AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Fast and Accurate Causal Parallel Decoding using Jacobi Forcing

Lanxiang Hu, Siqi Kou, Yichao Fu, Samyam Rajbhandari, Tajana Rosing, Yuxiong He, Zhijie Deng, Hao Zhang

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Efficient language transformers · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
use linear progressive noise schedule instead of random (window 8) (ablation)JF-LLM · HumanEval (10k instances)pass@1 accuracy: 82.9 -> 84.7Improved
use reverse progressive noise schedule instead of linear progressive (ablation)JF-LLM · HumanEval (10k instances)pass@1 accuracy: 84.7 -> < 80.0 (window 8)Hurt
increase progressive schedule window size from 8 to 16 (ablation)JF-LLM · HumanEval (10k instances)pass@1 accuracy: 84.7 -> 81.7Hurt
use full noise-conditioned mask rather than intra-window clean context (ablation)JF-LLM · HumanEval (100k examples)speedup: 1.9× -> 3.6×Improved
add multi-block decoding with rejection recyclingJF-LLM · HumanEval (A100)speedup: 3.86× -> 3.97×Improved
add multi-block decoding with rejection recyclingJF-LLM · MATH (A100)speedup: 3.65× -> 3.68×No clear effect
add second round of progressive distillation with larger block sizeJF-LLM · HumanEvalspeedup: further 20% speedup, minor performance degradationImproved
increase number of decoding blocks beyond 2 (ablation)JF-LLM (MR) · HumanEvalspeedupNo clear effect
replace AR-adapted dLLMs with Jacobi Forcing causal parallel decoderQwen2.5-Coder-7B-Instruct · HumanEvalspeedup: 1.77× (D2F) -> 3.86×Improved
Jacobi Forcing training vs lossless AR modelQwen2.5-Coder-7B-Instruct · HumanEvalpass@1 accuracy: 87.8 -> 83.5Hurt
Jacobi Forcing training vs EAGLE-3 speculative decodingQwen2.5-Coder-7B-Instruct · GSM8Kspeedup: 3.16× (EAGLE-3) vs 3.50× (JF-LLM)Mixed

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy