Reranking study: Zerank-2 + LLM second pass hits 0.47 nDCG@10 at ~$0.037/query
CShorten30 · x · 2026-11-16
CShorten30's reranking study (tweet 15/17) maps a clean Pareto frontier at August 2026 prices: the sweet spot is ZeroEntropy zerank-2 at k=2,000 recall plus a GPT-5.6 Luna listwise pass over its top-20, costing $0.037/query and lifting nDCG@10 from 0.181 to 0.470 — with the cross-encoder eating 94% of the cost. The author's original 'Mixture-of-Cross-Encoders' idea failed: models solve different queries, but fixed-weight blends don't convert that into better rankings, pointing toward per-query routing instead.
More from coding & agent
- Your agent's inference endpoint sees all its memory: treat it as a data boundary — krishnan · 2026-11-16
- Dev ships a gladiator management game built entirely with Claude Code — StarDrifterG · 2026-11-16
- Rust rewrite of mcbench-lite cuts first start from ~71s to under 1s with AI coding — adonis_singh · 2026-11-16
- Solo dev ships his son's game idea to iOS and Android for under $300 — morganinuk2 · 2026-11-16
- Enterprise AI coding fails on people, docs and process — not the codebase — _jaydeepkarale · 2026-11-16
- Anthropic: agent memory works best by forgetting selectively—notes beat 400+ archives 4x in Pokémon Red — The Decoder · 2026-11-16