mlsanityData to 7 Oct 2026

Lookahead Sample Reward Guidance for Test-Time Scaling of Diffusion Models

arXiv 2602.03211 ↗First posted 3 Feb 2026cs.LG cs.AIcited by 3

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

Diffusion models have demonstrated strong generative performance; however, generated samples often fail to fully align with human intent. This paper studies an efficient test-time scaling method for sampling from regions with higher human-aligned reward values. Existing methods for computing the expected future reward (EFR) face important limitations: backward rollout incurs prohibitively high sampling costs, while Tweedie-based approaches, including Sequential Monte Carlo and gradient guidance, suffer from bias and inherent sampling issues. We show that the EFR at any xt can be computed using only marginal samples from a pre-trained diffusion model, enabling closed-form reward guidance without neural backpropagation. To further improve efficiency, we introduce a few-step lookahead sampling and an accurate solver that guides particles toward high-reward lookahead samples. We refer to this sampling scheme as LiDAR sampling. LiDAR achieves the same GenEval performance as the latest gradient guidance method for SDXL with a 9.5x speedup. We release the code at https://github.com/aailab-kaist/Diffusion-LiDAR-Sampling.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 51 of its 66 references that are in the map. First posted on arXiv 3 Feb 2026.

Builds on

51 of 66 references.

Closest ancestors · 20

Same direction, newest first

  1. 2025A General Framework for Inference-time Scaling and Steering of Diffusion Models12 Jan
  2. 2024Simple Guidance Mechanisms for Discrete Diffusion Models13 Dec
  3. 2024Shielded Diffusion: Generating Novel and Diverse Images using Sparse Repellency8 Oct
  4. 2024Derivative-Free Guidance in Continuous and Discrete Diffusion Models with Soft Value-Based Decoding15 Aug
  5. 2024Improved Distribution Matching Distillation for Fast Image Synthesis23 May
  6. 2024Hyper-SD: Trajectory Segmented Consistency Model for Efficient Image Synthesis21 Apr
  7. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesis5 Mar
  8. 2023LCM-LoRA: A Universal Stable-Diffusion Acceleration Module9 Nov
  9. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis4 Jul
  10. 2023Consistency Models2 Mar
  11. 2023Universal Guidance for Diffusion Models14 Feb
  12. 2023Stable Target Field for Reduced Variance Score Estimation in Diffusion Models1 Feb
8 more
  1. 2022Refining Generative Process with Discriminator Guidance in Score-based Diffusion Models28 Nov
  2. 2022Classifier-Free Diffusion Guidance26 Jul
  3. 2022DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps2 Jun
  4. 2022Maximum Likelihood Training of Implicit Nonlinear Diffusion Models27 May
  5. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  6. 2021Maximum Likelihood Training of Score-Based Diffusion Models22 Jan
  7. 2020Score-Based Generative Modeling through Stochastic Differential Equations26 Nov
  8. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 31

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  3. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  4. 2022Flow Matching for Generative Modelingcited by 5,426
  5. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,067
  6. 2021CLIPScore: A Reference-free Evaluation Metric for Image Captioningcited by 2,388
  7. 2022Video Diffusion Modelscited by 2,280
  8. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,928
  9. 2022Diffusion Posterior Sampling for General Noisy Inverse Problemscited by 1,375
  10. 2025s1: Simple test-time scalingcited by 1,311
  11. 2023ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generationcited by 1,010
  12. 2023GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignmentcited by 946
19 more
  1. 2023Training Diffusion Models with Reinforcement Learningcited by 939
  2. 2025Large Language Diffusion Modelscited by 872
  3. 2023Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesiscited by 847
  4. 2023Diffusion Model Alignment Using Direct Preference Optimizationcited by 767
  5. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distributioncited by 712
  6. 2023Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervisioncited by 445
  7. 2023Human Preference Score: Better Aligning Text-to-Image Models with Human Preferencecited by 284
  8. 2024Adjoint Matching: Fine-tuning Flow and Diffusion Generative Models with Memoryless Stochastic Optimal Controlcited by 222
  9. 2025HPSv3: Towards Wide-Spectrum Human Preference Scorecited by 196
  10. 2024Fine-Tuning of Continuous-Time Diffusion Models as Entropy-Regularized Controlcited by 136
  11. 2024Tuning Language Models by Proxycited by 88
  12. 2024Deep Reward Supervisions for Tuning Text-to-Image Diffusion Modelscited by 59
  13. 2026Meta Flow Maps enable scalable reward alignmentcited by 32
  14. 2024Weak-to-Strong Preference Optimization: Stealing Reward from Weak Aligned Modelcited by 22
  15. 2025Training-Free Safe Denoisers for Safe Use of Diffusion Modelscited by 19
  16. 2024Improving Long-Text Alignment for Text-to-Image Diffusion Modelscited by 18
  17. 2025Tilt Matching for Scalable Sampling and Fine-Tuningcited by 17
  18. 2025Preference Optimization by Estimating the Ratio of the Data Distributioncited by 9
  19. 2025Diffusion Adaptive Text Embedding for Text-to-Image Diffusion Modelscited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Feature-Space Guided Diffusion for Realistic Ultrasound Image Synthesiscited by 0
  2. 2026RIDE: Reference-Anchored Inference-Time Diffusion Editing for Scaffold Hoppingcited by 0
  3. 2026FastGuide: Accelerating Reward Guidance for Diffusion Large Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Best-of-N Guidance for Test-time Diffusion Alignment
  2. 2025Test-time Alignment of Diffusion Models without Reward Over-optimization
  3. 2026Steering diffusion models with quadratic rewards: a fine-grained analysis
  4. 2025Effective Test-Time Scaling of Discrete Diffusion through Iterative Refinement
  5. 2026Are we really tilting? The mechanics of reward guidance in flow and diffusion models
  6. 2025Towards a unified framework for guided diffusion models
  7. 2026Hierarchical Variational Policies for Reward-Guided Diffusion
  8. 2026NoiseTilt: Noise-Tilted Reverse Kernels for Diffusion Reward Alignment