Fix DeepSeek-R1 Stalling on Contradictory Instructions

Thoughtology analysis found that while DeepSeek-R1 adapts to user input when contradictions arise, this comes at the cost of efficiency because the model spends significant compute time deliberating over user intentions. Eliminate contradictory constraints in your prompt before calling R1.

User Prompt: Fit a survival analysis Cox proportional hazards model in Python lifelines for the dataset schema below. Single Objective: Optimize strictly for concordance index (C-index) with L2 penalizer=0.1; ignore all legacy linear regression comments in the script.