Here's what stopped breaking, when you make LLM agents author in two formats

1. 1Separate judgement and invariants in your LLM agent's output using Markdown and TOML
2. 2Use Markdown to carry judgement and TOML to carry invariants
3. 3Test your LLM agent's output in both formats to identify potential failures
4. 4Configure your model to handle errors and inconsistencies in both formats
5. 5Apply this approach to your existing LLM agent projects to improve their robustness

Developers and data scientists working with LLM agents can benefit from this knowledge to improve their model's performance and reliability

💡 Separating judgement and invariants is crucial for reliable LLM agent output

💡 Separate judgement and invariants in LLM agent output using Markdown and TOML to improve reliability