AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Information Estimation with Discrete Diffusion

Alberto Foresti, Giulio Franzese, Pietro Michiardi

ICLR 2026 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Mutual information estimation · 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
use INFO-SEDD instead of embedding-trick neural estimators on high-dimensional discrete synthetic dataMI estimation · High-dim synthetic MI (MI=50, D=50)MI estimate (true 50): 47.77 vs best baseline 32.60 (MINDE)Improved
use INFO-SEDD in low-dimensional synthetic benchmarkMI estimation · Low-dim synthetic MI (D=10)MI estimateNo clear effect
use INFO-SEDD instead of GAN-DIME/SMILE etc. for training convergenceMI estimators · High-dim syntheticepochs to convergeImproved
use INFO-SEDD instead of variational estimators for memoryMI estimators · High-dim synthetic (MI=50, D=50)max memory (MB): 3180.49 -> 1666.30Improved
use INFO-SEDD instead of MINDE for runtimeMI estimators · High-dim synthetic (MI=50, D=50)runtime per epoch (s): 4.68 -> 7.06Hurt
use INFO-SEDD-C instead of INFO-SEDD-J for label-vs-DNA MI (ablation)INFO-SEDD · Human vs. worm consistency testMI consistencyImproved
use INFO-SEDD instead of variational estimators on text consistency testMI estimation · SummEval consistency testMI consistency with rhoImproved

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