SKU RLE-1207 · Sold by External
tau2-bench (τ²-bench)
Product specifications
| SKU | RLE-1207 |
|---|---|
| Data type | RL environments |
| Volume | 278 tasks (retail 114 + telecom 114 + airline 50) |
| Size on disk | Not published by source |
| Format | Python package + JSON domain data |
| Access model | PUBLIC LICENSE |
| Pricing | Free · open-source license |
| Quality score | — |
| License | MIT |
τ²-bench extends τ-bench into a dual-control setting: in the new telecom technical-support/troubleshooting domain both the agent and the simulated user can invoke tools, alongside the retained retail and airline domains. It ships a compositional task generator and evaluates agents on task-completion actions gated by a reward basis. Task counts are airline 50, retail 114, and telecom 114 (~278 total). Distributed as an MIT-licensed Python package (uv-based) by Sierra Research; a later τ³ line adds banking and voice modalities.