List your company

Image Classification on SVHN (test) (Top-1 and Top-5 Accuracy)

97.3Top-1 AccuracyFine-tuned (STL)
15.909637.039858.1779.3002Apr 25, 2026Apr 28, 2026May 1, 2026May 4, 2026May 7, 2026May 10, 2026May 13, 2026

Evaluation results

MethodLinks
Fine-tuned (STL)Backbone=CLIP-ViT-B/32, Evaluation Protocol=Fine-tuned (STL)2026.0597.3—
WUDI-Merging w/ FEATCALBackbone=CLIP-ViT-B/32, Upstream Merger=WUDI-Merging, Refinement=FEATCAL2026.0596.7—
WUDI-Merging w/ SurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=WUDI-Merging, Refinement=Surgery2026.0595.6—
WUDI-Merging w/ ProbSurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=WUDI-Merging, Refinement=ProbSurgery2026.0595.6—
Traditional MTLBackbone=CLIP-ViT-B/32, Evaluation Protocol=Traditional MTL2026.0595.5—
Task Arithmetic w/ FEATCALBackbone=CLIP-ViT-B/32, Upstream Merger=Task Arithmetic, Refinement=FEATCAL2026.0595—
Ours (DS-7B) SVHNSearch Cost=22 cycles, Generation=Delta, Training Protocol=LLM-Based NAS on LEMUR (fine-tuned, 1-epoch proxy)2026.0594.9—
Ours (Qwen-7B) SVHNSearch Cost=22 cycles, Generation=Delta, Training Protocol=LLM-Based NAS on LEMUR (fine-tuned, 1-epoch proxy)2026.0594.9—
Ours (Mistral-7B) SVHNSearch Cost=22 cycles, Generation=Delta, Training Protocol=LLM-Based NAS on LEMUR (fine-tuned, 1-epoch proxy)2026.0594.9—
WUDI-MergingBackbone=CLIP-ViT-B/32, Upstream Merger=WUDI-Merging2026.0594.8—
AdaMerging w/ FEATCALBackbone=CLIP-ViT-B/32, Upstream Merger=AdaMerging, Refinement=FEATCAL2026.0594.2—
Simple Averaging w/ FEATCALBackbone=CLIP-ViT-B/32, Upstream Merger=Simple Averaging, Refinement=FEATCAL2026.0592.6—
AdaMerging w/ SurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=AdaMerging, Refinement=Surgery2026.0590.5—
AdaMerging w/ ProbSurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=AdaMerging, Refinement=ProbSurgery2026.0590.2—
AdaMergingBackbone=CLIP-ViT-B/32, Upstream Merger=AdaMerging2026.0587.6—
Task Arithmetic w/ ProbSurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=Task Arithmetic, Refinement=ProbSurgery2026.0585.3—
Khalid v7 SVHNSearch Cost=22 cycles, Generation=Full code, Training Protocol=LLM-Based NAS on LEMUR (fine-tuned, 1-epoch proxy)2026.0584.5—
Task Arithmetic w/ SurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=Task Arithmetic, Refinement=Surgery2026.0583.6—
Task ArithmeticBackbone=CLIP-ViT-B/32, Upstream Merger=Task Arithmetic2026.0577.9—
Simple Averaging w/ SurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=Simple Averaging, Refinement=Surgery2026.0570.8—
Simple Averaging w/ ProbSurgeryBackbone=CLIP-ViT-B/32, Upstream Merger=Simple Averaging, Refinement=ProbSurgery2026.0570.4—
Simple AveragingBackbone=CLIP-ViT-B/32, Upstream Merger=Simple Averaging2026.0564.5—
Pre-trainedBackbone=CLIP-ViT-B/32, Evaluation Protocol=Zero-shot2026.0531.6—
MetaErrLabeled Training Data=10%2026.0422.8370.13
VATLabeled Training Data=10%2026.0421.1354.04
ICTLabeled Training Data=10%2026.0420.4167.21
LPLabeled Training Data=10%2026.0419.7164.52
CLLabeled Training Data=10%2026.0419.4764.24
PL-CBLabeled Training Data=10%2026.0419.0462.88