tils (page 2)

• Why Cross‑Entropy Still Rules Loss Functions

  Negative log‑likelihood beats MSE for classification—here’s an intuitive recap

  • ml
  • math
  • code

  2025-05-08
• Exploring Two-Tower Ranking Models

  Generating synthetic data and exploring the two-tower ranking model setup with Margin and InfoNCE loss

  • ml
  • code

  2025-05-04
• Setting Up Postgres for Cloudflare Hyperdrive

  Steps to set up a Postgres instance on Hetzner for Cloudflare Hyperdrive, pgvector, and Cloudflare Functions.

  • devops

  2025-04-28
• How To Classify Gmail Messages With LLMs (v2)

  An upgraded Google Apps Script that uses structured outputs and lean labels to keep your inbox zen

  • gmail
  • productivity

  2025-04-27
• Style vs Substance in Chatbot Arena

  How LMSYS corrects for "pretty formatting" bias when ranking LLMs.

  • ai
  • ml

  2025-03-24