Human Activity Evaluation on SceneTeract (test) benchmark leaderboard · SOTA2 Research

Human Activity Evaluation on SceneTeract (test)

69.9AccuracyGemini-3-Flash-Preview

Evaluation Results

Method	Links
Gemini-3-Flash-Preview2026.03		69.9	12	0.39	7.5	94.7	58.5
Qwen3-VL-4B-Instruct with GRPO2026.03		69.2	14.1	0.364	9.7	98.2	62.3
Claude-Sonnet-4-62026.03		62.8	21.9	0.208	12.8	97.9	81.4
Claude-Sonnet-4-62026.03		62.3	20.6	0.201	19.4	100.4	81.4
Qwen3-VL-8B-Instruct2026.03		62.3	20.1	0.204	6.6	99.2	69.5
Gemini-3-Flash-Preview2026.03		62.1	35.7	0.144	16.5	79.2	58.5
Gemini-3.1-Pro-Preview2026.03		61.7	22.4	0.182	4.5	89.3	53.3
Gemma3-12B-Instruct2026.03		61.7	36.1	0.116	17.2	78.2	93.7
Gemini-3.1-Pro-Preview2026.03		61.6	6	0.321	6.9	112.7	53.3
Qwen3-VL-8B-Instruct2026.03		61.5	32.9	0.13	17.6	82.7	69.5
Qwen3-VL-4B-Instruct with GRPO2026.03		61.5	32.9	0.13	17.2	83.8	62.3
Qwen3-VL-4B-Instruct2026.03		61.4	22.8	0.172	13.2	97.6	64.6
Qwen3-VL-4B-Instruct2026.03		60.8	32.6	0.111	20.3	87.3	64.6
Gemma3-4B-Instruct2026.03		60.1	38.8	0.01	18.9	71.9	97.9
Gemma3-4B-Instruct2026.03		59.8	40.2	0	22.5	72.3	97.9
Gemma3-12B-Instruct2026.03		59.3	40.2	-0.055	22.5	69.4	93.7
Ministral3-3B-Instruct2026.03		54	33.3	-0.049	15	80	21.7
Ministral3-3B-Instruct2026.03		41.7	0.7	0.069	21.6	144.7	21.7