Share your thoughts, 1 month free Claude Pro on usSee more

Task#1 on STAR

99.88Accuracy

Qwen3-VL-8B S2(KGRPO)

Updated 1mo ago

Evaluation Results

Method	Links
Qwen3-VL-8B S2(KGRPO) 2025.10		99.88	96.4
Qwen3-VL-4B S2(KGRPO) 2025.10		99.75	95.36
Qwen3-VL-2B S2(KGRPO) 2025.10		99.38	94.49
LLaVA-NEXT-34B S2(SimPO) 2025.10		98.25	92.96
GPT-4o 2025.10		94.25	-
KGRPO 2025.10		79.88	86.4
Qwen2.5-VL-7B S2(KGRPO) 2025.10		79.88	86.4
S1(Single) 2025.10		77.13	82.47
Qwen2.5-VL-32B S2(SimPO) 2025.10		76.37	82.01
GRPO 2025.10		75.25	83.07
KGRPO 2025.10		75	81.49
Qwen2.5-VL-3B S2(KGRPO) 2025.10		75	81.49
SimPO 2025.10		71	79.03
SimPO 2025.10		69.63	82.96
GPT-4o-mini 2025.10		67.5	-
GPT-4o-mini 2025.10		67.5	-
LLaVA-1.5/NEXT-7B S2(SimPO) 2025.10		67.38	48.86
DPO 2025.10		66.5	82.11
ORPO 2025.10		65.75	81.96
S1(Full) 2025.10		64.88	81.79
GRPO 2025.10		60	76.49
DPO 2025.10		55.5	73.28
S1(Single) 2025.10		51	62.07
GPT-4o 2025.10		43.75	-
Qwen3-VL-30B 2025.10		43.75	-
S1(Full) 2025.10		42.75	52.67
ORPO 2025.10		39	64.79
Qwen2.5VL-72B 2025.10		38.25	-
GPT-4v 2025.10		37.75	-
GPT-4v 2025.10		37.75	-
QVQ-72B 2025.10		30.75	-
Zero-shot 2025.10		18.25	-
Zero-shot 2025.10		6.13	-