AI Paper Atlas

Research map / Large language models / Alignment and preferences

Lost in Benchmarks? Rethinking Large Language Model Benchmarking with Item Response Theory

Hongli Zhou, Hui Huang, Ziqing Zhao, Lvyuan Han, Huicheng Wang, Kehai Chen, Muyun Yang, Wei Bao, Jian Dong, Bing Xu, Conghui Zhu, Hailong Cao and 1 more

AAAI 2026 · Large language models › Alignment and preferences · approach: LLM-as-judge and feedback data · 36 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 PSN-IRT instead of traditional IRT estimation (MLE, MCMC, VI, VIBO)IRT 4PL · response predictionaccuracy: 0.7211 -> 0.7998Improved
use PSN-IRT 4PL instead of Deep-IRTIRT estimation · rank stabilityKendall's tau: 0.9697 -> 1.0000Improved
replace one-hot inputs with Instructor semantic embeddings (ablation)PSN-IRT · response predictionaccuracy: 0.7998 -> 0.7808Hurt
replace MLP backbone with a GNN (ablation)PSN-IRT · rank stabilityKendall's tau: 1.0000 -> 0.7273Hurt
select 1000 items by Fisher information instead of using all items (ablation)benchmark construction · without weak modelsKendall's tau vs arena ranking: 0.2381 -> 0.9048Improved
select 1000 items by discriminability instead of Fisher information (ablation)benchmark construction · without weak modelsscore variance: 0.0030 -> 0.0000Hurt
select items by clustering success/failure vectors (ablation)benchmark construction · without weak modelsKendall's tau vs arena ranking: 0.2381 -> 0.2381No clear effect
select 1000 items at random instead of all items (ablation)benchmark construction · with weak modelsKendall's tau vs arena ranking: 0.6444 -> 0.6444No clear effect
increase Fisher-selected set size from 400 to 1000 items (ablation)benchmark construction · without weak modelsKendall's tau vs arena ranking: 0.7143 -> 0.9048Improved
train PSN-IRT on 50% random item subset (ablation)PSN-IRT · item parameter stabilityPearson correlation of difficulty with full-data estimate: 0.7437 (50%) vs 0.9519 (70%)No clear effect

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