List your company

Visual Question Answering on VRQABench

72.4AccuracyPF-OPSD
31.00841.75452.563.246Jun 2, 2026

Evaluation results

MethodLinks
PF-OPSDProtocol=Qwen3.5-9B training baseline2026.0672.4
GRPOProtocol=Qwen3.5-9B training baseline2026.0663.5
SFTProtocol=Qwen3.5-9B training baseline2026.0661.8
Workflow Agent (Gemini-3-Flash)Protocol=Workflow-agent2026.0647.2
Gemini-3-FlashProtocol=Zero-shot / no-simulation2026.0645.9
GPT-5.4Protocol=Zero-shot / no-simulation2026.0643.2
HY3Protocol=Zero-shot / no-simulation2026.0638.2
Qwen3.5-9BProtocol=Zero-shot / no-simulation2026.0633.2
Qwen3.6-27BProtocol=Zero-shot / no-simulation2026.0633
Qwen2.5-VL-7BProtocol=Zero-shot / no-simulation2026.0632.7
Workflow Agent (Qwen3.5-9B)Protocol=Workflow-agent2026.0632.6