name: PhoneToolBench description: > A held-out benchmark for phone tool calling: 520 items in English, Portuguese, Spanish and German, over the full 108-tool phone catalog of the LiquidAI phone-assistant tools dataset. Scoring is deterministic exact match of the predicted tool-call list against the gold calls (no LLM judge). The headline number is macro exact match, the mean over the seven item categories (single, multi, reasoning, lookalike, partial, no_tool, clarify). tasks: - id: phone_tool_bench config: default split: test