Summary: Researchers have developed a new framework called Automatic Reasoning and Tool-use (ART) that uses large language models (LLMs) to automatically generate intermediate reasoning steps as a program. ART selects demonstrations of multi-step reasoning and tool use from a task library, achieving substantial improvement over few-shot prompting and automatic chain of thought (CoT) on unseen tasks in the BigBench and MMLU benchmarks, and matching the performance of hand-crafted CoT prompts on a majority of these tasks.
Top Tweets

Comments ()

Log in or Sign Up to comment.