Research
Bullynx research and evaluation methods
Reproducible methods for evaluating AI chart readers: what to test, how to score it, what counts as a refusal, and which limitations must remain visible.
Protocol published · Results not yet published
AI chart-reader evaluation methodology
A fixed 100-point rubric for market structure, level fidelity, evidence, refusal behavior, and repeated-run consistency, with an explicit protocol-only publication status.
Read the methodology →Publication rule
Bullynx will separate protocols from results. A methodology can be published before a benchmark exists; a score cannot. Any future result must include the chart set, tested version, date, raw denominator, repeated runs, limitations, and conflict-of-interest disclosure.