Absurd solutions

Open
#200 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
1/5
Estimated time
Under an hour
Newbie friendliness
25/100
Issue type
Documentation
Clarity
Needs clarification
Activity status
Quiet
Domain
content

Research direction

The issue contains examples of AI generated MATLAB exam solutions and explains how they are graded. Start by identifying the intended change in the issue, since it does not specify a repository file or a concrete edit. Done would mean the examples and grading guidance are placed or updated in the intended location.

Written by the indexing model from the issue text.

Description

I am marking the exams and here I want to report the most absurd solutions

using the AI a student has created the required pivot table MANUALLY


willingn     = [1;1;2;2;3;3;4;4;5;5;6;6;7;7];         

disc_compete = {"[1, 4)"; "[4, 7)"; "[1, 4)"; "[4, 7)"; "[1, 4)"; "[4, 7)"; "[1, 4)"; "[4, 7)"; "[1, 4)"; "[4, 7)"; "[1, 4)"; "[4, 7)"; "[1, 4)"; "[4, 7)"};                   

XGlobal_1_5  = [35; 4; 33; 3; 24; 8; 22; 34; 23; 44; 10; 31; 9;  9];   
XGlobal_5_7  = [ 2; 1;  2; 2;  2; 1; 13;  7; 30;  9;264; 12;571;972];  

% crating a Table now 
T = table(willingn, disc_compete, XGlobal_1_5, XGlobal_5_7);
disp(T)

Of course this student had a penalization of -5 to the overall exam

The same student also got this very generic and wrong output from AI

% Comment on univariate outliers:
% Univariate outliers are observations that fall far from the main cluster
% in either age or xGlobal. Points with age > 30 or age < 20 may be outliers.
% Points with xGlobal values much higher than the main distribution

Look at this student



% The table needs to group:
% Rows: 'willing' (1 to 7) and discretized 'compete' into [1, 4) and [4, 7]
% Columns: discretized 'xGlobal' into [1, 5) and [5, 7]

% Discretize 'compete' 
% [1, 4) matches values 1, 2, 3. [4, 7] matches 4, 5, 6, 7.
disc_compete = cell(height(sm), 1);
disc_compete(sm.compete >= 1 & sm.compete < 4) = {'[1, 4)'};
disc_compete(sm.compete >= 4 & sm.compete <= 7) = {'[4, 7]'};

% Discretize 'xGlobal'
% [1, 5) matches 1, 2, 3, 4. [5, 7] matches 5, 6, 7.
disc_xGlobal = cell(height(sm), 1);
disc_xGlobal(sm.xGlobal >= 1 & sm.xGlobal < 5) = {'[1, 5)'};
disc_xGlobal(sm.xGlobal >= 5 & sm.xGlobal <= 7) = {'[5, 7]'};

% Create a temporary table for pivot grouping
tmpTable = table(sm.willingn, disc_compete, disc_xGlobal, ...
    'VariableNames', {'willingn', 'disc_compete', 'xGlobal_bins'});

% Compute the Pivot Table using unstack
TVP_raw = groupsummary(tmpTable, {'willingn', 'disc_compete', 'xGlobal_bins'});
TVP = unstack(TVP_raw, 'GroupCount', 'xGlobal_bins');

% Ensure column order matches the prompt: [1, 5) then [5, 7]
target_records = sm(sm.willingn == 1 & ...
                    sm.compete >= 4 & sm.compete <= 7 & ...
                    sm.xGlobal >= 1 & sm.xGlobal < 5, :);

disp('--- Task 2: First 3 columns of the 4 extracted records ---');
disp(target_records(1:4, 1:3));

disp('--- Task 3: Contingency Table (willingn vs compete) ---');
[ContingencyTable, chi2, p, labels] = crosstab(sm.willingn, sm.compete);
disp(ContingencyTable);

The incredible thing is that the dataset has not even been LOADED into memory!!!!!!!!!!!!!!!!!

SAL.... you have been penalized!

Now let us come to this solution which is incredible

%% Compute the pivot table shown in the text

KU.disc_compete = strings(height(KU),1);

KU.disc_compete(KU.compete >= 1 & KU.compete < 4) = "[1,4)";
KU.disc_compete(KU.compete >= 4 & KU.compete <= 7) = "[4,7]";

KU.disc_compete = categorical(KU.disc_compete);

KU.disc_xGlobal = strings(height(KU),1);

KU.disc_xGlobal(KU.xGlobal >= 1 & KU.xGlobal < 5) = "[1,5)";
KU.disc_xGlobal(KU.xGlobal >= 5 & KU.xGlobal <= 7) = "[5,7]";

KU.disc_xGlobal = categorical(KU.disc_xGlobal);

TVP = groupsummary( ...
KU,...
{'willingn','disc_compete','disc_xGlobal'} ...
);

TVP = unstack(TVP,"GroupCount","disc_xGlobal");
TVP.Properties.VariableNames{3} = '[1,5)';
TVP.Properties.VariableNames{4} = '[5,7]';

disp("Pivot table:")
disp(TVP)

This student however managed to get it right from the AI.
However the evaluation in this case is reduced

Look what this student has done later

%% Comment on both plots

fprintf("Chi-square statistic = %.4f\n", chi2);
fprintf("p-value = %.4f\n", pValue);

if pValue < 0.05
    disp("Variables are associated")
else
    disp("Variables are not associated")
end

disp("Cells with large values contribute more to the Chi-square statistic.")

disp("The Pareto plot highlights the most important cells.")

This is the typical very generic thing generated by AI.

Also this part has been fully created by AI

%% Create a plot showing the contribution of each cell
% to the Chi-square index and the associated Pareto plot

rowTotals = sum(ContTable,2);
colTotals = sum(ContTable,1);

grandTotal = sum(ContTable(:));

Expected = rowTotals * colTotals / grandTotal;

ChiContribution = (ContTable - Expected).^2 ./ Expected;

figure
heatmap(ChiContribution)

title("Chi-square contributions")

figure
pareto(ChiContribution(:))

KU... you got a PENALIZATION for this

This is another absurd solution

This code generated by AI

SI = readtable("univer.xlsx","ReadRowNames",true);

disp(SI.Properties.VariableNames)
disp("Dataset Variables displayed")

%% TASK 2


SI.compClass = categorical( ...
    discretize(SI.compete, [1 4 7], ...
    'categorical', ["Low", "High"]));

SI.globClass = categorical( ...
    discretize(SI.xGlobal, [1 5 7], ...
    'categorical', ["Small", "Large"]));

%% Summary Table

S = groupsummary( ...
    SI,...
    {'willingn','compClass','globClass'} ...
    );

S = unstack(S,"GroupCount","globClass");


disp(S)

produces

Image a table which is very different from the onw which was requested

Of course ZO... you have been strongly penalized

The summary of this is: you can use the copilot and the AI but

  1. if the solution generated by AI is correct but overcomplicated you will not have full score
  2. if the solution generated by AI is wrong and then you do not realize it is wrong your exam will be strongly penalized
Dominant language
MATLAB
Stars
30
Forks
1
PR merge metrics
No merged PRs in 30d

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from UniprJRC/DSwithMATLAB

All issues in UniprJRC/DSwithMATLAB

Similar issues

More MATLAB issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.