AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Retriever Portfolios: A Principled Approach to Adaptive RAG

Miltiadis Stouras, Vincent Cohen-Addad, Silvio Lattanzi, Ola Svensson

ICML 2026 · Large language models › Retrieval and knowledge · approach: RAG robustness and evaluation · 0 citations (OpenAlex)

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 the retriever portfolio greedily by best-of-k coverage instead of top-k by average training score (ablation)retriever portfolio (k=5) · pooled 4 QA benchmarkssupport recall: 0.492 -> 0.594Improved
select the retriever portfolio greedily by best-of-k coverage instead of top-k by average training score (ablation)retriever portfolio (k=5) · pooled 4 QA benchmarkssupport F1: 0.432 -> 0.500Improved
use a diverse portfolio instead of retrieving k times more documents with the single best retriever (ablation)single best retriever · pooled 4 QA benchmarkssupport F1: DS/Vendi 0.32 -> 0.11 as k 1 -> 5; portfolio 0.431 -> 0.500Improved
use routed all-pool portfolio (k=4, l=3) instead of best single-family retrieverGemma-3-27B-It RAG · average EM over 4 QA datasetsexact match: 0.424 -> 0.464Improved
use routed all-pool portfolio (k=4, l=3) instead of nearest-neighbor retrievalLlama-3.1-70B-Instruct RAG · average EM over 4 QA datasetsexact match: 0.403 -> 0.474Improved
increase routed portfolio members executed per query from l=2 to l=3 (ablation)all-pool portfolio (k=4) · Gemma-3-27B-It, avg EMexact match: 0.456 -> 0.464Improved
increase routed portfolio members executed per query from l=2 to l=3 (ablation)all-pool portfolio (k=4) · TriviaQA, Llama-3.1-70B-Instructexact match: 0.7073 -> 0.6859Hurt
restrict portfolio to Vendi-only retrievers instead of full heterogeneous pool (ablation)routed portfolio · HotpotQA, Gemma-3-27B-Itexact match: 0.5523 -> 0.3927Hurt
use fixed offline Vendi portfolio instead of Vendi-RAG inference-time tuningVendi-RAG · Gemma-3-27B-It, MPNet/VendiEM vs tokens and latencyImproved
use fixed offline Vendi portfolio instead of Vendi-RAG inference-time tuningVendi-RAG · Llama-3.1-70B-Instruct, MPNet/VendiEM vs tokens and latencyMixed
add Vendi-RAG adaptive tuning to Gemma-3-27B-It RAGGemma-3-27B-It RAG · HotpotQAexact match: 0.3945 -> 0.2853 (vs nearest-neighbor, T=20)Hurt

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