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

Truthful Question Answering on TruthfulQA (accuracy, intervention rate)

70Accuracy

Aligned base model

38.846.95563.1Apr 22, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
70-
2026.04
70-
2026.04
70-
2026.04
67-
2026.04
67-
2026.04
67-
2026.04
58-
2026.04
58-
2026.04
58-
2026.04
5120
2026.04
5120
2026.04
5031.9
2026.04
5019.2
2026.04
50-
2026.04
50-
2026.04
4917.2
2026.04
48-
2026.04
4831.9
2026.04
48-
2026.04
4717.2
2026.04
46-
2026.04
46-
2026.04
4519.2
2026.04
4320.9
2026.04
4320.9
2026.04
4231.5
2026.04
40-
2026.04
4031.5
2026.04
40-
2026.04
40-