Real-world issue resolution on tau2-bench
72Airline Success RateEDV
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EDVBackbone Model=(Ensemble)2026.06 | 72 | 88.6 | 99.1 | 86.6 | |
| RouterBackbone Model=(Ensemble)2026.06 | 68 | 85.2 | 97.2 | 83.5 | |
| ReasoningBank (RB)Backbone Model=Minimax-M2.12026.06 | 66 | 83.3 | 87.7 | 79 | |
| ReasoningBank (RB)Backbone Model=GLM-4.7-FP82026.06 | 66 | 82.5 | 97.2 | 81.9 | |
| JudgeBackbone Model=(Ensemble)2026.06 | 66 | 84.7 | 93.9 | 81.5 | |
| No Memory (NM)Backbone Model=Mimo-V2-Flash2026.06 | 64.5 | 79.2 | 91.7 | 78.5 | |
| No Memory (NM)Backbone Model=GLM-4.7-FP82026.06 | 64 | 78.1 | 96.6 | 79.6 | |
| ReasoningBank (RB)Backbone Model=Mimo-V2-Flash2026.06 | 64 | 83.3 | 94.7 | 80.7 | |
| No Memory (NM)Backbone Model=Minimax-M2.12026.06 | 61.5 | 82.2 | 85.5 | 76.4 |