Negotiation on diverse-buyer pool 5-buyer iter-50 60 70 T=9 inbox (OOD Evaluation)
0.54RewardQwen3-30B-A3B-Instruct-2507-trained
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen3-30B-A3B-Instruct-2507-trainedParams=30B2026.07 | 0.54 | 77 | 70.7 | 2.39 | 70.7 | 0 | |
| gpt-5.1-high-reasoningParams=closed-source2026.07 | 0.4 | 42.1 | 95.1 | 3.83 | 46.5 | 0 | |
| Qwen3-30B-A3B-nothinkParams=30B2026.07 | 0.369 | 21.4 | 44.5 | 2.04 | 61.4 | 45.3 | |
| Kimi-K2.6-thinkingParams=1T2026.07 | 0.3 | 31.3 | 96 | 1.51 | 58 | 0 | |
| DeepSeek-V3.1-nothinkParams=671B2026.07 | 0.258 | 35 | 88.3 | 1.89 | 63.7 | 1.2 | |
| DeepSeek-V3.1-thinkingParams=671B2026.07 | 0.246 | 30.7 | 91.4 | 2.23 | 57.3 | 0.8 | |
| gpt-5.1-no-reasoningParams=closed-source2026.07 | 0.241 | 26.2 | 91.8 | 1.11 | 69.1 | 0 | |
| Kimi-K2.6-nothinkParams=1T2026.07 | 0.237 | 32.5 | 88.7 | 1.99 | 57.3 | 2.7 | |
| Qwen3-30B-A3B-thinkParams=30B2026.07 | 0.232 | 28.1 | 84 | 1.46 | 63.3 | 0 | |
| Qwen3-235B-A22B-Instruct-2507Params=235B2026.07 | 0.2 | 26.4 | 90.6 | 1.48 | 59.5 | 2.3 | |
| Qwen3-30B-A3B-Instruct-2507-untrainedParams=30B2026.07 | 0.059 | 18.6 | 73.7 | 1.98 | 56.4 | 7.5 | |
| Qwen3-4B-Instruct-2507Params=4B2026.07 | 0.038 | 25.9 | 39.8 | 2.08 | 48 | 9.8 | |
| Llama-3.3-70B-InstructParams=70B2026.07 | 0.025 | 34.6 | 75 | 2 | 56.2 | 16.4 |