arXiv:2601.10421cs.CLcs.AI2026-01被引 68
语言模型不是认知模型，只是工具。
Are Language Models Models?
- 从马尔可夫三层次分析，语言模型在实现层不成立
- 算法表征层动机不足，计算理论层也存问题
- 适合研究工具而非认知模拟，避免夸大其作用
Futrell 和 Mahowald 声称语言模型（LMs）可作为模型系统，但从马尔可夫（Marr）的三个层次评估发现：在实现层，该说法显然不成立；在算法-表征层，动机薄弱；在计算理论层，也存在问题。语言模型更适合作为研究工具，称其为认知模型则过度夸大，且助长了大语言模型的炒作风气。
原文摘要 · Abstract (English)
Futrell and Mahowald claim LMs "serve as model systems", but an assessment at each of Marr's three levels suggests the claim is clearly not true at the implementation level, poorly motivated at the algorithmic-representational level, and problematic at the computational theory level. LMs are good candidates as tools; calling them cognitive models overstates the case and unnecessarily feeds LLM hype.
语言模型认知科学模型评估
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。