AI Paper Atlas

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

Harmonious Parameter Adaptation in Continual Visual Instruction Tuning for Safety-Aligned MLLMs

Ziqi Wang, Chang Che, Qi Wang, Hui Ma, Zenglin Shi, Cees G. M. Snoek, Meng Wang

CVPR 2026 · Continual learning › Continual fine-tuning of LLMs · approach: Mixture-of-experts for multimodal continual tuning · 4 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
retain top-p% safety-focused parameters from the previous model (ablation)HPA (with 0.1% harmful data) · CVIT 6 tasks, 0.1% harmful data injectedMASR (%): 58.22 -> 6.02Improved
retain safety-focused parameters at shared-focused positions only (ablation)HPA · CVIT 6 tasks, 0.1% harmful data injectedMASR (%): AP 66.69 -> 74.16; MASR 58.22 -> 11.51Mixed
combine safety-focused retention with balanced shared-position selection (ablation)HPA · CVIT 6 tasks, 0.1% harmful data injectedAP (%): 73.49 -> 74.82Improved
add orthogonal parameter adjustment (ablation)HPA · CVIT 6 tasks, 0.1% harmful data injectedBWT (%): -7.00 -> -3.88Improved
add orthogonal parameter adjustment (ablation)HPA · CVIT 6 tasks, 0.1% harmful data injectedAP (%): 74.82 -> 76.62Improved
add orthogonal parameter adjustment (ablation)HPA · CVIT 6 tasks, 0.1% harmful data injectedMASR (%): 9.67 -> 7.22Improved
add safety-focused retention (epsilon) in the full framework (ablation)HPA · CVIT 6 tasks, 0.1% harmful data injectedAP (%): 66.69 -> 73.49Improved
increase fixed retention rate p (ablation)HPA · CVIT 6 taskssafety vs task performanceMixed
use layer-wise dynamic retention ratio instead of a fixed p (ablation)HPA · CVIT 6 taskssafety/task trade-offImproved
increase a fixed alpha too much (ablation)HPA · CVIT 6 tasksoverall performanceHurt
use adaptive per-layer alpha instead of fixed alpha (ablation)HPA · CVIT 6 taskssafety/task trade-offImproved
HPA vs Safe Delta under 0.1% harmful data injectionHPA · CVIT 6 tasks, 0.1% harmful data injectedMASR (%): 24.26 -> 7.22Improved

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