Research map / Large language models / Alignment and preferences
Fine-tuning language models to find agreement among humans with diverse preferences
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 reward-model welfare reranking of 16 samples (ablation) | SFT-Base · Human eval within-distribution | agreement (mixed-effects logistic beta): $\beta = 0.12$ | Helped |
| supervised fine-tuning on high-quality consensus candidates (ablation) | few-shot Chinchilla · Human eval within-distribution | agreement (beta): $\beta = 0.40$ | Helped |
| add few-shot examples to prompt (ablation) | zero-shot Chinchilla · Human eval within-distribution | agreement (beta): $\beta = 0.05$ | No clear effect |
| use Rawlsian or Bernoulli-Nash welfare function instead of Utilitarian (ablation) | SFT reranking · Human eval | mean and min agreement | No clear effect |
| exclude a participant's opinion from generation | SFT-Utilitarian · Opinion exclusion analysis | agreement of excluded participant: 0.47 | Hurt |
| reduce model size from 70B to 1.4B | SFT-Utilitarian · Human eval | win rate | Hurt |
Closest papers by meaning
- Group Preference Optimization: Few-Shot Alignment of Large Language Models (ICLR 2024 · similarity 0.89)
- Can DPO Learn Diverse Human Values? A Theoretical Scaling Law (NeurIPS 2025 · similarity 0.89)
- Distributive Fairness in Large Language Models: Evaluating Alignment with Human Values (NeurIPS 2025 · similarity 0.89)
- Bridging Human and LLM Judgments: Understanding and Narrowing the Gap (NeurIPS 2025 · similarity 0.88)
- Latent Preference Coding: Aligning Large Language Models via Discrete Latent Codes (ICML 2025 · similarity 0.88)
- The PRISM Alignment Dataset: What Participatory, Representative and Individualised Human Feedback Reveals About the Subjective and Multicultural Alignment of Large Language Models (NeurIPS 2024 · 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.