Research map / Large language models / AI and society
Dropouts in Confidence: Moral Uncertainty in Human-LLM Alignment
Jea Kwon, Luiz Felipe Vecchietti, Sungwon Park, Meeyoung Cha
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 inference-time attention dropout 0.10 (ablation) | Llama-3.1-70B · Moral Machine alignment | L2 distance to human AMCE: 0.703 -> 0.550 | Improved |
| add inference-time attention dropout 0.10 (ablation) | Qwen3-1.7B · Moral Machine alignment | L2 distance to human AMCE: 1.808 -> 1.300 | Improved |
| add inference-time attention dropout 0.10 (ablation) | Llama-3.2-1B · Moral Machine alignment | L2 distance to human AMCE: 1.170 -> 1.291 | Hurt |
| add inference-time attention dropout 0.10 (ablation) | Qwen2.5-14B · Moral Machine alignment | L2 distance to human AMCE: 0.958 -> 1.066 | Hurt |
| add inference-time attention dropout 0.10 (ablation) | Gemma-3-27b · Moral Machine alignment | L2 distance to human AMCE: 1.312 -> 1.289 | No clear effect |
| add inference-time attention dropout across 32 LLMs (ablation) | 32 open-source LLMs · Moral Machine alignment | alignment (delta L2) | Improved |
| add attention dropout (0, 0.05, 0.1) (ablation) | 32 LLMs · UncertaintySet | mutual information / total entropy | Improved |
| add attention dropout (0, 0.05, 0.1) (ablation) | 32 LLMs · UncertaintySet | conditional entropy | No clear effect |
| use Option A/B prompt format instead of Case 1/2 with dropout (ablation) | Llama3.2-3B · Moral Machine alignment | delta L2: -0.28 -> -0.11 | Mixed |
| add attention dropout 0.10 (blind human eval) (ablation) | Qwen3-1.7B · Moral Machine alignment | MSE of human choices: 0.369 -> 0.263 | Improved |
Closest papers by meaning
- Language Model Alignment in Multilingual Trolley Problems (ICLR 2025 · similarity 0.89)
- Unveiling the Bias Impact on Symmetric Moral Consistency of Large Language Models (NeurIPS 2024 · similarity 0.89)
- MoralReason: Generalizable Moral Decision Alignment for LLM Agents Using Reasoning-Level Reinforcement Learning (AAAI 2026 · similarity 0.88)
- EAI: Emotional Decision-Making of LLMs in Strategic Games and Ethical Dilemmas (NeurIPS 2024 · similarity 0.88)
- Many LLMs Are More Utilitarian Than One (NeurIPS 2025 · similarity 0.88)
- Strong Preferences Affect the Robustness of Preference Models and Value Alignment (ICLR 2025 · similarity 0.88)
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.
