive generation",
   "api_url": "https://www.edgepedia.org/api/v1/topics/technology.computing.modern_ai.methods.generative_media_methods"
  }
 ],
 "geo": [
  {
   "id": "geo.nongeo.t2021.technology.computing",
   "label": "Non-geographic · 2021 and later: Computing and digital systems",
   "api_url": "https://www.edgepedia.org/api/v1/geo/geo.nongeo.t2021.technology.computing",
   "path": [
    {
     "id": "geo.nongeo",
     "label": "Non-geographic",
     "api_url": "https://www.edgepedia.org/api/v1/geo/geo.nongeo"
    },
    {
     "id": "geo.nongeo.t2021",
     "label": "Non-geographic · 2021 and later",
     "api_url": "https://www.edgepedia.org/api/v1/geo/geo.nongeo.t2021"
    },
    {
     "id": "geo.nongeo.t2021.technology",
     "label": "Technology and the built world",
     "api_url": "https://www.edgepedia.org/api/v1/geo/geo.nongeo.t2021.technology"
    },
    {
     "id": "geo.nongeo.t2021.technology.computing",
     "label": "Computing and digital systems",
     "api_url": "https://www.edgepedia.org/api/v1/geo/geo.nongeo.t2021.technology.computing"
    }
   ]
  }
 ],
 "excerpt": "Score distillation is a machine learning technique that uses a frozen diffusion model's denoising direction to optimize images or 3D scenes, best known from DreamFusion's 2022 text-to-3D.",
 "snippet": "Score distillation is a machine learning technique that uses a frozen diffusion model's denoising direction to optimize images or 3D scenes, best known from DreamFusion's 2022 text-to-3D.",
 "node": "technology.computing.modern_ai.methods.generative_media_methods",
 "markdown": "# Score distillation\n\nScore distillation is a machine learning technique that uses the score function (denoising direction) of a frozen diffusion model as a teacher signal to optimize a separate differentiable model or representation, such as an image or a 3D scene, without paired training data and without backpropagating through the teacher. Its best-known use is text-to-3D synthesis: a pretrained text-to-image diffusion model guides the optimization of a 3D scene whose renders are pushed toward what the 2D model considers a good match for a text prompt.<sup>[1](https://doi.org/10.48550/arxiv.2209.14988)</sup>\n\n| Key fact | Detail |\n|---|---|\n| What is optimized | A weighted KL divergence between noised renders of the optimized model and the distributions learned by a frozen diffusion model<sup>[2](https://proceedings.iclr.cc/paper_files/paper/2024/file/eb11d415557899106891f538de72569d-Paper-Conference.pdf)</sup> |\n| Tractability trick | The Jacobian term of the true gradient is omitted, avoiding backpropagation through the denoising network<sup>[3](https://www.ecva.net/papers/eccv_2024/papers_ECCV/papers/11754.pdf)</sup> |\n| Original application | DreamFusion (2022) optimized a NeRF-like 3D scene with a 2D text-to-image diffusion model<sup>[1](https://doi.org/10.48550/arxiv.2209.14988)</sup> |\n| Main variant | Variational Score Distillation (VSD) optimizes a whole distribution rather than a single point; SDS is its single-point Dirac special case<sup>[4](https://doi.org/10.48550/arxiv.2305.16213)</sup><sup> • </sup><sup>[5](https://openaccess.thecvf.com/content/ICCV2025/papers/Lu_Adversarial_Distribution_Matching_for_Diffusion_Distillation_Towards_Efficient_Image_and_ICCV_2025_paper.pdf)</sup> |\n| Typical artifacts | Over-saturated colors, over-smoothing, and the Janus multi-face problem<sup>[6](https://ar5iv.labs.arxiv.org/html/2312.02189)</sup> |\n| Cost benchmark | SDS: 66 min and 6.2 GB VRAM per 3D asset; VSD: 334 min and 47.9 GB (43-prompt, 50-view benchmark)<sup>[7](https://proceedings.neurips.cc/paper_files/paper/2024/file/2ded44d59f5094eed0d02132fe75b60d-Paper-Conference.pdf)</sup> |\n| CFG sensitivity | Small classifier-free guidance scales give over-smoothing; large scales give over-saturation<sup>[8](https://openaccess.thecvf.com/content/CVPR2024/papers/Wei_Adversarial_Score_Distillation_When_score_distillation_meets_GAN_CVPR_2024_paper.pdf)</sup> |\n\n## How it works\n\nLet \\( x = g(\\theta) \\) be an image rendered by a differentiable generator \\( g \\) with parameters \\( \\theta \\), which may be a NeRF, a mesh, or a pixel image. At each step, a timestep \\( t \\) and Gaussian noise \\( \\epsilon \\) are drawn, the render is noised to \\( z_t(x) \\), and a frozen diffusion model predicts the noise it would remove. The score distillation sampling (SDS) gradient is\n\n\\[ \\nabla_{\\theta} L_{\\mathrm{SDS}} = w(t) \\left( \\epsilon_{\\phi}^{\\omega}(z_t(x); y, t) - \\epsilon \\right) \\frac{\\partial x}{\\partial \\theta}, \\]\n\nwhere \\( y \\) is the conditioning text, \\( w(t) \\) a timestep-dependent weight, and \\( \\epsilon_{\\phi}^{\\omega} \\) the guided noise prediction.<sup>[2](https://proceedings.iclr.cc/paper_files/paper/2024/file/eb11d415557899106891f538de72569d-Paper-Conference.pdf)</sup> Poole and colleagues formally showed that this loss minimizes the KL divergence between a family of Gaussian distributions around \\( x \\) and the distributions \\( p(z_t, y, t) \\) learned by the pretrained diffusion model; [DreamFusion](https://www.edgepedia.org/dreamfusion) derived it from probability density distillation, which minimizes KL divergence between Gaussians with shared means under the forward diffusion process.<sup>[2](https://proceedings.iclr.cc/paper_files/paper/2024/file/eb11d415557899106891f538de72569d-Paper-Conference.pdf)</sup><sup> • </sup><sup>[1](https://doi.org/10.48550/arxiv.2209.14988)</sup>\n\nThe SDS update derived from that KL objective omits the U-Net Jacobian term of the full gradient through the denoising network.<sup>[16](https://proceedings.neurips.cc/paper_files/paper/2023/file/e7fd2c0a1a6f956c94024e955b34cc43-Paper-Conference.pdf)</sup> In practice this term is omitted to avoid backpropagating through the denoising model, giving the approximate gradient above; the DreamFusion authors found that omitting the poorly conditioned Jacobian term yields a more stable gradient for backpropagation to the scene model.<sup>[3](https://www.ecva.net/papers/eccv_2024/papers_ECCV/papers/11754.pdf)</sup><sup> • </sup><sup>[6](https://ar5iv.labs.arxiv.org/html/2312.02189)</sup> With classifier-free guidance, the predicted noise is the weighted sum of the conditioned and unconditioned predictions,\n\n\\[ \\epsilon_{\\phi}^{\\omega}(z_t, y, t) = \\omega \\, \\epsilon_{\\phi}(z_t, y, t) + (1 - \\omega) \\, \\epsilon_{\\phi}(z_t, t), \\]\n\nwith guidance weight \\( \\omega \\).<sup>[3](https://www.ecva.net/papers/eccv_2024/papers_ECCV/papers/11754.pdf)</sup> Writing the guided direction out shows that for small guidance strength \\( s \\) the \\( \\epsilon \\) term acts as an averaging term that regresses the image toward the mean.<sup>[9](https://proceedings.neurips.cc/paper_files/paper/2024/file/3b62bca132cf5c8973b09a2fc6dc8ca6-Paper-Conference.pdf)</sup>\n\n## How it is done\n\nA practitioner loop over a NeRF or image proceeds as follows. Each iteration, draw a random timestep \\( t \\) and Gaussian noise \\( \\epsilon \\); render the current scene from a random camera; noise the render; predict noise with the frozen, classifier-free-guided diffusion model; and backpropagate the weighted residual through the differentiable generator.<sup>[2](https://proceedings.iclr.cc/paper_files/paper/2024/file/eb11d415557899106891f538de72569d-Paper-Conference.pdf)</sup> A representative configuration (NFSD) uses 25,000 iterations of AdamW at learning rate 0.01, with the rendering resolution raised from 64×64 to 512×512 after 5,000 iterations, [Stable Diffusion](https://www.edgepedia.org/stable-diffusion) 2.1-base as the teacher, and the maximum diffusion time annealed to 500.<sup>[2](https://proceedings.iclr.cc/paper_files/paper/2024/file/eb11d415557899106891f538de72569d-Paper-Conference.pdf)</sup>\n\nTime annealing matters: high noise levels during training produce artifacts and multi-faced geometry, and as the model converges less noise should be added each step.<sup>[6](https://ar5iv.labs.arxiv.org/html/2312.02189)</sup> Gradually reducing the diffusion timesteps drawn during optimization is an effective approach for improving generation quality, and is used across several follow-up methods.<sup>[2](https://proceedings.iclr.cc/paper_files/paper/2024/file/eb11d415557899106891f538de72569d-Paper-Conference.pdf)</sup>\n\n## Origin\n\nDreamFusion (Poole and colleagues, 2022, arXiv) replaced earlier CLIP-based optimization with a loss distilled from a 2D diffusion model and applied it to a NeRF-like 3D scene parameterization, which SDS supports for any parameter space that maps back to images differentiably.<sup>[1](https://doi.org/10.48550/arxiv.2209.14988)</sup> A later benchmark paper describes score distillation as having been introduced concurrently in DreamFusion (SDS), Score Jacobian Chaining (SJC), and Magic3D, with the shared idea of distilling a frozen 2D diffusion model into 3D assets.<sup>[7](https://proceedings.neurips.cc/paper_files/paper/2024/file/2ded44d59f5094eed0d02132fe75b60d-Paper-Conference.pdf)</sup> The 3D scenes themselves build on Neural Radiance Fields (Mildenhall and colleagues, 2021, Communications of the ACM).<sup>[10](https://doi.org/10.1145/3503250)</sup>\n\n## Variants\n\n**VSD (ProlificDreamer).** Variational Score Distillation (Wang and colleagues, 2023, arXiv) formulates the objective as variational inference,\n\n\\[ L_{\\mathrm{SDS}}(\\theta) := \\mathbb{E}_{t,c} \\, \\frac{\\sigma_t}{\\alpha_t} \\, \\omega(t) \\, D_{\\mathrm{KL}} \\left( q_{\\theta}^{t}(x_t \\mid c) \\, \\| \\, p_t(x_t \\mid y^c) \\right), \\]\n\nbut optimizes a whole distribution \\( \\mu \\) from which parameters \\( \\theta \\) are sampled, rather than the single point \\( \\theta \\) of SDS.<sup>[4](https://doi.org/10.48550/arxiv.2305.16213)</sup> SDS is a specialized instance of VSD in which the variational distribution is a single-point Dirac distribution.<sup>[5](https://openaccess.thecvf.com/content/ICCV2025/papers/Lu_Adversarial_Distribution_Matching_for_Diffusion_Distillation_Towards_Efficient_Image_and_ICCV_2025_paper.pdf)</sup> VSD supplies its negative direction with a diffusion model fine-tuned concurrently on rendered images; CSD interprets this as an adaptive form of negative classifier score optimization.<sup>[11](https://doi.org/10.48550/arxiv.2310.19415)</sup>\n\n**Decomposition-based variants.** ISD (VividDreamer) decouples SDS into a weighted sum of a reconstruction term and a classifier-free guidance term, replacing the reconstruction term with an invariant score derived from DDIM sampling.<sup>[12](https://link.springer.com/chapter/10.1007/978-3-031-73223-2_8)</sup> NFSD decomposes the score into interpretable components, \\( \\nabla_{\\theta} L_{\\mathrm{NFSD}} = w(t)(\\delta D + s \\, \\delta C) \\, \\partial x / \\partial \\theta \\).<sup>[2](https://proceedings.iclr.cc/paper_files/paper/2024/file/eb11d415557899106891f538de72569d-Paper-Conference.pdf)</sup> LMC-SDS adds a learned manifold corrective loss to the SDS gradient.<sup>[3](https://www.ecva.net/papers/eccv_2024/papers_ECCV/papers/11754.pdf)</sup> SDI replaces SDS's random noise sampling with prompt-conditioned DDIM inversion, proving that per-view SDS guidance is a simplified reparameterization of DDIM sampling in which vanilla SDS draws fresh noise each step while DDIM keeps trajectories consistent with previously predicted noise.<sup>[7](https://proceedings.neurips.cc/paper_files/paper/2024/file/2ded44d59f5094eed0d02132fe75b60d-Paper-Conference.pdf)</sup>\n\n## Applications\n\nBeyond 3D, the transport-path view of score distillation has been applied to text-to-2D generation, text-based NeRF optimization, translating paintings to real images, optical illusion generation, and 3D sketch-to-real, matching or beating specialized methods.<sup>[9](https://proceedings.neurips.cc/paper_files/paper/2024/file/3b62bca132cf5c8973b09a2fc6dc8ca6-Paper-Conference.pdf)</sup> ProlificDreamer renders NeRFs at 512×512 resolution with complex effects such as smoke and drops.<sup>[4](https://doi.org/10.48550/arxiv.2305.16213)</sup>\n\n## Limitations and alternatives\n\nClassic SDS-generated 3D assets commonly show over-saturated colors, blurry appearance, omitted fine details, oversimplified geometry, and the Janus problem, in which the object contains multiple canonical views seen from different viewpoints.<sup>[6](https://ar5iv.labs.arxiv.org/html/2312.02189)</sup> Diagnoses differ in emphasis. ISD's decomposition attributes over-saturation to the large classifier-free guidance scale (SDS uses guidance scale 100.0) and over-smoothing to the reconstruction term.<sup>[12](https://link.springer.com/chapter/10.1007/978-3-031-73223-2_8)</sup> A transport-path analysis attributes the characteristic artifacts to a linear approximation of the optimal transport path from corrupted images to the natural image distribution, and to poor estimates of the source distribution.<sup>[9](https://proceedings.neurips.cc/paper_files/paper/2024/file/3b62bca132cf5c8973b09a2fc6dc8ca6-Paper-Conference.pdf)</sup> Original SDS also has mode-seeking behavior, so results from different random seeds look very similar; the NFSD authors attribute low diversity to diffusion scores being uncorrelated across successive iterations.<sup>[3](https://www.ecva.net/papers/eccv_2024/papers_ECCV/papers/11754.pdf)</sup><sup> • </sup><sup>[2](https://proceedings.iclr.cc/paper_files/paper/2024/file/eb11d415557899106891f538de72569d-Paper-Conference.pdf)</sup> ISD does not fully solve the Janus problem; its authors mitigate it by adding back-side descriptions to prompts and widening the camera field of view.<sup>[12](https://link.springer.com/chapter/10.1007/978-3-031-73223-2_8)</sup>\n\nSDS is highly sensitive to the classifier-free guidance scale: small scales give over-smoothing, large scales over-saturation, and worse Janus artifacts.<sup>[8](https://openaccess.thecvf.com/content/CVPR2024/papers/Wei_Adversarial_Score_Distillation_When_score_distillation_meets_GAN_CVPR_2024_paper.pdf)</sup> VSD works with small CFG weights such as 7.5, where SDS cannot generate plausible results; smaller CFG weights yield more diverse results, and ProlificDreamer sets CFG = 7.5 as a trade-off between diversity and optimization stability.<sup>[4](https://doi.org/10.48550/arxiv.2305.16213)</sup> Test-time optimization is slow, largely due to the NeRF representation.<sup>[6](https://ar5iv.labs.arxiv.org/html/2312.02189)</sup> In a 43-prompt, 50-view benchmark, SDS at 10k steps reached CLIP 29.81 ± 2.49 in 66 min using 6.2 GB of VRAM, while VSD at 25k steps reached CLIP 33.31 ± 2.39 in 334 min using 47.9 GB.<sup>[7](https://proceedings.neurips.cc/paper_files/paper/2024/file/2ded44d59f5094eed0d02132fe75b60d-Paper-Conference.pdf)</sup>\n\nAlternatives replace the per-sample SDS gradient with distribution-level objectives. Distribution Matching Distillation (DMD) trains a fast one-step generator with a gradient expressed as the difference of two score functions, modeled by two diffusion denoisers for the real and fake distributions, plus a regression loss on a fixed set of noise-image pairs.<sup>[13](https://doi.org/10.48550/arxiv.2311.18828)</sup> DMD's reverse KL minimization is zero-forcing and can cause mode collapse; DMD used an ODE-based regularizer and DMD2 a GAN-based regularizer with real data to counteract this.<sup>[5](https://openaccess.thecvf.com/content/ICCV2025/papers/Lu_Adversarial_Distribution_Matching_for_Diffusion_Distillation_Towards_Efficient_Image_and_ICCV_2025_paper.pdf)</sup> Score identity [Distillation](https://www.edgepedia.org/distillation) (SiD) is a data-free method that distills a pretrained diffusion model into a single-step generator, reformulating forward diffusion as semi-implicit distributions and using three score-related identities; it achieves an exponentially fast reduction in Fréchet inception distance and approaches or exceeds the teacher's FID.<sup>[14](https://doi.org/10.48550/arxiv.2404.04057)</sup> Later methods including Moment Matching Distillation, SiD, and Score Implicit Matching use variants of Fisher divergence, and Phased DMD performs few-step distillation by applying score matching within subintervals of the diffusion trajectory, re-initializing each phase's fake model from the pretrained teacher.<sup>[5](https://openaccess.thecvf.com/content/ICCV2025/papers/Lu_Adversarial_Distribution_Matching_for_Diffusion_Distillation_Towards_Efficient_Image_and_ICCV_2025_paper.pdf)</sup><sup> • </sup><sup>[15](https://openaccess.thecvf.com/content/CVPR2026/papers/Fan_Phased_DMD_Few-step_Distribution_Matching_Distillation_via_Score_Matching_within_CVPR_2026_paper.pdf)</sup> How score distillation compares with fine-tuning or feed-forward generators has not been settled by published head-to-head comparisons.\n\n## References\n\n1. [Poole, Ben and colleagues (2022). DreamFusion: Text-to-3D using 2D Diffusion. arXiv (Cornell University).](https://doi.org/10.48550/arxiv.2209.14988)\n2. [Noise-Free Score Distillation (NFSD)](https://proceedings.iclr.cc/paper_files/paper/2024/file/eb11d415557899106891f538de72569d-Paper-Conference.pdf)\n3. [Score Distillation Sampling with Learned Manifold Corrective (LMC-SDS)](https://www.ecva.net/papers/eccv_2024/papers_ECCV/papers/11754.pdf)\n4. [Wang, Zhengyi and colleagues (2023). ProlificDreamer: High-Fidelity and Diverse Text-to-3D Generation with Variational Score Distillation. arXiv (Cornell University).](https://doi.org/10.48550/arxiv.2305.16213)\n5. [Adversarial Distribution Matching for Diffusion Distillation Towards Efficient Image and Video Synthesis](https://openaccess.thecvf.com/content/ICCV2025/papers/Lu_Adversarial_Distribution_Matching_for_Diffusion_Distillation_Towards_Efficient_Image_and_ICCV_2025_paper.pdf)\n6. [StableDreamer: Taming Noisy Score Distillation Sampling for Text-to-3D](https://ar5iv.labs.arxiv.org/html/2312.02189)\n7. [Score Distillation via Reparametrized DDIM (SDI)](https://proceedings.neurips.cc/paper_files/paper/2024/file/2ded44d59f5094eed0d02132fe75b60d-Paper-Conference.pdf)\n8. [Adversarial Score Distillation: When score distillation meets GAN](https://openaccess.thecvf.com/content/CVPR2024/papers/Wei_Adversarial_Score_Distillation_When_score_distillation_meets_GAN_CVPR_2024_paper.pdf)\n9. [Rethinking Score Distillation as a Bridge Between Image Distributions](https://proceedings.neurips.cc/paper_files/paper/2024/file/3b62bca132cf5c8973b09a2fc6dc8ca6-Paper-Conference.pdf)\n10. [Ben Mildenhall and colleagues (2021). NeRF. Communications of the ACM.](https://doi.org/10.1145/3503250)\n11. [Yu, Xin and colleagues (2023). Text-to-3D with Classifier Score Distillation. arXiv (Cornell University).](https://doi.org/10.48550/arxiv.2310.19415)\n12. [VividDreamer: Invariant Score Distillation for Hyper-Realistic Text-to-3D Generation](https://link.springer.com/chapter/10.1007/978-3-031-73223-2_8)\n13. [Yin, Tianwei and colleagues (2023). One-step Diffusion with Distribution Matching Distillation. arXiv (Cornell University).](https://doi.org/10.48550/arxiv.2311.18828)\n14. [Zhou, Mingyuan and colleagues (2024). Score identity Distillation: Exponentially Fast Distillation of Pretrained Diffusion Models for One-Step Generation. arXiv (Cornell University).](https://doi.org/10.48550/arxiv.2404.04057)\n15. [Phased DMD: Few-step Distribution Matching Distillation via Score Matching within Subintervals](https://openaccess.thecvf.com/content/CVPR2026/papers/Fan_Phased_DMD_Few-step_Distribution_Matching_Distillation_via_Score_Matching_within_CVPR_2026_paper.pdf)\n16. [proceedings.neurips.cc](https://proceedings.neurips.cc/paper_files/paper/2023/file/e7fd2c0a1a6f956c94024e955b34cc43-Paper-Conference.pdf)\n\n---\n*Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › Foundation-model methods and training › Generative media methods: diffusion, flow, and autoregressive generation*\n\n*Initially written Sep 29, 2026 · Edited: Oct 15, 2026*\n\n*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*\n\nLicense: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license\n",
 "same_as": [],
 "url": "https://www.edgepedia.org/score-distillation",
 "markdown_url": "https://www.edgepedia.org/score-distillation.md",
 "license": {
  "name": "Edgepedia Community License 1.0",
  "url": "https://www.edgepedia.org/edgepedia/license",
  "summary": "Free with credit, commercial use included. AI training is open to everyone. For other uses, organizations over USD 100M in revenue or 100M monthly users license separately.",
  "spdx": "LicenseRef-Edgepedia-Community-1.0"
 },
 "credit": "\"Score distillation\", Edgepedia (EdgeChat), https://www.edgepedia.org/score-distillation. Edgepedia Community License 1.0.",
 "credit_md": "\"[Score distillation](https://www.edgepedia.org/score-distillation)\", Edgepedia (EdgeChat), [https://www.edgepedia.org/score-distillation](https://www.edgepedia.org/score-distillation). [Edgepedia Community License 1.0](https://www.edgepedia.org/edgepedia/license).",
 "credit_html": "\"<a href=\"https://www.edgepedia.org/score-distillation\">Score distillation</a>\", Edgepedia (EdgeChat), <a href=\"https://www.edgepedia.org/score-distillation\">https://www.edgepedia.org/score-distillation</a>. <a href=\"https://www.edgepedia.org/edgepedia/license\">Edgepedia Community License 1.0</a>.",
 "speakable": "Score distillation is a machine learning technique that uses a frozen diffusion model's denoising direction to optimize images or 3D scenes, best known from DreamFusion's 2022 text-to-3D."
}