AI Paper Atlas

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

Annotation-Efficient Honesty Alignment via Confidence Elicitation and Calibration

Shiyu Ni, Keping Bi, Jiafeng Guo, Minghao Tang, Jingtong wu, Zengxin Han, Xueqi Cheng

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Statistical frameworks for LLM evaluation

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 confidence elicitation stage before calibration (1k annotations) (ablation)Cal-Only · HonestyBench in-domain avg (Qwen2.5-7B)AUROC: 73.41 -> 84.36Improved
add confidence elicitation stage before calibration (1k annotations) (ablation)Cal-Only · HonestyBench OOD avg (Qwen2.5-7B)AUROC: 77.32 -> 84.47Improved
add confidence elicitation before calibration with all 560k annotations (ablation)Cal-Only (560k) · HonestyBench in-domain avg (Qwen2.5-7B)AUROC: 86.20 -> 86.49No clear effect
add calibration stage with 1k annotations after elicitation (ablation)Eli-Only · HonestyBench in-domain avg (Qwen2.5-7B)AUROC: 71.19 -> 84.36Improved
replace Consis-Sem inference-time sampling with trained Eli-Only one-shot confidenceConsis-Sem · HonestyBench in-domain avg (Qwen2.5-7B)AUROC: 73.62 -> 71.19No clear effect
add elicitation then 1k calibration at 32B scaleCal-Only (1k) · HonestyBench in-domain avg (Qwen2.5-32B)AUROC: 79.31 -> 84.39Improved
add elicitation then 1k calibration at 32B scaleCal-Only (1k) · HonestyBench OOD avg (Qwen2.5-32B)AUROC: 79.62 -> 84.01Improved
add EliCal 1k calibration versus best training-free Consis-SemConsis-Sem · HonestyBench in-domain avg (Qwen2.5-7B)AUROC: 73.62 -> 84.36Improved
add temperature scaling (Thermometer) instead of learned elicitationEliCal · HonestyBench in-domain avg (Qwen2.5-7B)AUROC: 84.36 -> 59.48Hurt
increase self-consistency samples k from 2 to 20 for Consis-Sem (ablation)Consis-Sem · HonestyBench in-domain avg (Qwen2.5-7B)AUROC: 68.04 -> 73.62Improved
increase self-consistency samples k from 2 to 20 for EliCal (ablation)EliCal (1k) · HonestyBench in-domain avg (Qwen2.5-7B)AUROC: 84.41 -> 84.36No clear effect
increase elicitation training data size (ablation)Eli-Only · HonestyBench in-domain avg (Qwen2.5-7B)AUROC: approaches Consis-Sem 73.620Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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