Current definition verify-v1. Charts stamped with it were computed by src/resolution/verify.py, and every value plotted for a past date could have been computed on that date.
Stage one (matching.py) proposes pairs by text similarity and is deliberately crude. This stage reads both contracts in full — question plus rules text on each side — and answers one question: do these resolve on the same underlying event?
This never sets verified. It writes llm_verdict, llm_confidence and llm_reason. verified is set by a human reading the top pairs by volume, which is the layer RESOLUTION_DB calls uncopyable. A test pins that separation.
Provider-agnostic over any OpenAI-compatible chat-completions endpoint. Groq and xAI are both supported; note these are different companies with confusingly similar names — Groq (gsk_ keys, api.groq.com) serves open models; xAI (xai- keys) serves Grok. Set CL_LLM_PROVIDER, or let the key prefix pick.
The reasons matter as much as the verdicts: "same event, but Kalshi settles on the BLS print and Polymarket on media consensus" is already a Phase 4 source flag, not just a rejected match.
Newest first. Each entry says what changed and whether values that had already been published moved with it, followed by the sentence from the module that decided the verdict.
xAI chat-completions, JSON verdicts, resumable cache.
The first version of this metric. There is no earlier definition to restate.
None on the site today. The definition is versioned and exported, and no chart currently carries its stamp — either a chart is missing from the export or the module is no longer drawn. The catalogue reports the gap rather than hiding the metric, because a definition nobody can see the output of is exactly the thing a reader should be told about.