Help with click beetle larvae

General
1

I know this type of issue has been brought up in multiple posts but I want to bring it up again.

Today’s issue is “click beetle” larva. It seems the CV is suggesting one specific species (Parallelostethus attenuatus) for anything that remotely looks like a larval click beetle. It seems the issue stems from a single observation ( https://www.inaturalist.org/observations/40589237 ) where a larva was reared to adulthood, and turned out to be this species. But now the CV has gone crazy suggesting this species for larva. There is no way to ID click beetle larva from photos (as far as I know), and I’m not really familiar enough with beetle larva to tell if one is an Elateridae or something similar, like a Tenebrionidae.

What can we do to fix this? Why has the CV latched on to this one observation? I’ve disagreed with all the research grade observations of larvae except the original post, but there were less than 10 of those. There are now over 800 observations ID’d as Parallelostethus attenuatus and I would guess close to 600 are incorrectly ID’d larvae.

There needs to be human input to the CV where we can tell it to back off.

2

I have now pushed every single one of those larval observations (except for the reared one) back to ‘beetles’.

In case you weren’t aware (or for others reading this thread), an important thing to note is that the CV doesn’t update in real time when corrections are made; each version of the model is based on an extract taken at a fixed point in time, and until the next model is released in 1-3 months time, suggestions will be based on that fixed point. So sometimes it requires a more extended period of making corrections so that you can span multiple of these fixed points

just noting it wasn’t only this one, there were a small handful of others which have been RG for a few years (that I have now corrected)

3

The CV is trained on all photos, not just RG.

This is what adding disagreeing IDs does.

4

Thanks. I must have missed a few of the RG observations.

I still think its crazy that one observation led to so many mistakes caused by the CV.

There must not have been as many larva as I thought, but still quite a few. Thanks for doing that. It seemed like a daunting task.

5

CV update published on 15 April

trained off data exported on March 8, 2026 Includes 373 insect sp across the world

Expect the next update when CNC dust has settled ?