Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Multi-turn Agentic Tool-use on TauBench Retail
Loading...
50.3
Success Rate (SR)
HERO
21.388
28.894
36.4
43.906
Jun 10, 2026
Success Rate (SR)
Average Turns
Updated 1mo ago
Evaluation Results
Method
Method
Links
Success Rate (SR)
Average Turns
HERO
Model=Qwen3-30B-A3B-In...
2026.06
50.3
10.7
GRPO
Model=Qwen3-30B-A3B-In...
2026.06
47.8
10.8
HERO
Model=Qwen3-4B-Instruct
2026.06
34.7
9.6
Environment Feedback Only
Model=Qwen3-30B-A3B-In...
2026.06
34.5
12.2
Base
Model=Qwen3-30B-A3B-In...
2026.06
33.6
11.9
GRPO
Model=Qwen3-4B-Instruct
2026.06
33.3
13.4
Environment Feedback Only
Model=Qwen3-4B-Instruct
2026.06
30.4
11.5
Base
Model=Qwen3-4B-Instruct
2026.06
27.2
10.5
Full-Demo Privileged Teacher
Model=Qwen3-30B-A3B-In...
2026.06
26.1
12.6
Full-Demo Privileged Teacher
Model=Qwen3-4B-Instruct
2026.06
22.5
12.4
Feedback
Search any
task
Search any
task