Masking in Multi-hop QA: An Analysis of How Language Models Perform with Context Permutation

TL;DR

This paper analyzes how different language models perform on multi-hop question answering when the order of context documents is permuted, revealing model differences, the impact of document order, and improvements via attention modifications.

Contribution

It provides a comparative analysis of encoder-decoder and decoder-only models in multi-hop QA, explores the effects of document permutation, and proposes attention-based enhancements.

Findings

Encoder-decoder models outperform decoder-only models in MHQA.

Document order aligned with reasoning improves performance.

Bi-directional attention boosts causal decoder-only models.

Abstract

Multi-hop Question Answering (MHQA) adds layers of complexity to question answering, making it more challenging. When Language Models (LMs) are prompted with multiple search results, they are tasked not only with retrieving relevant information but also employing multi-hop reasoning across the information sources. Although LMs perform well on traditional question-answering tasks, the causal mask can hinder their capacity to reason across complex contexts. In this paper, we explore how LMs respond to multi-hop questions by permuting search results (retrieved documents) under various configurations. Our study reveals interesting findings as follows: 1) Encoder-decoder models, such as the ones in the Flan-T5 family, generally outperform causal decoder-only LMs in MHQA tasks, despite being significantly smaller in size; 2) altering the order of gold documents reveals distinct trends in both Flan T5 models and fine-tuned decoder-only models, with optimal performance observed when the document order aligns with the reasoning chain order; 3) enhancing causal decoder-only models with bi-directional attention by modifying the causal mask can effectively boost their end performance. In addition to the above, we conduct a thorough investigation of the distribution of LM attention weights in the context of MHQA. Our experiments reveal that attention weights tend to peak at higher values when the resulting answer is correct. We leverage this finding to heuristically improve LMs' performance on this task. Our code is publicly available at https://github.com/hwy9855/MultiHopQA-Reasoning.