arXivSophie L. Wang, Amil Dravid, Rulin Shao, Kevin Farhat, Sewon Min, Alexei A. EfrosMon, Oct 5, 2026, 10:59 AM PDT
score 17.0
Simple text cues unlock reasoning in untrained AI models
Original: Base Models Can Reason By Taking a Cue From Training Data
Source: arxiv.org ↗
Writing ELI5 summary…