AI Paper Atlas

Research map / Continual learning / Continual fine-tuning of LLMs

On Token's Dilemma: Dynamic MoE with Drift-Aware Token Assignment for Continual Learning of Large Vision Language Models

Chongyang Zhao, Mingsong Li, Haodong Lu, Dong Gong

CVPR 2026 · Continual learning › Continual fine-tuning of LLMs · approach: Mixture-of-experts for multimodal continual tuning · 1 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 load-balancing auxiliary loss (ablation)IncMoELoRA · CoINMFN: 49.68 -> 50.76Improved
add Token Assignment Guidance (TAG) (ablation)IncMoELoRA + Laux · CoINMFN: 50.76 -> 54.44Improved
add Token Assignment Guidance (TAG) (ablation)IncMoELoRA + Laux · CoINBWT: -15.44 -> -7.04Improved
add exclusivity loss Lexc (ablation)IncMoELoRA + Laux + TAG · CoINMFN: 54.44 -> 55.18Improved
add specialization loss Lspe (ablation)IncMoELoRA + Laux + TAG + Lexc · CoINMAA: 54.38 -> 57.70Improved
raise ambiguity threshold from 20% to 50% (ablation)LLaVA-DyMoE · CoINMFN: 57.03 -> 55.32Hurt
lower ambiguity threshold from 20% to 10% (ablation)LLaVA-DyMoE · CoINMFN: 57.03 -> 56.87No clear effect
vary loss weight alpha from 1e-3 over 5e-4 to 1e-2 (ablation)LLaVA-DyMoE · CoINMFN: 57.03 -> 55.43 at 1e-2No clear effect
expand experts per task (IncMoELoRA) instead of static LoRA/MoELoRAMoELoRA · CoINMFN: 43.93 -> 49.68Improved
add LLaVA-DyMoE to ASD data paradigm instead of SEFESEFE · CoINBWT: -10.45 -> -4.75Improved
add LLaVA-DyMoE to ASD data paradigm instead of SEFESEFE · CoINMAA: 63.04 -> 62.26No clear effect
combine LLaVA-DyMoE with replay buffer instead of ProgLoRAProgLoRA · CoINBWT: -6.59 -> -1.55Improved

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.

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

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