luhengshiwo/LLMForEverybody

Provides structured paper-by-paper analysis tracing Transformer's evolution from foundational architectures (Transformer, BERT, GPT series) through multimodal (CLIP, LLaVA) and recent efficient variants (LLaMA, Mistral, Mixtral), with accompanying video walkthroughs and curated interview questions covering core concepts like self-attention, instruction tuning, and MoE routing. The platform integrates Bilibili and YouTube video tutorials alongside paper summaries, enabling learners to progressively build mental models of LLM development trajectories across language, vision, and code domains.

5,847 stars. Actively maintained with 18 commits in the last 30 days.