$τ$-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains
This addresses the problem of unreliable agent deployment in real-world applications for developers and researchers, though it is incremental as it builds on existing benchmarking efforts.
The paper tackles the lack of benchmarks for testing language agents on interactions with human users and adherence to domain-specific rules, proposing τ-bench, which shows that state-of-the-art agents like GPT-4o succeed on less than 50% of tasks and have low consistency (pass^8 <25% in retail).
Existing benchmarks do not test language agents on their interaction with human users or ability to follow domain-specific rules, both of which are vital for deploying them in real world applications. We propose $τ$-bench, a benchmark emulating dynamic conversations between a user (simulated by language models) and a language agent provided with domain-specific API tools and policy guidelines. We employ an efficient and faithful evaluation process that compares the database state at the end of a conversation with the annotated goal state. We also propose a new metric (pass^k) to evaluate the reliability of agent behavior over multiple trials. Our experiments show that even state-of-the-art function calling agents (like gpt-4o) succeed on <50% of the tasks, and are quite inconsistent (pass^8 <25% in retail). Our findings point to the need for methods that can improve the ability of agents to act consistently and follow rules reliably.