mlsanityData to 14 Oct 2026

Performance Plateaus in Inference-Time Scaling for Text-to-Image Diffusion Without External Models

arXiv 2506.12633 ↗First posted 14 Jun 2025cs.CV cs.LGcited by 0

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

Recently, it has been shown that investing computing resources in searching for good initial noise for a text-to-image diffusion model helps improve performance. However, previous studies required external models to evaluate the resulting images, which is impossible on GPUs with small VRAM. For these reasons, we apply Best-of-N inference-time scaling to algorithms that optimize the initial noise of a diffusion model without external models across multiple datasets and backbones. We demonstrate that inference-time scaling for text-to-image diffusion models in this setting quickly reaches a performance plateau, and a relatively small number of optimization steps suffices to achieve the maximum achievable performance with each algorithm.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 12 of its 14 references that are in the map. First posted on arXiv 14 Jun 2025.

Builds on

12 of 14 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Inference-Time Scaling for Diffusion Models beyond Scaling Denoising Steps16 Jan
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  3. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 9

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  2. 2023GPT-4 Technical Reportcited by 18,396
  3. 2018Representation Learning with Contrastive Predictive Codingcited by 7,777
  4. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  5. 2024InitNO: Boosting Text-to-Image Diffusion Models via Initial Noise Optimizationcited by 123
  6. 2025From Reflection to Perfection: Scaling Inference-Time Optimization for Text-to-Image Diffusion Models via Reflection Tuningcited by 56
  7. 2023CONFORM: Contrast is All You Need For High-Fidelity Text-to-Image Diffusion Modelscited by 54
  8. 2025Reflect-DiT: Inference-Time Scaling for Text-to-Image Diffusion Transformers via In-Context Reflectioncited by 43
  9. 2024Self-Cross Diffusion Guidance for Text-to-Image Synthesis of Similar Subjectscited by 16

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Energy Scaling Laws for Diffusion Models: Quantifying Compute in Image Generation
  2. 2026Flash-BoN: Instant Drafts for Inference-Time Scaling in Diffusion Models
  3. 2026Accelerating Diffusion via Hybrid Data-Pipeline Parallelism Based on Conditional Guidance Scheduling
  4. 2026Inference-Time Scaling of Diffusion Models via Progressive Seed Pruning
  5. 2026FlashDiff: Efficient Regional Execution and Scheduling for Diffusion Model Serving
  6. 2024AsyncDiff: Parallelizing Diffusion Models by Asynchronous Denoising
  7. 2024Partially Conditioned Patch Parallelism for Accelerated Diffusion Model Inference
  8. 2023Approximate Caching for Efficiently Serving Diffusion Models