AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

Adaptive Protein Tokenization

Rohit Dilip, Ayush Varshney, David Van Valen

ICML 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Generative protein structure design · 0 citations (OpenAlex)

Read the paper

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
drop the token tail (keep 16 instead of 128 tokens) (ablation)APT autoregressive generation · Unconditional generationdesignability: 0.232 -> 0.482Improved
add classifier annealing (w=0.8) to 16-token generation (ablation)APT · Unconditional generationdesignability: 0.482 -> 0.771Improved
entropy-based tail cutoff instead of fixed cutoff, without classifier annealing (ablation)APT · Unconditional generationdesignability: 0.482 -> 0.494No clear effect
entropy-based cutoff combined with classifier annealing (ablation)APT · Unconditional generationdesignability: 0.771 -> 0.871Improved
strengthen classifier annealing to w=0.3 (ablation)APT · Unconditional generationdesignability / TM to unannealed: designability 0.462 -> 0.669; TM 0.93 -> 0.15Mixed
keep more tokens in generation (ablation)APT · Unconditional generationdesignability vs gFIDMixed
scale FSQ codebook from 1000 to 4000+ (ablation)APT tokenizer · ReconstructionRMSDImproved
use wider (384) and deeper (4 layer) encoder (ablation)APT 4k codebook · ReconstructionRMSD and flow lossHurt
train AR model on tokenizer with 128 vs 64 max tokens (ablation)APT AR model · Unconditional generationdesignabilityNo clear effect
use MLP probe instead of linear probe on APT tokens (ablation)APT representations · CATH-T classificationtop-1 accuracyImproved
lower flow loss during training (ablation)APT training · ReconstructionRMSDHurt

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