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

Harmfulness Evaluation on Harmfulness Evaluation Suite (JBB, SR, WJ, GCG, JBC, PAIR) (test)

91JBB

No train

Updated 1mo ago

Evaluation Results

Method	Links
No train 2026.04		91	87.2	91.6	83	90	95.3	89.7
No train 2026.04		90	84.4	94	81	75	93.8	86.4
No train 2026.04		78	73.5	91.6	76	77	96.9	82.2
SafeChain 2026.04		76	70.6	83.2	58	59	82.8	71.6
SafeChain 2026.04		76	78.9	80.8	54	59	84.4	72.2
SafeChain 2026.04		76	78	86.8	73	80	84.4	79.7
No train 2026.04		74	65.8	74	68	56	81.3	69.8
No train 2026.04		74	78	90.8	72	88	98.4	83.5
SafeChain 2026.04		73	56.9	65.6	53	28	64.1	56.8
No train 2026.04		73	81.2	91.6	74	75	0	82.5
SafeChain 2026.04		73	77.6	87.2	61	67	93.8	76.6
SafeChain 2026.04		72	73.5	88.4	60	73	92.2	76.5
SafeChain 2026.04		71	71.6	80	69	49	81.3	70.3
No train 2026.04		71	78.9	90.8	70	67	95.3	78.8
SafeChain 2026.04		71	78.3	88.4	65	60	96.9	76.6
SafeChain 2026.04		70	71.9	80	66	43	85.9	69.5
SafeChain 2026.04		69	67.7	80.8	69	54	87.5	71.3
SafeChain 2026.04		69	76.4	86.4	69	55	89.1	74.1
SafeChain 2026.04		60	62.3	65.6	50	37	76.6	58.6
S1-ALT 2026.04		13	11.8	37.6	7	16	31.3	19.4
STAR-1 2026.04		12	10.9	52.4	6	0	31.3	18.8
STAR-1 2026.04		12	6.7	42	7	5	32.8	17.6
STAR-1 2026.04		12	4.8	47.6	10	3	37.5	19.1
STAR-1 2026.04		11.6	7.2	52.3	10.2	7.8	46.3	22.6
R1-ALT 2026.04		10	6.4	36.8	7	2	23.4	14.3
STAR-1 2026.04		9	9	52.4	8	7	42.2	21.3
DirectRefusal 2026.04		9	11.5	20.4	8	0	20.3	11.5
DirectRefusal 2026.04		9	10.2	12	3	0	6.3	6.7
DirectRefusal 2026.04		8	12.5	6.4	4	0	14.1	7.5
DirectRefusal 2026.04		7	13.4	18	6	5	14.1	10.6
DirectRefusal 2026.04		5	11.5	14.4	5	0	4.7	6.8
R1-ALT 2026.04		1	2.6	14.8	1	0	7.8	4.5
S1-ALT 2026.04		1	1.9	8	2	13	14.1	6.7
R1-ALT 2026.04		0	1	21.2	2	0	4.7	4.8
R1-ALT 2026.04		0	0.6	26	1	0	10.9	6.4
R1-ALT 2026.04		0	1.9	13.2	1	0	6.3	3.7