REVIEW 32 cited by
Aligning Text-to-Image Diffusion Models with Reward Backpropagation
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Text-to-image diffusion models have recently emerged at the forefront of image generation, powered by very large-scale unsupervised or weakly supervised text-to-image training datasets. Due to their unsupervised training, controlling their behavior in downstream tasks, such as maximizing human-perceived image quality, image-text alignment, or ethical image generation, is difficult. Recent works finetune diffusion models to downstream reward functions using vanilla reinforcement learning, notorious for the high variance of the gradient estimators. In this paper, we propose AlignProp, a method that aligns diffusion models to downstream reward functions using end-to-end backpropagation of the reward gradient through the denoising process. While naive implementation of such backpropagation would require prohibitive memory resources for storing the partial derivatives of modern text-to-image models, AlignProp finetunes low-rank adapter weight modules and uses gradient checkpointing, to render its memory usage viable. We test AlignProp in finetuning diffusion models to various objectives, such as image-text semantic alignment, aesthetics, compressibility and controllability of the number of objects present, as well as their combinations. We show AlignProp achieves higher rewards in fewer training steps than alternatives, while being conceptually simpler, making it a straightforward choice for optimizing diffusion models for differentiable reward functions of interest. Code and Visualization results are available at https://align-prop.github.io/.
Forward citations
Cited by 32 Pith papers
-
Flow-GRPO: Training Flow Matching Models via Online RL
Flow-GRPO is the first online RL method for flow matching models, raising GenEval accuracy from 63% to 95% and text-rendering accuracy from 59% to 92% with little reward hacking.
-
DiT-Reward: Generative Representations for Text-to-Image Reward Modeling
DiT-Reward converts pretrained DiT models into reward predictors that outperform HPSv3 on four benchmarks while providing 1.65x inference speedup.
-
The Reward Was in Your Data All Along: Correcting Flow Matching with Discriminator-Guided RL
DRL trains a discriminator on data versus base-model samples in pretrained representation space and uses its logit as reward in KL-regularized RL, cutting guidance-free FID from 9.38 to 2.62 on SiT and similar gains o...
-
RankE: End-to-End Post-Training for Discrete Text-to-Image Generation with Decoder Co-Evolution
RankE co-evolves AR policy and decoder via alternating ranking optimization, improving both FID and CLIP scores on LlamaGen-XL and Janus-Pro where policy-only RL degrades FID.
-
Efficient Adjoint Matching for Fine-tuning Diffusion Models
EAM speeds up adjoint matching for diffusion model reward fine-tuning by switching to linear base drift, allowing deterministic few-step solvers and closed-form adjoints with up to 4x faster convergence on text-to-ima...
-
Efficient Adjoint Matching for Fine-tuning Diffusion Models
EAM reformulates adjoint matching for diffusion fine-tuning with linear base drift to allow efficient deterministic sampling and closed-form adjoints while matching or exceeding prior performance.
-
Reinforce Adjoint Matching: Scaling RL Post-Training of Diffusion and Flow-Matching Models
Reinforce Adjoint Matching derives a simple consistency loss for RL post-training of diffusion models by tilting the clean distribution toward higher-reward samples under KL regularization while keeping the noising pr...
-
Personalizing Text-to-Image Generation to Individual Taste
PAMELA provides a multi-user rating dataset and personalized reward model that predicts individual image preferences more accurately than prior population-level aesthetic models.
-
Beyond VLM-Based Rewards: Diffusion-Native Latent Reward Modeling
DiNa-LRM introduces a diffusion-native latent reward model using a noise-calibrated Thurstone likelihood on noisy states, matching VLM performance at lower compute in image alignment and preference optimization.
-
WorldCycle: Self-Verifiable Reinforcement Learning for Long-Horizon Video World Models
WorldCycle post-trains interactive video world models with reinforcement learning rewards for spatial closure and temporal consistency on reversible action cycles, reducing long-horizon drift and improving composite-a...
-
DanceOPD: On-Policy Generative Field Distillation
Hard-routed, single low-noise on-policy velocity matching composes conflicting image-generation capabilities into one flow student better than joint training, merging, or dense OPD baselines.
-
DiffusionBench: On Holistic Evaluation of Diffusion Transformers
NanoGen unifies DiT training on ImageNet and T2I, reveals negative Pearson correlations (-0.377 to -0.580) in method rankings across metrics from 21 models, and motivates DiffusionBench for holistic evaluation.
-
STAR: SpatioTemporal Adaptive Reward Allocation for Text-to-Image RL Post-Training
STAR uses text-image attention to create dynamic spatial allocation maps that vary across denoising steps and applies the same advantage more strongly to relevant latent regions in RL post-training of diffusion models.
-
Plug-and-Play Guidance for Discrete Diffusion Models via Gradient-Informed Logit Correction
Introduces GILC, a training-free plug-and-play guidance framework for discrete diffusion models that uses Jacobian-free logit correction to achieve SOTA results on DNA, protein, and molecular generation tasks.
-
Plug-and-Play Guidance for Discrete Diffusion Models via Gradient-Informed Logit Correction
GILC guides discrete diffusion at inference time by adding reward gradients to the prediction logits, matching or beating fine-tuned baselines on DNA, protein, and molecule generation tasks without retraining.
-
Parallel Tempering Initial Sampling in Inference-Time Reward Alignment
PATHS applies parallel tempering to improve initial particle sampling for SMC reward alignment, yielding better results on layout-to-image and quantity-aware generation tasks.
-
AdvantageFlow: Advantage-Weighted Least Squares for RL in Flow Models
AdvantageFlow proposes an advantage-weighted forward-process least-squares loss for RL in rectified flow models, stabilized by rollout policy regularization, and reports better image generation performance than Flow-G...
-
Reinforce Adjoint Matching: Scaling RL Post-Training of Diffusion and Flow-Matching Models
Derives RAM, a reward-adjusted consistency loss extending diffusion pretraining regression to efficient KL-regularized RL post-training, achieving peak rewards up to 50x faster than Flow-GRPO on Stable Diffusion 3.5M.
-
Adjoint Matching through the Lens of the Stochastic Maximum Principle in Optimal Control
Adjoint Matching is derived from the Stochastic Maximum Principle as a Hamiltonian objective whose critical points solve stochastic optimal control problems for generative dynamics.
-
Adjoint Matching through the Lens of the Stochastic Maximum Principle in Optimal Control
Adjoint matching objectives derived from the Stochastic Maximum Principle have critical points satisfying HJB stationarity conditions for SOC problems with control-dependent drift and diffusion.
-
Direct Diffusion Score Preference Optimization via Stepwise Contrastive Policy-Pair Supervision
Diffusion image models can be aligned without human labels by supervising every denoising step with score targets from original versus degraded prompts.
-
Test-Time Alignment of Text-to-Image Diffusion Models via Null-Text Embedding Optimisation
Optimizing the null-text embedding in classifier-free guidance aligns diffusion outputs to a target reward while preserving cross-reward quality.
-
A novel method and dataset for depth-guided image deblurring from smartphone Lidar
Lidar depth-guided deblurring via a zero-shot diffusion method, evaluated on a new 45-scene dataset, achieves the best perceptual quality (LPIPS).
-
ShortFT: Diffusion Model Alignment via Shortcut-based Fine-Tuning
ShortFT fine-tunes Stable Diffusion by backpropagating reward gradients through a distilled few-step shortcut denoising chain, improving alignment scores over DRaFT-LV and DRTune.
-
Improving Video Generation with Human Feedback
A human preference dataset and VideoReward model enable Flow-DPO and Flow-NRG to produce smoother, better-aligned videos from text prompts in flow-based generators.
-
FlowAWR: Online Adaptive Flow Reinforcement via Advantage-Weighted Rectification
FlowAWR derives an advantage-weighted rectification for optimal velocity fields in flow models, claiming 2-5x faster convergence than DiffusionNFT on SD3.5-Medium.
-
DanceOPD: On-Policy Generative Field Distillation
DanceOPD routes samples across capability velocity fields in flow-matching models and trains via on-policy student-induced states to compose T2I, local editing, and global editing without mutual interference.
-
Visual Generation in the New Era: An Evolution from Atomic Mapping to Agentic World Modeling
Visual generation models are evolving from passive renderers to interactive agentic world modelers, but current systems lack spatial reasoning, temporal consistency, and causal understanding, with evaluations overemph...
-
Dichotomous Diffusion Policy Optimization
DIPOLE decomposes a KL-regularized RL objective into a pair of sigmoid-weighted diffusion policies whose score combination (CFG-like) yields stable and controllable policy improvement.
-
PSR: Scaling Multi-Subject Personalized Image Generation with Pairwise Subject-Consistency Rewards
A data-generation pipeline plus pairwise subject-consistency rewards in RL improve consistency and prompt adherence for multi-subject personalized image generation.
-
Instant Preference Alignment for Text-to-Image Diffusion Models
An MLLM-driven, training-free pipeline extracts preference keywords from a reference image and modulates diffusion cross-attention at global and regional levels for instant, multi-round preference-aligned image generation.
-
Alignment and Safety of Diffusion Models via Reinforcement Learning and Reward Modeling: A Survey
A literature survey that organizes diffusion model alignment methods along five axes (feedback source, reward form, optimization mechanism, distribution shift handling, and explicit safety constraints) and identifies ...
Discussion (0). Sign in to comment.