A Logic-Driven Framework for Consistency of Neural Models

By Tao Li, Vivek Gupta, Maitrey Mehta, Vivek Srikumar. 2019

120 citations recorded by the source corpus.

While neural models show remarkable accuracy on individual predictions, their internal beliefs can be inconsistent across examples. In this paper, we formalize such inconsistency as a generalization of prediction error. We propose a learning framework for constraining models using logic rules to regularize them away from inconsistency. Our framework can leverage both labeled and unlabeled examples and is directly compatible with off-the-shelf learning schemes without model redesign. We instantiate our framework on natural language inference, where experiments show that enforcing invariants stated in logic can help make the predictions of neural models both accurate and consistent.

Subfields: Artificial Intelligence, Safety, Robustness & Interpretability, Knowledge Graphs & Representation, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Ashish Sabharwal
2. Peter Clark
3. Tushar Khot
4. Hannaneh Hajishirzi
5. Kyle Richardson
6. Daniel Khashabi
7. Dan Roth
8. Mohit Bansal

In Safety, Robustness & Interpretability

1. Hannaneh Hajishirzi
2. Peter Clark
3. Oyvind Tafjord
4. Kyle Richardson
5. Hinrich Schütze
6. Dan Roth
7. Tushar Khot
8. Akari Asai

In Knowledge Graphs & Representation

1. Hinrich Schütze
2. Dan Roth
3. Muhao Chen
4. Tushar Khot
5. Hannaneh Hajishirzi
6. Oyvind Tafjord
7. Akari Asai
8. Kyle Richardson

In Signal Processing

1. Qiang Ning
2. Dan Roth
3. Hinrich Schütze
4. Hannaneh Hajishirzi
5. Oyvind Tafjord
6. Akari Asai
7. Peter Clark
8. Nora Kassner

Citations in the corpus

• Logic-Guided Data Augmentation and Regularization for Consistent Question Answering (2020, methodology): “Qualitative Analysis Table 4 shows qualitative examples, comparing our method with RoBERTa baseline.”
• Joint Constrained Learning for Event-Event Relation Extraction (2020, methodology): “Motivated by the logic-driven framework proposed by Li et al. (2019), the declarative logical constraints are converted into differentiable functions that can be incorporated into the learning objective for relation extraction tasks.”
• Enriching a Model's Notion of Belief using a Persistent Memory (2021, methodology): “Prior work on reducing inconsistency has focused on retraining the model itself to be more consistent in its answers, e.g., (Ribeiro et al., 2019; Li et al., 2019), but with imperfect results.”
• BeliefBank: Adding Memory to a Pre-Trained Language Model for a Systematic Notion of Belief (2021, methodology): “Most significantly, we find that improving consistency in this way improves accuracy, while earlier finetuning-based approaches report either no accuracy gains (Ribeiro et al., 2019; Minervini and Riedel, 2018; Li et al., 2019) or only slight gains (Asai and Hajishirzi, 2020).”
• Temporal Reasoning on Implicit Events from Distant Supervision (2020, background): “, 2019) and other decomposition-based approaches (Talmor and Berant, 2018; Khashabi et al., 2018; Li et al., 2019; Wolfson et al., 2020; Khot et al., 2021).”