Personalized Summarization on Summarize from Human Feedback (test)
78.23Win RateEXACT
Evaluation Results
| Method | Links | |
|---|---|---|
| EXACTBackbone=Gemma-2-9B-it2026.02 | 78.23 | |
| EXACTBackbone=Llama-3.1-8B2026.02 | 73.81 | |
| EXACTBackbone=Qwen2.5-7B-Instruct2026.02 | 72.12 | |
| RewardBackbone=Qwen2.5-7B-Instruct2026.02 | 71.28 | |
| DriftBackbone=Qwen2.5-7B-Instruct2026.02 | 68.88 | |
| DriftBackbone=Llama-3.1-8B2026.02 | 68.32 | |
| RewardBackbone=Gemma-2-9B-it2026.02 | 67.32 | |
| DriftBackbone=Gemma-2-9B-it2026.02 | 66.32 | |
| RewardBackbone=Llama-3.1-8B2026.02 | 64.32 |