AI Paper Atlas

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

Enhancing Uncertainty Estimation in LLMs with Expectation of Aggregated Internal Belief

Zeguan Xiao, Diyang Dou, Boya Xiong, Yun Chen, Guanhua Chen

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Statistical frameworks for LLM evaluation · 5 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
aggregate hidden states from the last n layers instead of only the last layer (ablation)EAGLE (Llama3 8B) · TriviaQAECE x100: 18.8 -> 1.7Improved
aggregate hidden states from the last n layers instead of only the last layer (ablation)EAGLE (Qwen2.5 7B) · GSM8KECE x100: 8.5 -> 3.0Improved
aggregate logits across layers rather than probabilities (ablation)EAGLE (Llama3 8B) · TriviaQAECE x100: 63.5 -> 1.7Improved
use expectation over score distribution instead of the max-probability score (ablation)EAGLE (Llama3 8B) · TriviaQAECE x100: 28.3 -> 1.7Improved
use expectation over score distribution instead of the max-probability score (ablation)EAGLE (Qwen2.5 7B) · TriviaQAECE x100: 31.8 -> 22.1Improved
use EAGLE confidence to pick among 5 sampled answersLlama3 8B · TriviaQAaccuracy: 71.3 -> 74.8Improved
use EAGLE confidence to pick among 5 sampled answersQwen2.5 7B · TriviaQAaccuracy: 67.5 -> 69.4Improved
replace the score-criteria prompt with a CoT explanation prompt (ablation)EAGLE (Llama3 8B) · TriviaQAECE x100: 1.7 -> 13.6Hurt
remove score criteria from the self-evaluation prompt (ablation)EAGLE (Llama3 8B) · TriviaQAECE x100: 1.7 -> 4.4Hurt
vary prompt variant (ablation)EAGLE (Llama3 8B) · GSM8KECE x100: 7.6 / 6.6 / 5.8No clear effect
narrow the confidence score range from 0-9 to 0-1 (ablation)EAGLE (Llama3 8B) · GSM8KECE x100: 7.6 -> 11.1Hurt
narrow the confidence score range from 0-9 to 0-4 (ablation)EAGLE (Llama3 8B) · GSM8KAUROC x100: 68.3 -> 64.7Hurt

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