BFCL v-3

The 'BFCL-v-3' dataset/benchmark is used to evaluate the performance of models in executing complex, multi-step tasks by providing a structured set of data that reflects various capabilities and challenges faced by large language models.

Papers using BFCL v-3 (7)

D-CORE: Incentivizing Task Decomposition in Large Reasoning Models for Complex Tool Use2026EnvFactory: Scaling Tool-Use Agents via Executable Environments Synthesis and Robust RL2026SkillX: Automatically Constructing Skill Knowledge Bases for Agents2026TOUCAN: Synthesizing 1.5M Tool-Agentic Data from Real-World MCP Environments2025DaMo: Data Mixing Optimizer in Fine-tuning Multimodal LLMs for Mobile Phone Agents2025Can a Single Model Master Both Multi-turn Conversations and Tool Use? CALM: A Unified Conversational Agentic Language Model2025LoopTool: Closing the Data-Training Loop for Robust LLM Tool Calls2025