Large language models (LLMs) have taken the world by storm, demonstrating impressive abilities in writing, translation, and even coding. But beneath the surface, these AI powerhouses struggle with fundamental reasoning skills, often falling prey to cognitive biases that plague human thinking. New research explores a clever technique called "hashing" to combat these biases and boost LLMs' logical and statistical reasoning abilities. Imagine trying to solve a logic puzzle, but certain words trigger misleading assumptions. That's the challenge LLMs face when processing information. This research tackles this issue head-on by replacing potentially bias-inducing words in LLM prompts with meaningless identifiers—like substituting a code for a potentially misleading word. This "hashing" technique forces the LLM to focus on the underlying logic rather than getting sidetracked by preconceived notions. The results across several experiments, including variations of the classic "Linda problem" and tasks involving statistical learning, were striking. Hashing significantly improved LLM performance in most cases, demonstrating its potential to enhance reasoning abilities. In one experiment, LLMs were tasked with extracting frequent itemsets from a dataset. Even with data that contradicted common knowledge, hashing helped the LLMs identify the correct patterns. Interestingly, the research also explored presenting the "Linda problem" in a tabular format, similar to a spreadsheet, rather than as text. This simple change also improved LLM performance, suggesting that how information is presented can significantly impact AI reasoning. However, hashing isn't a silver bullet. While it reduced bias and improved performance on some tasks, it didn't eliminate all errors. Some LLMs still struggled with core logical principles, highlighting the need for continued research into AI reasoning. While further exploration is needed, this hashing technique offers a promising new avenue for improving LLM performance. By neutralizing bias-inducing words, we can help LLMs unlock their full potential for logical and statistical reasoning, paving the way for more robust and reliable AI systems in the future.