AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

Anytime-Valid Inference for Online Ranking of Large Language Models

Runzhe Gu, Wenguang Sun, Bowen Gang, Xintao Xia

ICML 2026 · Large language models › Reasoning and chain of thought · approach: Test-time compute scaling

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes SERPANT, an anytime-valid sequential procedure for online pairwise-preference ranking of LLMs that remains statistically valid under adaptive sampling and continuous monitoring.

Problem

Online LLM evaluation from sequentially collected pairwise preferences: adaptive sampling and continuous monitoring invalidate classical fixed-sample inference, so existing ranking procedures are largely heuristic.

Approach

Applies anytime-valid sequential inference to online pairwise-preference ranking so that model rankings can be stopped and monitored at any time with valid error guarantees. Details beyond the abstract are not available.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add transitivity-based pruning (ablation)SERPANT (random sampling) · Databricks, t=2000partial-order pairs discovered: 35 (78%) -> 41 (91%)Improved
add transitivity-based pruning (ablation)SERPANT (random sampling) · Databricks, t=1000partial-order pairs discovered: 24 (53%) -> 28 (62%)Improved
add transitivity-based pruning (ablation)SERPANT (random sampling) · Databricks, t=3000partial-order pairs discovered: 38 (84%) -> 43 (96%)Improved
replace random pair sampling with tournament samplingSERPANT · Databricks, t=500partial-order pairs discovered: 5 (11%) -> 12 (27%)Improved
replace random pair sampling with tournament samplingSERPANT · TriviaQA, t=2000partial-order pairs discovered: 23 (51%) -> 32 (71%)Improved
replace random pair sampling with tournament samplingSERPANT · SQuAD2.0, t=2000partial-order pairs discovered: 29 (64%) -> 30 (67%)Improved
replace random pair sampling with tournament samplingSERPANT · MMLU, t=2000partial-order pairs discovered: 33 (73%) -> 32 (71%)No clear effect
replace random pair sampling with tournament samplingSERPANT · Databricks, t=2000partial-order pairs discovered: 41 (91%) -> 41 (91%)No clear effect
replace random pair sampling with tournament samplingSERPANT · simulation, m=20, w1=0.2, w2=0.8powerImproved
replace random pair sampling with tournament samplingSERPANT top-10 sets · simulation, m=20top-10 set sizeImproved
replace empirical win-rate ranking with SERPANT e-process testingranking of 10 LLMs · Databricks, t=4000false discoveries: 4 (9%) -> 1 (2%)Improved
replace empirical win-rate ranking with SERPANT e-process testingranking of 10 LLMs · Databricks, t=4000true discoveries: 41 (91%) -> 32 (71%)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