BFCL Multi-Turn

The 'BFCL Multi-Turn' dataset/benchmark contains multi-turn tool-call trajectories grounded in live-sampled server state and is used to evaluate the performance of models in orchestrating multi-step tool calls in realistic execution environments.

Papers using BFCL Multi-Turn (3)

Synthesize and Reward -- Reinforcement Learning for Multi-Step Tool Use in Live Environments2026Robust Tool Use via Fission-GRPO: Learning to Recover from Execution Errors2026MUA-RL: Multi-turn User-interacting Agent Reinforcement Learning for agentic tool use2025