# Self-Assembling Modular Networks for Interpretable Multi-Hop Reasoning

arXiv: https://arxiv.org/abs/1909.05803 · HTML: https://mlsanity.com/paper/1909.05803/

## Key facts

- Cited by 28 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 12 Sep 2019.

## Built on by (most cited first, 28 of 28)

- [Decomposed Prompting: A Modular Approach for Solving Complex Tasks](https://mlsanity.com/paper/2210.02406/) (2022)
- [ADaPT: As-Needed Decomposition and Planning with Language Models](https://mlsanity.com/paper/2311.05772/) (2023)
- [Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks](https://mlsanity.com/paper/2207.13243/) (2022)
- [Soft Merging of Experts with Adaptive Routing](https://mlsanity.com/paper/2306.03745/) (2023)
- [SelfExplain: A Self-Explaining Architecture for Neural Text Classifiers](https://mlsanity.com/paper/2103.12279/) (2021)
- [Reason first, then respond: Modular Generation for Knowledge-infused Dialogue](https://mlsanity.com/paper/2111.05204/) (2021)
- [Interpreting Deep Learning Models in Natural Language Processing: A Review](https://mlsanity.com/paper/2110.10470/) (2021)
- [OpenViDial 2.0: A Larger-Scale, Open-Domain Dialogue Generation Dataset with Visual Contexts](https://mlsanity.com/paper/2109.12761/) (2021)
- [Enhancing Ethical Explanations of Large Language Models through Iterative Symbolic Refinement](https://mlsanity.com/paper/2402.00745/) (2024)
- [Summarization Programs: Interpretable Abstractive Summarization with Neural Modular Trees](https://mlsanity.com/paper/2209.10492/) (2022)
- [Hopper: Multi-hop Transformer for Spatiotemporal Reasoning](https://mlsanity.com/paper/2103.10574/) (2021)
- [VGNMN: Video-grounded Neural Module Network to Video-Grounded Language Tasks](https://mlsanity.com/paper/2104.07921/) (2021)
- [MURMUR: Modular Multi-Step Reasoning for Semi-Structured Data-to-Text Generation](https://mlsanity.com/paper/2212.08607/) (2022)
- [Locate Then Ask: Interpretable Stepwise Reasoning for Multi-hop Question Answering](https://mlsanity.com/paper/2208.10297/) (2022)
- [Coarse-grained decomposition and fine-grained interaction for multi-hop question answering](https://mlsanity.com/paper/2101.05988/) (2021)
- [Explaining the Road Not Taken](https://mlsanity.com/paper/2103.14973/) (2021)
- [Towards Interpretable and Reliable Reading Comprehension: A Pipeline Model with Unanswerability Prediction](https://mlsanity.com/paper/2111.09029/) (2021)
- [Better Retrieval May Not Lead to Better Question Answering](https://mlsanity.com/paper/2205.03685/) (2022)
- [Learning Label Modular Prompts for Text Classification in the Wild](https://mlsanity.com/paper/2211.17142/) (2022)
- [Divide & Conquer for Entailment-aware Multi-hop Evidence Retrieval](https://mlsanity.com/paper/2311.02616/) (2023)
- [Ask to Understand: Question Generation for Multi-hop Question Answering](https://mlsanity.com/paper/2203.09073/) (2022)
- [Analyzing the Effectiveness of the Underlying Reasoning Tasks in Multi-hop Question Answering](https://mlsanity.com/paper/2302.05963/) (2023)
- [A Comprehensive Survey on Multi-hop Machine Reading Comprehension Datasets and Metrics](https://mlsanity.com/paper/2212.04070/) (2022)
- [A Comprehensive Survey on Multi-hop Machine Reading Comprehension Approaches](https://mlsanity.com/paper/2212.04072/) (2022)
- [Dynamic MOdularized Reasoning for Compositional Structured Explanation Generation](https://mlsanity.com/paper/2309.07624/) (2023)
- [AS-ES Learning: Towards Efficient CoT Learning in Small Models](https://mlsanity.com/paper/2403.01969/) (2024)
- [Prediction or Comparison: Toward Interpretable Qualitative Reasoning](https://mlsanity.com/paper/2106.02399/) (2021)
- [Perturbation-based Active Learning for Question Answering](https://mlsanity.com/paper/2311.02345/) (2023)

## Abstract (arXiv)

Multi-hop QA requires a model to connect multiple pieces of evidence scattered in a long context to answer the question. The recently proposed HotpotQA (Yang et al., 2018) dataset is comprised of questions embodying four different multi-hop reasoning paradigms (two bridge entity setups, checking multiple properties, and comparing two entities), making it challenging for a single neural network to handle all four. In this work, we present an interpretable, controller-based Self-Assembling Neural Modular Network (Hu et al., 2017, 2018) for multi-hop reasoning, where we design four novel modules (Find, Relocate, Compare, NoOp) to perform unique types of language reasoning. Based on a question, our layout controller RNN dynamically infers a series of reasoning modules to construct the entire network. Empirically, we show that our dynamic, multi-hop modular network achieves significant improvements over the static, single-hop baseline (on both regular and adversarial evaluation). We further demonstrate the interpretability of our model via three analyses. First, the controller can softly decompose the multi-hop question into multiple single-hop sub-questions to promote compositional reasoning behavior of the main network. Second, the controller can predict layouts that conform to the layouts designed by human experts. Finally, the intermediate module can infer the entity that connects two distantly-located supporting facts by addressing the sub-question from the controller.