September 9, 2024 Dataset Restricted Access

Völschow, Marcel; Buczek, P.; Carreno-Mosquera, P.; Mousavias, C.; Reganova, S.; Roldan-Rodriguez, E.; Steinbach, Peter; Strube, A.

MARC21 XML Export

<?xml version='1.0' encoding='UTF-8'?>
<record xmlns="http://www.loc.gov/MARC21/slim">
  <leader>00000nmm##2200000uu#4500</leader>
  <controlfield tag="001">3137</controlfield>
  <datafield tag="542" ind1=" " ind2=" ">
    <subfield code="l">restricted</subfield>
  </datafield>
  <controlfield tag="005">20240912092419.0</controlfield>
  <datafield tag="100" ind1=" " ind2=" ">
    <subfield code="a">Völschow, Marcel</subfield>
  </datafield>
  <datafield tag="980" ind1=" " ind2=" ">
    <subfield code="a">user-rodare</subfield>
  </datafield>
  <datafield tag="500" ind1=" " ind2=" ">
    <subfield code="a">The dataset is available in English here only and will be removed, once the mlphys101 publication was accepted and released to the public.</subfield>
  </datafield>
  <datafield tag="980" ind1=" " ind2=" ">
    <subfield code="a">dataset</subfield>
  </datafield>
  <datafield tag="773" ind1=" " ind2=" ">
    <subfield code="a">https://www.hzdr.de/publications/Publ-39561</subfield>
    <subfield code="i">isIdenticalTo</subfield>
    <subfield code="n">url</subfield>
  </datafield>
  <datafield tag="773" ind1=" " ind2=" ">
    <subfield code="a">10.14278/rodare.3136</subfield>
    <subfield code="i">isVersionOf</subfield>
    <subfield code="n">doi</subfield>
  </datafield>
  <datafield tag="041" ind1=" " ind2=" ">
    <subfield code="a">eng</subfield>
  </datafield>
  <datafield tag="245" ind1=" " ind2=" ">
    <subfield code="a">mlphys101 - Exploring the performance of Large-Language Models in multilingual undergraduate physics education</subfield>
  </datafield>
  <datafield tag="260" ind1=" " ind2=" ">
    <subfield code="c">2024-09-09</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">machine learning</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">deep learning</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">large language models</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">chatgpt</subfield>
  </datafield>
  <datafield tag="653" ind1=" " ind2=" ">
    <subfield code="a">blablador</subfield>
  </datafield>
  <datafield tag="024" ind1=" " ind2=" ">
    <subfield code="a">10.14278/rodare.3137</subfield>
    <subfield code="2">doi</subfield>
  </datafield>
  <datafield tag="520" ind1=" " ind2=" ">
    <subfield code="a">&lt;p&gt;Large-Language Models such as ChatGPT have the potential to revo-&lt;br&gt;&#13;
lutionize academic teaching in physics in a similar way the electronic calculator,&lt;br&gt;&#13;
the home computer or the internet did. AI models are patient, produce answers&lt;br&gt;&#13;
tailored to a student’s needs and are accessible whenever needed. Those involved&lt;br&gt;&#13;
in academic teaching are facing a number of questions: Just how reliable are pub-&lt;br&gt;&#13;
licly accessible models in answering, how does the question’s language affect the&lt;br&gt;&#13;
models’ performance and how well do the models perform with more difficult tasks&lt;br&gt;&#13;
beyond retrieval? To adress these questions, we benchmark a number of publicly&lt;br&gt;&#13;
available models on the mlphys101 dataset, a new set of 823 university level MC5&lt;br&gt;&#13;
questions and answers released alongside this work. While the original questions&lt;br&gt;&#13;
are in English, we employ GPT-4 to translate them into various other languages,&lt;br&gt;&#13;
followed by revision and refinement by native speakers. Our findings indicate that&lt;br&gt;&#13;
state-of-the-art models perform well on questions involving the replication of facts,&lt;br&gt;&#13;
definitions, and basic concepts, but struggle with multi-step quantitative reason-&lt;br&gt;&#13;
ing. This aligns with existing literature that highlights the challenges LLMs face&lt;br&gt;&#13;
in mathematical and logical reasoning tasks. We conclude that the most advanced&lt;br&gt;&#13;
current LLMs are a valuable addition to the academic curriculum and LLM pow-&lt;br&gt;&#13;
ered translations are a viable method to increase the accessibility of materials, but&lt;br&gt;&#13;
their utility for more difficult quantitative tasks remains limited.&lt;/p&gt;&#13;
&#13;
&lt;p&gt;The dataset is available in English here only and will be removed, once the mlphys101 publication was accepted and released to the public.&lt;/p&gt;</subfield>
  </datafield>
  <datafield tag="909" ind1="C" ind2="O">
    <subfield code="o">oai:rodare.hzdr.de:3137</subfield>
    <subfield code="p">openaire_data</subfield>
    <subfield code="p">user-rodare</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="a">Buczek, P.</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="a">Carreno-Mosquera, P.</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="a">Mousavias, C.</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="a">Reganova, S.</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="a">Roldan-Rodriguez, E.</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="a">Steinbach, Peter</subfield>
    <subfield code="0">(orcid)0000-0002-4974-230X</subfield>
  </datafield>
  <datafield tag="700" ind1=" " ind2=" ">
    <subfield code="a">Strube, A.</subfield>
  </datafield>
</record>