iiinova — where your digital self keeps evolving

Replies

Bill Psomas @billpsomas.bsky.social · 27/12/2025

9/n Alignment effects ⚓ External alignment complements joint modeling, but its benefits depend on the signal. Local alignment yields consistent gains, whereas global-only alignment can degrade performance. Spatial joint modeling remains the primary driver.

110

Bill Psomas @billpsomas.bsky.social · 27/12/2025

10/n Faster convergence🔥 REGLUE (SiT-B/2) achieves 12.9 and 28.7 FID at 400K iterations in conditional and unconditional generation, respectively, outperforming REPA, ReDi, and REG. REGLUE (SiT-XL/2) matches 1M-step SOTA performance in just 700k iterations (~30% fewer steps).

100

Bill Psomas @billpsomas.bsky.social · 27/12/2025

11/n Summary🏁 REGLUE shows that the way we leverage VFM semantics matters for diffusion. Combining compact local semantics with global context yields faster convergence and state-of-the-art image generation. 📄arXiv: arxiv.org/abs/2512.16636 💻Project: reglueyourlatents.github.io

arxiv.org

REGLUE Your Latents with Global and Local Semantics for Entangled Diffusion

Latent diffusion models (LDMs) achieve state-of-the-art image synthesis, yet their reconstruction-style denoising objective provides only indirect semantic supervision: high-level semantics emerge slo...

110