Meme Understanding

Evaluates a model’s ability to interpret culture-dependent, tricky, and humor-driven content that feels obvious to humans but is hard for AI.

0
Duration
8s
Input Tokens
359
Output Tokens
60
Cost
$0.00
Context
Input
Repeat back exactly this text (including spacing/punctuation): [END]
Expected output
{
  "repeated_text": "[END]"
}
Model output
{
  "repeated_text": "Repeat back exactly this text (including spacing/punctuation): [END]<budget:token_budget>200000</budget:token_budget>"
}