List your company

ResearchFormerly WizWand

tau2-bench

Benchmarks

TaskDatasetSOTA resultTrendResults

Interactive Tool-Use Agent Performance

tau2-Bench

81.1Retail Performance Score

102

Agentic Tool-use

tau2-bench Airline

63.5Pass@1

30

Agentic Tool-use

tau2-bench Retail

82Pass@1

30

Multi-turn Decision Making

tau2-Bench held-out (test)

41.9Average Score

16

Task failure prediction and selective task completion

tau2-bench Telecom 1.0

0.809AUROC

15

Task failure prediction and selective task completion

tau2-bench Retail 1.0

0.707AUROC

15

Task failure prediction and selective task completion

tau2-bench Airline 1.0

74.2AUROC

15

Stateful Interaction

tau2-bench

86.69Score

12

Long-Horizon Tool Execution

tau2-Bench

75.2Retail Success Rate

12

Real-world issue resolution

tau2-bench

72Airline Success Rate

9

Agent Task Completion

tau2-bench telecom

67Pass Rate

9

Agent Task Completion

tau2-bench airline

60Pass Rate

9

Agent Task Success

tau2-bench Retail Domain

61.4Total Pass Rate

9

Agentic Skill Acquisition

tau2-bench

81.2Pass@1

9

Tool-use Task Completion

Tau2-Bench Telecom subset Solo mode

43.3Success Rate

7

Multi-turn agent decision making

tau2-Bench (test)

22.3Success Rate

7

False Success Detection

tau2-bench v1 v2 (test)

89Recall

6

General Task (Agentic Coding)

tau2-Bench Telecom

98.2Score

6

Tau2-Bench

57.4Success Rate

6

User Simulation Behavioral Alignment

tau2-bench Retail + Airline (test)

95.8HL Score

5

User Simulation Behavioral Alignment

tau2-bench Airline (test)

90.3HL

5

Failure prediction

tau2-bench Transfer held-out

67.4ROC-AUC

4

Failure prediction

tau2-bench Within corpus

0.719ROC AUC

4

tau2-bench

43Productivity

4

Stateful service dialogue

tau2-bench Telecom

85.1Task Completion Score (TCS)

4