AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

Relevance-Based Embeddings: Lightweight Candidate Retrieval via Heavy-Ranker Calls

Kirill Sergeevich Shevkunov, Andrey Ploskonosov, Liudmila Prokhorenkova

ICML 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Metric and ordinal embedding

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
select support items by l2-greedy instead of random (ablation)AnnCUR · YugiohHitRate(100): 0.4724 -> 0.5618Improved
select support items by l2-greedy instead of random (ablation)AnnCUR · Dr.WhoHitRate(100): 0.1919 -> 0.2960Improved
select support items by l2-greedy instead of random (ablation)AnnCUR · QA.SmallHitRate(100): 0.5828 -> 0.6100Improved
select support items by l2-greedy instead of random (ablation)AnnCUR · RecMusicHitRate(100): 0.1478 -> 0.1478No clear effect
select popular items as support items instead of random (ablation)AnnCUR · YugiohHitRate(100): 0.4724 -> 0.2429Hurt
select popular items as support items instead of random (ablation)AnnCUR · RecGames2HitRate(100): 0.6697 -> 0.7623Improved
select KMeans cluster centers as support items instead of random (ablation)AnnCUR · StarTrekHitRate(100): 0.2287 -> 0.3226Improved
select KMeans cluster centers as support items instead of random (ablation)AnnCUR · QA.SmallHitRate(100): 0.5828 -> 0.5578Hurt
add trainable relevance mappings (RBE) to KMeans support itemsAnnCUR+KMeans · RecMusicHitRate(100): 0.1661 -> 0.3729Improved
add trainable relevance mappings (RBE) to l2-greedy support itemsAnnCUR+l2-greedy · QAHitRate(100): 0.5700 -> 0.6022Improved
add trainable relevance mappings (RBE) to l2-greedy support itemsAnnCUR+l2-greedy · MilitaryHitRate(100): 0.3357 -> 0.3349No clear effect
use RBE instead of dual encoderRecGames1 · RecGames1 (X=100)HitRate(X,X): 0.7048 -> 0.6682Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy