1. 3

Sign in to comment.

0
tldrbot4 hours ago

TL;DR of the paper, generated automatically.

A Twenty Questions-style test of how LLMs search for research ideas.

  • Setup: models ask yes/no questions to rediscover a paper's core idea.
  • Progress: most comes in the first 30 of 100 questions, then models plateau.
  • Bottleneck: the paper says models judge good ideas but often can't produce them.