AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Complementing Self-Consistency with Cross-Model Disagreement for Uncertainty Quantification

Kimia Hamidieh, Veronika Thost, Walter Gerych, Mikhail Yurochkin, Marzyeh Ghassemi

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Statistical frameworks for LLM evaluation · 9 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
add epistemic (cross-model disagreement) term to aleatoric self-consistencyAU (self-consistency) · 8 datasets, Mistral-7B refAUROC: 0.617 -> 0.718Improved
add EU to AU on HotpotQAAU · HotpotQA (5 models avg)AUROC: +0.15Improved
add EU to AU on pooled dataAU · All datasets pooledAUROC: 0.707 -> 0.746Improved
add EU to AU for API modelsAU · SimpleQA (GPT-4o)AUROC: 0.59 -> 0.70Improved
add EU to AUAU · SimpleQA (Claude 3.7 Sonnet)AUROC: 0.53 -> 0.58Improved
add EU to AU on strongest models for XSumAU · XSum (Llama, Qwen)AUROC: 0.708 -> 0.693 (Llama)Hurt
add EU to AU on TruthfulQAAU · TruthfulQAAUROC: 0.528 -> 0.579 (Mistral); 0.47 -> 0.456 (Llama)Mixed
use TU for selective abstentionAU · HotpotQA, XSumAURC: 0.304 -> 0.206 (HotpotQA)Improved
increase number of auxiliary models from 2 to 5 (ablation)TU · Mistral-7B refAUROCImproved
increase samples per model from 5 to 50 (ablation)TU · multiple datasetsAUROCNo clear effect
increase reference model size (Gemma3, Qwen2.5) (ablation)TU and AU · TriviaQAAUROCHurt
replace diverse auxiliary models with noise-perturbed copies of reference (ablation)TU · multiple datasetsTU-AU AUROC gainHurt

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