Research map / Image generation / Fast diffusion sampling
Parallel Sampling from Masked Diffusion Models via Conditional Independence Testing
Iskander Azangulov, Teodora Pandeva, Niranjani Prasad, Javier Zazo, Sushrut Karmalkar
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 |
|---|---|---|---|
| use PUNT conditional-independence-tested parallel unmasking instead of top-k, EB-sampler and Dilated samplers | Dream 7B and LLaDA 1.5 · IFEval and MTBench (long-form) | accuracy vs forward evaluations: up to 16% higher accuracy on IFEval | Improved |
| use PUNT on short-answer tasks (GSM8K, HumanEval, MBPP) | LLaDA, MBPP temperature 0.7 · MBPP pass@1 vs NFE | pass@1 | No clear effect |
| measure by denoising steps instead of NFE for short-answer tasks | PUNT on MBPP (LLaDA) · MBPP | pass@1 vs steps | Improved |
| vary the independence threshold epsilon (0.01 to 0.32) (ablation) | PUNT · MTBench 1024-token generation (LLaDA) | parallel sampling KL error | No clear effect |
| vary hyperparameters of PUNT (ablation) | PUNT · alignment benchmarks | accuracy | No clear effect |
| apply PUNT to protein generation with MemDLM | MemDLM (ESM-2 150M masked diffusion) · unconditional membrane protein generation | mean pLDDT | Improved |
Closest papers by meaning
- Masked Diffusion Models are Secretly Time-Agnostic Masked Models and Exploit Inaccurate Categorical Sampling (ICLR 2025 · similarity 0.92)
- Accelerated Sampling from Masked Diffusion Models via Entropy Bounded Unmasking (NeurIPS 2025 · similarity 0.92)
- Learn from Your Mistakes: Self-Correcting Masked Diffusion Models (NeurIPS 2026 · similarity 0.91)
- Scaling up Masked Diffusion Models on Text (ICLR 2025 · similarity 0.91)
- Beyond Masked and Unmasked: Discrete Diffusion Models via Partial Masking (NeurIPS 2025 · similarity 0.90)
- Remasking Discrete Diffusion Models with Inference-Time Scaling (NeurIPS 2025 · similarity 0.89)
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.
