AI Paper Atlas

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

Little By Little: Continual Learning via Incremental Mixture of Rank-1 Associative Memory Experts

Haodong Lu, Chongyang Zhao, Jason Xue, Lina Yao, Kristen Moore, Dong Gong

ICML 2026 · Continual learning › Continual fine-tuning of LLMs · approach: Low-rank and model-merging adaptation · 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
replace learned external router with self-activation over rank-1 atoms (ablation)MoRAM · X-TAIL Lastaccuracy (%): 69.76 -> 69.85No clear effect
use rank-1 atoms with external router instead of rank-r MoE-LoRA (ablation)MoE-LoRA · X-TAIL Lastaccuracy (%): 74.53 -> 69.76Hurt
add top-k sparsity to self-activation (ablation)MoRAM · X-TAIL Lastaccuracy (%): 69.85 -> 70.62Improved
add temperature scaling (tau_MoRAM) to self-activation (ablation)MoRAM · X-TAIL Lastaccuracy (%): 69.85 -> 79.62Improved
add threshold-based selection delta (ablation)MoRAM · X-TAIL Lastaccuracy (%): 69.85 -> 71.08Improved
combine top-k, temperature and threshold (ablation)MoRAM · X-TAIL Lastaccuracy (%): 69.85 -> 80.90Improved
add temperature scaling to coarse MoE-LoRA (ablation)MoE-LoRA · X-TAIL Lastaccuracy (%): 74.53 -> 74.57No clear effect
increase rank activation budget from 2 to 16 atoms (ablation)MoRAM · 10-task checkpointaccuracy (%)Improved
use higher temperature tau_MoRAM (ablation)MoRAM · X-TAIL Transfertransfer accuracyImproved
use MoRAM instead of LoRA r=32Llama-3.1-8B on CodeAlpaca · MMLU out-of-domainaccuracy (%): 63.59 -> 63.70Improved
use MoRAM instead of TreeLoRALLaMA-2-7B-Chat on TRACE · TRACEBWT (%): 3.46 -> 1.37Improved
add continual learning with MoRAM over CLIP zero-shot (Transfer score)CLIP on X-TAIL · X-TAIL Transferaccuracy (%): 63.2 (CoDyRA) -> 63.3No clear effect

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