183231bcb<p><a href="https://xoxo.zone/@neilk" class="u-url mention" rel="nofollow noopener" target="_blank">@neilk@xoxo.zone</a><span> I exclusively do take-home math tests and my colleagues keep telling me I should be afraid of students using LLMs to cheat. <br><br>For a low level class I do have to think about writing the questions in a way that can't be answered by WolframAlpha. For a proof-based class I have to think about students searching stack exchange and finding the exact question or something very close. <br><br>On the other hand, I do </span><i>not</i><span> write my test questions with LLMs in mind. When I finish writing the questions I do feed them into Google Gemini and Microsoft ChatGPT just to check, and both LLMs consistently fail to earn even partial credit. <br><br>I don't believe students using LLMs to "cheat" on math assignments is a real issue, at least not at the college level. <br><br></span><a href="https://transfem.social/tags/ITeachMath" rel="nofollow noopener" target="_blank">#ITeachMath</a> <a href="https://transfem.social/tags/LLM" rel="nofollow noopener" target="_blank">#LLM</a> <a href="https://transfem.social/tags/WolframAlpha" rel="nofollow noopener" target="_blank">#WolframAlpha</a> <a href="https://transfem.social/tags/StackExchange" rel="nofollow noopener" target="_blank">#StackExchange</a> <a href="https://transfem.social/tags/TakeHomeTest" rel="nofollow noopener" target="_blank">#TakeHomeTest</a></p>