Search papers, labs, and topics across Lattice.
1
3
5
Free-form Chain-of-Thought prompting actually *hurts* LLMs when it comes to structured function-calling tasks, but a guided-structured template approach boosts tool-use accuracy by up to 12%.
