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

Leakage Analysis in LLM-based Tutoring on GSM8K

0Student Leakage

Manually Defined Prompts

Updated 1mo ago

Evaluation Results

Method
Manually Defined Prompts 2026.04	0	-	40	8.92
Manually Defined Prompts 2026.04	0	-	46	10.03
Manually Defined Prompts 2026.04	0	-	75	5.53
Manually Defined Prompts 2026.04	0	-	9	11.56
Manually Defined Prompts 2026.04	0	-	4	10.68
Manually Defined Prompts 2026.04	0	-	49	7.98
Manually Defined Prompts 2026.04	0	-	10	10.6
Manually Defined Prompts 2026.04	0	-	2	8.38
Manually Defined Prompts 2026.04	0	-	22	9.32
Finetuned Adv. Agent 2026.04	2	8.73	82	6.3
Finetuned Adv. Agent 2026.04	3	10.95	70	10.78
Finetuned Adv. Agent 2026.04	3	12.53	4	12.2
Finetuned Adv. Agent 2026.04	3	13.18	61	9.86
Finetuned Adv. Agent 2026.04	3	15.53	3	10.65
Finetuned Adv. Agent 2026.04	3	11.39	38	12.01
Finetuned Adv. Agent 2026.04	4	11.52	34	10.69
Finetuned Adv. Agent 2026.04	4	12.66	6	10.39
Finetuned Adv. Agent 2026.04	4	13.99	3	9.08
LLM-Generated Attacks 2026.04	7	5.13	52	8.1
LLM-Generated Attacks 2026.04	7	3.95	65	5.77
LLM-Generated Attacks 2026.04	8	5.17	39	7.82
LLM-Generated Attacks 2026.04	9	5.54	23	9.77
LLM-Generated Attacks 2026.04	9	6.04	11	10.74
LLM-Generated Attacks 2026.04	9	6.31	2	10.61
LLM-Generated Attacks 2026.04	9	6.02	12	13.07
LLM-Generated Attacks 2026.04	9	6.06	3	7.27
LLM-Generated Attacks 2026.04	9	6.27	6	10.01
Multi-Agent Student 2026.04	27	5.69	48	7.04
Multi-Agent Student 2026.04	41	7.43	27	9.54
Multi-Agent Student 2026.04	41	6.76	24	9.3
Multi-Agent Student 2026.04	42	7.67	14	10.68
Student w/ Reasoning 2026.04	45	5.27	25	7.21
Multi-Agent Student 2026.04	50	8.22	5	12.73
Multi-Agent Student 2026.04	50	9.03	7	12.22
Student w/ Reasoning 2026.04	51	6.9	6	9.81
Student w/ Reasoning 2026.04	52	5.57	12	8.44
Student w/ Reasoning 2026.04	55	6.16	18	10.74
Multi-Agent Student 2026.04	55	7.13	5	8.86
Multi-Agent Student 2026.04	56	7.18	13	10.73
Student w/ Reasoning 2026.04	58	6.83	3	10.99
Base Student Adv. Agent 2026.04	61	5.19	14	6.44
Base Student Adv. Agent 2026.04	61	6.48	8	9.69
Student w/ Reasoning 2026.04	61	7.21	5	12.17
Student w/ Reasoning 2026.04	62	6.39	1	13.42
Student w/ Reasoning 2026.04	63	6.19	9	11.09
Student w/ Reasoning 2026.04	63	6.21	4	9.63
Multi-Agent Student 2026.04	63	7.67	0	8.5
Base Student Adv. Agent 2026.04	64	6.08	10	10.04
Base Student Adv. Agent 2026.04	66	5.44	6	8.75
Base Student Adv. Agent 2026.04	66	6.76	3	11.64
Base Student Adv. Agent 2026.04	68	6.5	1	12.97
Base Student Adv. Agent 2026.04	73	5.75	2	7.88
Base Student Adv. Agent 2026.04	75	5.76	4	9.64
Base Student Adv. Agent 2026.04	76	5.77	0	5