Research map / Image generation / Fast diffusion sampling
Any-Order Flexible Length Masked Diffusion
Jaeyeon Kim, Lee Cheuk Kit, Carles Domingo-Enrich, Yilun Du, Sham M. Kakade, Timothy Ngotiaoco, Sitan Chen, Michael Samuel Albergo
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 changed | Where | Measured | Effect |
|---|---|---|---|
| model variable length with insertion expectation (FlexMDM) instead of padded fixed-length MDM | Masked diffusion from scratch (175M, OpenWebText) · OpenWebText length distribution | length distribution fidelity | Improved |
| use FlexMDM instead of MDM | 175M model, OpenWebText · OpenWebText generative perplexity | generative perplexity | No clear effect |
| use FlexMDM instead of MDM on subgoal maze planning | Masked diffusion from scratch · Maze planning, Hard (K=12) | success rate (%): 24.2% -> 90.0% | Improved |
| use FlexMDM instead of MDM on subgoal maze planning | Masked diffusion from scratch · Maze planning, Easy (K=2) | success rate (%): 68.4% -> 92.3% | Improved |
| retrofit LLaDA-8B into FlexMDM, then instruction fine-tune | LLaDA-Base 8B · GSM8K | accuracy (%): 58% -> 67% | Improved |
| retrofit LLaDA-8B into FlexMDM, then instruction fine-tune | LLaDA-Base 8B · HumanEval-infill (single line) | pass rate (%): 52% -> 65% | Improved |
| increase number of sampling steps | FlexMDM 8B vs IFT-ed LLaDA · GSM8K / code infilling | accuracy | Mixed |
| use adaptive (confidence-based) unmasking in place of vanilla CTMC inference | FlexMDM · downstream tasks | performance | Improved |
Closest papers by meaning
- Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions (ICML 2025 · similarity 0.91)
- Accelerated Sampling from Masked Diffusion Models via Entropy Bounded Unmasking (NeurIPS 2025 · similarity 0.90)
- Learn from Your Mistakes: Self-Correcting Masked Diffusion Models (NeurIPS 2026 · similarity 0.89)
- Simplified and Generalized Masked Diffusion for Discrete Data (NeurIPS 2024 · similarity 0.89)
- Beyond Masked and Unmasked: Discrete Diffusion Models via Partial Masking (NeurIPS 2025 · similarity 0.89)
- Looped Diffusion Language Models (NeurIPS 2026 · similarity 0.88)
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.
