List your company
ResearchFormerly WizWand

Pairwise relationship classification on Street-view fine-tuning dataset (test)

72F1 Score (Yes)Qwen2-VL
-2.8816.563655.44Sep 16, 2025
Updated 5mo ago

Evaluation results

MethodLinks
Qwen2-VLModel Size=7B, Fine-tuning Method=LoRA (rank 64), Fine-tuning Data=61k-Balanced-YN2025.09727065787963
Qwen2-VLModel Size=7B, Fine-tuning Method=LoRA (rank 64), Fine-tuning Data=76k-Unbalanced2025.09707267757369
Qwen2-VLModel Size=7B, Fine-tuning Method=LoRA (rank 64), Fine-tuning Data=61k-Balanced2025.09686259758053
Qwen2-VLModel Size=7B, Fine-tuning Method=LoRA (rank 64), Fine-tuning Data=76k-Unbalanced-YN2025.09686561757857
Qwen2.5-VLModel Size=7B, Fine-tuning Method=LoRA (rank 128), Fine-tuning Data=29k-YN2025.09663952789126
Qwen2.5-VLModel Size=7B, Fine-tuning Method=LoRA (rank 64), Fine-tuning Data=29k-YN2025.09655355728242
Qwen2-VLModel Size=7B, Fine-tuning Method=LoRA (rank 64), Fine-tuning Data=29k-YN2025.09645756697549
Qwen2-VLModel Size=72B, Fine-tuning Method=—, Fine-tuning Data=—2025.09625856677151
Qwen2.5-VLModel Size=72B, Fine-tuning Method=—, Fine-tuning Data=—2025.09595654646849
Phi 3.5Model Size=4.2B, Fine-tuning Method=—, Fine-tuning Data=—2025.09572445477916
Qwen2-VLModel Size=7B, Fine-tuning Method=—, Fine-tuning Data=—2025.09535450585850
PixtralModel Size=12B, Fine-tuning Method=—, Fine-tuning Data=—2025.09525148564657
LLama3.2-VisionModel Size=11B, Fine-tuning Method=—, Fine-tuning Data=—2025.09463040385525
Gemma 3Model Size=24B, Fine-tuning Method=—, Fine-tuning Data=—2025.09276749541884
Claude-SonnetModel Size=—, Fine-tuning Method=—, Fine-tuning Data=—2025.09246647541685
Qwen2.5-VLModel Size=7B, Fine-tuning Method=—, Fine-tuning Data=—2025.0910695754597
GPT 4oModel Size=—, Fine-tuning Method=—, Fine-tuning Data=—2025.090700540100