AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Memory-Free Continual Learning with Null Space Adaptation for Zero-Shot Vision-Language Models

Yujin Jo, Taesup Kim

ICLR 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Adapting VLMs for zero-shot classification

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
make null-space bases U_n and V_n trainable instead of freezing them (ablation)NuSA-CL · MTIL full-shotLast: 82.79 -> 77.32Hurt
adapt both text and vision encoders instead of one (ablation)NuSA-CL · MTIL full-shotAvg.: 72.62 / 70.49 -> 75.08Improved
update Tail (null-like) directions instead of Top or Random (ablation)NuSA-CL · MTILforgetting: 4.44% / 4.57% -> 2.57%Improved
increase update rank from 32 to 256 (ablation)NuSA-CL · MTILforgetting and target accuracy: forgetting 1.46% -> 5.11%Mixed
raise energy cutoff rho to 0.999 (ablation)NuSA-CL · MTIL full-shotLast: 82.79 -> 79.16Hurt
vary energy cutoff rho between 0.80 and 0.99 (ablation)NuSA-CL · MTIL full-shotAvg.: 74.87 to 75.08No clear effect
use data-agnostic SVD null space instead of InfLoRA gradient projectionNuSA-CL · MTILinit time and Avg.: ~81 min -> <1 min; Avg 74.2 -> 75.1Improved
use null-space LoRA instead of standard LoRACLIP continual learning · MTIL full-shotAvg.: 70.1 -> 75.1Improved
initialize LoRA in minor singular components (MiLoRA) without persistent constraintLoRA · MTIL full-shotAvg.: 70.1 -> 68.7No clear effect
NuSA-CL vs storage-based MoE-AdaptersNuSA-CL · MTIL full-shotAvg. / efficiency: Avg 75.1 vs 76.7; 1.5M vs 59.8M paramsMixed

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