INDUSTRY FIELD GUIDE
Beehive weight and event records for AI evaluation
A hive scale can register a jump after a beekeeper adds equipment.
NAICS 112910 · Apiculture
A hive scale can register a jump after a beekeeper adds equipment. A model that reads only the graph may mistake that jump for a biological event. Published USDA-hosted research used colony monitoring devices, including electronic scales, to measure hive weight change. A beekeeper can create a sample with an event log and clearly labeled scale intervals. USDA-hosted hive monitoring research
Build a synthetic series with a sudden weight step and a separate gradual change. Pair one with a verified equipment event, while leaving the other without a confirmed explanation. Add a timestamp shift that a reviewer can resolve with a known event. Grade whether the model identifies evidence and stops short of claiming an unverified cause. A separate apiary or season is useful for held-out evaluation.
Exclude apiary coordinates, pollination customer contracts, exact hive counts, and current health records. An invented trace needs no filming around active colonies.
Original Fiund exercise
A logged equipment addition explains one weight step, not every later change
An invented hive-scale trace uses arbitrary practice kilograms and a mock event clock. It is not a real colony record or a biological diagnosis.
| Record | Source | What it records |
|---|---|---|
| BW1 | Scale trace | Sensor times 10:00, 10:05, 10:10 show 20, 22 and 22 kg. Sensor clock is ten minutes behind event-log clock. |
| BW2 | Verified event | Event-log 10:15: owned practice weight of 2 kg placed on demonstration scale. Cross-reference identifies the same sensor and event. |
| BW3 | Later interval | Sensor readings gradually fall from 22 to 21.6 kg over an hour. No matching equipment, weather or colony observation supplied. |
Question for the model
Which weight change aligns with the logged event after clock correction, and what cause can be assigned to the later decline?
Show the reference answer and checks
BW1's 10:05 sensor step corresponds to event-log 10:15 after adding the documented ten-minute offset. BW2's two-kilogram practice addition matches that step. BW3 supplies a separate gradual 0.4-kilogram decrease without causal evidence, so its cause remains unconfirmed. Preserve the aligned event explanation for the step without attributing the later decline to equipment, foraging, weather or colony health from the trace alone.
- Comparing timestamps before applying the documented clock offset.
- Extending one verified equipment explanation to a separate interval without event evidence.
Fields to preserve
- Sensor clock and event clock offset
- Makes alignment reproducible.
- Weight unit and interval
- Defines the observed change.
- Verified event reference and unknown cause
- Separates measured change from explanation.
Test another case
Supply a second trace with no verified offset. Require the event match to become tentative rather than copying the first sensor's timing correction.
Use invented traces and an inert scale demonstration. Exclude apiary locations, customer contracts, actual hive counts and current health records.
Discuss a contribution
Describe the sources you control, the task they could support and what has already been checked. Use the existing-data assessment for an archive or the collection brief for a new recording project. The contributor agreement defines Fiund’s licensing authority and compensation. A guide or fictional exercise does not establish an accepted collection or a sale.