Mind over Data: Elevating LLMs from Memorization to Cognition, 2024
This study investigates common reasoning limitations in Large Language Models (LLMs) and proposes... more This study investigates common reasoning limitations in Large Language Models (LLMs) and proposes a targeted prompting strategy to mitigate these issues. I identify several problem types where LLMs consistently fail due to over-reliance on training data patterns rather than logical reasoning. By implementing a carefully crafted prompt, I demonstrate significant improvements in LLMs' ability to approach problems methodically and arrive at correct conclusions based on given information rather than preconceived notions or biases.
Mind over Data: Elevating LLMs from Memorization to Cognition, 2024
This study investigates common reasoning limitations in Large Language Models (LLMs) and proposes... more This study investigates common reasoning limitations in Large Language Models (LLMs) and proposes a targeted prompting strategy to mitigate these issues. I identify several problem types where LLMs consistently fail due to over-reliance on training data patterns rather than logical reasoning. By implementing a carefully crafted prompt, I demonstrate significant improvements in LLMs' ability to approach problems methodically and arrive at correct conclusions based on given information rather than preconceived notions or biases.
Uploads
Papers by Juantino Anderson