Heads up This site is currently under heavy development.
← all datasets
◆ Evals & benchmarks

Sierra tau2-bench

MIT

Sierra's τ²-bench — agents must follow policy while using tools with a simulated user in airline, retail and telecom domains.

Summary

🚀 τ³-bench is here! From text-only to multimodal, knowledge-aware agent evaluation. Voice full-duplex · Knowledge retrieval · 75+ task fixes τ-Voice paper · τ-Knowledge paper · Task fixes paper · Release notes How do you say $\tau^3$-bench? We just say "tau three," but you do you!
from the project README
Formats
json
License
MIT
Added
2026-09-10

Get it

From the project’s own instructions where it documents any; otherwise a plain clone.

  1. shell
    $ git clone https://github.com/sierra-research/tau2-bench
  2. shell
    $ cd tau2-bench && uv sync
  3. shell
    $ tau2 run --domain airline --agent-llm gpt-4.1 --user-llm gpt-4.1 --num-trials 1 --num-tasks 5

    Put provider keys in .env first (cp .env.example .env). Models go through LiteLLM, so any supported provider works.

Contents

1,570 files 359.8 MB repository

  • .json 927
  • .py 335
  • .md 80
  • .png 34
  • .pdf 28
  • .jpg 23
  • .wav 23
  • .html 20

Use it with

Commands are curated, not yet run by us.

jq — Stream the records one per line to see their shape.

sample.json
$ jq -c . sample.json

DuckDB — Query the file in place with SQL — DuckDB detects CSV, JSON and JSONL by extension.

sample.json
$ duckdb -c 'SELECT * FROM '\''sample.json'\'' LIMIT 10'
my-toolchain — 0 tools
paste an install list to detect your tools

A brew list, a Brewfile, requirements.txt, a Dockerfile — or just the product names, free-form. Nothing leaves your browser.

    browse all tools →

    Send us feedback