Triggering Multi-Hop Reasoning for Question Answering in Language Models using Soft Prompts and Random Walks
This addresses a key limitation in language models for question answering, though it is incremental as it builds on existing methods like T5 and knowledge graphs.
The paper tackles the problem of language models struggling with multi-hop reasoning in question answering by using soft prompts and random walks over knowledge graphs, resulting in substantial improvements on 2-hop reasoning tasks over standard tuning approaches.
Despite readily memorizing world knowledge about entities, pre-trained language models (LMs) struggle to compose together two or more facts to perform multi-hop reasoning in question-answering tasks. In this work, we propose techniques that improve upon this limitation by relying on random walks over structured knowledge graphs. Specifically, we use soft prompts to guide LMs to chain together their encoded knowledge by learning to map multi-hop questions to random walk paths that lead to the answer. Applying our methods on two T5 LMs shows substantial improvements over standard tuning approaches in answering questions that require 2-hop reasoning.