List your company

Task#2 on STAR

94.88AccuracyKGRPO
17.1437.322557.50577.6875Oct 22, 2025

Evaluation results

MethodLinks
KGRPOBackbone=Qwen2.5-VL-7B, Settings=S22025.1094.8897.75
GPT-4o-miniBackbone=GPT-4o-mini2025.1072.25—
S1(Single)Backbone=Qwen2.5-VL-3B2025.1056.6374.75
GPT-4oBackbone=GPT-4o2025.1056.33—
GPT-4vBackbone=GPT-4v2025.1041.25—
Zero-shotBackbone=Qwen2.5-VL-3B2025.1020.13—