📊 Results:
✅ MDPO matches previous SOTA with 60× fewer updates and improves over SOTA +9.6% on MATH-500, +54.2% on Countdown when trained with the same number of gradient updates.
✅ MDPO + RCR consistently outperforms all baselines

We see this as a step toward more sampling and data-efficient DLMs.

🔹 Running Confidence Remasking (RCR) – a training-free decoding strategy that allows revising low-confidence tokens flexibly.

🔹 Masked Diffusion Policy Optimization – the first policy gradient method that optimizes MDLMs as a sequential decision-making process. MDPO exploits the nice property that MDLMs yield text completions at every inference step and optimizes the model with with intermediate-step rewards.

👉 Rigid Remasking: The remasking schedules used in inference usually 'freeze' tokens once predicted and not remasked immediately, making it impossible to revise low-confidence tokens predicted at early steps.
💡 To address these, we propose:

👉 Training–Inference Divide: MDLMs are trained to predict all randomly masked tokens in a single pass, while at inference, MDLMs follow a model-dependent, confidence-guided unmasking schedule that progressively reveals structure of the generated sequence.

Masked diffusion language models (MDLMs) are rising as powerful alternatives to autoregressive LMs, yet face two fundamental but overlooked problems:

🚀 Introducing our new paper, MDPO: Overcoming the Training-Inference Divide of Masked Diffusion Language Models.

📄 Paper: www.scholar-inbox.com/papers/He202...
arxiv.org/pdf/2508.13148
💻 Code: github.com/autonomousvi...
🌐 Project Page: cli212.github.io/MDPO/

📢 New paper CVPR 25!
Can meshes capture fuzzy geometry? Volumetric Surfaces uses adaptive textured shells to model hair, fur without the splatting / volume overhead. It’s fast, looks great, and runs in real time even on budget phones.
🔗 autonomousvision.github.io/volsurfs/
📄 arxiv.org/pdf/2409.02482

Excited to share that today our paper recommender platform www.scholar-inbox.com has reached 20k users! We hope to reach 100k by the end of the year.. Lots of new features are being worked on currently and rolled out soon.