AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Coevolutionary Continuous Discrete Diffusion: Make Your Diffusion Language Model a Latent Reasoner

Cai Zhou, Chenxiao Yang, Yi Hu, Chenyu Wang, Chubin Zhang, Muhan Zhang, Lester Mackey, Tommi Jaakkola, Stephen Bates, Dinghuai Zhang

ICML 2026 · Image generation › Autoregressive and latent generation · approach: Latent diffusion with learned tokenizers · 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 contextualized Qwen3-Embedding latents with joint continuous-discrete diffusion instead of discrete-only masked diffusionCCDD-MDiT (same params and FLOPs) · LM1B validationvalidation perplexity: 39.17 -> 29.22Improved
scale architecture from MDiT to MMDiT (ablation)CCDD with Qwen3-Embedding · LM1B validationvalidation perplexity: 29.22 -> 25.76Improved
use MoEDiT architecture instead of MDiT (ablation)CCDD with Qwen3-Embedding · OWT validation (Qwen-2 tokenizer)validation perplexity: 29.18 -> 21.90Improved
use MMDiT instead of MoEDiT (ablation)CCDD with Qwen3-Embedding · OWT validation (Qwen-2 tokenizer)validation perplexity: 21.90 -> 26.59Hurt
use even simple RoBERTa embeddings as continuous spaceCCDD-MoEDiT · OWT validation (GPT-2 tokenizer)validation perplexity: 27.39 -> 24.56Improved
add representation classifier-free guidance, w=1.0 (ablation)CCDD-MoEDiT · OWT generationgenerative NLL: 9.06 -> 8.38Improved
increase guidance scale from 1.0 to 1.5 (ablation)CCDD-MoEDiT · OWT generationgenerative NLL: 8.38 -> 8.25Improved
use CCDD with only 8 denoising steps versus MDLM with 256 stepsCCDD · OWT few-step generationgenerative NLLImproved
use CCDD training to reach MDLM final perplexityCCDD-MDiT · LM1B validationtraining iterations to 39.17 PPL: 1000k -> 40k iterationsImproved
use contextualized last-layer latents instead of first-layer token-wise latents (ablation)continuous diffusion on Qwen3-Embedding-0.6B · LM1B (500k steps)representation MSE vs token cross-entropyMixed
add looped depth to LT (T=2 to T=3) (ablation)looped transformer 6M · Countdownaccuracy: 60.6 -> 68.2Improved
add looped depth to CCDD (T=2 to T=3) (ablation)CCDD 6M · Countdownaccuracy: 67.8 -> 73.7Improved

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