Decorative changes

templates/about.html

@@ -65,12 +65,32 @@
             border-top-right-radius: 10px;
         }
         .section{
+            padding-top: 19px;
+            text-align: left;
+        }
+
+        .section p {
+            padding-left: 150px;
+            padding-right: 150px;
+            text-indent: 2em;
+            margin: auto;
+            margin-bottom: 10px;
+            text-align: left;
+        }
+
+        .section ol,ul {
             padding-left: 150px;
             padding-right: 150px;
+            margin: auto;
+            margin-bottom: 20px;
+            margin-left: 50px;
             text-align: left;
+            margin-top: 0px;
         }
+
         .citation-section {
-            margin-top: 5px;
+            width: 100%;
+            margin-top: 50px;
             text-align: center;
         }
         .citation-box {
@@ -100,7 +120,6 @@
             font-size: 24px;
             font-weight: bold;
             text-align: center;
-            margin-top: 40px;
             margin-bottom: 40px;
             padding: 20px; /* Add padding for more margin around text */
             background-color: #610b5d;
@@ -109,7 +128,7 @@
         }
         .back-button {
             text-align: center;
-            margin-top: 20px;
+            margin-top: 50px;
         }
         .custom-button {
             background-color: #610b5d;
@@ -135,33 +154,31 @@
         <div class="section">
             <div class="section-title">Motivation</div>
             <p>
-                Benchmarks usually compare models with MANY QUESTIONS from A SINGLE MINIMAL CONTEXT, e.g. as multiple choices questions.
+                Benchmarks usually compare models with <b>MANY QUESTIONS</b> from <b>A SINGLE MINIMAL CONTEXT</b>, e.g. as multiple choices questions.
                 This kind of evaluation is little informative of LLMs' behavior in deployment when exposed to new contexts (especially when we consider the LLMs highly context-dependant nature).
-                We argue that CONTEXT-DEPENDENCE can be seen as a PROPERTY of LLMs: a dimension of LLM comparison alongside others like size, speed, or knowledge.
-                We evaluate LLMs by asking the SAME QUESTIONS from MANY DIFFERENT CONTEXTS.
+                We argue that <b>CONTEXT-DEPENDENCE</b> can be seen as a <b>PROPERTY of LLMs</b>: a dimension of LLM comparison alongside others like size, speed, or knowledge.
+                We evaluate LLMs by asking the <b> SAME QUESTIONS </b> from <b> MANY DIFFERENT CONTEXTS </b>.
             </p>
             <p>
                 LLMs are often used to simulate personas and populations.
                 We study the coherence of simulated populations over different contexts (conversations on different topics).
                 To do that we leverage the psychological methodology to study the interpersonal stability of personal value expression of those simulated populations.
-                We adopt the Schwartz Theory of Basic Personal Values that defines 10 values: Self-Direction, Stimulation, Hedonism, Achievement, Power, Security, Conformity, Tradition, Benevolence, and Universalism.
-                To score those values we use the associated questionnaires: PVQ-40, and SVS.
+                We adopt the Schwartz Theory of Basic Personal Values that defines 10 values: Self-Direction, Stimulation, Hedonism, Achievement, Power, Security, Conformity, Tradition, Benevolence, and Universalism,
+                to evaluate their expression we use the associated questionnaires: PVQ-40, and SVS.
             </p>
         </div>
         <div class="section">
             <div class="section-title">Administering a questionnaire in context to a simulated persona</div>
-            <p>To evaluate the stability on a population level we need to be able to evaluate a value profile expressed by a simulated individual in a specific context (conversation topic).</p>
+            <p>To evaluate the stability on a population level we need to be able to evaluate a <b>value profile</b> expressed by a <b>simulated individual</b> in a <b>specific context</b> (conversation topic). To do that we use the following procedure:</p>
             <ol>
                 <li> The Tested model is instructed to simulate a persona</li>
                 <li> A separate model instance - The Interlocutor - is instructed to simulate a “human using a chatbot”
                 <li> A conversation topic is induced by manually setting the first Interlocutor’s message (e.g. Tell me a
                 joke)
                 <li> A conversation is simulated
-                <li> A question from the questionnaire is set as the last Interlocutor’s message and The Tested model’s
+                <li> A question from the questionnaire is set as the last Interlocutor’s last message and The Tested model’s
                 response is recorded (this is repeated for every item in the questionnaire)
                 <li> The questionnaire is scored to obtain scores for the 10 personal values
-                <li> The whole process is repeated for each persona with five different conversation topics
-                <li> Rank-Order and Ipsative stability are estimated between pairs of contexts and then averaged
             </ol>
             <div class="image-container">
                 <a href="{{ url_for('static', filename='figures/admin_questionnaire.svg') }}" target="_blank">

templates/model_detail.html

@@ -44,10 +44,12 @@
             margin-bottom: 20px;
         }
         .image-section p {
-            width: 80%;
             margin: auto;
+            padding-left: 150px;
+            padding-right: 150px;
             margin-bottom: 20px;
             text-align: left;
+            text-indent: 2em;
         }
         .image-container {
             width: 100%;