Collaborative Interaction on CollabBench CWAH-MultiPlayer
90.03StepQwen2.5-7B-Instruct
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen2.5-7B-InstructMethod Category=Base, Agent Role=Agent 22026.06 | 90.03 | 31.62 | 0.24 | 1.04 | 2.19 | 2.3 | |
| Qwen2.5-7B-InstructMethod Category=Base, Agent Role=Agent 12026.06 | 84.51 | 33.23 | 0.24 | 1.22 | 2.58 | 2.5 | |
| Qwen2.5-7B-InstructMethod Category=Trained, Agent Role=Agent 12026.06 | 71.64 | 25.16 | 0.23 | 1.43 | 3.03 | 3.33 | |
| DeepSeek-V3.1Method Category=Base, Agent Role=Agent 12026.06 | 69.26 | 25.17 | 0.41 | 2.32 | 3.35 | 2.87 | |
| Qwen2.5-72B-InstructMethod Category=Base, Agent Role=Agent 12026.06 | 68.68 | 24.36 | 0.29 | 2.41 | 3.61 | 3.23 | |
| GPT-5.2Method Category=Base, Agent Role=Agent 12026.06 | 67.49 | 25.84 | 0.21 | 2.66 | 3.74 | 3.07 | |
| Qwen2.5-72B-InstructMethod Category=Base, Agent Role=Agent 22026.06 | 66.54 | 23.65 | 0.29 | 2.51 | 3.71 | 3.39 | |
| DeepSeek-V3.1Method Category=Base, Agent Role=Agent 22026.06 | 65.75 | 26.28 | 0.43 | 2.45 | 3.5 | 3.3 | |
| Qwen2.5-7B-InstructMethod Category=Trained, Agent Role=Agent 22026.06 | 63.65 | 22.8 | 0.23 | 1.45 | 3.02 | 3.02 | |
| GPT-5.2Method Category=Base, Agent Role=Agent 22026.06 | 62.98 | 20.7 | 0.21 | 2.72 | 3.66 | 3.32 | |
| GPT-5.2Method Category=Oracle, Agent Role=Agent 12026.06 | 60.91 | 22.26 | 0.16 | 2.81 | 3.77 | 3.69 | |
| GPT-5.2Method Category=Oracle, Agent Role=Agent 22026.06 | 60.29 | 25.01 | 0.16 | 2.96 | 4.01 | 3.53 |