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

Harmlessness Evaluation on LINGUASAFE (test)

100Tie Rate

SHARD

Updated 1mo ago

Evaluation Results

Method	Links
SHARD 2026.06		100	0	0
SHARD w/Teacher 2026.06		100	0	0
HHTUNE 2026.06		100	0	0
SHARD 2026.06		99.8	0	0.2
SHARD w/Teacher 2026.06		99.8	0.2	0
HHTUNE 2026.06		99.8	0	0.2
RATIONAL 2026.06		99.8	0.2	0
RATIONAL 2026.06		99.8	0	0.2
SHARD 2026.06		99.8	-	-
Benign Intent 2026.06		99.8	-	-
SHARD 2026.06		99.8	-	-
Benign Intent 2026.06		99.8	-	-
Benign Intent 2026.06		99.7	-	-
SHARD 2026.06		99.5	-	-
SHARD w/Teacher 2026.06		99.4	0.3	0.3
RATIONAL 2026.06		99.4	0.6	0
SHARD 2026.06		99	1	0
HHTUNE 2026.06		99	0.8	0.2
RATIONAL 2026.06		98.5	1.5	0
Benign Intent 2026.06		98.3	-	-
Benign Intent 2026.06		98.3	-	-
SHARD w/Teacher 2026.06		98.2	0	1.8
HHTUNE 2026.06		98.2	1.8	0
SHARD 2026.06		97.5	-	-
SHARD 2026.06		97.3	-	-
Benign Intent 2026.06		97.2	-	-
RATIONAL 2026.06		96.7	2.7	0.5
SHARD 2026.06		96.4	3.6	0
HHTUNE 2026.06		96.4	3	0.5
SHARD 2026.06		96.1	-	-
SHARD 2026.06		95.7	0.6	3.7
RATIONAL 2026.06		95.7	1.9	2.5
HHTUNE 2026.06		94.4	2.5	3.1
SHARD w/Teacher 2026.06		93.2	6.6	0.2
Benign Intent 2026.06		92.4	-	-
SHARD 2026.06		91	9	0
SHARD 2026.06		90.6	-	-
SHARD w/Teacher 2026.06		83.7	7.1	9.2