<!DOCTYPE html>
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body text="#0432ff" bgcolor="#feffff">
<p>Derek</p>
<p>The paper emphasises that it's not N-version programming as
originally conceived, and the author makes no claims about
achieving high dependability or fault-tolerance. </p>
<p>Knight and Levsen's explanation of the common-mode errors in
their experiment was that people taught in much the same way make
the same sort of oversights and errors. It may be that multiple
LLMs suffer from a similar weaknesses. I hope further research is
done before anyone claims that the technique is adequate for
high-dependability in safety applications</p>
<p>Martyn</p>
<div class="moz-cite-prefix">On 10/06/2026 13:20, Derek M Jones
wrote:<br>
</div>
<blockquote type="cite"
cite="mid:78fa89f6-4f4c-4f76-875c-1b26c66da7b9@knosof.co.uk">All,
<br>
<br>
N-version program is a great idea, but it is is also a lot
<br>
more expensive. At least until LLMs came along.
<br>
<br>
This paper
<br>
<a class="moz-txt-link-freetext" href="https://arxiv.org/abs/2606.07828">https://arxiv.org/abs/2606.07828</a>
<br>
<br>
creates five implementations of a 23k line specification.
<br>
Five different languages were used for the implementation,
<br>
but that is just for bragging rights.
<br>
<br>
So the correctness issue has now shifted from the code to
<br>
the specification.
<br>
<br>
A few months ago an LLM generated a C compiler. The compiler
<br>
did very well on correct code (which is the vast majority of
<br>
the training data), but very badly on incorrect code.
<br>
<a class="moz-txt-link-freetext" href="https://shape-of-code.com/2026/02/22/investigating-an-llm-generated-c-compiler/">https://shape-of-code.com/2026/02/22/investigating-an-llm-generated-c-compiler/</a>
<br>
<br>
</blockquote>
</body>
</html>