{
  "id": "eval-harness/golden-dataset-curation",
  "slug": "golden-dataset-curation",
  "title": "Session: golden-dataset-curation — 2026-09-08",
  "type": "source",
  "project": "eval-harness",
  "date": "2026-09-08",
  "started": "2026-09-08T12:10:00+00:00",
  "ended": "2026-09-08T13:34:00+00:00",
  "model": "gemini-2.5-pro",
  "cwd": "/home/dev/code/eval-harness",
  "git_branch": "data/golden-set",
  "permission_mode": "default",
  "user_messages": 1,
  "tool_calls": 2,
  "tools_used": [
    "Bash",
    "Write"
  ],
  "is_subagent": false,
  "wikilinks_out": [
    "Datasets",
    "Embeddings",
    "eval-harness"
  ],
  "body_text": "Project: eval-harness · Branch: data/golden-set · Mode: default · Model: gemini-2.5-pro\n\nSummary\n\nCurated a 300-case golden dataset from support tickets: de-duplicated near-identical questions with embeddings, stratified by product area, and stored expected answers with source citations.\n\nConversation\n\nTurn 1 — User\n\nOur eval set is 80% password-reset questions.\n\nTurn 1 — Assistant\n\nCluster ticket questions by embedding similarity, sample per product area, and keep the citation for each expected answer so faithfulness can be checked.\n\nTools used in this turn:\n\n- Write: scripts/curate_golden.py\n- Bash: python scripts/curate_golden.py --target 300 — suites/golden.yaml\n\nConnections\n\n- eval-harness\n- Datasets\n- Embeddings",
  "sha256": "b3e0e2d0009b83cc",
  "source_url": "sessions/eval-harness/2026-09-08-golden-dataset-curation.html"
}