AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

LoKI: Low-Damage Knowledge Implanting of Large Language Models

Runyu Wang, Peng Ping, Zhengyu Guo, Xiaoye Zhang, Quan Shi, Liting Zhou, Tianbo Ji

AAAI 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning for reasoning data · 9 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 layer-balanced selection instead of global highest-attribution selection (ablation)LoKI (q=10) · LB Reranker, 6 general benchmarksavg degradation (%): 39.04% -> 8.86%Improved
use layer-balanced selection instead of global lowest-attribution selection (ablation)LoKI (q=10) · LB Reranker, 6 general benchmarksavg degradation (%): 30.48% -> 8.86%Improved
suppress highest-attribution KVA vectors vs lowest-attribution vectors (validates KVA) (ablation)Llama3.1-8B-Instruct · 6 general benchmarksavg degradation (%): 36.73% (S-H) vs 11.35% (S-L)Improved
fine-tune with LoKI (q=30) instead of DoRALlama3.1-8B-Instruct on ToolACE · 6 general benchmarksavg degradation (%): 4.92% -> 1.23%Improved
increase q from 10 to 30 (ablation)LoKI on ToolACE · BFCL V3 overalloverall accuracy (%): 56.76 -> 58.93Improved
increase q from 10 to 30 (ablation)LoKI on ToolACE · 6 general benchmarksavg degradation (%): 0.34% -> 1.23%Hurt
increase q from 5 to 30 (ablation)LoKI on LB Reranker · BEIR MAP@1 diff vs full fine-tunerelative difference (%): -3.1 -> +1.0Improved
increase q from 5 to 30 (ablation)LoKI on LB Reranker · 6 general benchmarksavg degradation (%): 6.12% -> 0.46%Mixed
add LoRA (r=32) on top of LoKI q=30 (ablation)LoKI (q=30) · ToolACE, BFCL overall and forgettingoverall accuracy / avg degradation: 58.93 -> 57.16; 1.23% -> 1.26%; params 563M -> 16MNo clear effect
use LoKI instead of full fine-tuningQwen2.5-0.5B-Instruct on LB Reranker · 6 general benchmarksavg degradation (%): 84.13% -> 0.46%Improved

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