AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Discrete Diffusion with Physical Mass Constraints for \emph{De Novo} Peptide Sequencing

An Zeyu, Wanyu Lin

ICML 2026 · Image generation › Autoregressive and latent generation · approach: Latent diffusion with learned tokenizers

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
add MS1 hard knapsack mass constraint (ablation)PhysNovo · Nine-speciespeptide precision: 0.577 -> 0.588Improved
add MS2 spectral soft constraint (cross-attention + remasking) (ablation)PhysNovo · Nine-speciespeptide precision: 0.583 -> 0.588Improved
replace one-shot NAR decoding with iterative discrete diffusionπ-PrimeNovo · Nine-speciesamino acid precision: 0.790 -> 0.815Improved
reduce denoising steps from T=100 to T=50 (ablation)PhysNovo · Nine-speciesspeed / amino acid precision: 46 -> 99 spectra/s; 0.815 -> 0.801Mixed
raise remasking threshold tau to 0.9 (ablation)PhysNovo · Nine-speciesamino acid precision: 0.815 -> 0.799Hurt
raise length loss weight lambda to 1.5 (ablation)PhysNovo · Nine-speciesamino acid precision: 0.815 -> 0.808Hurt
lower length loss weight lambda to 0.1 (ablation)PhysNovo · Nine-speciesamino acid precision: 0.815 -> 0.810Hurt
use physics-constrained diffusion instead of LIPNovoLIPNovo · Seven-species (low resolution)peptide precision: 0.327 -> 0.346Improved
use physics-constrained diffusion instead of ReNovo on HC-PT amino acid precisionReNovo · HC-PTamino acid precision: 0.651 vs 0.649No clear effect
replace LIPNovo with PhysNovo in leave-one-out cross-validationLIPNovo · Nine-species leave-one-outmean PTM precision: 0.790 -> 0.816Improved
use oracle length predictor (ablation)PhysNovo · Nine-speciespeptide precision: 0.588 -> 0.624Improved

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