AvaTaR: Optimizing LLM Agents for Tool Usage via Contrastive Reasoning

Open in new window