What are special tokens used for function calling?

• A. <TOOLS>, <END_TOOLS>, [EXECUTE_TOOL], [RESULTS], </RESULTS>
• B. {TOOLS}, {/TOOLS}, {CALLS}, {OUTPUT}, {/OUTPUT}
• C. [TOOL_LIST], [LIST_TOOLS], [TOOL_EXEC], [TOOL_OUTPUT]
• D. [AVAILABLE_TOOLS], [/AVAILABLE_TOOLS], [TOOL_CALLS], [TOOL_RESULTS], [/TOOL_RESULTS]

Answer: (D)

Special tokens for function calling are structured tags that mark the different stages where tools are listed, invoked, and where results appear. The best choice uses opening and closing tags for each section: [AVAILABLE_TOOLS] … [/AVAILABLE_TOOLS] to declare what tools exist, [TOOL_CALLS] … [/TOOL_CALLS] to request a tool to run, and [TOOL_RESULTS] … [/TOOL_RESULTS] to receive the output. This clear, paired naming makes boundaries explicit and easy to parse, so the system knows exactly when tools can be used and where to place the results.

Other options mix different bracket styles (like angle brackets or braces) or omit closing tags, which can create confusion for the parser and disrupt the tool-calling flow. The consistent, self-describing tokens in the correct choice keep the function-calling mechanism reliable.