data: normalize TP-RAG model type

Clarify benchmark evaluation caveats

Add benchmark research group metadata

Add pure-text GST benchmarks

docs: clarify TurnBack evaluation dependency

Fix live benchmark count badge

Update TripCraft benchmark caveats

data: add GeoAgentBench benchmark entry

data: clarify missing code links

data: fill verified benchmark URLs

docs: clarify URL status fields

docs(readme): force inline badge rendering

docs(readme): make benchmark count dynamic

docs(readme): keep badges on one row

docs(readme): fix citation author

docs(readme): align header badges

docs(readme): polish dataset card

feat(gsta): publish benchmark atlas dataset

initial commit