Literature Retrieval on ScholarGym Fast (test) benchmark leaderboard · SOTA2 Research

Literature Retrieval on ScholarGym Fast (test)

95RecallGemini3-Pro

Evaluation Results

Method	Links
Gemini3-Pro2026.01		95	19.9	32.9	95.8	1.1	2.2
DeepSeek-V3.22026.01		85.5	13.5	23.3	86.2	0.7	1.4
GPT-5.22026.01		83.7	30.5	44.7	88.3	0.9	1.8
DeepSeek-V3.22026.01		81.2	28.7	42.3	87.2	0.9	1.8
GLM-4.72026.01		75.4	11.1	19.4	81.4	1	2
Qwen3-30B2026.01		67.3	18.1	28.5	72	1.1	2.1
Qwen3-8B2026.01		48.3	15.2	23.1	55	2.2	4.2
Qwen3-30B2026.01		48.2	29	36.2	59	2.5	4.8
Qwen3-8B2026.01		45.8	21.6	29.3	58	2.3	4.5
Qwen3-30B2026.01		31.2	5.8	9.8	35	5.2	9.1
Qwen3-8B2026.01		18.5	4.2	6.9	21	3.6	6.1