mlsanityData to 14 Oct 2026

RewardSDS: Aligning Score Distillation via Reward-Weighted Sampling

arXiv 2503.09601 ↗First posted 12 Mar 2025cs.CVcited by 4

On the map: Diffusion models → 3D, point cloud and scene generation

also in 1 more
Abstract (arXiv)

Score Distillation Sampling (SDS) has emerged as an effective technique for leveraging 2D diffusion priors for tasks such as text-to-3D generation. While powerful, SDS struggles with achieving fine-grained alignment to user intent. To overcome this, we introduce RewardSDS, a novel approach that weights noise samples based on alignment scores from a reward model, producing a weighted SDS loss. This loss prioritizes gradients from noise samples that yield aligned high-reward output. Our approach is broadly applicable and can extend SDS-based methods. In particular, we demonstrate its applicability to Variational Score Distillation (VSD) by introducing RewardVSD. We evaluate RewardSDS and RewardVSD on text-to-image, 2D editing, and text-to-3D generation tasks, showing significant improvements over SDS and VSD on a diverse set of metrics measuring generation quality and alignment to desired reward models, enabling state-of-the-art performance. Project page is available at https://itaychachy.github.io/reward-sds/.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Placed in 3D, point cloud and scene generation, one of 14 directions in Diffusion models. Builds on 47 of its 53 references that are in the map. First posted on arXiv 12 Mar 2025.

Builds on

47 of 53 references.

Closest ancestors · 10

Same direction, newest first

  1. 2024JointDreamer: Ensuring Geometry Consistency and Text Congruence in Text-to-3D Generation via Joint Score Distillation17 Jul
  2. 2024Rethinking Score Distillation as a Bridge Between Image Distributions13 Jun
  3. 2024DreamFlow: High-Quality Text-to-3D Generation by Approximating Probability Flow22 Mar
  4. 2023DreamControl: Control-Based Text-to-3D Generation with 3D Self-Prior11 Dec
  5. 2023Text-to-3D with Classifier Score Distillation30 Oct
  6. 2023DreamTime: An Improved Optimization Strategy for Diffusion-Guided 3D Generation21 Jun
  7. 2023ProlificDreamer: High-Fidelity and Diverse Text-to-3D Generation with Variational Score Distillation25 May
  8. 2023Debiasing Scores and Prompts of 2D Diffusion for View-consistent Text-to-3D Generation27 Mar
  9. 2022Score Jacobian Chaining: Lifting Pretrained 2D Diffusion Models for 3D Generation1 Dec
  10. 2022DreamFusion: Text-to-3D using 2D Diffusion29 Sep

Other references · 37

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  3. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  4. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  5. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,585
  6. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,355
  7. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,226
  8. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,152
  9. 2023Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasetscited by 2,581
  10. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,986
  11. 2022Imagen Video: High Definition Video Generation with Diffusion Modelscited by 1,778
  12. 2022Make-A-Video: Text-to-Video Generation without Text-Video Datacited by 1,742
25 more
  1. 2024Large Language Monkeys: Scaling Inference Compute with Repeated Samplingcited by 991
  2. 2023Training Diffusion Models with Reinforcement Learningcited by 982
  3. 2023DreamGaussian: Generative Gaussian Splatting for Efficient 3D Content Creationcited by 843
  4. 2023Diffusion Model Alignment Using Direct Preference Optimizationcited by 790
  5. 2022Latent-NeRF for Shape-Guided Generation of 3D Shapes and Texturescited by 506
  6. 2025Inference-Time Scaling for Diffusion Models beyond Scaling Denoising Stepscited by 234
  7. 2024Adjoint Matching: Fine-tuning Flow and Diffusion Generative Models with Memoryless Stochastic Optimal Controlcited by 233
  8. 2023Text-To-4D Dynamic Scene Generationcited by 205
  9. 20234D-fy: Text-to-4D Generation Using Hybrid Score Distillation Samplingcited by 185
  10. 2023Re-imagine the Negative Prompt Algorithm: Transform 2D Diffusion into 3D, alleviate Janus problem and Beyondcited by 139
  11. 2022VectorFusion: Text-to-SVG by Abstracting Pixel-Based Diffusion Modelscited by 112
  12. 2023Noise-Free Score Distillationcited by 99
  13. 2024Golden Noise for Diffusion Models: A Learning Frameworkcited by 92
  14. 2023Generating images of rare concepts using pre-trained diffusion modelscited by 79
  15. 2023Word-As-Image for Semantic Typographycited by 63
  16. 2023RealFill: Reference-Driven Generation for Authentic Image Completioncited by 59
  17. 2024DreamReward: Text-to-3D Generation with Human Preferencecited by 56
  18. 2023DiffSketcher: Text Guided Vector Sketch Synthesis through Latent Diffusion Modelscited by 54
  19. 2024A Noise is Worth Diffusion Guidancecited by 49
  20. 2023Posterior Distillation Samplingcited by 46
  21. 2023If at First You Don't Succeed, Try, Try Again: Faithful Diffusion-based Text-to-Image Generation by Selectioncited by 40
  22. 2023Taming Mode Collapse in Score Distillation for Text-to-3D Generationcited by 36
  23. 2023SteinDreamer: Variance Reduction for Text-to-3D Score Distillation via Stein Identitycited by 23
  24. 2023Correcting Diffusion Generation through Resamplingcited by 15
  25. 2025RecDreamer: Consistent Text-to-3D Generation via Uniform Score Distillationcited by 9

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 3; 2026: 1.

4 papers.

  1. 2025DyPE: Dynamic Position Extrapolation for Ultra High Resolution Diffusioncited by 14
  2. 2025Vision-Language Models as Differentiable Semantic and Spatial Rewards for Text-to-3D Generationcited by 6
  3. 2025Track, Inpaint, Resplat: Subject-driven 3D and 4D Generation with Progressive Texture Infillingcited by 2
  4. 2026Variance Reduction on the Camera Axis: Multi-View Score Distillation for 3Dcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Rethinking Score Distilling Sampling for 3D Editing and Generation
  2. 2024Score Distillation Sampling with Learned Manifold Corrective
  3. 2024A Quantitative Evaluation of Score Distillation Sampling Based Text-to-3D
  4. 2025Target-Balanced Score Distillation
  5. 2026Reward Score Matching: Unifying Reward-based Fine-tuning for Flow and Diffusion Models
  6. 2026Preference Score Distillation: Leveraging 2D Rewards to Align Text-to-3D Generation with Human Preference
  7. 2025DiffExp: Efficient Exploration in Reward Fine-tuning for Text-to-Image Diffusion Models
  8. 2024Geometry-Aware Score Distillation via 3D Consistent Noising and Gradient Consistency Modeling