RESEARCH · RESEARCH · #2017
Which Metadata Matters for LLM Tool Selection in MCP (arXiv:2610.14334v1)
This arXiv paper analyzes metadata for 367,174 MCP tools and runs controlled, single-factor metadata interventions across five LLMs and five tasks, estimating position-adjusted effects with conditional logistic regression. It finds parameter-schema differences drive the largest and most consistent shifts in which tool an LLM selects, descriptions that explain parameters increase selection, names have weaker effects, and explicit request-compatibility can override position biases.
KEY POINTS
- This arXiv paper analyzes metadata for 367,174 MCP tools and runs controlled, single-factor metadata interventions across five LLMs and five tasks, estimating position-adjusted effects with conditional logistic regression.
- It finds parameter-schema differences drive the largest and most consistent shifts in which tool an LLM selects, descriptions that explain parameters increase selection, names have weaker effects, and explicit request-compatibility can override position biases.
- Shows that parameter-schema design (not just names or descriptions) can change which implementation an LLM invokes, so schema and ordering choices materially affect MCP behavior.
WHY IT MATTERS
Shows that parameter-schema design (not just names or descriptions) can change which implementation an LLM invokes, so schema and ordering choices materially affect MCP behavior.