Research map / Large language models / Language, safety and interpretability
Rethinking Uncertainty Estimation in LLMs: A Principled Single-Sequence Measure
Lukas Aichberger, Kajetan Schweighofer, Sepp Hochreiter
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Generate the approximating sequence with greedy decoding instead of multinomial sampling (ablation) | NLL of most likely sequence (MSP) · TriviaQA | AUROC | Improved |
| Use beam search instead of greedy decoding to approximate the most likely sequence (ablation) | G-NLL · TriviaQA | AUROC | Improved |
| Replace sampling-based baselines by single greedy sequence (G-NLL) averaged over all scenarios | PE/SE family · QA average over 18 scenarios | AUROC: .707 -> .721 | Improved |
| Add length normalization to the single-sequence measure (ablation) | G-NLL · QA datasets | AUROC | Hurt |
| Use G-NLL instead of sampling-based measures on 70B IT long answers | Llama-3.1-70B IT · QA AUROC, long answers, LLM correctness | AUROC: .600 -> .562 | Hurt |
Closest papers by meaning
- Graph-based Uncertainty Metrics for Long-form Language Model Generations (NeurIPS 2024 · similarity 0.92)
- SPANUQ: Span-Level Uncertainty Quantification for Large Language Model Generation (NeurIPS 2026 · similarity 0.92)
- Kernel Language Entropy: Fine-grained Uncertainty Quantification for LLMs from Semantic Similarities (NeurIPS 2024 · similarity 0.92)
- Benchmarking LLMs via Uncertainty Quantification (NeurIPS 2024 · similarity 0.91)
- FLUE: Streamlined Uncertainty Estimation for Large Language Models (AAAI 2025 · similarity 0.91)
- Improving Uncertainty Estimation through Semantically Diverse Language Generation (ICLR 2025 · similarity 0.91)
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.
