MATCH: Mapping Axis Transparency and Continuity cHeck — A Minimal Numeric Probe for Contextual Inertia in Large Language Models

MATCH (Mapping Axis Transparency and Continuity cHeck) is a minimal, reproducible probe for observing contextual inertia in Large Language Models (LLMs). This protocol introduces the concept of a "Mapping Axis" — a latent interpretive framework that an LLM implicitly activates when processing sequential inputs. When an LLM continues to interpret new inputs using a previously activated axis, this phenomenon is termed "Mapping Axis Continuity" or "Contextual Inertia." MATCH uses only numerical inputs to observe these behaviors, providing a language-agnostic, reproducible protocol for studying LLM interpretive behavior. This is a probe, not a benchmark. It is not intended for ranking or evaluating model performance. ## Files included ### PDF (Recommended)- MATCH_v1.0_EN.pdf — Full documentation (English)- MATCH_v1.0_JP.pdf — Full documentation (Japanese) ### Markdown- MATCH_README.md — Full documentation (English)- MATCH_README_JP.md — Full documentation (Japanese)- MATCH_protocol.md — Input sequence only (English)- MATCH_protocol_JP.md — Input sequence only (Japanese) ## How to Cite **Cite this version (v1.0):**Watanabe, T. (2026). MATCH: Mapping Axis Transparency and Continuity cHeck — A Minimal Numeric Probe for Contextual Inertia in Large Language Models (Version 1.0) [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.18406544 **Cite all versions (concept DOI, always resolves to latest):**https://doi.org/10.5281/zenodo.18406543 ## KeywordsLLM, Large Language Model, Context, Attention, Mapping Axis, Contextual Inertia, Reproducibility, Probe