Share your thoughts, 1 month free Claude Pro on usSee more

Scientific Reasoning on Principia OOD (test)

83.78Average Score @8

CoT Self-Instruct

Updated 1mo ago

Evaluation Results

Method	Links
CoT Self-Instruct 2026.06		83.78	91.49	1.87
Combined (2× data) 2026.06		83.24	93.62	1.33
Qwen3.5-4B 2026.06		81.91	91.49	-
Agentic Self-Instruct 2026.06		80.32	91.49	-1.59
Agentic Self-Instruct 2026.06		67.05	79.09	0.83
Combined (2× data) 2026.06		66.25	72.73	0.03
Qwen3.5-4B 2026.06		66.22	76.36	-
CoT Self-Instruct 2026.06		65.91	77.27	-0.31
Agentic Self-Instruct 2026.06		56.82	64.95	0.82
Combined (2× data) 2026.06		56.5	65.11	0.5
CoT Self-Instruct 2026.06		56.28	65.33	0.28
Qwen3.5-4B 2026.06		56	65.41	-
Agentic Self-Instruct 2026.06		51.47	59.91	1.04
Combined (2× data) 2026.06		51.17	59.91	0.74
CoT Self-Instruct 2026.06		51.1	60.06	0.67
Qwen3.5-4B 2026.06		50.43	59.54	-
Agentic Self-Instruct 2026.06		35.43	43.67	1.75
CoT Self-Instruct 2026.06		35.25	43.67	1.57
Combined (2× data) 2026.06		34.97	44.3	1.29
Qwen3.5-4B 2026.06		33.68	41.93	-