KTO: Model Alignment as Prospect Theoretic Optimization
Paper
• 2402.01306 • Published
• 21
KTO Fine tuning!
A KTO version EpistemeAI2/Fireball-Alpaca-Llama3.1.07-8B-Philos-Math
Uploaded model
- Developed by: EpistemeAI2
- License: apache-2.0
- Finetuned from model : EpistemeAI2/Fireball-Alpaca-Llama3.1.07-8B-Philos-Math
This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.
Open LLM Leaderboard Evaluation Results
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 24.90 |
| IFEval (0-Shot) | 72.74 |
| BBH (3-Shot) | 26.90 |
| MATH Lvl 5 (4-Shot) | 13.22 |
| GPQA (0-shot) | 4.03 |
| MuSR (0-shot) | 4.28 |
| MMLU-PRO (5-shot) | 28.26 |
- Downloads last month
- 24
Model tree for EpistemeAI/Fireball-Alpaca-Llama3.1.07-8B-Philos-Math-KTO-beta
Base model
meta-llama/Llama-3.1-8B
Finetuned
meta-llama/Llama-3.1-8B-Instruct Spaces using EpistemeAI/Fireball-Alpaca-Llama3.1.07-8B-Philos-Math-KTO-beta 2
Paper for EpistemeAI/Fireball-Alpaca-Llama3.1.07-8B-Philos-Math-KTO-beta
Evaluation results
- strict accuracy on IFEval (0-Shot)Open LLM Leaderboard72.740
- normalized accuracy on BBH (3-Shot)Open LLM Leaderboard26.900
- exact match on MATH Lvl 5 (4-Shot)Open LLM Leaderboard13.220
- acc_norm on GPQA (0-shot)Open LLM Leaderboard4.030
- acc_norm on MuSR (0-shot)Open LLM Leaderboard4.280
- accuracy on MMLU-PRO (5-shot)test set Open LLM Leaderboard28.260