Numerical Symbolic Regression on LLM-SRBench 129-task synthetic (test) benchmark leaderboard · SOTA2 Research

Numerical Symbolic Regression on LLM-SRBench 129-task synthetic (test)

35Chemistry 95% Acc (Tol=0.01)Claude Opus 4.6

Updated 1mo ago

Evaluation Results

Method	Links
Claude Opus 4.62026.06		35	27	23	34	15	23	9	23	31	34	27	34	21	25	19	22	94	117	78	113
GPT-5.2-medium2026.06		33	26	22	32	14	23	9	23	28	36	24	32	20	25	18	24	88	117	73	111
GPT-OSS-120B2026.06		29	—	—	23	—	16	—	10	—	18	—	15	—	19	—	15	—	82	—	64
GPT-OSS-20B2026.06		18	—	—	8	—	8	—	5	—	13	—	7	—	16	—	10	—	55	—	30
GLM-4.5-Air2026.06		16	—	—	4	—	10	—	4	—	16	—	7	—	20	—	18	—	62	—	32
Qwen3-Coder-480B-A35B-Instruct2026.06		15	—	—	2	—	12	—	6	—	15	—	6	—	22	—	17	—	63	—	32
Qwen3-Coder-30B-A3B-Instruct2026.06		13	—	—	3	—	7	—	3	—	11	—	5	—	21	—	16	—	53	—	27
Claude Opus 4.62026.06		9	5	1	1	1	6	1	1	0	7	0	2	7	19	3	10	13	41	5	14
Claude Opus 4.62026.06		8	5	0	2	1	2	0	0	0	9	0	2	2	11	1	1	8	30	1	5
Qwen3-Coder-30B-A3B-Instruct2026.06		8	—	—	2	—	5	—	2	—	8	—	4	—	20	—	13	—	42	—	21
Claude Opus 4.62026.06		1	0	0	0	2	4	1	2	2	3	2	2	3	10	1	5	7	18	4	9