AI Paper Atlas

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

Don't Throw Away Your Beams: Improving Consistency-based Uncertainties in LLMs via Beam Search

Ekaterina Fadeeva, Maiya Goloburda, Aleksandr Rubashevskii, Roman Vashurin, Artem Shelmanov, Preslav Nakov, Mrinmaya Sachan, Maxim Panov

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Uncertainty quantification for LLMs · 2 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
replace multinomial sampling with probability-weighted beam search candidatesDissimilarity · 6 QA datasets avg, Gemma 3 4B instructPRR: .206 -> .252Improved
replace multinomial sampling with beam search candidatesEccentricity · 6 QA datasets avg, Qwen 3 8B instructPRR: .251 -> .345Improved
replace multinomial sampling with beam search candidatesCocoaPPL · 6 QA datasets avg, Qwen 3 8B basePRR: .461 -> .461No clear effect
replace multinomial sampling with beam search candidatesDissimilarity · Qwen 3 8B base, 6 QA datasets avgPRR: .477 -> .478No clear effect
increase number of candidates M under beam search (ablation)Dissimilarity, Eccentricity, EigVecDissimilarity · TriviaQA, Gemma 3 4B basePRRImproved
use beam search with M=1 (greedy) (ablation)Dissimilarity · TriviaQA, Gemma 3 4B basePRRHurt
use beam search instead of multinomial sampling on longer generations (ablation)Dissimilarity · TriviaQA and CoQA, Gemma 3 4B basePRRNo clear effect
use diverse beam search instead of plain beam search (ablation)Dissimilarity, Eccentricity, EigVecDissimilarity · TriviaQA and CoQA, Gemma 3 4B base and Llama 3.1 8B basePRRHurt
mix beam candidates with multinomial samples (hybrid) (ablation)Dissimilarity · TriviaQA and CoQA, Gemma 3 4B base and Llama 3.1 8B basePRR: .771 -> .784Mixed
add a probability floor epsilon to beam weights (toward uniform weights) (ablation)beam-weighted estimators · TriviaQA and CoQA, Gemma 3 4B base and Llama 3.1 8B basePRRNo clear effect
apply beam candidates to Semantic Entropy (ablation)Semantic Entropy · 6 QA datasets avg, Gemma 3 4B basePR-AUC: .401 -> .472Improved
apply beam candidates to Degree Matrix and SAR (ablation)Degree Matrix, SAR · 6 QA datasets avg, Gemma 3 4B basePR-AUC: .469 -> .467No 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