List your company

Visual Question Answering on VDR

37.8AccuracyVision-DeepResearch-30B-A3B
0.56810.23419.929.566Jan 29, 2026

Evaluation results

MethodLinks
Vision-DeepResearch-30B-A3BWorkflow=Agent Workflow2026.0137.8
Vision-DeepResearch-8BWorkflow=Agent Workflow2026.0129.2
Claude-3.7-SonnetWorkflow=Agent Workflow2026.0127.2
Qwen3-VL-30B-A3B-ThinkingWorkflow=Agent Workflow2026.0123.2
GPT-5Workflow=Agent Workflow2026.0120.4
Qwen3-VL-30B-A3B-InstructWorkflow=Agent Workflow2026.0120.2
Gemini-2.5 ProWorkflow=Agent Workflow2026.0118.8
Qwen3-VL-8B-ThinkingWorkflow=Agent Workflow2026.0117.6
Qwen3-VL-8B-InstructWorkflow=Agent Workflow2026.0117
Gemini-2.5 FlashWorkflow=Agent Workflow2026.0116.3
Claude-4-SonnetWorkflow=Agent Workflow2026.0113.6
GPT-5Workflow=Direct Answer2026.019.8
Gemini-2.5 ProWorkflow=Direct Answer2026.018
Gemini-2.5 FlashWorkflow=Direct Answer2026.016.2
Qwen3-VL-8B-ThinkingWorkflow=Direct Answer2026.015.6
Claude-3.7-SonnetWorkflow=Direct Answer2026.014.6
Qwen3-VL-30B-A3B-ThinkingWorkflow=Direct Answer2026.014.4
Qwen3-VL-30B-A3B-InstructWorkflow=Direct Answer2026.013.8
Qwen3-VL-8B-InstructWorkflow=Direct Answer2026.012.8
Claude-4-SonnetWorkflow=Direct Answer2026.012