{"slug": "compression-asymmetry-and-trajectory-binding-in-noise-anchored-diffusion", "title": "Compression Asymmetry and Trajectory Binding in Noise-Anchored Diffusion Inversion", "summary": "Researchers identified two mechanisms behind effective stored-noise diffusion inversion: element-wise compression asymmetry, where int8 full-dimensional anchors preserve reconstruction but low-dimensional subspace summaries fail, and trajectory binding, where both the matched forward anchor and a trained score network are necessary. They introduced Noise-Anchored Reverse Correction (NARC), a training-free inversion method that stores a single int8 latent anchor and uses a noise-level-dependent anchor-weight schedule, achieving +3.24 dB PSNR improvement over PnP DirectInv on PIE-Bench++ with Stable Diffusion 1.5 while using 400x less storage.", "body_md": "arXiv:2607.09784v1 Announce Type: new\nAbstract: Real-image diffusion inversion is governed by a tight quality-cost trade-off, with costs incurred in computation, storage, or per-image optimization. We study this trade-off through the forward Gaussian noise anchor that defines a diffusion trajectory and isolate two mechanisms behind effective stored-noise inversion. First, diffusion noise exhibits an element-wise compression asymmetry: int8 full-dimensional anchors preserve reconstruction, whereas low-dimensional subspace summaries are much less reliable, often collapsing even at comparable or smaller payloads; the element-wise over subspace ordering persists across five stored-noise inversion methods. Second, inversion is trajectory-bound and score-prior coupled: the matched forward anchor and a trained score network are both necessary, arguing against a purely algebraic-identity explanation. Together, these findings specify what to store and how to use it. They lead to Noise-Anchored Reverse Correction (NARC), a training-free inversion primitive that stores a single int8 latent anchor and reuses it with a fixed, noise-level-dependent anchor-weight schedule: strong anchoring when the reverse trajectory is noise-dominated, then relaxed anchoring as image detail emerges. On PIE-Bench++ with Stable Diffusion 1.5, NARC outperforms five modern non-exact baselines and improves PSNR by +3.24 dB over PnP DirectInv while using about 400x less inversion storage than PnP DirectInv. The compression asymmetry, anchor specificity, and editing plug-in also transfer to SDXL 1024^2.", "url": "https://wpnews.pro/news/compression-asymmetry-and-trajectory-binding-in-noise-anchored-diffusion", "canonical_source": "https://arxiv.org/abs/2607.09784", "published_at": "2026-07-14 04:00:00+00:00", "updated_at": "2026-07-14 04:03:48.146059+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "computer-vision", "ai-research"], "entities": ["Stable Diffusion", "PIE-Bench++", "SDXL", "PnP DirectInv", "Noise-Anchored Reverse Correction"], "alternates": {"html": "https://wpnews.pro/news/compression-asymmetry-and-trajectory-binding-in-noise-anchored-diffusion", "markdown": "https://wpnews.pro/news/compression-asymmetry-and-trajectory-binding-in-noise-anchored-diffusion.md", "text": "https://wpnews.pro/news/compression-asymmetry-and-trajectory-binding-in-noise-anchored-diffusion.txt", "jsonld": "https://wpnews.pro/news/compression-asymmetry-and-trajectory-binding-in-noise-anchored-diffusion.jsonld"}}