Show HN: LLM 100k portfolio management benchmark
github.comPoC for something some the potential to yield some interesting results eventually. LLM:s for trading is so stupid. They would only be useful for a small set of tasks that are not normal retail trading either way (for example trying to trade the news by being slightly faster). >LLM:s for trading is so stupid. That's the hypothesis this experiment is trying to validate but so far I have no reasons to believe they will behave much worse than human portfolio managers. I don't get why? like the llms wouldn't know the latest earnings reports or news? so what do they bring to the table? To optimize their portfolio, the primary objective defined for the LLMs, it is imperative to evaluate the risk-reward ratio, formulate cogent assumptions about future market conditions, and leverage tools and their understanding of human psychology and financial market dynamics. This task may be a good proxy to measure how well LLMs are able to coordinate the aforementioned efforts. * List current holdings and recent context * Update portfolios based on model decisions In the project overview at the top of the readme