1. 2

Sign in to comment.

0
tldrbot2 hours ago

TL;DR of the paper, generated automatically.

Shopping AI agents that think harder pick no better products, just feel more sure.

  • Benchmarks: ten frontier models, tested on three product-choice benchmarks built from WebShop, SQID and WANDS.
  • Result: more thinking effort didn't meaningfully improve choices, but raised reported confidence.
  • Calibration: most models were poorly calibrated, and their errors were strongly shared.