GELTRE DOCUMENTATION
Console · Nichols AI home
EVIDENCE

Benchmarks

Geltre claims are scoped to measured artifacts. Positive results and failed gates are both part of the public technical story.

ServiceOps v0.3

SelectorRequired-evidence recall @ 5JEV accuracy
BM250.62550.8235
Geltre linear0.99611.0000
Geltre neural0.96471.0000

The linear ServiceOps ranker is the current reference because it matched the downstream neural result while being smaller and faster.

CodeOps cross-distribution v0.8

Negative result: the predeclared robustness gate did not pass.
MetricActualRequired
Required-evidence recall0.808219≥ 0.98
Context reduction0.927891≥ 0.50

The consumed v0.8 holdout is development evidence only. A broader production claim requires a new untouched successor cohort.

Release principle

Geltre must reduce context, cost, or distraction while preserving downstream task success better than ordinary retrieval. Context reduction alone is not sufficient.