Research map / Image generation / Fast diffusion sampling
Efficient Diffusion LLMs via Temporal-Spatial Parallel Decoding and Confidence Extrapolation
Zekai Li, Ji Liu, Yiqing Huang, Ziqiong Liu, Dong Li, Emad Barsoum
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 |
|---|---|---|---|
| add TSPD trace-aware controller (ablation) | TSPD+CE · GSM8K, gen len 256 | speedup: 2.9× -> 5.0× | Improved |
| add Confidence Extrapolation (ablation) | TSPD · GSM8K, gen len 256 | speedup: 4.4× -> 5.0× | Improved |
| add CE to other controllers (ablation) | Learn2PD, Fast-dLLM, Credit Decoding · GSM8K, gen len 256 | speedup: Learn2PD 4.2× -> 4.7×; Fast-dLLM 7.8× -> 9.3×; CD 7.9× -> 8.4× | Improved |
| add CE (ablation) | TSPD · GSM8K, gen len 256 | accuracy: 79.5 -> 79.4 | No clear effect |
| add confidence feature to TSPD (ablation) | TSPD · GSM8K, gen len 256 | speedup: 2.8× -> 4.4× | Improved |
| add entropy feature to TSPD (ablation) | TSPD · GSM8K, gen len 256 | speedup: 4.1× -> 4.4× | Improved |
| add momentum feature to TSPD (ablation) | TSPD · GSM8K, gen len 256 | speedup: 4.3× -> 4.4× | Improved |
| add position feature to TSPD (ablation) | TSPD · GSM8K, gen len 256 | accuracy: 78.5 -> 79.5 | Improved |
| add position feature to TSPD (ablation) | TSPD · GSM8K, gen len 256 | speedup: 4.5× -> 4.4× | Hurt |
| use 2-layer LSTM instead of MLP or deeper LSTM (ablation) | TSPD controller · GSM8K, gen len 256 | accuracy: 2-layer MLP 78.8 -> 2-layer LSTM 79.5 | Improved |
| add Dual Cache (ablation) | TSPD+CE · GSM8K, gen len 256 | speedup: 5.0× -> 11.2× | Improved |
| add Prefix Cache (ablation) | TSPD+CE · GSM8K, gen len 256 | speedup: 5.0× -> 5.4× | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Fast-dLLM — Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding (ICLR 2026) · better on 5 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Accelerating Diffusion LLMs via Adaptive Parallel Decoding (NeurIPS 2025 · similarity 0.92)
- Denoising Time Matters: Diverse Generation in Diffusion Language Models (NeurIPS 2026 · similarity 0.91)
- Mean-Field Parallel Decoding for Discrete Diffusion Language Models (NeurIPS 2026 · similarity 0.91)
- Polestar: Drift-Aware Cache Calibration and Token Commitment for Efficient Inference of Diffusion LLMs (NeurIPS 2026 · similarity 0.91)
- Fast and Fluent Diffusion Language Models via Convolutional Decoding and Rejective Fine-tuning (NeurIPS 2025 · similarity 0.91)
- Confidence-Based Decoding is Provably Efficient for Diffusion Language Models (NeurIPS 2026 · similarity 0.91)
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.
