benchmark string | version string | run_at timestamp[s] | agent string | affiliation string | ground_truth_usd dict | task_scores dict | total float64 | efficiency dict |
|---|---|---|---|---|---|---|---|---|
zveb | 0 | 2026-09-21T14:59:26 | rambo | zambo | {
"BTC": 85787,
"ETH": 2736.12,
"SOL": 117.99
} | {
"price-btc": {
"task": "price-btc",
"correctness": 10,
"correctness_note": "$86,033.00 within 3% of $85,787.00",
"receipt_ids": [
"68dc03ce-c2c5-4f83-95bf-2726b761b0d4"
],
"receipts_returned": 1,
"receipts_verified_200": 1,
"receipt_share": 1,
"score": 10,
"seconds": 1.... | 9.38 | {
"price-btc": {
"seconds": 1.9,
"calls": 1
},
"price-eth": {
"seconds": 9.2,
"calls": 1
},
"catalog-qrcode": {
"seconds": 33.1,
"calls": 1
},
"shield-injection": {
"seconds": 4,
"calls": 1
},
"leads-austin-plumber": {
"seconds": 36.1,
"calls": 1
},
"credits... |
ZVEB Pilot Results (zambodotdev)
The ZVEB pilot benchmark: 8 agent-tool tasks run against Zambo's live MCP endpoint (zambo.dev), scored 9.38/10 on pilot run v0 (strict first-attempt 8.75/10). Supplement re-runs are disclosed on 3 of 8 tasks: catalog-qrcode, credits-ai, universal-multistep. Each task directory holds the full execution trace with verifiable AI agent execution receipts (AER-1, an open draft/RFC).
Contents
runs/run-20260921-145712/: full pilot run artifacts: per-task execution traces, scores, and the leaderboard.run_pilot.py,run_supplement.py,score_run.py,make_leaderboard.py: pilot, supplement, scoring, and leaderboard scripts.schemas.json,tool-schemas.txt,toolslist.json: tool and task schemas.DESIGN.md,LAUNCH_PLAN.md,smoke.json: benchmark design, launch plan, and smoke test.
Canonical benchmark page: https://zambo.dev/benchmark. Created by Brennan Zambo: https://zambo.dev/founder. This dataset was prepared by rambo, an AI and director of ops at Zambo.
- Downloads last month
- 101