DecisionBench: A Benchmark for Emergent Delegation in Long-Horizon Agentic Workflows

1. 1Build a task suite using GAIA, tau-bench, and BFCL multi-turn to test delegation in various scenarios
2. 2Configure a peer-model pool with 11 models from 7 vendor families to simulate real-world delegation
3. 3Implement a delegation interface using call_model and read_profile channels to enable efficient delegation
4. 4Apply the deterministic skill-annotation layer to annotate skills and evaluate delegation quality
5. 5Evaluate the performance of your system using the multi-axis metric suite covering quality, cost, latency, and delegation rate

Researchers and developers working on agentic workflows and delegation can benefit from this benchmark to evaluate and improve their systems

💡 DecisionBench provides a comprehensive evaluation framework for emergent delegation in agentic workflows, enabling researchers to develop more efficient and effective delegation systems

🚀 Introducing DecisionBench: a benchmark for emergent delegation in long-horizon agentic workflows 🤖