AI Paper Atlas

Research map / Continual learning / Online and replay-based

Grow-on-Demand: Sparse and Adaptive Expert Expansion for Continual Instruction Tuning

Ying Zhang, Xingyue Guo, Yu Zhao, Xuhui Sui, Baohang Zhou, Xinying Qian, Xiaojie Yuan

AAAI 2026 · Continual learning › Online and replay-based · approach: Plasticity-preserving continual reinforcement learning · 2 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 adaptive Expert Demand Detector instead of always adding an expert (ablation)GoD-MoE · TRACE (Order-1)Avg score: 56.1 -> 55.7No clear effect
use adaptive expansion instead of always adding an expert (ablation)GoD-MoE · TRACE (Order-1)trainable parameters: reduces trainable parameters by 35.3%Improved
replace static MoE-LoRA with GoD-MoE adaptive expert expansion and freezingMoE-LoRA · TRACE (Order-1)Avg score: 35.6 -> 55.7Improved
split LoRA into multiple experts (MoE-LoRA)LoRA · TRACE (Order-1)Avg score: 23.1 -> 35.6Improved
add 1% replay data to LoRALoRA · TRACE (Order-1)Avg score: 23.1 -> 50.0Improved
GoD-MoE without replay versus replay-based PMoEPMoE · TRACE (Order-1)Avg score: 51.1 -> 55.7Improved
GoD-MoE versus replay-based PMoE on forgettingPMoE · TRACE (Order-1)BWT: +12.2 -> -3.6Hurt
GoD-MoE versus second-best baseline on ScienceQAGoD-MoE · TRACE ScienceQAscore: by 9.3%Improved
GoD-MoE versus second-best baseline on MeetingBankGoD-MoE · TRACE MeetingBankscore: over 7.7%Improved
change task order from Order-1 to Order-2 and Order-3 (ablation)GoD-MoE · TRACEAvg score: 55.7 -> 56.4 and 55.0No clear effect
continual tuning on TRACE then evaluating GSMGoD-MoE · GSMaccuracy: 26.1 -> 22.1Hurt
continual tuning on TRACE then evaluating MMLUGoD-MoE · MMLUaccuracy: 46.6 -> 48.3Improved

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