# Weak-to-Strong Diffusion with Reflection

arXiv: https://arxiv.org/abs/2502.00473 · HTML: https://mlsanity.com/paper/2502.00473/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models.
- First posted on arXiv 1 Feb 2025.

## Placed in

- Diffusion models / [Diffusion foundations and sampling](https://mlsanity.com/diffusion/diffusion-foundations-and-sampling/)
- Mixture-of-experts / [Vision mixture-of-experts](https://mlsanity.com/moe/vision-mixture-of-experts/)

## Built on by (most cited first, 13 of 13)

- [NoiseAR: AutoRegressing Initial Noise Prior for Diffusion Models](https://mlsanity.com/paper/2506.01337/) (2025)
- [Noise Matters: Optimizing Matching Noise for Diffusion Classifiers](https://mlsanity.com/paper/2508.11330/) (2025)
- [FastLightGen: Fast and Light Video Generation with Fewer Steps and Parameters](https://mlsanity.com/paper/2603.01685/) (2026)
- [Guidance Matters: Rethinking the Evaluation Pitfall for Text-to-Image Generation](https://mlsanity.com/paper/2602.22570/) (2026)
- [CRAFT: Aligning Diffusion Models with Fine-Tuning Is Easier Than You Think](https://mlsanity.com/paper/2603.18991/) (2026)
- [Consistent Story Generation: Unlocking the Potential of Zigzag Sampling](https://mlsanity.com/paper/2506.09612/) (2025)
- [Improving Diffusion Generalization with Weak-to-Strong Segmented Guidance](https://mlsanity.com/paper/2603.20584/) (2026)
- [Step Back to Move Forward: Reflection-Aware Preference Optimization for Visual Generation](https://mlsanity.com/paper/2609.04282/) (2026)
- [Test-Time Weak-to-Strong Alignment: Transferring Implicit Rewards from Weak to Strong Flow Models](https://mlsanity.com/paper/2609.05968/) (2026)
- [TeEFusion: Blending Text Embeddings to Distill Classifier-Free Guidance](https://mlsanity.com/paper/2507.18192/) (2025)
- [Reflective Flow Sampling Enhancement](https://mlsanity.com/paper/2603.06165/) (2026)
- [Guiding Diffusion Models with Semantically Degraded Conditions](https://mlsanity.com/paper/2603.10780/) (2026)
- [G3-LoRA: Organizing Reward-Weighted Video Data with Gradient-Guided Grouped LoRA](https://mlsanity.com/paper/2609.35189/) (2026)

## Abstract (arXiv)

The goal of diffusion generative models is to align the learned distribution with the real data distribution through gradient score matching. However, inherent limitations in training data quality, modeling strategies, and architectural design lead to inevitable gap between generated outputs and real data. To reduce this gap, we propose Weak-to-Strong Diffusion (W2SD), a novel framework that utilizes the estimated difference between existing weak and strong models (i.e., weak-to-strong difference) to bridge the gap between an ideal model and a strong model. By employing a reflective operation that alternates between denoising and inversion with weak-to-strong difference, we theoretically understand that W2SD steers latent variables along sampling trajectories toward regions of the real data distribution. W2SD is highly flexible and broadly applicable, enabling diverse improvements through the strategic selection of weak-to-strong model pairs (e.g., DreamShaper vs. SD1.5, good experts vs. bad experts in MoE). Extensive experiments demonstrate that W2SD significantly improves human preference, aesthetic quality, and prompt adherence, achieving SOTA performance across various modalities (e.g., image, video), architectures (e.g., UNet-based, DiT-based, MoE), and benchmarks. For example, Juggernaut-XL with W2SD can improve with the HPSv2 winning rate up to 90% over the original results. Moreover, the performance gains achieved by W2SD markedly outweigh its additional computational overhead, while the cumulative improvements from different weak-to-strong difference further solidify its practical utility and deployability.