eedi-mining-misconceptions-in-mathematics

[hollow basin]

hey there, nice to meet you all here

[lofty sorrel]

Hey hey 👋

[hollow basin]

anyone wants to team up for this one?

[tame trellis]

Hi everyone. I'm not going to lie to you. I'm not sure if I completely grasp what we're trying to achieve here. My understanding is that we're trying to link a misconception to an incorrect answer. Is that correct?

[hollow basin]

yes I also understand that's what we need to do

[lofty sorrel]

@golden quail Should the submission contain only one Misconception ID, or multiple?

The sample submission contains multiple Misconception IDs: https://www.kaggle.com/competitions/eedi-mining-misconceptions-in-mathematics/data?select=sample_submission.csv

While the Kaggle staff seems to be saying that we should only predict one Misconception ID: https://www.kaggle.com/competitions/eedi-mining-misconceptions-in-mathematics/discussion/533744#2990576

Would you mind clarifying?

[golden quail]

I'm a host for the #um-game-playing-strength-of-mcts-variants competition, I don't know anything about this one 😄

[lofty sorrel]

Oh, sorry, I misunderstood the Discord role.

[indigo jolt]

There isn't much test data, but is there anything anyone can do?

[near ore]

I'm looking for 5 members to join our team (Math Bashers). DM me ASAP if you're interested!

[solemn iron]

Hey there, anyone still looking for members to join their team. I just joined the competition and am looking for teams.

[urban vine]

goog

[brisk helm]

Feel free to contact me if anyone need group member

[obtuse jacinth]

let's Team Up ?!

[muted cairn]

I was participating in the EEDI competition, any idea why the cell randomly stops running while loading the model?

[supple sage]

Is there any restriction on the total parameters of the model? @golden quail

[digital basin]

MalAlgoQA: a new dataset that might be useful for finetuning for this task

Paper: https://www.arxiv.org/abs/2407.00938
Dataset: https://forms.gle/D6ayHHrZFQWUApaM8

This is our dataset to complement our findings in this paper (https://arxiv.org/abs/2310.02439) with EEDI team -- a precursor to this challenge.
Section 2.2: Using LLMs to simulate an expert tutor

In this paper study if LLMs can follow a malgorithm which leads to an incorrect distractor choice (spin on word algorithm using the “mal-rules” word coined by Payne & Squibb).

[topaz pumice]

thanks

[supple sage]

Since Kaggle requires us to rerun our script to obtain a score, but their free QPU only has 16GB of memory, can I use a larger model?

[clever pelican]

Hello! I have a quick question about this competition. Are we allowed to use self-generated datasets or models trained on self-generated datasets as long as we make them free and public? Since the rule says "Freely & publicly available external data is allowed, including pre-trained models"

[winged sorrel]

Well, someone has to make that external dataset available, correct? So why not you?