Research map / Large language models / Language, safety and interpretability
Gradients with Respect to Semantics Preserving Embeddings Tell the Uncertainty of Large Language Model
Mingda Li, Rundong Lv, Xinyu Li, Weinan Zhang, Ting Liu
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 |
|---|---|---|---|
| add entropy re-weighting of token log-likelihoods (ablation) | SemGrad Qwen3-4B · TriviaQA | AUROC: 76.83 -> 80.40 | Improved |
| add entropy re-weighting of token log-likelihoods (ablation) | SemGrad Llama3.1-8B · TriviaQA | AUROC: 81.28 -> 84.72 | Improved |
| use semantic preserving token t* instead of last input token (ablation) | SemGrad Qwen3-4B · TruthfulQA | AUROC: 63.35 -> 69.06 | Improved |
| use semantic preserving token t* instead of last input token (ablation) | SemGrad Qwen3-4B · SciQ | AUROC: 70.49 -> 72.20 | Improved |
| use l1 norm instead of l2 norm (ablation) | SemGrad · SciQ/TriviaQA/TruthfulQA | AUROC: 72.07 -> 72.20 | No clear effect |
| use top-half layer span instead of all layers 1:(L-1) (ablation) | SemGrad Qwen3-4B · TruthfulQA | AUROC: 66.60 -> 69.06 | Improved |
| use top-half layer span instead of single layer L-1 (ablation) | SemGrad Llama3.1-8B · SciQ | AUROC: 74.64 -> 75.76 | Improved |
| use SemGrad instead of SAR on multi-answer data | SAR · TruthfulQA | AUROC: +3.27 points | Improved |
| use SemGrad instead of ExGrad on multi-answer data | ExGrad · TruthfulQA | AUROC: +6.82 points | Improved |
| use SemGrad instead of ParaGrad on multi-answer data | ParaGrad · TruthfulQA | AUROC: +3.3 | Improved |
| use SemGrad instead of ParaGrad on single-answer data | ParaGrad · TriviaQA Qwen3-4B | AUROC: 82.02 -> 80.40 | Hurt |
| combine SemGrad and ParaGrad into HybridGrad | SemGrad · Average over 9 pairs | AUROC: 75.75 -> 76.59 | Improved |
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.
- Semantic Density — Semantic Density: Uncertainty Quantification for Large Language Models through Confidence Measurement in Semantic Space (NeurIPS 2024) · better on 1 of 1 shared result
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
- Inv-Entropy: A Fully Probabilistic Framework for Uncertainty Quantification in Language Models (NeurIPS 2025 · similarity 0.88)
- SPANUQ: Span-Level Uncertainty Quantification for Large Language Model Generation (NeurIPS 2026 · similarity 0.88)
- Conformal Prediction with Paraphrase-Aware Scoring for LLM Uncertainty Quantification (NeurIPS 2026 · similarity 0.87)
- Efficient semantic uncertainty quantification in language models via diversity-steered sampling (NeurIPS 2025 · similarity 0.87)
- Language Model Cascades: Token-Level Uncertainty And Beyond (ICLR 2024 · similarity 0.87)
- Conformal Prediction Beyond the Seen: A Missing Mass Perspective for Uncertainty Quantification in Generative Models (NeurIPS 2025 · similarity 0.87)
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.
