labels: experimental, llm, publication, public_good
Challenge task for an agentic system: detect the presence of fraud in the Enron dataset.
- Almost certainly would require test-time training or some kind of fine-tuning
- Strong possibility this dataset is included in common pre-training corpuses already