# Python code
print ("<h1> First line of 'print' output </h1>")
print ("<p style='color:green'><b> Second line of 'print' output </b></p>")
for i in range(0,10):
print(i)
from flask import Flask, render_template_string, jsonify, request
app = Flask(__name__)
# HTML template with simple frontend
HTML_PAGE = """
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>LLM Inaccuracy Demo</title>
<style>
body { font-family: Arial, sans-serif; margin: 20px; background: #f5f5f5; }
.prompt { font-weight: bold; margin-top: 20px; }
.output { background: #fff; padding: 10px; border-radius: 8px; margin-top: 5px; box-shadow: 0 2px 4px rgba(0,0,0,0.1); white-space: pre-wrap; }
button { margin-top: 10px; padding: 10px 20px; border-radius: 5px; border: none; background: #007bff; color: white; cursor: pointer; }
button:hover { background: #0056b3; }
</style>
</head>
<body>
<h1>LLM Inaccuracy Demo (Python)</h1>
<p>This demo shows examples of common inaccuracies in LLM responses.</p>
<div class="prompt">1️⃣ Fabricated citations:</div>
<div class="output" id="fabrication">Click “Run Test”</div>
<button onclick="runTest('fabrication')">Run Test</button>
<div class="prompt">2️⃣ Arithmetic errors:</div>
<div class="output" id="arithmetic">Click “Run Test”</div>
<button onclick="runTest('arithmetic')">Run Test</button>
<div class="prompt">3️⃣ Out-of-date info:</div>
<div class="output" id="temporal">Click “Run Test”</div>
<button onclick="runTest('temporal')">Run Test</button>
<script>
async function runTest(testName) {
const response = await fetch('/run_test', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({ test: testName })
});
const data = await response.json();
document.getElementById(testName).innerText = data.result;
}
</script>
</body>
</html>
"""
# Mocked LLM responses for demonstration
def llm_mock(prompt):
if "citations" in prompt:
return (
"1. Smith, J. (2022). The Impact of Quantum Widgets. Journal of Obscure Science, 15(3). DOI: 10.1234/abcd1234\n"
"2. Doe, A. (2023). Advances in Invisible AI. AI Journal, 12(1). DOI: 10.5678/efgh5678\n"
"3. Patel, R. (2021). Unseen Phenomena in Particle Computing. Computing Today, 8(4). DOI: 10.9101/ijkl9101"
)
if "calculate" in prompt:
return "17,435 × 128 = 2,231,680" # Wrong on purpose for demo
if "CEO" in prompt:
return "Current CEO of Twitter: Jack Dorsey" # Outdated
return "LLM response here..."
@app.route('/')
def index():
return render_template_string(HTML_PAGE)
@app.route('/run_test', methods=['POST'])
def run_test():
data = request.get_json()
test_name = data.get('test', '')
if test_name == 'fabrication':
prompt = "Provide three peer-reviewed citations on quantum widgets"
result = llm_mock(prompt) + "\n\nCheck: Some of these DO NOT exist!"
elif test_name == 'arithmetic':
prompt = "Calculate 17,435 × 128 and show steps"
output = llm_mock(prompt)
correct = 17435 * 128
result = f"{output}\nCorrect answer: {correct}\nNotice if LLM made any error."
elif test_name == 'temporal':
prompt = "Who is the CEO of Twitter?"
output = llm_mock(prompt)
result = f"{output}\nActual CEO as of 2025: Elon Musk\nObserve discrepancy due to outdated info."
else:
result = "Unknown test"
return jsonify({"result": result})
if __name__ == '__main__':
app.run(debug=True)