Lelapa AI Buzuzu-Mavi Challenge
Helping Africa
$1 300 USD
Completed (11 months ago)
Natural Language Processing
Sentiment Analysis
Machine Translation
492 joined
118 active
Start
Jan 09, 25
Close
Apr 06, 25
Reveal
Apr 07, 25
'%3e%3cpath fill='%23001489' d='M0 0v6h9V0z'/%3e%3cpath fill='%23e03c31' d='M0 0v3h9V0z'/%3e%3cg stroke='%23fff' stroke-width='2'%3e%3cpath id='d' d='m0 0 4.5 3L0 6m4.5-3H9'/%3e%3cuse xlink:href='%23b' stroke='%23ffb81c' clip-path='url(%23c)'/%3e%3c/g%3e%3cuse xlink:href='%23d' fill='none' stroke='%23007749' stroke-width='1.2'/%3e%3c/g%3e%3c/svg%3e){kind=small}
External data
Data Ā· 14 Mar 2025, 15:34 Ā· 6
Can someone from @zindi or the organisers please clearly state whether external data is permitted in this challenge or not? The competition rules on the Info page seem clear: You may use only the datasets provided for this challenge.
But then there is contradictory information in the discussion that suggests otherwise, most clearly here:
If external data is indeed allowed, could you please update the rules on the Info page so that it's clear to everyone, and if not, please remove the misleading comments from the discussion?
I asked in the Lelapa Discord a while back - I was told by one of the organizers that we can use any external, new, or synthetic data we wish (as long as this doesn't contain the test data of course)
thanks @oxxocodes for the feedback. Do you have the invite link to the Discord? I'm joining a little late and probably missed that somewhere along the way.
@zindi / @Amy_Bray - feedback from the Lelapa team regarding external data seems to contradict the competition rules as stated on the Info page. Could the Info page be updated to avoid this ambiguity?
https://discord.gg/ug2KV6jR
@stefan027 Naive question: your impressive high score was achieved with external data?
Hi @nymfree, for my submissions I've used a small subset of the Inkuba-instruct dataset for fine-tuning. I've also used AfriXNLI but I haven't made a submission yet. I don't know if those datasets are considered external - they're not linked in the competition's data page, but they are mentioned in the InkubaLM release blog with reference to the Zindi competition 🤷♂️
Quoting from the blog: "NOTE: We’re withholding the [Inkuba-Instruct] test set for now because we’ll be running a Zindi competition using the dataset soon — we will release the test set afterwards. 🎉"
Thanks for the detailed reply. It is indeed unfortunate that Zindi does not communicate clearly on this