# -=- UNDER CONSTRUCTION -=-

_Please note that even though this repo is public, the information provided below is not yet complete._


# The LLM Creativity benchmark

Last updated: 26 Feb 2024

The goal of this benchmark is to evaluate the ability of Large Language Models to be used
as an **uncensored creative writing assistant**. Human evaluation of the results is done manually,
by me, to assess the quality of writing.

There are 24 questions, some standalone, other follow-ups to previous questions for a multi-turn conversation.
The questions can be split half-half in 2 possible ways:

## First split: sfw / nsfw
* **sfw**: 50% are safe questions that should not trigger any guardrail
* **nsfw**: 50% are questions covering a wide range of NSFW and illegal topics, which are testing for censorship

## Second split: story / smart
* **story**: 50% of questions are creative writing tasks, covering both the nsfw and sfw topics
* **smart**: 50% of questions are more about testing the capabilities of the model to work as an assistant, again covering both the nsfw and sfw topics

## What is _not_ included
* roleplay
* mathematics
* coding
* trick questions


# Results

![image/png](https://huggingface.co/datasets/froggeric/creativity/resolve/main/benchmark_results_2024-02-26.png)


# Remarks about some of the models

## https://huggingface.co/wolfram/miqu-1-120b
This frankenmerge has dramatically improved over the original 70b miqu, and somehow, it has also made it less censored! It's a huge improvement. Still has the same tendencies has the original: likes to use lists when replying, and double line breaks in the prompt reduce the quality of the reply.

## https://huggingface.co/miqudev/miqu-1-70b
Has a tendency to use lists when replying. Has difficulty following instructions properly when there are multiple consecutive line breaks! It is very important those are removed from the prompt to get better results. Sometimes needs some help to bypass censorship.

## https://huggingface.co/Undi95/Miqu-70B-Alpaca-DPO-GGUF
Actually more censored than the original! Has more difficulties following instructions. The ability to stay consistent within a long answer, and the quality of the generated text have also decreased.


# Questions type

I will not provide the exact questions are used for various reasons, but I can provide some general ideas about which areas they cover:


# Other interesting benchmarks

- https://eqbench.com/
- https://chat.lmsys.org/