mlsanityData to 7 Oct 2026

Decoupled Rationalization with Asymmetric Learning Rates: A Flexible Lipschitz Restraint

arXiv 2305.13599 ↗First posted 23 May 2023cs.LG cs.CLcited by 14

Not on the map. Why?

Abstract (arXiv)

A self-explaining rationalization model is generally constructed by a cooperative game where a generator selects the most human-intelligible pieces from the input text as rationales, followed by a predictor that makes predictions based on the selected rationales. However, such a cooperative game may incur the degeneration problem where the predictor overfits to the uninformative pieces generated by a not yet well-trained generator and in turn, leads the generator to converge to a sub-optimal model that tends to select senseless pieces. In this paper, we theoretically bridge degeneration with the predictor's Lipschitz continuity. Then, we empirically propose a simple but effective method named DR, which can naturally and flexibly restrain the Lipschitz constant of the predictor, to address the problem of degeneration. The main idea of DR is to decouple the generator and predictor to allocate them with asymmetric learning rates. A series of experiments conducted on two widely used benchmarks have verified the effectiveness of the proposed method. Codes: hrefhttps://github.com/jugechengzi/Rationalization-DRhttps://github.com/jugechengzi/Rationalization-DR.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Builds on 32 of its 45 references that are in the map. First posted on arXiv 23 May 2023.

Builds on

32 of 45 references.

References · 32

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2013Intriguing properties of neural networkscited by 5,114
  4. 2014Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translationcited by 4,016
  5. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,708
  6. 2017Improved Training of Wasserstein GANscited by 2,223
  7. 2018Spectral Normalization for Generative Adversarial Networkscited by 1,618
  8. 2018Lipschitz regularity of deep neural networks: analysis and efficient estimationcited by 413
  9. 2019Efficient and Accurate Estimation of Lipschitz Constants for Deep Neural Networkscited by 290
  10. 2016Rationalizing Neural Predictionscited by 281
  11. 2018Evaluating the Robustness of Neural Networks: An Extreme Value Theory Approachcited by 181
  12. 2020Invariant Rationalizationcited by 131
20 more
  1. 2019Interpretable Neural Predictions with Differentiable Binary Variablescited by 126
  2. 2020Learning to Faithfully Rationalize by Constructioncited by 113
  3. 2019Rethinking Cooperative Rationalization: Introspective Extraction and Complement Controlcited by 96
  4. 2020An Information Bottleneck Approach for Controlling Conciseness in Rationale Extractioncited by 69
  5. 2018Deriving Machine Attention from Human Rationalescited by 54
  6. 2022Rethinking Lipschitz Neural Networks and Certified Robustness: A Boolean Function Perspectivecited by 48
  7. 2019Bi-level Actor-Critic for Multi-agent Coordinationcited by 38
  8. 2021Understanding Interlocking Dynamics of Cooperative Rationalizationcited by 36
  9. 2022Can Rationalization Improve Robustness?cited by 36
  10. 2019A Game Theoretic Approach to Class-wise Selective Rationalizationcited by 32
  11. 2021UNIREX: A Unified Learning Framework for Language Model Rationale Extractioncited by 32
  12. 2019Cooperative Learning of Disjoint Syntax and Semanticscited by 27
  13. 2021Gradient Normalization for Generative Adversarial Networkscited by 25
  14. 2019Lipschitz Generative Adversarial Netscited by 24
  15. 2022Learning to Scaffold: Optimizing Model Explanations for Teachingcited by 20
  16. 2022FR: Folded Rationalization with a Unified Encodercited by 18
  17. 2021The Irrationality of Neural Rationale Modelscited by 16
  18. 2021Distribution Matching for Rationalizationcited by 14
  19. 2023MGR: Multi-generator Based Rationalizationcited by 12
  20. 2022Making a (Counterfactual) Difference One Rationale at a Timecited by 10

Built on by

Cited by year

Citing papers per year: 2023: 2; 2024: 7; 2025: 5.

14 papers.

Sort:

  1. 2023D-Separation for Causal Self-Explanationcited by 21
  2. 2024Multi-Level Explanations for Generative Language Modelscited by 11
  3. 2024Is the MMI Criterion Necessary for Interpretability? Degenerating Non-causal Features to Plain Noise for Self-Rationalizationcited by 11
  4. 2025Breaking Free from MMI: A New Frontier in Rationalization by Probing Input Utilizationcited by 6
  5. 2025Adversarial Cooperative Rationalization: The Risk of Spurious Correlations in Even Clean Datasetscited by 6
  6. 2024MARE: Multi-Aspect Rationale Extractor on Unsupervised Rationale Extractioncited by 4
  7. 2023PDE+: Enhancing Generalization via PDE with Adaptive Distributional Diffusioncited by 3
  8. 2024CAVE: Controllable Authorship Verification Explanationscited by 3
  9. 2024Graph-Guided Textual Explanation Generation Frameworkcited by 3
  10. 2025Boosting Explainability through Selective Rationalization in Pre-trained Language Modelscited by 2
  11. 2024Adversarial Attack for Explanation Robustness of Rationalization Modelscited by 1
  12. 2025Learnable Game-theoretic Policy Optimization for Data-centric Self-explanation Rationalizationcited by 1
2 more
  1. 2024Interlocking-free Selective Rationalization Through Genetic-based Learningcited by 0
  2. 2025Enhancing Graph Neural Networks: A Mutual Learning Approachcited by 0