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

Multi-attribute steering on Multi-attribute steering S+T (test)

100S Score

Uniform+E

Updated 21d ago

Evaluation Results

Method	Links
Uniform+E 2026.05		100	100	21	77.4
Uniform 2026.05		99.8	99.9	26	78.6
Adaptive+E 2026.05		99.6	99.9	17	76.7
Contrastive Vec. 2026.05		99.5	71.7	15	79.9
Contrastive Vec. 2026.05		99	100	91	92.1
Adaptive 2026.05		98.6	100	50	56.8
Uniform+E 2026.05		98.5	99.9	96	93.3
Uniform+E 2026.05		95.3	98.4	95	91.1
Adaptive 2026.05		95.2	99.9	71	92.5
Uniform 2026.05		94.7	99.9	6	43.8
Adaptive+E 2026.05		93.9	97.4	58	92.9
Contrastive Vec. 2026.05		93.4	26.4	8	81.9
Uniform 2026.05		93.2	99.9	91	93.1
Adaptive 2026.05		92.8	100	67	87.2
Uniform 2026.05		92.7	100	98	89.3
PCA 2026.05		92.5	23	7	61.5
Adaptive+E 2026.05		91.7	87.5	9	65
Probe 2026.05		91.6	78.6	98	92.3
Adaptive 2026.05		91.1	98.2	8	56.9
Uniform+E 2026.05		89.4	95.4	7	54.4
Prompt 2026.05		88.7	50.5	17	82.7
Contrastive Vec. 2026.05		84.4	64	376	91.8
Adaptive+E 2026.05		82.8	77.4	59	91.5
Probe 2026.05		81.9	49.5	15	88.1
Prompt 2026.05		80.3	63.2	12	80.2
Probe 2026.05		78.1	32.3	8	74.3
Probe 2026.05		72.8	99.8	324	94.1
No Steering 2026.05		69.1	13.9	16	85.1
PCA 2026.05		67.9	29.4	13	79.4
No Steering 2026.05		57.7	10.5	14	87.9
No Steering 2026.05		29	20.3	44	92.5
PCA 2026.05		28.3	21.9	48	92.8
No Steering 2026.05		25.5	18.6	43	92.6
PCA 2026.05		20.1	11.2	310	93.5