Citation Generation on Citation Generation N=5 benchmark leaderboard · SOTA2 Research

Citation Generation on Citation Generation N=5

20.4Author AccuracyDeepSeek-R1-Distill-Qwen3-8B

Evaluation Results

Method	Links
DeepSeek-R1-Distill-Qwen3-8B2026.04		20.4	39.8	44.1	29	23.7	3.2
DeepSeek-R1-Distill-Qwen2.5-32B2026.04		19.8	45.5	47.5	38.6	23.8	4
Qwen3-30B-A3B-Base2026.04		19.2	57.1	51.9	38.9	27.7	5
Qwen2.5-32B-Instruct2026.04		16	45.8	43	39.8	54.9	10.5
Qwen3-30B-A3B-Instruct2026.04		15.5	55.2	47.3	33	33.3	6.8
Mistral-Small-24B-Instruct-25012026.04		15.4	52.7	45.5	33.2	32.6	5
Moonlight-16B-A3B-Instruct2026.04		9.6	44.8	35.2	26.2	38.6	4.3
Qwen2.5-14B-Instruct2026.04		9.3	39.1	32.8	25.4	22.7	4.3
DeepSeek-R1-Distill-Qwen2.5-14B2026.04		7.7	22	26.4	19.8	13.2	1.1