Read-only view: browsing is open, changes need a signed-in user.
Staging: test environment for new releases. Data may be reset; do not share from here.
WatchtowerSignicat competitor intelligence

Evaluation

Labels

Human 0gemini-3.8-flash 8

Latest report

Latest completed runtime evaluation job · metrics based on model:gemini-3.8-flash

Level rank correlation (Spearman)0.66
Daily precision at 5n/a (ranking selection missing or fewer than k items)
Weekly precision at 10n/a (ranking selection missing or fewer than k items)
Tag F1 · Geos0.89
Tag F1 · Themes0.80
Tag F1 · Verticals0.86
Tag F1 · Event types0.83
Model cost per analysed article (USD)0.02
Raw report
{
  "acceptance": {
    "precision_at_5_ge_0_7": false,
    "spearman_ge_0_6": true
  },
  "candidate_count": 140,
  "human_label_counts": {
    "inclusion": 0,
    "level": 0,
    "tags": 0
  },
  "inputs": {
    "article_count": 140,
    "completed_digest_count": 7,
    "scope": "current_database"
  },
  "label_basis": "model:gemini-3.8-flash",
  "label_sources": {
    "human": 0,
    "model": {
      "gemini-3.8-flash": 8
    }
  },
  "ledger_cost_usd_per_article": {
    "analyzed_article_count": 8,
    "call_count": 27,
    "reason": null,
    "unknown_attempts": 0,
    "value": 0.015074875
  },
  "ledger_latency_ms_mean": {
    "call_count": 27,
    "reason": null,
    "value": 6180.0
  },
  "precision_at_10_weekly": {
    "count": 0,
    "missing_label_windows": 0,
    "missing_labels": 0,
    "reason": "ranking selection missing or fewer than k items",
    "scored_windows": 0,
    "selected": 1,
    "short_windows": 1,
    "value": null,
    "window_count": 2,
    "zero_windows": 1
  },
  "precision_at_5_daily": {
    "count": 0,
    "missing_label_windows": 0,
    "missing_labels": 0,
    "reason": "ranking selection missing or fewer than k items",
    "scored_windows": 0,
    "selected": 1,
    "short_windows": 1,
    "value": null,
    "window_count": 5,
    "zero_windows": 4
  },
  "readiness": "scored",
  "recorded_prediction_rows": 8,
  "recorded_selection_rows": 2,
  "same_model_self_review": [],
  "selected_ids": {
    "daily": [
      {
        "selected_ids": [],
        "window_id": "1"
      },
      {
        "selected_ids": [],
        "window_id": "2"
      },
      {
        "selected_ids": [],
        "window_id": "3"
      },
      {
        "selected_ids": [],
        "window_id": "4"
      },
      {
        "selected_ids": [
          "5"
        ],
        "window_id": "6"
      }
    ],
    "weekly": [
      {
        "selected_ids": [],
        "window_id": "5"
      },
      {
        "selected_ids": [
          "5"
        ],
        "window_id": "7"
      }
    ]
  },
  "spearman_level": {
    "count": 8,
    "reason": null,
    "value": 0.6561880335523164
  },
  "tag_f1": {
    "event_types": {
      "count": 8,
      "fn": 2,
      "fp": 3,
      "reason": null,
      "tp": 12,
      "value": 0.8275862068965517
    },
    "geos": {
      "count": 8,
      "fn": 2,
      "fp": 0,
      "reason": null,
      "tp": 8,
      "value": 0.8888888888888888
    },
    "themes": {
      "count": 8,
      "fn": 2,
      "fp": 6,
      "reason": null,
      "tp": 16,
      "value": 0.8
    },
    "verticals": {
      "count": 8,
      "fn": 1,
      "fp": 2,
      "reason": null,
      "tp": 9,
      "value": 0.8571428571428571
    }
  },
  "verification_pass_rate": {
    "count": 8,
    "original_facts": 29,
    "pending_articles": 132,
    "reason": null,
    "retained_facts": 22,
    "value": 0.7586206896551724
  }
}

Precision and rank correlation are shown only when supported by labelled examples and recorded selections.