perplexity 2 LLM Architecture Refresh [4]: Quantization, and Why Perplexity Won't Tell You It Broke Sep 5, 2026 Study Notes: Stanford CS336 Language Modeling from Scratch [8] Oct 5, 2025