Ride-hailing adjudication on Appeal benchmark leaderboard · SOTA2 Research

Ride-hailing adjudication on Appeal

94.6Precision (Normal)RideJudge-8B

Evaluation Results

Method	Links
RideJudge-8B2026.03		94.6	94.11	91.52	91.93	91.86	88.41
RideJudge-4B2026.03		94.55	93.24	93.06	90.13	90.86	87.25
Qwen3-VL-8B-Thinking2026.03		91.3	3.64	100	5.83	3.48	8.82
QwQ-32B2026.03		76.88	77.82	66.66	36.77	61.57	35.18
Qwen3-VL-32B-Thinking2026.03		76.18	78.16	75.93	36.77	57.3	56.86
Qwen3-VL-32B-Thinking2026.03		75.85	69.67	75	36.32	49.75	45.94
Qwen3-VL-8B-Thinking2026.03		75	4.16	0.89	77.78	2.78	7.37
QwQ-32B2026.03		72.77	75.04	58.6	48.88	60.28	35.82
DeepSeek-V3.12026.03		71.15	93.59	86.79	20.63	58.09	75.36
DeepSeek-V3.12026.03		70.91	95.49	80.7	20.63	57.89	75.25
MiniCPM-V2026.03		70.18	6.93	0	0	16.48	12.96
MiniCPM-V2026.03		67.44	5.03	0	0	14.9	19.68
Qwen3-VL-8B-Instruct2026.03		65.57	24.09	51.35	8.52	41.81	62.09
Qwen3-32B-Instruct2026.03		64.83	88.39	30.24	62.78	40.81	26.21
Qwen3-VL-8B-Instruct2026.03		63.09	25.48	27.34	15.7	39.82	60.61
Qwen3-32B-Instruct2026.03		62.76	89.08	27.07	63.23	38.93	26.77
Qwen3-VL-32B-Instruct2026.03		60.61	79.2	41.88	52.02	43.2	65.55
Qwen3-VL-32B-Instruct2026.03		60.57	83.36	38.29	16.14	41.71	59.14