3. Hackathon motivation

• Performing trade entry reliably requires attention to detail and highly specialized knowledge

  • Trader and quant jargon

  • Conventions and defaults

  • Interaction between different parameters

• Well-known LLM limitations such as variability of response from one call to the next and hallucinations are additional challenges

• Unlike Python coding assistance and business document writing, optimizing LLMs for this task was not a priority for the foundation model developers (OpenAI, Meta, Anthropic, Google)

• This is why the quant community has the best chance of developing a reliable solution to trade entry for what-if analysis

• The goal of this hackathon is to advance the state of the art in LLM-based trade entry, with focus on achieving near 100% reliability across a wide range of trade definition formats