Social Intelligence Analysis on SocialBench (test) benchmark leaderboard · SOTA2 Research

Social Intelligence Analysis on SocialBench (test)

91.7KnowledgeQwen2.5-7B-Instruct + Neeko

Evaluation Results

Method	Links
Qwen2.5-7B-Instruct + Neeko2026.01		91.7	73.8	33.8	38.3	86.9	72.6	82.9	90.1	79.8	72.2
Qwen2.5-7B-Instruct2026.01		90.4	79.3	34.1	30.8	87	75.2	86.9	90.3	80.2	72.7
Qwen2.5-7B-Instruct + SFT2026.01		90.3	77.7	39.4	33.3	90	76	87.2	91.1	82.4	74.2
Humanish-Roleplay2026.01		89.9	80.8	34.3	29.2	92	84.2	84.3	89.9	74.2	73.2
Qwen2.5-7B-Instruct + Character-R12026.01		89.6	79.5	40.8	43.3	93	84.1	87.5	92	82.4	76.9
Peach-2.0-9B-8k-Roleplay2026.01		89.3	72.8	46.7	26.3	87	48.4	83.7	90.3	85.7	70
Qwen2.5-7B-Instruct + RAR2026.01		88.1	72.8	40	33.9	93.7	87.7	86.1	90.4	83.5	75.1
Qwen2.5-7B-RoleMRC2026.01		86.8	75.7	43.4	68.3	84	45.9	81.4	83.1	76.4	71.7
Llama-3.2-3B-Instruct + Character-R12026.01		84.7	72.7	40.5	26.7	73	56.4	77.6	81.6	67.6	64.5
CoSER-Llama-3.1-8B2026.01		82.2	62.2	39.4	45.8	75	69.8	50.6	73.4	15.9	57.1
Llama-3.1-8B-RoleMRC2026.01		81.6	74.9	42.4	33.8	65	65.7	75.6	81.9	65.4	65.1
Llama-3.2-3B-Instruct + RAR2026.01		78.1	70.5	41.5	29.6	73.5	51.2	73.7	79.7	63.5	62.4
Crab2026.01		77.9	64.1	38.9	52.5	72	59.1	69.9	82.3	46.7	62.6
Llama-3.2-3B-Instruct + Neeko2026.01		77.1	65	39.8	29.6	57.6	44.6	72.8	81.6	60	58.7
Llama-3.2-3B-Instruct + SFT2026.01		74.7	73.5	40	30.4	73	49.4	76.9	83.5	66.2	63.1
CharacterGLM-6B2026.01		74.7	79.4	26.2	41.2	81.1	68.2	84.4	70.4	36.3	62.5
Haruhi-Zero-7B2026.01		74.2	71.5	29	40.8	57	43.1	57.7	72.2	29.1	52.7
Llama-3.2-3B-Instruct2026.01		73.5	72.5	40.4	35.4	73	50.4	76.9	80.5	66.9	63.3
Peach-9B-8k-Roleplay2026.01		58	25.3	40.8	31.3	67	46.4	60.9	67.1	48.4	49.4