AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Inheriting Generalizable Knowledge from LLMs to Diverse Vertical Tasks

Chang Liu, boyu shi, Xu Yang, Qiufeng Wang, Xin Geng

ICLR 2026 · Large language models › Retrieval and knowledge · approach: Knowledge editing in LLMs

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 spectral alignment (ablation)MASA 877M · DollyEvalRouge-L: 23.38 -> 24.46Improved
add spectral alignment (ablation)MASA 877M · S-NIRouge-L: 17.35 -> 18.51Improved
add adaptive scaling (ablation)MASA 877M · DollyEvalRouge-L: 23.58 -> 24.46Improved
add adaptive scaling (ablation)MASA 877M · S-NIRouge-L: 17.20 -> 18.51Improved
add adaptive scaling (ablation)MASA 877M · BoolQaccuracy: 72.14 -> 73.36Improved
increase gene matrix ratio from 3% to 23% (ablation)MASA 877M · avg of 6 tasksaverage score: 45.52 -> 46.91Improved
increase gene matrix ratio from 23% to 40% (ablation)MASA 877M · avg of 6 tasksaverage score: 46.91 -> 45.28Hurt
align gene matrices with MASA-OLMo initialization instead of random init16L-877M · Dialogue avg Rouge-LRouge-L: 17.53 -> 18.42Improved
initialize with MASA-OLMo instead of distillation12L-709M · Dialogue avg Rouge-LRouge-L: 16.88 -> 18.71Improved
initialize with MASA-DeepSeek instead of Scratch12L-267M · Language understanding avgaverage accuracy: 52.53 -> 54.40Improved
use EEP expert pruning instead of Scratch12L-267M · Language understanding avgaverage accuracy: 52.53 -> 41.85Hurt
use knowledge distillation instead of Scratch16L-877M · Dialogue avg Rouge-LRouge-L: 17.53 -> 17.06No 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