ML Research Engineering on PaperBench benchmark leaderboard · SOTA2 Research

ML Research Engineering on PaperBench

33.26Adaptive Pruning ScoreAiScientist

Evaluation Results

Method	Links
AiScientist2026.04		33.26	49.47	61.11	30.02	26.46	28.98	8.34	30.1	28.53	40.55	58.76	10.18	28.66	44.13	31.69	49.32	14.81	42.1	27.33	30.82	33.73	12.2
BasicAgent2026.04		30.82	33.78	51.45	23.55	9.8	21.6	3.71	20.68	22.55	32.49	22.18	6.56	22.43	36.93	6.99	27.2	3.69	32.18	17.81	25.14	22.58	4.9
AiScientist2026.04		27.25	46.29	56.59	33.79	23.09	35.21	10.11	27.9	30.23	29.95	49.92	10.87	18.28	36.77	19.85	64.94	20.13	18.81	32.45	17.87	30.52	15.67
BasicAgent2026.04		24.53	20.86	48.46	15.43	12.59	21.67	5.87	17.75	12.97	14.86	26.63	10.43	15.45	25.39	11.45	53.51	8.37	17.04	15.27	6.61	19.26	6.25
IterAgent2026.04		11.93	44.43	47.91	19.28	12.5	16.67	6.7	22.74	26.15	34.96	25.77	0.27	27.56	41.26	4.95	35.53	8.81	28.06	21.19	10.75	22.37	54.9
IterAgent2026.04		3.05	45.13	45.04	8.3	12.44	23.89	4.15	15.26	18.3	21.89	30.81	8.88	10.43	33.34	12.65	60.24	13.69	17.37	18.13	8.99	20.6	27.44