notesum.ai
Published at October 22Optimizing LLMs with Direct Preferences: A Data Efficiency Perspective
cs.AI
cs.LG
Released Date: October 22, 2024
Authors: Pietro Bernardelle1, Gianluca Demartini
Aff.: 1The University of Queensland, Brisbane, Australia

| Dataset | Size | Training (80%) | Evaluation (10%) | Testing (10%) | Prompt Type |
|---|---|---|---|---|---|
| Dataset Aa | 7,560 | 6,048 | 756 | 756 | Conversational |
| Dataset Bb | 12,900 | 10,320 | 1,290 | 1,290 | Question-Answering |
| Dataset Cc | 63,600 | 50,880 | 6,360 | 6,360 | Question-Answering |
| Combination | 84,060 | 67,248 | 8,406 | 8,406 | Conversational & |
| Question-Answering |