latimal/benchmarks

refactor: tables format, dark theme, food-embed.latimal.com

Honest numbers: clean LLM expansions v5 (no benchmark leakage)

Final update: multi-vector retrieval (global 0.932, concept 0.844)

Update Category Search to 0.843 (concept-specific retrieval fix)

Update to v9+adaptive routing production scores (May 2026)

Update benchmarks with LoRA-bev-v9 results (May 2026)

Add Latimal Menu Intelligence benchmark chart

initial commit