τ³
An extended version of tau-bench that focuses on three-way tool-agent-user interaction, simulates realistic customer service scenarios, and tests the impact of context compression on policy compliance and retention of user information.
Timeline (last 90 days)
On the two benchmarks AppWorld and τ³, tested with two target models, SpecCompact achieved compression speedups of 2.04 to 5.63 times compared with Hermes-style summarization.
τ³ (an extended version of tau-bench) focuses on three-way tool-agent-user interaction, simulating real customer-service scenarios and requiring agents to follow domain policies and conduct multi-turn dialogue with simulated users.
AA's newly launched proprietary evaluation replaces the previous τ³ (tau-cubed) benchmark.
All Facts (2)
On the two benchmarks AppWorld and τ³, tested with two target models, SpecCompact achieved compression speedups of 2.04 to 5.63 times compared with Hermes-style summarization.
50%Unverifiedτ³ (an extended version of tau-bench) focuses on three-way tool-agent-user interaction, simulating real customer-service scenarios and requiring agents to follow domain policies and conduct multi-turn dialogue with simulated users.
50%