Summarization on Summarization (Custom Metrics)
95Win RateCurriculum-RLAIF
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Curriculum-RLAIFBase Model=Qwen2.5-32B, Category=Curriculum2025.05 | 95 | — | — | — | |
| External Eval.Base Model=Qwen2.5-32B, Category=Curriculum2025.05 | 93 | — | — | — | |
| Curriculum-RLAIFBase Model=LLaMA-3-8B, Category=Curriculum2025.05 | 92 | — | — | — | |
| Internal Eval.Base Model=Qwen2.5-32B, Category=Curriculum2025.05 | 92 | — | — | — | |
| Implicit Eval. (DPO)Base Model=Qwen2.5-32B, Category=Curriculum2025.05 | 91 | — | — | — | |
| Internal Eval.Base Model=LLaMA-3-8B, Category=Curriculum2025.05 | 90 | — | — | — | |
| Conventional RLAIFBase Model=Qwen2.5-32B, Category=Non-Curriculum2025.05 | 90 | — | — | — | |
| External Eval.Base Model=LLaMA-3-8B, Category=Curriculum2025.05 | 89 | — | — | — | |
| Curriculum-RLAIFBase Model=Gemma-1-2B, Category=Curriculum2025.05 | 88 | — | — | — | |
| Implicit Eval. (DPO)Base Model=LLaMA-3-8B, Category=Curriculum2025.05 | 87 | — | — | — | |
| RLCDBase Model=Qwen2.5-32B, Category=Non-Curriculum2025.05 | 87 | — | — | — | |
| External Eval.Base Model=Gemma-1-2B, Category=Curriculum2025.05 | 86 | — | — | — | |
| CAIBase Model=Qwen2.5-32B, Category=Non-Curriculum2025.05 | 86 | — | — | — | |
| Internal Eval.Base Model=Gemma-1-2B, Category=Curriculum2025.05 | 85 | — | — | — | |
| Conventional RLAIFBase Model=LLaMA-3-8B, Category=Non-Curriculum2025.05 | 85 | — | — | — | |
| Implicit Eval. (DPO)Base Model=Gemma-1-2B, Category=Curriculum2025.05 | 82 | — | — | — | |
| RLCDBase Model=LLaMA-3-8B, Category=Non-Curriculum2025.05 | 81 | — | — | — | |
| Conventional RLAIFBase Model=Gemma-1-2B, Category=Non-Curriculum2025.05 | 80 | — | — | — | |
| CAIBase Model=LLaMA-3-8B, Category=Non-Curriculum2025.05 | 79 | — | — | — | |
| HBAT#Param=13B, PPO=true, DPO=false2024.06 | 78.81 | 26.77 | -3.51 | 6.41 | |
| HBAT-Freeze#Param=13B, PPO=true, DPO=false2024.06 | 77.22 | 25.8 | -3.63 | 6.18 | |
| RLCDBase Model=Gemma-1-2B, Category=Non-Curriculum2025.05 | 77 | — | — | — | |
| CAIBase Model=Gemma-1-2B, Category=Non-Curriculum2025.05 | 75 | — | — | — | |
| HBAT#Param=13B, PPO=false, DPO=true2024.06 | 74.92 | 24.12 | -4.05 | 5.4 | |
| HBAT#Param=7B, PPO=true, DPO=false2024.06 | 72.52 | 26.18 | -3.82 | 5.74 | |
| RLHF#Param=13B, PPO=true, DPO=false2024.06 | 71.67 | 24.51 | -3.96 | 5.55 | |
| HBAT-Freeze#Param=13B, PPO=false, DPO=true2024.06 | 71.44 | 23.1 | -5.08 | 4.85 | |
| HBAT#Param=7B, PPO=false, DPO=true2024.06 | 70.58 | 23.14 | -4.18 | 4.95 | |
| DPO#Param=13B, PPO=false, DPO=true2024.06 | 69.4 | 23.02 | -5.39 | 4.55 | |
| HBAT-Freeze#Param=7B, PPO=true, DPO=false2024.06 | 64.79 | 25.33 | -4.28 | 5.26 | |
| RLHF+pt#Param=13B, PPO=true, DPO=false2024.06 | 64.42 | 22.92 | -5.49 | 3.97 | |
| HBAT-Freeze#Param=7B, PPO=false, DPO=true2024.06 | 64.18 | 23.01 | -5.05 | 4.45 | |
| RLHF#Param=7B, PPO=true, DPO=false2024.06 | 63.8 | 25.85 | -4.27 | 4.43 | |
| DPO#Param=7B, PPO=false, DPO=true2024.06 | 61.37 | 22.96 | -5.13 | 4.27 | |
| SFT#Param=13B, PPO=false, DPO=false2024.06 | 57.91 | 23.27 | -5.12 | 4.01 | |
| RLHF+pt#Param=7B, PPO=true, DPO=false2024.06 | 56.26 | 22.25 | -5.64 | 3.74 | |
| SFT#Param=7B, PPO=false, DPO=false2024.06 | 53.2 | 22.6 | -5.46 | 3.72 | |
| SFT+ppo#Param=13B, PPO=true, DPO=false2024.06 | 28.97 | 13.84 | -5.97 | 2.53 | |
| SFT+ppo#Param=7B, PPO=true, DPO=false2024.06 | 18.91 | 13.75 | -5.78 | 2.4 |