List your company

Multimodal Role-playing on MMRole (out-of-domain)

1.164IAClaude-Opus-4
1.014241.053121.0921.13088May 10, 2026

Evaluation results

MethodLinks
Claude-Opus-4Backbone=Closed large MLLMs2026.051.1641.121.1951.2011.1851.2391.1891.2461.192
SFTBackbone=Qwen3-VL-8B2026.051.1361.1461.1411.1231.1251.0781.1091.051.114
SFTBackbone=Gemma-3n-E4B2026.051.121.1211.1181.111.1111.0881.1041.0711.105
CAVIBackbone=Qwen3-VL-8B2026.051.1111.0471.1431.1781.1261.2271.1771.2211.154
VCDBackbone=Qwen3-VL-8B2026.051.0951.0271.121.1351.1131.2061.1521.2061.132
VCDBackbone=Gemma-3n-E4B2026.051.0951.0341.1261.1291.111.1791.1271.1971.125
GPT-5Backbone=Closed large MLLMs2026.051.0921.0931.1361.1461.1271.2051.1391.2011.142
CAVIBackbone=Gemma-3n-E4B2026.051.0861.0471.1431.1571.1321.21.1521.1911.139
AIRBackbone=Qwen3-VL-8B2026.051.0851.0041.11.141.1141.2041.1461.1951.124
Gemini-Pro-3Backbone=Closed large MLLMs2026.051.0761.1031.1161.0891.0821.0871.0841.0931.091
ClearSightBackbone=Gemma-3n-E4B2026.051.0751.0181.1081.1171.1041.1771.121.1671.111
MemVRBackbone=Gemma-3n-E4B2026.051.0731.051.1161.1241.1071.171.1251.1741.117
AIRBackbone=Gemma-3n-E4B2026.051.0731.051.1161.1241.1071.171.1251.1741.117
MemVRBackbone=Qwen3-VL-8B2026.051.0660.9921.0711.1181.0941.2131.1381.1971.111
ClearSightBackbone=Qwen3-VL-8B2026.051.0560.9981.0731.1131.0781.21.1341.2041.107
BaseBackbone=Qwen3-VL-8B2026.051.0390.961.0421.0781.0481.1551.0971.1441.07
BaseBackbone=Gemma-3n-E4B2026.051.020.9961.0631.071.051.1181.0631.1091.061