01.2 / RUN COMPARISON
Compare before guessing.
Choose a successful baseline, check the test conditions, and inspect what changed. The engineer decides what to investigate.
Showing 10 of 10 fields. Select a field name to inspect its evidence.
On narrow screens, scroll the table horizontally to see both runs and the result.
| Configuration | run-042 · Baseline | run-043 · Selected run | Result |
|---|---|---|---|
| Manually reported | G-01 · revision A | G-02 · revision B | Changed |
| a31c9e2000000000000000000000000000000000 · clean | f72b4d1000000000000000000000000000000000 · clean | Changed | |
| gripper-fw-1.8 | gripper-fw-1.9 | Changed | |
| pick-policy-v3 | pick-policy-v4 | Changed | |
| teleop-block-A-v12 | teleop-block-A-v13 | Changed | |
| camera-cal-07 | camera-cal-08 | Changed | |
| Unknown — Force was not recorded | Unknown — Force was not recorded | Unknown | |
| pick-v2 | pick-v2 | Unchanged | |
| Pick block-A | Pick block-A | Unchanged | |
| 22 °C · 0.25 m/s | 22 °C · 0.25 m/s | Unchanged |
Changes are starting points, not proven causes. Missing values remain unknown. These invented runs demonstrate the actual comparison engine; no customer results are shown.
Inspect the example’s context and provenance
Baseline run-042 is fixed for this illustration; choose your own baseline in the full demo. Each run references the exact snapshot revision shown above. All entries were manually supplied by “Demo operator.”
Baseline observed: 2026-09-01T09:00:00Z; recorded: 2026-09-01T09:02:00Z. Selected hardware observed: 2026-09-01T10:00:00Z; recorded: 2026-09-01T10:02:00Z.
Evidence references and underlying model/dataset artifacts are unavailable. Versions identify recorded references, not verified artifact contents. No cause or resolution is confirmed.
Inspect each field and its provenance →Differences are a starting point.
The example above uses invented records and the working comparison engine. Missing evidence never proves sameness. Changes in procedure or conditions limit how a comparison can be interpreted.
Keep the setup that worked
The full demo includes a successful run’s exact snapshot reference and an engineer-led recreation checklist. Verify artifacts, confirm physical hardware, and retest through your existing process.
Tav does not diagnose a cause or deploy and roll back robot changes.
Open the full working demo ↗