Best-of-N Alignment on HH-RLHF
53BWRAdaBoN
Evaluation Results
| Method | Links | |
|---|---|---|
| AdaBoNLM=Qwen, RM=Armo, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 53 | |
| AdaBoNLM=Gemma, RM=FsfairX, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 53 | |
| AdaBoNLM=Qwen, RM=FsfairX, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 54 | |
| AdaBoNLM=Gemma, RM=Mistral, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 54 | |
| AdaBoNLM=Mistral, RM=Armo, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 55 | |
| AdaBoNLM=Qwen, RM=Mistral, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 55 | |
| AdaBoNLM=Gemma, RM=Armo, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 55 | |
| AdaBoNLM=Llama, RM=FsfairX, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 57 | |
| AdaBoNLM=Llama, RM=Armo, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 57 | |
| AdaBoNLM=Mistral, RM=Mistral, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 58 | |
| AdaBoNLM=Llama, RM=Mistral, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 59 | |
| AdaBoNLM=Mistral, RM=FsfairX, batch size K=5, budget B=120, exploration budget d=0.75B2025.05 | 60 |