TY  - RPRT
TI  - Biasless Language Models Learn Unnaturally: How LLMs Fail to Distinguish the Possible from the Impossible
AU  - Imry Ziv
AU  - Nur Lan
AU  - Emmanuel Chemla
PY  - 2026
DO  - 10.18653/v1/2026.eacl-long.249
UR  - https://arxiv.org/abs/2510.07178
ID  - 2510.07178
ER  -