AI Paper Atlas

Research map / Large language models / Alignment and preferences

Search Arena: Analyzing Search-Augmented LLMs

Mihran Miroyan, Tsung-Han Wu, Logan King, Tianle Li, Jiayi Pan, Xinyan Hu, Wei-Lin Chiang, Anastasios Nikolas Angelopoulos, Trevor Darrell, Narges Norouzi, Joseph E. Gonzalez

ICLR 2026 · Large language models › Alignment and preferences · approach: LLM-as-judge and feedback data · 34 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
use a reasoning model for search-augmented chatsearch-augmented LLMs · Search Arenaaverage win rate: top three models over 60.0% average win rateImproved
increase search context size from medium to high (ablation)sonar-pro · Search Arenaaverage win rate: 57.6% -> 63.9%Improved
increase search context size from medium to high (ablation)GPT-4o search models · Search Arenaaverage win rateNo clear effect
longer responsesuser preference (Bradley-Terry) · Search ArenaBT coefficient: beta_length = 0.334Improved
longer responses on Factual Lookup promptsuser preference (Bradley-Terry) · Search Arena (Factual Lookup)BT coefficient: 0.334 -> 0.156Mixed
more cited sourcesuser preference (Bradley-Terry) · Search ArenaBT coefficient: beta_citations = 0.209Improved
cite Wikipediauser preference (Bradley-Terry) · Search ArenaBT coefficient: beta_wiki = -0.071Hurt
cite tech, community or social platformsuser preference (Bradley-Terry) · Search ArenaBT coefficient: 0.073, 0.061, 0.057Improved
add irrelevant (non-supporting) citationsuser preference (Bradley-Terry) · Search Arena (citation subset)BT coefficient: beta_irrelevant = 0.27 vs beta_support = 0.29Improved
add contradicting claim-citation pairsuser preference (Bradley-Terry) · Search Arena (citation subset)BT coefficientNo clear effect
add web search to a non-search LLM (ablation)Gemini-2.5 Pro Experimental · Text Arenavote preference: 143 search vs 156 non-search, p=0.244No clear effect
add web search to a non-search LLM (ablation)Gemini-2.5 Pro Experimental · Text Arena (Factual Lookup)vote preference: p-value=0.012Improved

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