LLMs as High-Dimensional Nonlinear Autoregressive Models with Attention: Training, Alignment and Inference
基于注意力机制的高维非线性自回归模型:训练、对齐与推理
机构 * Cornell University(康奈尔大学)
专题命中 偏好对齐 :alignment(title,abstract);RLHF(abstract);DPO(abstract);分类 cs.CL、cs.AI、cs.LG
AI总结 本文将LLMs表述为具有注意力依赖的高维非线性自回归模型,探讨了训练、对齐与推理的原理及方法。
Comments 27 pages, 12 figures. Mathematical survey framing LLMs as high-dimensional nonlinear autoregressive models with attention, covering training, alignment, and inference, with nanoGPT/nanochat-style code examples. Feedback welcome