AI Paper Atlas

Research map / Architectures and efficiency / Mixture of experts

TEAM: Temporal–Spatial Consistency Guided Expert Activation for MoE Diffusion Language Model Acceleration

Linye Wei, Zixiang Luo, Pingzhi Tang, Meng Li

ICML 2026 · Architectures and efficiency › Mixture of experts · approach: Efficient MoE language model systems · 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
add speculative exploration for hot tokens (SEH) (ablation)vanilla SDAR 30B-A3B · HumanEvalSpeedup / APT: 1.00 -> 1.43 speedup; APT 18.33 -> 12.23Improved
add delayed caching for decoded tokens (DCD) on top of SEH (ablation)SDAR + SEH · HumanEvalSpeedup / APT: 1.43 -> 2.01 speedup; APT 12.23 -> 7.79Improved
add limited activation for cold tokens (LAC) on top of SEH+DCD (ablation)SDAR + SEH + DCD · HumanEvalSpeedup / APT: 2.01 -> 2.20 speedup; APT 7.79 -> 6.80Improved
add LAC on top of SEH+DCD (ablation)SDAR + SEH + DCD · MBPPSpeedup: 1.84 -> 2.08Improved
add full TEAM to SDARSDAR 30B-A3B · Average of 4 benchmarksScore: 77.91 -> 77.84No clear effect
add full TEAM to SDARSDAR 30B-A3B · Average of 4 benchmarksSpeedup / activated experts per forward pass: 1x -> 1.94x; APF 54.99 -> 34.48Improved
never refresh the block cache (refresh-free) instead of refreshing every 4 steps (ablation)DCD · Average of 4 benchmarksSpeedup: 1.29x -> 1.47xImproved
never refresh the block cache instead of refreshing every 8 steps (ablation)DCD · Average of 4 benchmarksScore: 77.10 -> 77.57No clear effect
narrow hot-token set from (tau_h 0.7, L_h 3) to (tau_h 0.8, L_h 2) (ablation)TEAM hot-token classification · HumanEval/MBPP averageScore: 73.10 -> 70.05Hurt
raise tau_h from 0.4 to 0.7 and lower L_h from 6 to 3 (ablation)TEAM hot-token classification · HumanEval/MBPP averageAPF: 23.35 -> 22.37Improved
construct branches from aligned top-1/top-2 candidates rather than also accepting the 3rd-confidence token (ablation)SEH · acceptance probability of candidatesacceptance rateImproved
cache decoded tokens at every iteration like dKV-Cache (for block diffusion)dLLMs with block diffusion · SDARaccelerationNo clear effect

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