ReWOO: Decoupling Reasoning from Observations for Efficient Augmented Language Models
Authors:
- Binfeng XuBXBinfeng Xu
- Zhiyuan PengZPZhiyuan Peng
- Bowen LeiBLBowen Lei
- Subhabrata MukherjeeSMSubhabrata Mukherjee
- Yuchen LiuYLYuchen Liu
- Dongkuan XuDXDongkuan Xu
Preprint
Abstract
Augmented Language Models (ALMs) blend the reasoning capabilities of Large Language Models (LLMs) with tools that allow for knowledge retrieval and action execution. Existing ALM systems trigger LLM thought processes while pulling observations from these tools in an interleaved fashion. Specifically, an LLM reasons to call an external tool, gets halted to fetch the tool's response, and then decides the next action based on all preceding response tokens. Such a paradigm, though straightforward and easy to implement, often leads to huge computation complexity from redundant prompts and repeated execution. This study addresses such challenges for the first time, proposing a modular paradigm ReWOO (Reasoning WithOut Observation) that detaches the reasoning process from external observations, thus significantly reducing token consumption. Comprehensive evaluations across six public NLP benchmarks and a curated dataset reveal consistent performance enhancements with our proposed methodology. Notably, ReWOO achieves 5x token efficiency and 4% accuracy improvement on HotpotQA, a multi-step reasoning benchmark. Furthermore, ReWOO demonstrates robustness under tool-failure scenarios. Beyond prompt efficiency, decoupling parametric modules from non-parametric tool calls enables instruction fine-tuning to offload LLMs into smaller language models, thus substantially reducing model parameters. Our illustrative work offloads reasoning ability from 175B GPT3.5 into 7B LLaMA, demonstrating the significant potential for truly efficient and scalable ALM systems.Keywords
All Related Versions
This publication has 10 references indexed in Scilit:
- ChatGPT for Robotics: Design Principles and Model AbilitiesIEEE Access, 2024
- TaskMatrix.AI: Completing Tasks by Connecting Foundation Models with Millions of APIsIntelligent Computing, 2024
- Code as Policies: Language Model Programs for Embodied Control2023
- Batch Prompting: Efficient Inference with Large Language Model APIs2023
- Prompt Compression and Contrastive Conditioning for Controllability and Toxicity Reduction in Language Models2022
- Did Aristotle Use a Laptop?A Question Answering Benchmark with Implicit Reasoning StrategiesTransactions of the Association for Computational Linguistics, 2021
- The Power of Scale for Parameter-Efficient Prompt Tuning2021
- HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering2018
- TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension2017
- Designing effective questions for classroom response system teachingAmerican Journal of Physics, 2006