Research map / Large language models / Language, safety and interpretability
Scheduling Thoughts: Learning the Order of Thought in Diffusion Language Models
Jiawei Xu, Minghui Liu, Aakriti Agrawal, Yifan Chen, Furong Huang
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 learned order policy instead of confidence ordering | SMDM-1B Sudoku decoding · Sudoku test | puzzle accuracy (%): 82.0 -> 91.8 | Improved |
| use learned order policy instead of human expert order | SMDM-1B Sudoku decoding · Sudoku test | puzzle accuracy (%): 76.6 -> 91.8 | Improved |
| use confidence ordering instead of random | SMDM-1B Sudoku decoding · Sudoku test | puzzle accuracy (%): 62.7 -> 82.0 | Improved |
| second-stage fine-tuning of diffusion head along learned order (ablation) | SAS order policy · Sudoku test | puzzle accuracy (%): 91.8 -> 97.5 | Improved |
| second-stage fine-tuning along human expert order (ablation) | Human order · Sudoku test | puzzle accuracy (%): 76.6 -> 97.8 | Improved |
| replace default LLaDA schedule with learned order policy | LLaDA-8B-Instruct · GSM8K | pass@1 (%): 64% -> 76% | Improved |
| replace default schedule with learned order policy | LLaDA-8B-Instruct · MBPP | pass@1 (%): 39.5% -> 41% | Improved |
| transfer order policy trained on code to math, full diffusion 256-256 | SAS policy · GSM8K | accuracy: +8.5% | Improved |
| transfer order policy trained on math to code, 256-256 | SAS policy · MBPP | accuracy: +1.5% | Improved |
| transfer order policy trained on math to code, 256-32 | SAS policy · MBPP | accuracy: +0.33% | Improved |
| dense self-aware reward instead of 0/1 terminal reward for order policy (ablation) | SAS order policy · GSM8K (5-shot) | accuracy | Mixed |
Closest papers by meaning
- The Flexibility Trap: Rethinking the Value of Arbitrary Order in Diffusion Language Models (ICML 2026 · similarity 0.91)
- Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions (ICML 2025 · similarity 0.90)
- From Next-Token to Next-Block: A Principled Adaptation Path for Diffusion LLMs (NeurIPS 2026 · similarity 0.90)
- UnMaskFork: Test-Time Scaling for Masked Diffusion via Deterministic Action Branching (ICML 2026 · similarity 0.90)
- Latent Refinement Decoding: Enhancing Diffusion Language Models by Refining Belief States (NeurIPS 2026 · similarity 0.89)
- Efficient Test-Time Scaling via Hierarchical Search and Self-Verification for Discrete Diffusion Language Models (ICML 2026 · 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.
