AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Planner Aware Path Learning in Diffusion Language Models Training

Fred Zhangzhi Peng, Zachary Bezemek, Jarrid Rector-Brooks, Shuibai Zhang, Michael M. Bronstein, Anru Zhang, Joey Bose, Alexander Tong

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Masked diffusion language model decoding

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
train with PAPL instead of standard masked diffusion lossDLM-150M protein model · protein generationfoldability (%): 42.43 -> 59.40Improved
train with PAPL instead of standard lossDLM-150M protein model · protein generationdiversity (%): 92.45 -> 91.73No clear effect
train with PAPL instead of standard MDLM lossMDLM on OpenWebText · OWT unconditional, T=32generative perplexity: 100.45 -> 40.19Improved
train with PAPL instead of standard MDLM lossMDLM on OpenWebText · OWT unconditional, T=128MAUVE: 0.015 -> 0.067Improved
train with PAPL instead of standard lossDLM 0.5B code model · HumanEvalpass@10: 31.1 -> 38.4Improved
train with PAPL instead of standard lossDLM 0.5B code model · MBPPpass@1: 17.6 -> 16.7Hurt
train with PAPL instead of standard lossDLM 0.5B code model · MBPP+pass@10: 32.6 -> 53.6Improved
train with PAPL instead of standard lossDLM 0.5B code model · HumanEval-Infillpass@1: 30.0 -> 32.5Improved
lower softmax temperature tau below default 1 (ablation)PAPL · protein foldabilityfoldabilityImproved
increase path-learning weight alpha up to 5 (ablation)PAPL · protein foldabilityfoldabilityImproved
increase alpha beyond 5 (ablation)PAPL · protein foldabilityfoldabilityHurt
interpolate planner-weighted loss with vanilla DLM loss (ablation)PAPL · training stabilitytraining varianceImproved

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