Legal Question Answering on LawBench revised (test) benchmark leaderboard · SOTA2 Research

Legal Question Answering on LawBench revised (test)

44.68ROUGE-1JurisMA

Evaluation Results

Method	Links
JurisMA2026.04		44.68	23.42	31.14	32.54	16.18	14.25	75.05	58.63	3.93
MMEP2026.04		42.6	20.78	27.27	30.64	14.8	12.41	73	56.28	—
Qwen3-14B2026.04		42.55	19.24	27.27	21.93	10.25	8.1	74.64	62.38	3.43
Qwen2.5-14B2026.04		40.96	16.1	23.7	34.24	13.52	9.9	73.15	56.22	—
GPT4o2026.04		40.24	15.27	24.5	34.94	12.34	8.89	73.22	55.16	3.36
ReAct2026.04		38.65	14.03	22.41	26.86	9.89	7.05	72.3	57.55	—
AutoGen-Tree2026.04		37.7	13.9	21.38	27.37	10.44	7.57	71.62	54.36	—
LexRAG2026.04		36.85	13.86	21.51	31.75	11	8.45	71.45	54.59	—
Wisdom2026.04		36.6	15.74	23.12	34.89	10.16	8.04	71.97	56.09	2.86
Fuzi-mingcha2026.04		32.31	9.92	17.41	23.13	7.07	5.19	70.47	52.46	2.78
AutoGen-Flat2026.04		31.29	12.11	19.27	24.42	9.8	7.25	66.92	48.45	—
LexiLaw2026.04		31.23	9.61	18.15	14.4	4.76	3.33	70	58.12	2.66
Hanfei2026.04		30.79	10.21	18.37	13.47	4.26	2.68	69.68	58.37	2.69
Lawyer-LLaMA2026.04		30.61	9.5	18.82	24.65	6.28	3.35	69.17	58.47	2.74
ChatLaw-33B2026.04		27.78	7.4	18.15	17.84	3.81	1.46	67.77	57.54	2.3
Parser2026.04		26.72	5.89	16.06	18.73	3.69	1.79	65.35	56.93	—
LawGPT2026.04		20.52	5.22	7.26	5.15	1.01	0.49	63.44	46.94	1.52