ARI: A Retriever That Learns From Reasoning-Model Preferences at Inference, Boosting Recall@100 by 7x
lateinteraction · x · 2026-11-07
Diane Tchuindjo, Devavrat Shah and Omar Khattab introduce the Amortized Reasoning Index (ARI), a new kind of retriever that updates its weights during search using gradients derived at inference time from a reasoning model's ranking preferences.
- Motivation: agents spend lots of reasoning judging which search results matter, then throw that judgment away on compaction. ARI "amortizes" it, so reasoning spent on past queries keeps improving future search.
- Results: with no relevance labels, a 400M cross-encoder that starts weak grows Recall@100 on held-out queries by up to 7x, ending above Qwen3-Embedding-8B on oblique, reasoning-intensive search tasks.
A blog post is up, with preliminary results and discussions to follow.
More from coding & agent
- Claude hands user a Lovecraftian Bash script then says 'things to look for are…' — teodorio · 2026-11-07
- iOS 27.3 reveals Siri's memory graph: memory atoms fed by Fastpass and agent recipes — alexisgallagher · 2026-11-07
- Indie dev adds AI agent contact editing with per-change undo and approval gates — SeanOliver · 2026-11-07
- Designing an MCP server for 60+ financial datasets: entity keys, per-connector auth, context budgets — vianexus_official · 2026-11-07
- zeeg to livestream Junior in action, breaking down technical details of the agent — zeeg · 2026-11-07
- Sketching musical interfaces with ASCII art using Gemini in Google Antigravity — alexanderchen · 2026-11-07