AhaBulbStuff Everyone Should Know
AhaBulb: Stuff Everyone Should Know · ahabulb.com/ai-hallucination-rewarded-guessing

A 2025 paper by OpenAI researchers says hallucinations persist partly because most evaluations reward what?

Start playing free

  1. Guessing over admitting uncertainty
  2. Short answers over long ones
  3. Polite answers over rude ones

🔒 The answer is for AhaBulb members. Members see the answer, why it's right and the official sources for every card of Stuff Every AI Expert Should Know, plus timed practice exams.

The first 50 cards of every test are free to play.

Play the free cards See membership

Like this card?
Play the game

References

  1. Kalai et al., Why Language Models Hallucinate (arXiv 2509.04664) arxiv.org/abs/2509.04664 Checked 2026-10-10.

Think this answer is wrong? Tap "Challenge this answer" on the card in the game and tell us why.

Test yourself in the game

Artificial IntelligenceIn generative AI, what is a hallucination?Artificial IntelligenceWhat term does NIST's generative AI risk profile use for AI hallucinations?Artificial IntelligenceWhat is benchmark contamination for a language model?

That's stuff everyone should know.