Spaces:
Running
Running
Commit History
-
Add HERB computation comparison split
Verified signatureLabel all HERB questions, add category filters across question views, and publish the per-run performance breakdown. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 953d9143-c31e-4a5b-9b68-d945c9e5303a8b27dd6 -
Add E2E v3 + rawtext views
Verified signatureAdd the production overlay trajectory with canonical type-aware scoring to run and Compare views, plus a structure alias for the cached E2E v3 scaffold. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: c2fe83fc-cc75-4d0b-b5ba-4a8def928fdc1bb926a
-
Add full-run baseline and E2E tabs
Verified signatureAdds c1, c2, c6 Agentic-DCI, naive-search, E2E raw, E2E combined, and Compare views over all 815 answerable questions. c1/c2 recovery overlays, explicit judge failures, bounded trajectory caches, and compact streamed shards are included. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: c2fe83fc-cc75-4d0b-b5ba-4a8def928fdcbfe91f0
-
HERB viewer: Corpus (38,600 heterogeneous artifacts) + Eval (1514 Q) static Space
herb-viewer bot +1: CopilotTwo-tab static viewer. Corpus tab: product + artifact-type filters, per-artifact dropdown; renders slack/documents/meeting transcripts/chats/urls/PRs in native shape with eid/customer ids resolved to names; product JSONs streamed on demand from the Salesforce/HERB dataset CDN (nothing heavy bundled). Eval tab: product / answerable-vs-unanswerable / type filters; shows ground-truth answer and clickable evidence-citation chips that jump to the cited artifact in the Corpus tab. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>23866a2