AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Plug-and-Play Guidance for Discrete Diffusion Models via Gradient-Informed Logit Correction

Hongkun Dou, Zike Chen, Fengji Li, Hongjue Li, Yue Deng

ICML 2026 · Image generation › Fast diffusion sampling · approach: Discrete diffusion solvers · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
guide the clean-prediction logits instead of the noisy state (ablation)GILC-DB · DNA diffusionmean rewardImproved
use Monte Carlo value estimation (n>1 samples) instead of deterministic estimation (ablation)GILC value function · DNA diffusion value-estimation errorL1 errorImproved
drop the model Jacobian (Jacobian-free logit correction) (ablation)GILC · DNA diffusionguidance stability: Jacobian condition number about 10^4-10^5Improved
use GILC-DB instead of DRAKES fine-tuningdiscrete diffusion DNA design · HepG2 enhancer designPred-Activity: 5.61 -> 7.04Improved
use GILC-DB instead of SVDDdiscrete diffusion DNA design · HepG2 enhancer designATAC-Acc (%): 51.9 -> 95.2Improved
use GILC-DB instead of DRAKESinverse protein folding · Success rateSuccess Rate (%): 78.6 -> 82.4Improved
use GILC-DB instead of DRAKESDNA design · HepG2 enhancer designApp-Log-Lik: -264 -> -267Hurt
use GILC-PG (policy gradient) instead of GILC-DB (ablation)GILC · HepG2 enhancer designPred-Activity: 7.04 -> 5.21 Pred-Activity; 0.900 -> 0.910 3-mer CorrMixed
use GILC-PG instead of GILC-DB (ablation)GILC · QM9 guidanceMAE: Cv 2.23 -> 1.53; ϵLUMO 988 -> 875Mixed
use GILC-DB (differentiable reward) instead of multi-trajectory baselinesDNA design · reward evaluations per stepreward calls: 20 -> 5Improved
remove KL regularization from DRAKES (ablation)DRAKES · inverse protein foldingSuccess Rate (%): 78.6 -> 34.1Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy