LLM : Your Enhanced Code Reviewer

Ideogram prompt :word cloud art of a Nerdy Harry potter man form with text:”LLM : Code Review & Critical Reasoning” black background and white words., cinematic, fashion, poster, illustration, typography, painting, conceptual art

The LLM (Large Language Model) Revolution is changing how we teach and evaluate skills. It’s not here to replace us but to be our digital Robin.

In this new way of coding, we’re not blindly accepting code, but we’re more like code chefs, carefully crafting each piece to fit our project. It’s about finding the right ingredients for success and avoiding coding mishaps.

Embracing this change is like getting a new tool to your toolbox. It’s a step forward in the world of software, and anyone can use it to make their coding life easier. Let’s dive into our discussion for today.

To start, here’s my generated Buddhabrot fractal code :

GPT4 output

Now let’s examine another code generated by codellama70b, recently released by Meta :

From an initial standpoint, we can see that one uses random number to map the complex plane, while the second code maps it through given width and height.

Now, let’s see what GPT-4 provides when prompted :

“What is the difference between the code ?”

My given input

Here’s the output, and you can form your own reasoning:

For a better distinction and more critical reasoning, let’s see what other open-source LLM offers:

Ref: lmsys

From the above discussions, we can see significant distinctions between the outputs. In GPT-4, the second code snippet discusses the Buddhabrot fractal, whereas Mixtral and Llama2 by Meta provide a completely different result, referring to it as a “Julia” set. What we can infer from this discussion is that LLM can help trouble differentiate complex logical reasoning and help you kickstart your projects much quicker than before.

If you enjoyed my post above, feel free to connect to me on linkedin or support my research and studies on patreon ! — !

“Opensource to the win !”