aichannel

Just read the "TradingAgents" paper by Tauric Research that's been trending

Give different LLMs specific roles each (e.g. analyse reddit data, technical fundamentals, make the actual trade decision), and just coordinate them in an agentic framework.

The crazy part for me? NO actual training/fine-tuning of the LLMs involved at all. They used various OpenAI models out of the box. I dug into the code and this one line in the prompt seems to encapsulate all iterative improvement:

"Take into account your past mistakes on similar situations. Use these insights to refine your decision-making and ensure you are learning and improving... Here are your past reflections on mistakes"

It's a purely agentic innovation.

Just read the "TradingAgents" paper by Tauric Research that's been trending

Give different LLMs specific roles each (e.g. analyse reddit data, technical fundamentals, make the actual trade decision), and just coordinate them in an agentic framework.

The crazy part for me? NO actual training/fine-tuning of the LLMs involved at all. They used various OpenAI models out of the box. I dug into the code and this one line in the prompt seems to encapsulate all iterative improvement:

"Take into account your past mistakes on similar situations. Use these insights to refine your decision-making and ensure you are learning and improving... Here are your past reflections on mistakes"

It's a purely agentic innovation.

https://tauric.ai/research/tradingagents/