Gaming TruthfulQA: Simple Heuristics Exposed Dataset Weaknesses

Link post

(Explanation. Also I have no reason to think they hate me.)

Do not use the original TruthfulQA multiple-choice or the HaluEval benchmark. We show that a simple decision tree can theoretically game multiple-choice TruthfulQA to 79.6% accuracy—even while hiding the question being asked! In response, the TruthfulQA authors created a new multiple-choice condition which avoids the vulnerabilities we highlight.

https://​​turntrout.com/​​original-truthfulqa-weaknesses