Skip to yearly menu bar Skip to main content


Poster

Take a Step Back: Evoking Reasoning via Abstraction in Large Language Models

Huaixiu Steven Zheng · Swaroop Mishra · Xinyun Chen · Heng-Tze Cheng · Ed H. Chi · Quoc V Le · Denny Zhou

Halle B #131
[ ]
Wed 8 May 1:45 a.m. PDT — 3:45 a.m. PDT

Abstract:

We present STEP-BACK PROMPTING, a simple prompting technique that enables LLMs to do abstractions to derive high-level concepts and first principles from instances containing specific details. Using the concepts and principles to guide reasoning, LLMs significantly improve their abilities in following a correct reasoning path towards the solution. We conduct experiments of STEP-BACK PROMPTING with PaLM-2L, GPT-4 and Llama2-70B models, and observe substantial performance gains on various challenging reasoning-intensive tasks including STEM, Knowledge QA, and Multi-Hop Reasoning. For instance, STEP-BACK PROMPTING improves PaLM-2L performance on MMLU (Physics and Chemistry) by 7% and 11% respectively, TimeQA by 27%, and MuSiQue by 7%.

Chat is not available.