BFCL v-3

The 'BFCL v-3' dataset/benchmark contains tool-use trajectories designed to evaluate the performance of reinforcement learning algorithms in executing tasks that involve tool usage and decision-making.

Papers using BFCL v-3 (8)

TIER: Trajectory-Invariant Execution Rewards for Multi-Step Tool Composition2026HINT-SD: Targeted Hindsight Self-Distillation for Long-Horizon Agents2026EnvFactory: Scaling Tool-Use Agents via Executable Environments Synthesis and Robust RL2026Controllable and Verifiable Tool-Use Data Synthesis for Agentic Reinforcement Learning2026TopoCurate:Modeling Interaction Topology for Tool-Use Agent Training2026MagicAgent: Towards Generalized Agent Planning2026ToolSample: Dual Dynamic Sampling Methods with Curriculum Learning for RL-based Tool Learning2025ShiQ: Bringing back Bellman to LLMs2025