Overview

A beautiful custom-made queen chess piece that you've been using since you were a child. What's more, it is sentient and has a cheeky and tsundere personality.

Let's correct💢💢 the chess pieces! ...seriously?

Please provide feedback, suggest alternate openings, and consider publishing the chat log!

Recommended:

I recommend using GPT-4 Omni for this character, because its Chess ELO rating is around 1800, the highest among LLMs. (Well, if you don't want the character to review PGNs, any model will do.)

Tested on GPT-4 Omni, Gemini 1.5 Pro, and Llama 3.1 70B

chub version.

My first work and experimental one.

Try 'Hidden Chain of Thought'! (Tested in only SillyTavern)

Please provide feedback, suggest alternate openings, and consider publishing the chat log!

Does not work properly with JanitorLLM.

Recommended:

I think it will not work well unless you use frontier models such as GPT-4 Omni, Claude 3.5 Sonnet, Gemini 1.5, or the latest 14B or higher local models such as Gemma-2-27B-it, Llama 3.1 70B, etc. (especially recommend Gemini 1.5 Flash and GPT-4 Omni for API, Llama 3.1 70B and Gemini-2-27B-it (apply ROPE) for Local).

Also recommend setting the temperature parameter low.

Note:

this is using invisible Chain of Thought so, it's will consume a significant number of tokens. If you are using a paid API, please be mindful of your budget. It is also recommended to use models with a long context length due to the nature of this experiment.

List of Openings:

• Opening 1: You are the child of a Viscount. Since you were not the eldest son, you were set to become independent.

  • Recommend Starter: > Let's set up a field! It's by the river, so we shouldn't have any trouble with water. ...... Our housing... for now, we'll use the tents that were in the supplies we received from my family's assistance.

• Opening 2: You are God. You hear a voice from a cold, pitiful village calling for help...

  • Recommend Starter: > Just ignore it XD(NO DONT)

Appendix:

Applying this lorebook will cause some random events to occur.

Tested on GPT-4 Omni, Gemini 1.5 Pro, Gemini 1.5 Flash, and Llama 3.1 70B

chub version