About State-of-the-Art AI: How Modern Models Actually Work

Retrieval-Augmented Generation (RAG) Direct Preference Optimization (DPO) Diffusion Models: Score-Based Image Generation via Reversed Noise Processes Mixture-of-Experts Layers FlashAttention Neural Scaling Laws Low-Rank Adaptation (LoRA) Byte-Pair Encoding Tokenization LLM Weight Quantization Rotary Position Embeddings (RoPE) State-Space Models (Mamba) Speculative Decoding Reinforcement Learning from Human Feedback (RLHF) Reasoning Models and Test-Time Compute Contrastive Language-Image Pretraining (CLIP) Transformer Architecture: Attention, Embeddings, and Feed-Forward Blocks KV Caching and Grouped-Query Attention