Citation Generation N=10 benchmark leaderboard · SOTA2 Research

Citation Generation N=10

3.9Author Component ScoreDeepSeek-R1-Distill-Qwen2.5-14B

Evaluation Results

Method	Links
DeepSeek-R1-Distill-Qwen2.5-14B2026.04		3.9	12.7	21.5	13.8	8.3	0.6
Moonlight-16B-A3B-Instruct2026.04		7.1	37	29.3	21.4	32.7	2.5
Qwen2.5-14B-Instruct2026.04		7.7	31.6	28.9	23.7	21	3.2
Mistral-Small-24B-Instruct-25012026.04		10.2	42.5	38.5	27.5	25	2.9
Qwen3-30B-A3B-Instruct2026.04		11.5	43.1	39.1	28.6	25.4	4.5
DeepSeek-R1-Distill-Qwen3-8B2026.04		12	28.3	31	21.2	25	2.2
Qwen3-30B-A3B-Base2026.04		13.3	42.1	41.3	31.7	20.7	3.1
Qwen2.5-32B-Instruct2026.04		13.6	43.4	41.6	39.3	53.8	8.4
DeepSeek-R1-Distill-Qwen2.5-32B2026.04		14.4	39.6	38.5	31	21.4	3.2