Tau-Bench Airline

τ-bench Airline is the airline variant of the τ-family: the model works as an assistant for an airline support team, booking flights and applying policies with the provided API tools while a simulated customer pushes the conversation around. Correct tool calls and respecting policy are what is graded.

Tau-Bench Airline measures how well a model acts as a general-purpose agent. In detail, it checks:

• complex booking calculations and rules
• strict adherence to airline policies
• safe tool use in a customer-facing dialogue
• multi-turn consistency across a long conversation

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.

What is the Tau-Bench Airline benchmark?

τ-bench Airline is the airline variant of the τ-family: the model works as an assistant for an airline support team, booking flights and applying policies with the provided API tools while a simulated customer pushes the conversation around. Correct tool calls and respecting policy are what is graded.

What does Tau-Bench Airline test?

Tau-Bench Airline mainly measures how well a model acts as a general-purpose agent. Concretely, it checks: complex booking calculations and rules; strict adherence to airline policies; safe tool use in a customer-facing dialogue.

How is the Tau-Bench Airline score calculated?

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.

Where do I get Tau-Bench Airline data and scores?

The official source is https://github.com/sierra-research/tau-bench. The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/tau-bench-airline/scores) or pull it directly from the Open LLM Radar API with any HTTP client.