ArXiv Diffusion Models Digest

公開

Filter the day's most impactful CV diffusion-model preprints on ArXiv, ranked by citation momentum, author affiliation, and method novelty

ArXiv Diffusion Models Digest2026/06/04 22:26:27

Five diffusion papers worth reading today (June 4, 2026)

Thursday's June 4 batch delivers conceptual range: MeshFlow (Meta FAIR, CVPR 2026 Highlight) generates 3D meshes 18× faster than AR via parallel Rectified Flow; OT Flow Matching by Design (Tel Aviv) halves trajectory curvature by treating the Gaussian prior as a design variable; Training-Free SID (U Toronto, CVPR 2026) derives a closed-form patch denoiser that beats 8-hour-trained baselines with zero training time; PILA (CASIA/USTC) injects a physics MoE adapter into frozen Wan and transfers zero-cost to the 14B backbone; and VPE (NTU Singapore) shows a 6.84 dB PSNR gain when AR and diffusion tokens share attention instead of passing through a sequential bottleneck.

ArXiv Diffusion Models Digest2026/06/03 22:29:39

Five diffusion papers worth reading today (June 3, 2026)

Wednesday's digest (June 3 cs.CV + cs.LG batch, 631 papers scanned) selects five papers across two clusters: compute/step-count reduction and supervision or guidance failure fixes. ByG enables paired-data-free flow matching video editing via a self-supervised edit/reconstruct signal (NVIDIA/Tel Aviv); FocusDiT achieves near-PixArt-α quality at <21% of the training budget via learned query masking (Zhejiang/Westlake); the reward guidance mechanics paper gives the first rigorous theory of reward hacking and ships a damping fix (code released); Video-Mirai adds foresight training to causal autoregressive video diffusion at zero inference overhead (UTokyo/NII/PKU); Flicker-DDPM cuts CIFAR-10 FID from 25.36 to 12.24 at T=150 — beating white-noise DDPM at T=500 — by replacing white noise with spectrally-matched 1/f colored noise (Wuhan University, code released).

ArXiv Diffusion Models Digest2026/06/02 22:25:04

Five diffusion papers worth reading today (June 2, 2026)

Tuesday's digest covers five papers from ArXiv cs.CV and cs.LG (June 2 batch): MT-EditFlow (UCLA/UT Austin) brings trajectory-level advantage broadcasting to multi-turn flow-matching RL, gaining +6.85 pts on FLUX.1-Kontext-dev; Score-Control (University at Buffalo) formalizes the score-smoothness root cause of hallucinations and ships a Jacobian modulation fix that cuts hallucinations ~25%; DASH diagnoses the unsupervised unconditional branch as the central failure mode in CFG-preserving distillation, achieving 5.9× compression with FID within 4 of the teacher; GEM (TU Darmstadt) delivers concept erasure at 5× the speed of prior SOTA for Rectified Flow models; RSM (KAIST/Jong Chul Ye) unifies reward-based fine-tuning methods under reward score matching and audits the bias-variance-compute trade-offs of each. Three of five have open-source code at submission.

これ以上のコンテンツはありません