Agentic Tool-use on τ²-bench telecom v1 (test)
85.9Pass@1RL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RLBackbone=Qwen3-8B, Protocol=Single-Task2026.06 | 85.9 | 67.8 | |
| Off. + On.Backbone=Qwen3-8B, Protocol=Multi-Task2026.06 | 85.8 | 69.4 | |
| Off-policy Distill.Backbone=Qwen3-8B, Protocol=Multi-Task2026.06 | 84.8 | 64.9 | |
| Off-policy Distill.Backbone=Qwen3-8B, Protocol=Single-Task2026.06 | 84.3 | 67.7 | |
| On-policy Distill.Backbone=Qwen3-8B, Protocol=Multi-Task2026.06 | 78.2 | 58.3 | |
| Pre-RLBackbone=Qwen3-8B, Protocol=Single-Task2026.06 | 66.2 | 44.7 |