List your company

Multimodal Understanding on SEED (Metrics: SEED, Average)

81.8SEED ScoreBAGEL (Ours)
25.01639.75854.569.242Dec 27, 2024Mar 27, 2025Jun 26, 2025Sep 25, 2025Dec 25, 2025Mar 26, 2026Jun 25, 2026

Evaluation results

MethodLinks
BAGEL (Ours)Baseline=BAGEL, Params=7B active2026.0681.8—
BLIP3o-8B (Ours)Baseline=BLIP3o-8B, Params=8B2026.0679.4—
BAGELBaseline=BAGEL, Params=7B active2026.0679.3—
VARGPT-v1.1 (Ours)Baseline=VARGPT-v1.1, Params=7B+2B2026.0679.2—
FP16Model=LLaVA-onevision-72B, Bitwidth=FP162024.1278.174.3
MBQModel=Qwen2-VL-72B, Bitwidth=W3A162024.1277.777.6
MBQModel=LLaVA-onevision-72B, Bitwidth=W3A162024.1277.673.6
FP16Model=Qwen2-VL-72B, Bitwidth=FP162024.1277.678.1
AWQModel=Qwen2-VL-72B, Bitwidth=W3A162024.1277.677.5
RTNModel=Qwen2-VL-72B, Bitwidth=W3A162024.1277.575
BLIP3o-8BBaseline=BLIP3o-8B, Params=8B2026.0677.5—
RTNModel=LLaVA-onevision-72B, Bitwidth=W3A162024.1277.472.1
GPTQModel=Qwen2-VL-72B, Bitwidth=W3A162024.1277.276.6
FP16Model=InternVL2-26B, Bitwidth=FP162024.1276.874.6
RTNModel=LLaVA-onevision-72B, Bitwidth=W4A82024.1276.670.7
SQModel=LLaVA-onevision-72B, Bitwidth=W4A82024.1276.670.3
RTNModel=Qwen2-VL-72B, Bitwidth=W4A82024.1276.673.2
MBQModel=LLaVA-onevision-72B, Bitwidth=W4A82024.1276.570.8
SQModel=Qwen2-VL-72B, Bitwidth=W4A82024.1276.472.2
MBQModel=InternVL2-26B, Bitwidth=W3A162024.1276.373.8
MBQModel=Qwen2-VL-72B, Bitwidth=W4A82024.1276.375.8
AWQModel=InternVL2-26B, Bitwidth=W3A162024.1276.273.5
VARGPT-v1.1Baseline=VARGPT-v1.1, Params=7B+2B2026.0676.1—
GPTQModel=LLaVA-onevision-72B, Bitwidth=W3A162024.127672.3
RTNModel=InternVL2-26B, Bitwidth=W4A82024.127672.7
GPTQModel=InternVL2-26B, Bitwidth=W3A162024.1275.872.3
RTNModel=InternVL2-26B, Bitwidth=W3A162024.1275.773.3
MBQModel=InternVL2-26B, Bitwidth=W4A82024.1275.772.7
Lumina-DiMOO (ReMDM)2026.0274.3—
FreeCorrectionbackbone=Lumina-DiMOO2026.0274.2—
ILLUMEBaseline=Vicuna-7B, Params=7B2026.0672.9—
TokenFlowBaseline=TokenFlow2026.0672.6—
Janus-Pro-7BBaseline=Janus-Pro-7B, Params=7B2026.0672.1—
SUDERBaseline=Janus-Pro-7B, Params=7B2026.0671.9—
MetaMorphBaseline=MetaMorph2026.0671.8—
Lumina-DiMOOgeneration step=64, generated sequence length=2562026.0271.4—
AWQModel=LLaVA-onevision-72B, Bitwidth=W3A162024.1271.255.1
SQModel=InternVL2-26B, Bitwidth=W4A82024.1270.665.5
Emu3Baseline=Emu32026.0668.2—
SEED-XBaseline=SEED-X2026.0666.5—
JanusBaseline=Janus2026.0663.7—
VILA-UBaseline=VILA-U2026.0659—
LLaVA-1.5-7BTokens=5762026.0557.9100
ETCTokens=42026.0556.6695.33
VoCoTokens=42026.0556—
ETCTokens=22026.0555.7694.55
VoCoTokens=22026.0555—
ETCTokens=12026.0554.1892.22
VoCoTokens=12026.0553.7—
LLaVA-1.5-7BTokens=02026.0529.2—
ChameleonBaseline=Chameleon2026.0627.2—