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

Sycophancy Mitigation on Held-out

43.6BRR

BCT

-1.6410.10521.8533.595Jun 4, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.06
43.6
2026.06
43.6
2026.06
43.6
2026.06
43.6
2026.06
41.8
2026.06
33.1
2026.06
30.9
2026.06
30.9
2026.06
30.9
2026.06
30.9
2026.06
26.7
2026.06
26.7
2026.06
26.7
2026.06
26.7
2026.06
25.2
2026.06
25.2
2026.06
25.2
2026.06
25.2
2026.06
18.3
2026.06
18.3
2026.06
18.3
2026.06
18.3
2026.06
17.7
2026.06
13.3
2026.06
12.1
2026.06
10.4
2026.06
8.6
2026.06
4.1
2026.06
3.9
2026.06
2.5
2026.06
2.1
2026.06
1.5
2026.06
1.1
2026.06
0.9
2026.06
0.8
2026.06
0.6
2026.06
0.5
2026.06
0.4
2026.06
0.2
2026.06
0.1