Tutor Robustness on MMLU (test) benchmark leaderboard · SOTA2 Research

Tutor Robustness on MMLU (test)

0Student Information LeakageManually Defined Prompts

Evaluation Results

Method	Links
Manually Defined Prompts2026.04		0	—	81	3.3
Manually Defined Prompts2026.04		0	—	42	5.37
Manually Defined Prompts2026.04		0	—	25	5.71
Finetuned Adv. Agent2026.04		0	—	88	4.1
Finetuned Adv. Agent2026.04		0	—	61	7.15
Finetuned Adv. Agent2026.04		0	—	10	7.92
Multi-Agent Student2026.04		15	5	60	4.63
Student w/ Reasoning2026.04		23	4.28	56	3.37
Multi-Agent Student2026.04		23	6.11	30	5.71
Multi-Agent Student2026.04		31	6.89	17	6.98
Student w/ Reasoning2026.04		32	4.73	26	4.34
Base Student Adv. Agent2026.04		34	4.35	46	4.37
Student w/ Reasoning2026.04		48	5.72	11	5.33
Base Student Adv. Agent2026.04		49	5.18	22	4.64
Base Student Adv. Agent2026.04		61	5.12	7	7.24