AI Paper Atlas

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

Calibrating Verbalized Confidence with Self-Generated Distractors

Victor Wang, Elias Stengel-Eskin

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Uncertainty quantification for LLMs · 7 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
Normalize verbalized confidence over self-generated distractors and combine with self-consistency (DiNCo)LLM verbalized confidence · TriviaQA and SimpleQAECE: avg ECE gain 0.077 (TriviaQA), 0.092 (SimpleQA) over best baselineImproved
Add self-consistency to NVC (DiNCo vs NVC alone) (ablation)DiNCo · SimpleQA GPT-4.1AUC: 0.729 -> 0.786Improved
Add distractor normalization to verbalized confidence (NVC vs VC) (ablation)Qwen3-8B · TriviaQAECE: 0.240 -> 0.171Improved
Add distractor normalization on Qwen3-1.7B (NVC vs DiNCo) (ablation)DiNCo · TriviaQA Qwen3-1.7BECE: 0.084 (NVC) vs 0.177 (DiNCo)Hurt
Scale self-consistency to 100 samples to match DiNCo (ablation)Self-consistency · TriviaQA Qwen3-8BcalibrationNo clear effect
Remove NLI-based weighting of distractors (ablation)DiNCo · TriviaQA Qwen3-8BcalibrationHurt
Black-box setting (no logit access) vs logit access (ablation)DiNCo · SimpleQA GPT-4.1AUC: 0.786 -> 0.605Hurt
Apply DiNCo to long-form biography generationQwen3-8B · FactScoreECE: 0.162 (SC) -> 0.076Improved
Apply DiNCo vs SC on FactScore correlationDiNCo · FactScorePearson / Spearman: +0.072 / +0.074 avg over SCImproved
Use DiNCo instead of MSP on GPT-4.1 SimpleQA AUCDiNCo · SimpleQA GPT-4.1AUC: 0.800 (MSP) vs 0.786Hurt
Use DiNCo vs Gemma-3-4B FactScore ECE of NVC (ablation)DiNCo · FactScore Gemma-3-4B-ITECE: 0.123 (NVC) vs 0.172Hurt

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