Negotiation task (Sell&Buy) on NegotiationArena Sell&Buy
17.94Self ProfitAMPO
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| AMPOBackbone=Qwen2.5-7B-Instruct2025.05 | 17.94 | 23.96 | 54.6 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2025.05 | 16.75 | 23.67 | 52.6 | |
| AMPOBackbone=Llama3.1-8B-Instruct2025.05 | 15.57 | 24 | 50.2 | |
| GRPOBackbone=Llama3.1-8B-Instruct2025.05 | 14.99 | 23.58 | 48.7 | |
| Qwen2.5-7B-InstructTraining Mode=Vanilla2025.05 | 11.9 | 13.87 | 38.1 | |
| Llama3.1-8B-InstructTraining Mode=Vanilla2025.05 | 3.12 | 3.82 | 9.4 |