TY  - RPRT
TI  - Large Language Models Can Take False First Steps at Inference-time Planning
AU  - Haijiang Yan
AU  - Jian-Qiao Zhu
AU  - Adam Sanborn
PY  - 2026
UR  - https://arxiv.org/abs/2602.02991
ID  - 2602.02991
ER  -