Citation Generation on Citation Generation N=15 benchmark leaderboard · SOTA2 Research

Citation Generation on Citation Generation N=15

13.3Author AccuracyDeepSeek-R1-Distill-Qwen2.5-32B

Evaluation Results

Method	Links
DeepSeek-R1-Distill-Qwen2.5-32B2026.04		13.3	35.8	36.5	29.2	18.8	3
Qwen2.5-32B-Instruct2026.04		13.1	40.2	40.6	38.6	53.5	8
Qwen3-30B-A3B-Base2026.04		10.3	32.5	34	26.6	17.7	2.3
DeepSeek-R1-Distill-Qwen3-8B2026.04		10.2	22.5	25.5	17.5	25.5	2.2
Qwen3-30B-A3B-Instruct2026.04		9.1	34.4	33.8	25.9	20.2	3.1
Mistral-Small-24B-Instruct-25012026.04		8.2	34.1	33.6	25.5	20.4	1.9
Qwen2.5-14B-Instruct2026.04		7.1	29.1	27.4	23	20	2.9
Moonlight-16B-A3B-Instruct2026.04		6.2	31.2	27.8	20.4	29.8	2
DeepSeek-R1-Distill-Qwen2.5-14B2026.04		3.4	10.9	21.5	12.5	6.4	0.4