Milestone papers from DeepSeek's model lineage — from the original DeepSeek LLM, DeepSeek Coder, DeepSeekMath (which introduced GRPO), and DeepSeekMoE, through the V2 and V3 architectures, to the R1 reasoning model and V4.

Ilya Sutskever, former chief scientist of OpenAI, recommended a reading list to John Carmack and said, "If you really learn all of these, you’ll know 90% of what matters today.". This paper list is not verified by Ilya, but maybe a good start for LLM.

Milestone papers from OpenAI's GPT lineage — from GPT-1's generative pretraining, through GPT-2 and GPT-3 scaling, Codex's code generation, and InstructGPT's RLHF alignment, to GPT-4.