Negotiation vs GPT-5.4 High Reasoning Seller on Standard Held-Out Test Set
0.4081RewardQwen3-30B-A3B-Instruct-2507-trained
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-30B-A3B-Instruct-2507-trainedParams=30B2026.04 | 0.4081 | 75 | 40.81 | 0 | |
| Qwen3-30B-A3B-Instruct-2507-untrainedParams=30B2026.04 | 0.2744 | 60.5 | 3.89 | 32.4 | |
| gpt-5.4-high-reasoningParams=-2026.04 | 0.1823 | 91.4 | 18.23 | 0 | |
| gpt-5.4-no-reasoningParams=-2026.04 | 0.1458 | 84.8 | 16.14 | 1.6 | |
| DeepSeek-V3.1-thinkingParams=671B2026.04 | 0.1223 | 92.6 | 13.02 | 0.8 | |
| DeepSeek-V3.1-nothinkParams=671B2026.04 | 0.1204 | 90.6 | 14.77 | 2.7 | |
| Kimi-K2-ThinkingParams=1T2026.04 | 0.106 | 90.6 | 12.56 | 1.2 |