mlsanityData to 14 Oct 2026

ToPO: Token-Conditioned Preference Routing for Attention-Based Latent Diffusion Models

arXiv 2609.03688 ↗First posted 3 Sep 2026cs.CVcited by 0

On the map: Diffusion models → Diffusion and flow policies for robotics

also in 1 more
Abstract (arXiv)

Pairwise preference labels rank complete images, yet Diffusion-DPO applies their effect over many spatial and denoising-time coordinates. For attention-based, noise-prediction latent diffusion, ToPO (Token-Oriented Preference Optimization) constructs a per-minibatch, detached, separable spatial-temporal route from branchwise squared-residual contrast in a frozen reference denoiser. Preferred-branch cross-attention uses content tokens to modulate the spatial factor, and an auxiliary pixel-midpoint ordering term is added without local labels or a learned reward model. In matched three-seed retrainings with a shared update schedule, ToPO has higher endpoint estimates than Diffusion-DPO on all five reported SD-1.5 metrics and on HPSv2, ImageReward, and CLIP for SDXL. It also receives larger raw win shares in an aggregate blind SDXL A/B study. These findings are scoped to the reported equal-update U-Net protocols rather than an equal-compute comparison.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Diffusion and flow policies for robotics, one of 14 directions in Diffusion models. Builds on 33 of its 42 references that are in the map. First posted on arXiv 3 Sep 2026.

Builds on

33 of 42 references.

Closest ancestors · 15

Same direction, newest first

  1. 2026STAR: SpatioTemporal Adaptive Reward Allocation for Text-to-Image RL Post-Training16 Jun
  2. 2025Inversion-DPO: Precise and Efficient Post-Training for Diffusion Models14 Jul
  3. 2025D-Fusion: Direct Preference Optimization for Aligning Diffusion Models with Visually Consistent Samples28 May
  4. 2025Rethinking Direct Preference Optimization in Diffusion Models24 May
  5. 2025InPO: Inversion Preference Optimization with Reparametrized DDIM for Efficient Diffusion Model Alignment24 Mar
  6. 2025Diffusion Model as a Noise-Aware Latent Reward Model for Step-Level Preference Optimization3 Feb
  7. 2024Margin-aware Preference Optimization for Aligning Diffusion Models without Reference10 Jun
  8. 2024Aesthetic Post-Training Diffusion Models from Generic Preferences with Step-by-step Preference Optimization6 Jun
  9. 2024Curriculum Direct Preference Optimization for Diffusion and Consistency Models22 May
  10. 2024Aligning Diffusion Models by Optimizing Human Utility6 Apr
  11. 2024A Dense Reward View on Aligning Text-to-Image Diffusion with Preference13 Feb
  12. 2023Using Human Feedback to Fine-tune Diffusion Models without Any Reward Model22 Nov
3 more
  1. 2023Diffusion Model Alignment Using Direct Preference Optimization21 Nov
  2. 2023Aligning Text-to-Image Diffusion Models with Reward Backpropagation5 Oct
  3. 2023Training Diffusion Models with Reinforcement Learning22 May

Other references · 18

  1. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  3. 2020Denoising Diffusion Implicit Modelscited by 9,781
  4. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  5. 2022Prompt-to-Prompt Image Editing with Cross Attention Controlcited by 2,431
  6. 2023Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generationcited by 977
  7. 2023GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignmentcited by 972
  8. 2023Attend-and-Excite: Attention-Based Semantic Guidance for Text-to-Image Diffusion Modelscited by 702
  9. 2024KTO: Model Alignment as Prospect Theoretic Optimizationcited by 264
  10. 2023Freestyle Layout-to-Image Synthesiscited by 67
  11. 2024PatchDPO: Patch-level DPO for Finetuning-free Personalized Image Generationcited by 18
  12. 2025Seeing What Matters: Visual Preference Policy Optimization for Visual Generationcited by 10
6 more
  1. 2026Learning from Noisy Preferences: A Semi-Supervised Learning Approach to Direct Preference Optimizationcited by 4
  2. 2026ViPO: Visual Preference Optimization at Scalecited by 2
  3. 2026Compositional Text-to-Image Generation Via Region-aware Bimodal Direct Preference Optimizationcited by 2
  4. 2024Reliability Assessment of Information Sources Based on Random Permutation Setcited by 1
  5. 2024Evaluating Evidential Reliability In Pattern Recognition Based On Intuitionistic Fuzzy Setscited by 1
  6. 2026Rectify Then Diffuse: Disentangling Concepts Before Denoising Trajectory Unfoldscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Diffusion-SDPO: Safeguarded Direct Preference Optimization for Diffusion Models
  2. 2025SUDO: Enhancing Text-to-Image Diffusion Models with Self-Supervised Direct Preference Optimization
  3. 2025Direct Diffusion Score Preference Optimization via Stepwise Contrastive Policy-Pair Supervision
  4. 2024Aligning Diffusion Models with Noise-Conditioned Perception
  5. 2026Temporal Concentration from Rollout Errors: Implicit Preference Optimization for Text-to-Video Diffusion
  6. 2025Ranking-based Preference Optimization for Diffusion Models from Implicit User Feedback
  7. 2024CoCoNO: Attention Contrast-and-Complete for Initial Noise Optimization in Text-to-Image Synthesis
  8. 2026A Decomposable Probe for Few-Step Diffusion Models: Prompt, Latent, and Score Selectivity across Backbone Families and Distillation Paradigms