Introduction

Reference solution for LLM Prompt Recovery in Kaggle competition.

Code and comments are available in baseline.ipynb.

Ideas

Ours

By writing the prompts and examples by ourselves, In-context Learning is used to enhance performance in few-shot scenarios.

Inspried by Others

• 1st: So interesting! Enhance the prompt for sentence-t5 by incorporating the lucrare token in a manner that is similar to the </s> token, thereby improving its effectiveness.