Dhagash Mehta

Position: Evaluating LLMs in Finance Requires Explicit Bias Consideration

Large Language Models (LLMs) are increasingly integrated into financial workflows, but evaluation practice has not kept up. Finance-specific biases can inflate performance, contaminate backtests, and make reported results useless for any deployment claim. We identify five recurring biases in financi…