Evolution and compression in LLMs: On the emergence of human-aligned categorization
语言模型中的进化与压缩:关于人类对齐分类的出现
机构 * New York University(纽约大学)
AI总结 研究探讨LLM是否能进化出高效的人类对齐语义系统,通过颜色分类实验发现,大模型在复杂性和英语对齐性上表现各异,仅最强模型能复现人类近最优的信息瓶颈贸易-offs。
Comments Published as a conference paper at ICLR 2026 (The Fourteenth International Conference on Learning Representations). OpenReview: https://openreview.net/forum?id=s7gSTR2AqA¬eId=s7gSTR2AqA