Research map / Large language models / Instruction and fine-tuning
Process for Adapting Language Models to Society (PALMS) with Values-Targeted Datasets
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 |
|---|---|---|---|
| fine-tune GPT-3 on small values-targeted dataset (80 samples) instead of base model | GPT-3 (125M to 175B) · Test prompts, Perspective API | mean toxicity score | Helped |
| fine-tune on values-targeted dataset instead of control dataset (books and Wikipedia) (ablation) | GPT-3 · Test prompts, Perspective API | mean toxicity score | Helped |
| fine-tune on control dataset instead of base model (ablation) | GPT-3 · Test prompts, Perspective API | mean toxicity score | Helped |
| fine-tune GPT-3 on values-targeted dataset instead of base model | GPT-3 (125M to 175B) · Test prompts, human evaluation | mean adherence rating (1-5) | Helped |
| increase model size under values-targeted fine-tuning | PALMS · Test prompts, human evaluation | mean adherence rating (1-5) | Helped |
| fine-tune on values-targeted dataset instead of base model | GPT-3 · Co-occurrence (gender, religion, race) | top descriptive word neutrality (qualitative) | Mixed |
| fine-tune on values-targeted dataset instead of base model | GPT-3 175B · Capability evaluations (as in GPT-3 paper) | accuracy | No clear effect |
Closest papers by meaning
- Raising the Bar: Investigating the Values of Large Language Models via Generative Evolving Testing (ICML 2025 · similarity 0.86)
- CultureLLM: Incorporating Cultural Differences into Large Language Models (NeurIPS 2024 · similarity 0.86)
- CLAVE: An Adaptive Framework for Evaluating Values of LLM Generated Responses (NeurIPS 2024 · similarity 0.86)
- Can DPO Learn Diverse Human Values? A Theoretical Scaling Law (NeurIPS 2025 · similarity 0.85)
- SocialStigmaQA: A Benchmark to Uncover Stigma Amplification in Generative Language Models (AAAI 2024 · similarity 0.85)
- CulturePark: Boosting Cross-cultural Understanding in Large Language Models (NeurIPS 2024 · similarity 0.85)
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.