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

Truthfulness on TruthfulQA (test)

48Accuracy

NUDGING

17.52825.43933.3541.261Feb 16, 2025Apr 28, 2025Jul 9, 2025Sep 19, 2025Nov 29, 2025Feb 9, 2026Apr 22, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
48
2026.04
48
2026.04
47
2026.04
46.4
2026.04
46
2026.04
46
2026.04
46
2026.04
45.8
2026.04
45.3
2026.04
45.2
2026.04
45.1
2026.04
44
2026.04
44
2025.02
35.02
2025.02
30.73
2025.02
28.29
2025.02
24.92
2025.02
23.39
2026.04
22.7
2025.02
22.02
2026.04
21.5
2026.04
21.5
2026.04
21.4
2026.04
19.9
2026.04
19
2026.04
18.7