Reproduce “Eliciting Thinking Hierarchy without a Prior” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yuqing Kong, Yunqi Li, Yubo Zhang, Zhihuan Huang, Jinzhao Wu
Researchers at The Center on Frontiers of Computing Studies, Peking University, introduce a method to elicit a thinking hierarchy from crowd-sourced open responses without prior knowledge. Their approach consistently outperforms traditional plurality voting by analyzing the joint distribution of self-reported answers and predictions of others' answers, successfully identifying correct answers even when they are a minority choice.
Key insights
- More sophisticated thinkers can predict the responses of less sophisticated ones, but generally not vice versa, a principle used to infer thinking hierarchies.
- A latent thinking hierarchy can be learned by analyzing the joint distribution of self-reported answers and predictions of others' answers, without requiring prior knowledge.
- The
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.