Interactive explainer makes TF-IDF, BM25 and NDCG click, BM25 by example
JnBrymn · x · 2026-10-06
A shared interactive tutorial (in the spirit of Bret Victor's "explorable explanations") teaches the core concepts of search and ranking evaluation:
- TF-IDF vs BM25: BM25 adds term-frequency saturation (k1 caps repeat value at k1+1) and length normalization (b), so 5 mentions of ECONNRESET score 1.55 rather than 5, and a 2x-longer page takes a 13% penalty;
- NDCG: the metric for evaluating how well results are ordered, with worked interactive examples.
More from Research
- Spark creator Josh Rosen: AI systems should steal far more from data engineering — colinmcnamara · 2026-10-06
- COLM kicks off its biggest edition ever with a keynote from Chelsea Finn — gregd_nlp · 2026-10-06
- LLM-assisted 22-page preprint settles Hilbert's 12th problem and Zauner's conjecture — basedjensen · 2026-10-06
- 18,000+ runs: new study shows how agent configuration shapes performance on 4 scientific tasks — zeynepakata · 2026-10-06
- Marin team heads to COLM to discuss data, scaling laws, and open source — dlwh · 2026-10-06
- Hamel Husain: similarity metrics like ROUGE don't work for LLM output evals — HamelHusain · 2026-10-06