List your company

Image Captioning on CapArena-Auto

91.7Average Win-Rate MarginReflectCAP
2.2625.4848.771.92Apr 14, 2026

Evaluation results

MethodLinks
ReflectCAPBase Model=Qwen3-VL-32B2026.0491.7
ReflectCAPBase Model=GPT-4.1-mini2026.0490
Qwen3-VL-32BStrategy=Zero-shot2026.0487.3
CapMASBase Model=Qwen3-VL-32B2026.0487
ReflectCAPBase Model=Qwen3-VL-8B2026.0485.3
Qwen3-VL-8BStrategy=Zero-shot2026.0476
CapMASBase Model=Qwen3-VL-8B2026.0474
GPT-5.2Strategy=Zero-shot2026.0470
Gemini-1.5-Pro†Strategy=Zero-shot2026.0462.3
GPT-4.1-miniStrategy=Zero-shot2026.0457.7
InternVL3.5-4BStrategy=Zero-shot2026.0454.3
CapMASBase Model=GPT-4.1-mini2026.0453.7
ReflectCAPBase Model=GPT-4.1-nano2026.0451.3
CapMASBase Model=InternVL3.5-4B2026.0446.7
GPT-4o-0806†Strategy=Zero-shot2026.0444.3
Qwen2.5VL-72B†Strategy=Zero-shot2026.0439.7
Qwen2.5-VL-7BStrategy=Zero-shot2026.0433.7
Claude-3.5-Sonnet†Strategy=Zero-shot2026.0429.7
CapMASBase Model=Qwen2.5-VL-7B2026.0428
ReflectCAPBase Model=Qwen2.5-VL-32B2026.0425.3
InternVL3.5-38BStrategy=Zero-shot2026.0424
GPT-4.1-nanoStrategy=Zero-shot2026.0421.2
ReflectCAPBase Model=InternVL3.5-4B2026.0418.7
CapMASBase Model=InternVL3.5-38B2026.0415
CapMASBase Model=GPT-4.1-nano2026.0414.7
ReflectCAPBase Model=InternVL3.5-38B2026.0412
CapMASBase Model=Qwen2.5-VL-32B2026.048
ReflectCAPBase Model=Qwen2.5-VL-7B2026.047.3
Qwen2.5-VL-32BStrategy=Zero-shot2026.045.7